Report No: ACS23224
.
Democratic Socialist Republic of Sri Lanka
Non-communicable disease burden in the
Western Province, Sri Lanka
.
October 2017
.
GHN06
SOUTH ASIA
.
.
.
This volume is a product of the staff of the International Bank for Reconstruction and Development/ The World
Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views
of the Executive Directors of The World Bank or the governments they represent. The World Bank does not
guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other
information shown on any map in this work do not imply any judgment on the part of The World Bank
concerning the legal status of any territory or the endorsement or acceptance of such boundaries.
.
Copyright Statement:
.
The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without
permission may be a violation of applicable law. The International Bank for Reconstruction and Development/
The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions
of the work promptly.
For permission to photocopy or reprint any part of this work, please send a request with complete information
to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-
8400, fax 978-750-4470, http://www.copyright.com/.
All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the
Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail
pubrights@worldbank.org.
ii
Non-communicable Disease Burden in the Western
Province, Sri Lanka
Health Nutrition and Population Global Practice
October, 2017
iii
Table of Contents
Acknowledgements......................................................................................................................... viii
Acronyms ......................................................................................................................................... ix
Executive Summary............................................................................................................................1
Chapter 1. Introduction .....................................................................................................................9
1.1. The Context ................................................................................................................................... 9
1.2. Objectives.................................................................................................................................... 13
1.3. Data and Methods ...................................................................................................................... 14
Chapter 2. Health Status and Physiological Risk Factors .................................................................... 19
2.1 Prevalence and Distribution of Self-reported and Diagnosed NCDs .......................................... 20
2.2. Hypertension (Elevated Blood Pressure) .................................................................................... 24
2.3. Obesity ........................................................................................................................................ 30
2.4. Obesity as a Risk Factor for Diabetes and Hypertension ............................................................ 37
Chapter 3. Behavioral and Environmental Risk Factors ..................................................................... 40
3.1. Behavioral Risk Factors ............................................................................................................... 40
3.2. Environmental Risk Factors......................................................................................................... 47
Chapter 4. Use of Health Services .................................................................................................... 54
4.1. Use of Healthcare........................................................................................................................ 54
4.2. Use of Healthcare by Source of Care .......................................................................................... 57
4.3. Perceived Gaps in the Package and Quality of Health Services .................................................. 63
Chapter 5. Out-of-pocket Health Expenditures ................................................................................. 67
5.1. Health Costs and Out-of-pocket Payments: Magnitude and Composition................................. 68
5.2. Factors Associated with OOP Spending ...................................................................................... 72
Chapter 6. An Overview of the Health System of Sri Lanka ............................................................... 76
Chapter 7. Conclusion and Recommendations.................................................................................. 82
7.1. Summary and Conclusion ........................................................................................................... 82
7.2. Recommendations ...................................................................................................................... 84
References....................................................................................................................................... 98
iv
List of Boxes:
Box 1.1. Using the SWIFT Approach to Predict Per Capita Household Consumption ................................ 18
Box 3.1. Premature Deaths from Air Pollution in Sri Lanka ........................................................................ 48
Box 5.1. Estimation technique for OOP Payments………………………………………………………………………………….73
Box 7.1. Costa Rica’s EBAIS Primary Care Model ........................................................................................ 94
List of Figures:
Figure 1.1. Global Comparison of Infant and Under-5 Mortality Rates ........................................................ 9
Figure 1.2. Causes of Premature Mortality in Sri Lanka.............................................................................. 10
Figure 1.3. Age Composition in Sri Lanka in 2015 and 2050 ....................................................................... 11
Figure 1.4. Population Aged Over 65 ……………………………………………………………………………………………….. 12
Figure 1.5. Urban Population ...................................................................................................................... 12
Figure 1.6. Lifetime Internal Migration by Province ................................................................................... 13
Figure 2.1. Prevalence of Self-reported NCDs ............................................................................................ 20
Figure 2.2. Prevalence of Diagnosed Chronic NCDs .................................................................................... 21
Figure 2.3. Prevalence of Diabetes in South Asia ....................................................................................... 21
Figure 2.4. Self-reported NCDs by Economic Status and Level of Education ............................................. 22
Figure 2.5. Diagnosed NCDs by Education Level ......................................................................................... 23
Figure 2.6. Prevalence of Diagnosed NCDs across Age Groups .................................................................. 23
Figure 2.7. Prevalence of Self-reported and Diagnosed NCDs by Gender .................................................. 24
Figure 2.8. Observed Hypertension (% of adults) ....................................................................................... 25
Figure 2.9. Observed Hypertension by Age Groups .................................................................................... 25
Figure 2.10. Observed Hypertension by Education Level and Economic Status ........................................ 26
Figure 2.11. Diagnosed versus Observed Prevalence of Hypertension ...................................................... 27
Figure 2.12. Diagnosis, Treatment, and Control of Hypertension .............................................................. 28
Figure 2.13. Comparing the Western Province with the Average for Sri Lanka ......................................... 32
Figure 2.14. Abdominal Obesity Distribution across BMI Categories ......................................................... 33
Figure 2.15. General and Abdominal Obesity (% of women and men) ...................................................... 34
Figure 2.16. Obesity by Age Groups............................................................................................................ 35
Figure 2.17. Obesity by Economic Status .................................................................................................... 35
Figure 2.18. Underweight by Economic Status ........................................................................................... 36
Figure 2.19. Obesity as a Risk Factor for Diabetes ...................................................................................... 37
Figure 2.20. Obesity as a Risk Factor for Hypertension .............................................................................. 38
Figure 3.1. Daily Smoking Prevalence and Age of Initiation of Daily Smoking by Age Cohort.................... 41
Figure 3.2. Smoking Prevalence by Economic Status and Education Level (men only) .............................. 42
Figure 3.3. Prevalence of Betel Chewing by Gender
Figure 3.4. Chewing Frequency ................................................................................................................... 44
Figure 3.5. Betel Chewing Prevalence by Socioeconomic Groups .............................................................. 45
Figure 3.6. Daily Betel Chewing by Age Groups .......................................................................................... 45
v
Figure 3.7. Daily Betel Chewing in Rural and Urban Area……………………………………………………………………….45
Figure 3.8. Multiple Behavioral Risk Factors (%) ........................................................................................ 46
Figure 3.9. Hypertension Prevalence by Behavioral Risk Factors ............................................................... 46
Figure 3.10. Risk from Outdoor Sources of Pollution ................................................................................. 47
Figure 3.11. Primary Sources of Fuel in the Western Province .................................................................. 50
Figure 3.12. Use of Unclean Fuels (Biomass or Kerosene) by Households’ Economic Status .................... 51
Figure 3.13. Risky Practices among Households Who Mainly Use Biomass ............................................... 52
Figure 3.14. Percentage of Households with No Functional Chimney Using Biomass ............................... 52
Figure 4.1. Use of Outpatient and Inpatient Healthcare ............................................................................ 55
Figure 4.2. Source of Outpatient Care by Households’ Economic Status ................................................... 56
Figure 4.3. Use of Healthcare by Gender .................................................................................................... 56
Figure 4.4. Those Sought Private Outpatient Care as a Percentage of All Who Used Outpatient Care ..... 58
Figure 4.5. Those Chose Private Outpatient Care as a Percentage of All Sought Outpatient Care ............ 59
Figure 4.6. Preventive MCH and Adult Services.......................................................................................... 59
Figure 4.7. Curative MCH and Adult Services ............................................................................................. 60
Figure 4.8. Level of Care Chosen for Preventive MCH and Adult Health Services ...................................... 61
Figure 4.9. Level of Care Chosen for Curative MCH and Adult Health Services ......................................... 62
Figure 4.10. Demographic Profile of Healthcare Users by Different Levels of Care .................................. 63
Figure 5.1. Out-of-pocket Payments as a Percentage of Total Health Expenditures ................................. 68
Figure 5.2. Health Costs, Health Expenditures, and OOPP for All Who Received Care .............................. 69
Figure 5.3. Share of Reimbursements in Total Health Expenditure ........................................................... 69
Figure 5.4. OOPP Among Entire Sample ..................................................................................................... 70
Figure 5.5. OOPP by Type and Source of Care ............................................................................................ 70
Figure 5.6. Composition of Health Expenditures by Type and Source of Care ........................................... 71
Figure 5.7. Sources of Finance for Healthcare ............................................................................................ 72
Figure 5.8. OOPP by Households’ Economic Status .................................................................................... 72
Figure 5.9. OOPP by Age, Physiological Risk Factors, and Diabetes ........................................................... 74
Figure 5.10. OOPP by Urban or Rural Location ........................................................................................... 75
Figure 6.1 Hospital Beds in Sri Lanka, 2015 ................................................................................................ 77
Figure 6.2. Organization of Sri Lanka’s Health System ............................................................................... 78
Figure 6.3. Numbers of Preventive and Curative Government Healthcare Facilities 2011-2015............... 79
Figure 7.1. Chile’s Front of Package Warning Food Labeling……………………………………………………………………88
Figure 7.2. Example of plain packaging for tobacco products……………………………………………………………….. 90
List of Tables
Table 1.1. Cut-off Points for Classifying Hypertension Status .................................................................... 16
Table 1.2. BMI, WC and WHR classification ................................................................................................ 17
vi
Table 2.1. Distribution of Diagnosed and Observed Hypertension ............................................................ 29
Table 2.2. Prevalence of General and Abdominal Obesity ......................................................................... 31
Table 2.3. Diet and Physical Activity in Sri Lanka (Age 18-69) .................................................................... 31
Table 3.1. Current Smoking Prevalence ...................................................................................................... 41
Table 3.2. Prevalence of Alcohol Use, Binge Drinking, and Excessive Alcohol Use .................................... 43
Table 3.3. Indoor PM2.5 Concentrations under Different Conditions ......................................................... 50
Table 4.1. Perceived Gaps in Public Health Service Quality and Completeness ......................................... 64
Table 6.1. Key Health Personnel in Sri Lanka, 2015 .................................................................................... 77
Table 7.1. Effectiveness of fiscal policies on diet………………………………………………………………………………….. 86
Table 7.2. Examples of taxes on drinks and foods in other countries……………………………………………………..87
Table 7.3. Examples of countries with legislation on salt reduction……………………………………………………….89
Table 7.4. Interventions on Diet and Physical Activity: Summary Results from a Systematic Review……..95
List of Annex Tables
Table A1. Distribution of Most Prevalent NCDs Across Socioeconomic Groups ...................................... 105
Table A2. Probit (Marginal Effects) for the Probability of Self-reported and Diagnosed NCDs ............... 106
Table A3. Probit (Marginal Effects) for the Probability of Self-reported and Diagnosed NCDs ............... 107
Table A4. Distribution of Observed Hypertension .................................................................................... 108
Table A5. Probit for Observed Hypertension and Awareness of Being Hypertensive .............................. 109
Table A6. Distribution of Obesity and Underweight Across Population Groups ...................................... 111
Table A7. Probit for Obesity and Underweight ......................................................................................... 112
Table A8. Correlates of Diagnosed Diabetes............................................................................................. 114
Table A9. Correlates of Observed Hypertension (probit) ......................................................................... 115
Table A10. Distribution of Current Smoking Among Groups (both sex) ................................................... 116
Table A11. Correlates of Daily Smoking .................................................................................................... 117
Table A12. Socioeconomic and Demographic Correlates of Excessive Drinking ...................................... 118
Table A 13. Distribution of Betel Chewing ................................................................................................ 119
Table A14. Socioeconomic and Demographic Correlates of Betel Chewing............................................. 120
Table A15. Households’ Primary Source of Fuel ....................................................................................... 121
Table A16. Risky Cooking Practices ........................................................................................................... 122
Table A17. Hygiene, Sanitation, and Working Conditions (percentage of households) ........................... 123
Table A18. Correlates of Outpatient Healthcare Use (by source of care) ................................................ 124
Table A19. Correlates of Healthcare Use .................................................................................................. 125
Table A20. Correlates of Healthcare Use (adult sample) .......................................................................... 126
Table A21. Correlates of Outpatient Healthcare Use by Source of Care (adult sample) .......................... 128
Table A22. Probability of Using Private Care Among Those who Used Care during Reference Period .... 130
Table A23. Costs of Healthcare ................................................................................................................. 131
Table A24. Distribution of OOP Spending ................................................................................................. 132
Table A25. Predictors of OOP Spending.................................................................................................... 133
Table A26. Predictors of OOP Spending (adult sample) ........................................................................... 134
vii
Acknowledgements
This report was prepared by the World Bank.
This task was led by Preeti Kudesia (Senior Health Specialist) and prepared by Zelalem Yilma Debebe
(Health Economist) with inputs from Owen Smith (Senior Economist), Deepika Eranjanie Attygalle (Senior
Health Specialist) and Iryna Postolovska (Young Professional). Mickey Chopra (Global Lead, Service
Delivery), Federica Secci (Health Specialist), David Locke Newhouse (Senior Economist), Yi-Kyoung Lee
(Senior Health Specialist), Phoebe M. Folger (Operations Officer) and Martha P. Vargas (Program
Assistant) provided valuable inputs. Editorial support was provided by Fiona Mackintosh (Consultant)
and Minh Thi Hoang Trinh (Program Assistant).
The initial study design, data collection and formulation of survey report was led by Kumari Vinodhani
Navaratne (Senior Health Specialist) and supported by Dr. Champika Wikramasinghe; Dr. Sumal
Nandasena and Nayana Fernando. Indika Samarakoon (Consultant) prepared draft background papers for
this study. The quantitative and qualitative surveys were conducted by Sri Lanka Business Development
Center and Quantum Consumer Solutions. Dung Doan (Consultant) helped identify the bottom 40 percent
households using the SWIFT approach.
The peer reviewers for this study were Son Nam Nguyen (Lead Health Specialist) and Sameh El-Saharty
(Program Leader). This task was prepared under the guidance of Francoise Clottes (Country Director
2013-2016, Sri Lanka), Idah Z. Pswarayi-Riddihough (current Country Director, Sri Lanka) and Rekha
Menon (Practice Manager, Health, Nutrition and Population, South Asia).
The report also benefited from information made available by several institutions including the Urban
Development Authority, Central Environment Authority; and Urban and Estate Health Unit of the Ministry
of Health and their field staff.
We recognize and appreciate the support and cooperation extended by the personnel of Ministry of
Health & Nutrition, Office of the Provincial Director of Health Services - Western Province,
Relevant Divisional Secretariat Offices, Ministry of Health Offices, National Institute of Health Sciences
and Grama Niladhari offices during the data collection process.
viii
Acronyms
BMI Body Mass Index
CVD Cardiovascular Diseases
DALY Disability-Adjusted Life Years
GDP Gross Domestic Product.
GHNDR Health, Nutrition and Population Global Practice
GND Grama Niladhari Divisions
HIES Household Income and Expenditure Survey
IFD International Federation of Diabetes
IHD Ischemic Heart Disease
LKR Sri Lankan Rupee
MCH Maternal and Child Health Care
NCD Non-Communicable Diseases
OOP Out-Of-Pocket
STEPS Stepwise Approach to Surveillance
SWIFT Survey of Wellbeing Via Instant and Frequent Tracking
UHC Universal Health Coverage
UN United Nations
USA United States of America
WC Waist Circumference
WDI World Development Indicator
WHO World Health Organization
WHR Waist-Hip Ratio
ix
Executive Summary
Why this study?
Having achieved impressive maternal and child health outcomes, Sri Lanka’s main health s ector
challenge is increasingly becoming the need to address non-communicable diseases (NCDs). Its
population is aging at a faster rate than the average for lower-middle-income countries and the average
for South Asia. According to the latest Global Burden of Disease estimates (IHME, 2017), NCDs account
for 81 percent of total deaths in the country. The increasing importance of NCDs is also evident from the
major causes of disability-adjusted life years (DALYs), a summary measure of years of life lost to death and
disease. The trend in DALYs between 1990 and 2016 shows that DALYs resulting from ischemic heart
disease, chronic obstructive pulmonary disease, and particularly diabetes mellitus are on the rise. The
share of DALYs attributable to diabetes mellitus and chronic obstructive pulmonary disease increased by
51 and 25 percent, respectively between 2005 and 2016 (IHME, 2017). Monitoring the prevalence and
distribution of these NCDs and their risk factors is key to detecting vulnerable groups and providing early
treatment.
The primary objectives of this study are: i) to examine the prevalence and distribution of NCDs and risk
factors across socioeconomic and demographic groups, and ii) to assess the performance of the health
system with regards to NCDs. The focus is on adults in the Western province of Sri Lanka. The Western
province is selected because it is the most urbanized and increasingly urbanizing province that is likely to
be home to a growing share of Sri Lanka’s aging population.
Data and Methods
The study mainly draws on household survey data collected specifically for the purposes of this study
in 2015. In addition, we examine qualitative data on health seeking behavior and data on indoor air
quality from 50 households. The quantitative household survey consisted of 10,107 people from 3,300
households, about 64 percent of whom were adults above the age of 20. Detailed information on health
status, healthcare utilization and out-of-pocket payments was collected from these individuals. We
combined three sets of data to measure the prevalence of NCDs/risk factors among individuals in the
province: i) self-reported, ii) diagnosed, and iii) observed/measured. The latter included measurement of
blood pressure, body mass index (BMI), waist circumference (WC) and waist-hip ratio (WHR). We
compared the data on diagnoses and on observed health status (in the case of hypertension) to
understand people’s knowledge of their own health status, the characteristics of those who are not aware
of their status, and the ability of patients to manage chronic conditions.
Key Findings
The findings show that the most common NCDs (diagnosed) posing a threat to healthy adult life in the
province include hypertension, diabetes, cataracts, ischemic heart disease, and asthma. The less
educated appear to have a higher burden of these conditions, but there is no systematic difference by
economic status. More concerning is the fact that the onset of these conditions is early, suggesting that
young adults live with health conditions for a substantial part of their lives. A special inquiry into two
physiological risk factors (hypertension and obesity) for various NCDs revealed substantial concerns.
1
First, more than one in four adults (26 percent) are hypertensive, but as many as 70 percent of them
are unaware of their status. Equally concerning is the finding that, of the 15 percent of adults who were
once diagnosed with hypertension, only less than half (7 percent) were found to have it under control (i.e.
systolic blood pressure ≤ 140 mmHg), suggesting a gap in managing the disease. The lack of awareness
about one’s hypertensive status is common across both genders, all places of residence, and all
socioeconomic groups but is more extensive among some groups than others. Despite having a higher
probability of being hypertensive, men are less likely to be aware of their hypertensive status, suggesting
that use of preventive care differs by gender. While hypertension does not seem to be associated with
socioeconomic status, the poorest and richest quintiles are more likely to be aware of their conditions
than the middle 60 percent. As expected, groups less aware of their status are those that have the least
contact with the health system (as suggested by the low utilization of outpatient care). For example, men
are less likely to seek outpatient care than women.1 Men and relatively younger people also seem to have
the lowest rates of preventive care utilization.
Diagnosis, Treatment, and Control of Hypertension
30% 26%
20% 15%
11%
10% 7%
0%
Observed Diagnosed Medically treated Controlled
hypertension hypertension hypertension hypertension
Second, the prevalence of general obesity in the province (15 percent) is substantially higher than the
average for Sri Lanka (6 percent). Combined with the prevalence of abdominal obesity2 (57 percent),
this poses a significant risk for NCDs. Obesity disproportionately affects women and adults between 30
and 60 years of age. Although the rate of obesity is the highest among the richest quintile, it is alarmingly
high even among the poor. Consistent with evidence from other countries, obesity in the province was
found to be significantly associated with the risk of hypertension and diabetes.
1
Men, however, are more likely to seek inpatient care, which may be a result of choosing to forgo preventive
outpatient care.
2
Abdominal obesity is based on waist-circumference.
2
Obesity by Economic Status
Poorest quintile 2nd poorest quintile Middle quintile 2nd richest quintile Richest
70% 65%
58%
52% 54% 56%
50%
30%
14% 14% 13% 14% 16%
10%
-10% General obesity (BMI≥30) WC > IFD cut-off
Moreover, taking public health actions solely based on BMI-based risk classifications excludes a large
proportion of adults, especially women, who are at a substantial risk of metabolic complications. Waist
circumference appears to predict the risk of diabetes better than BMI-based obesity. A significant
proportion of people in the normal BMI weight category is at risk of complications related to abdominal
obesity. Given that the health risk of a given level of body fat in the Asian population is much higher than
in other regions, policymakers need to give this issue a considerable amount of attention.
Abdominal Obesity Distribution Across BMI Categories (WC cut-off)
100% 96%
83%
80%
60%
35%
40%
20% 9%
0%
Underweight Normal Overweight Obese (I,II,III)
Behavioral risk factors such as smoking, betel chewing, and harmful alcohol consumption are generally
not very widespread, but their concentration among older men and lower socioeconomic groups should
be of concern for policymakers. The clustering of these risk factors among these vulnerable groups may
worsen the existing economic disadvantages for such groups. Furthermore, the risk of hypertension is
higher among those who practice these habits, which suggests that the potential health benefits of
expanding preventive counseling services, especially to men, are large.
Environmental risk factors, particularly indoor air pollution, also increase the vulnerability of lower
socioeconomic groups and those living in the rural parts of the province. About 47 percent of households
in the Western province use unclean sources of fuel (43 percent use biomass and 4 percent use kerosene).
Our readings of indoor air pollution in 50 households showed levels of pollution that were substantially
3
higher than WHO’s interim target-1. Indoor air pollution is a greater risk factor for the health of poorer
households, as they are more likely to depend on biomass for fuel and to have no functional chimney.
Almost one-third of households in the poorest quintile cook with biomass but have no functional chimney,
compared to only 4 percent of households in the richest quintile. Children from households that primarily
use unclean sources of fuel are about 2 percentage points more likely (than children from households that
use clean sources of fuel) to have symptoms of wheezing and whistling in the chest.
The poorer populations tend to use public facilities, while private facilities are largely used by the
better-off, mainly for reasons of convenience and shorter waiting times. Among those who used some
outpatient or inpatient care, the probability of choosing private care rather than public care is higher for
the better-off, the better educated, and the elderly. Analysis of health-seeking behavior suggests that
much of this preference is driven by differences in waiting times and the soft skills of health providers
between public and private providers of care rather than by any major differences in infrastructure,
amenities, or perceived clinical quality. Quantitative data, however, also show that individuals prefer to
use public facilities for more serious healthcare concerns.
The data also revealed that a high proportion of people seeking care bypass primary care facilities in
favor of higher-level services. This is particularly true for adult preventive services but is rarer for
preventive MCH services. Consumers also often bypass primary care facilities for curative needs that could
be met at the primary level. For example, 63 percent of respondents reported that they would go to a
secondary facility for curative care for chronic conditions, while only 23 percent responded that they
would seek care from primary facilities. Furthermore, a significant proportion of households did not know
where they could seek care for adult preventive health services, such as counseling on nutrition. This an
issue on which policymakers need to focus.
Out-of-pocket (OOP) payments are modest given Sri Lanka’s per capita income, but ensuring this does
not escalate will require public health actions to control NCDs. Most out-of-pocket spending in the
province is for private care, but part of it is likely due to the unavailability of laboratory tests and
medications in public facilities, forcing patients to use private laboratories and pharmacies. Considering
the increasing prevalence of NCDs and physiological risk factors that are associated with higher healthcare
use and spending, containing OOP spending will require reducing the prevalence of these risk factors and
removing the quality constraints that drive people away from public care, especially for routine
conditions. Early diagnoses and management of risk factors can reduce an otherwise costly treatment of
NCDs (both for individuals and the government).
The curative side of the public health system is not well suited to deal with the overwhelming burden
of non-communicable diseases. Primary level facilities provide facility based episodic NCD care, and they
do not routinely initiate or coordinate such care. A culture of self-referral and lack of an effective
gatekeeping mechanism produce discontinuity in client information between providers and constrain the
doctor-patient relationship. The ability to choose doctors appears to be an important factor driving
patients to utilize private facilities. Given that NCDs require long-term integrated care, however, a strong
doctor-patient relationship is crucial for the effectiveness of treatment. In the current system, NCD patients
cannot be tracked as they receive care from different facilities at different times, oftentimes resulting in lack
of continuity of care. Tracking these patients is further complicated by the absence of an electronic health
information system.
4
Recommendations
While the analyses were based on the Western province, recommendations are made for the country as
a whole. There are two reasons for this. First, the focus on the Western province is a strategic one. Given
the rate of urbanization and the relative homogeneity and size of the country, it can be argued that other
provinces will follow the trends exhibited by the Western province. As other provinces urbanize and their
socio-economic conditions and life styles change, they will be faced with similar challenges that the
Western province is facing now. Therefore, reorienting the health system in anticipation of these
challenges could lead to prevention of risk factors and early detection of NCDs, which would result in
significant health gains nationwide. Second, several of the recommendations are systemic and
institutional, and therefore would apply to the whole country.
A multi-pronged approach, consisting of multi-sectoral preventive interventions, health system
reorientation and strengthening, and a targeted approach aimed at those most vulnerable to NCDs and
NCD risk-factors, is required to address the challenges posed by the behavioral, physiological, and
environmental risk factors for NCDs, as identified in this study. Vulnerable populations include men,
people with multiple risk factors, and the poor (who suffer more from smoking, betel chewing, indoors
pollution, etc.).
Even though the health sector bears most of the burden of prevention and treatment of NCDs, most
interventions that could create health promoting environments lie outside the health sector.
Acknowledging this, Sri Lanka has recently approved a National Multi-sectoral Action Plan for the
prevention and control of NCDs (2016-2020) focusing on the following four strategic areas: i) leadership,
advocacy, and partnership; ii) health promotion and risk reduction; iii) reorientation of the health system
for early detection and management of NCDs and risk factors; and iv) surveillance, monitoring and
evaluation, and research. The recommendations listed below are consistent with this Action Plan and are
based on the findings presented in this study.
A. Interventions to control risk factors and prevent the onset of NCDs
I) Introducing and expanding population-based interventions: Population-based
interventions, such as community-wide campaigns and national NCD literacy campaigns
together with regulations and corporate social responsibility, can effectively reduce the trend
in unhealthy aging populations. Such interventions are key for primary prevention of NCDs
and are affordable even in low income settings. They do not require health system
strengthening and have a low cost of implementation. These interventions target not only
those already suffering from NCDs but also those at risk of NCDs. The Lancet NCD Action
Group and the NCD Alliance propose the delivery of five priority interventions based on
their health effects, cost effectiveness, low cost of implementation, and political and
financial feasibility (Beaglehole et al. 2011). Among this set of five interventions, four are
population-based.
5
Recommendations for population-based interventions
Recommendation Actions
Reduce unhealthy - Mass media campaigns
diet and promote - Fiscal measures to discourage consumption of unhealthy foods (i.e. taxes) and
physical activity promote consumption of fruits and vegetables (i.e. subsidies)
- Food labelling and marketing restrictions to reduce consumption of unhealthy
foods (such as saturated and trans-fat and sugar in sweetened beverages)
Reduce - Mass media campaigns to inform households about health risks of dietary salt
consumption of - Nudge and/or regulate the private sector to change industry norms with
dietary salt regards to salt and fat content of processed foods
Control tobacco - Accelerate implementation of the WHO Framework Convention on Tobacco
use Control (FCTC)
o Harmonization of taxes and further tax increases
o Prohibition of illicit trade of tobacco products
o Ban point of sale displays and all other forms of advertising
- Regulating the content and emissions of tobacco products
Reduce harmful - Tax increases
alcohol - Ban advertisements
consumption - Restrict access
- Enforce the National Alcohol Policy
- Establish mechanisms to reduce the production and sale of illicit alcohol.
II) Targeted campaigns promoting healthy behavior: To maximize impact, Sri Lanka should
customize campaign messages for different target groups and use a tailored platform to
communicate messages.
a) Campaigns on behavioral risk factors: The relatively high prevalence of smoking,
excessive alcohol use, and betel chewing among the socio-economically
disadvantaged, men, and the elderly suggest that these groups of people may not be
fully aware of the health risks of such lifestyle choices. Campaign messages with hard-
hitting evidence on health effects of these lifestyle related risk factors, such as
smoking, could be designed such that they are appealing to these groups of
population. The platform used for such campaigns could also be tailored to these
population groups. In addition, to deter early initiation of unhealthy behaviors, school
health programs should be designed and implemented.
b) Campaigns on utilization of preventive check-ups and counseling services:
Campaigns that motivate people to have regular preventive check-ups and counseling
services should especially target young adult men who are found to forgo such health
services. This would help to delay the onset of NCDs and provide early treatment. It
is, however, important to ensure that the elderly are not left behind, as they continue
to be the most vulnerable due to their age and high prevalence of NCDs and
6
behavioral risk factors. This is particularly important considering Sri Lanka’s rapid
aging of the population and the implications this will have on future healthcare costs.
c) Campaigns on healthy weight: Campaigns on the health benefits of maintaining a
healthy weight and how to achieve it should target women and younger adults who
are at a higher risk of being overweight. This effort should raise awareness not only
about the BMI-based risk of body fat but also about the health risks of abdominal
obesity. These campaigns could involve messages on healthy foods and food based
dietary guidelines, unhealthy diet (both for food prepared at home and purchased
processed foods), and the health benefits of physical activity.
B. Health system reorientation and strengthening
I) Introducing integrated and continuous care with primary care as default first contact: The
study’s findings of widespread risk factors, low awareness of health conditions and ineffective
management of diagnosed conditions suggest that Sri Lanka’s health system is not as effective
at dealing with NCDs as it has been for maternal and child health. The system provides facility
based episodic care, but there is no routine initiation and coordination of NCD care at the
primary level. A new NCD case is typically diagnosed within an outpatient department or in
the hospital during inpatient admission, and its management is usually centered around a
single disease by a specialist rather than patient-centered care that primary care providers
could provide. Effective management of NCDs is also constrained by a culture of self-referral,
which limits doctor-patient familiarity. As such, it is essential to establish an integrated NCD
care system that goes beyond facility based episodic care to reach and screen those who forgo
preventive care and ensure necessary follow-up. By institutionalizing primary care as the first
point of contact, a more productive doctor-patient relationship can be established. Table
below presents the recommendations and actions needed for the introduction of an effective
and integrated chronic care model as a model of primary care service delivery.
7
Recommendations for health system reorientation and strengthening
Recommendation Actions
Introduce integrated and - Constitute primary care teams to enable provision of comprehensive NCD
continuous care with care
primary care as default first o Train providers to meet complex NCD needs (including facility
contact based health promotion and behavior change services)
- Regularly assess the capacity of the health system the pillars of health service
delivery to provide high quality integrated primary NCD care services
o Evaluate the availability of human resource, facilities, and
drugs to ensure adequate access to quality care
- Develop referral chains to ensure the continuum of care
o Geographic mapping of facilities and primary care teams
o Establish feedback mechanisms between different providers
and levels of care
- Invest in an electronic information system to transfer patient information
between providers in the integrated delivery of care model
Institutionalize primary - Develop and implement basic health services such as screening services for
care level opportunistic blood pressure, cholesterol and diabetes, and interpersonal communication
NCD screening and program (for improved diet and life style)
counseling
Improve soft skills of - Training on communication
providers at government - Performance based incentives to encourage better communication with
health facilities to ensure patients.
patient comfort and trust
Implementing these recommendations will have economic benefits both at the micro and macro levels.
At the micro level, the prevention and early management of NCDs protects households from loss of
productivity and financial risk due to high OOP payment for medical care. At the macro level, the fiscal
implications of rising NCDs could be substantial. While the aging of the population will result in higher
costs for public provision of health care, NCDs could reduce the tax base of the economy by affecting
productivity and labor supply. Given the large societal costs of premature mortality and morbidity due to
NCDs, there is a strong case for investment in prevention and management of NCDs. Yet, the question of
fiscal sustainability of these reforms remains to be explored, especially if these interventions are to be
financed from the existing health budget. Efficiency gains from the gate-keeper system could be one
source of fiscal space to strengthen the health system and launch population-wide interventions.
Notwithstanding this, proper examination of the fiscal implications of these interventions requires
scrutiny. It is also worth noting that given the developmental threats that NCDs pose in aging populations,
there is a case for the Ministry of Finance to allocate more resources to the health sector.
8
Chapter 1. Introduction
1.1. The Context
With its impressive maternal and child health outcomes and control of communicable diseases,
Sri Lanka is often depicted as a success story. It has better child health outcomes than would be predicted
by its income level (Figure 1.1). In 2013, the country’s under-5 mortality stood at 10 per 1,000 live births
while maternal mortality ratio was 29 per 100,000 live births (WHO, 2015a). The country is also close to
eliminating communicable diseases such as malaria, polio, tetanus, and measles. Life expectancy at birth
increased from 70 in 1990 to 75 in 2014 and compares favorably with both the 2014 average for South
Asia (68) and countries in Sri Lanka’s income group (67) (WDI, 2015). However, because of years lived with
morbidity and disability, healthy life expectancy at birth in Sri Lanka in 2012 was 10 years lower than life
expectancy at birth (75) (WHO, 2015a). This is partly a result of non-communicable diseases (NCDs).
Figure 1.1. Global Comparison of Infant and Under-5 Mortality Rates
100
150
80
U5 mortality per 1000 live birth
100
Afghanistan Pakistan
60
Afghanistan
Pakistan
40
India
50
Bangladesh India
Bangladesh
20
Sri Lanka Sri Lanka
0
0
6 7 8 9 10 11 6 7 8 9 10 11
Per capita GDP (log) Per capita GDP (log)
Source: World Development Indicators (2015)
As is typical of a middle-income country, Sri Lanka’s main health system challenge is increasingly
becoming the need to address non-communicable diseases (NCDs). NCDs, also known as chronic
diseases,3 are medical conditions that are not caused by infectious agents but result from a combination
of physiological, behavioral, environmental, and genetic factors (WHO, 2017). According to the latest
Global Burden of Disease estimates (IHME, 2017), NCDs account for 81 percent of total deaths in Sri Lanka
(Figure 1.2). These include cardiovascular diseases (CVDs), cancers, chronic respiratory diseases, diabetes
3
Not all NCDs are chronic, and some communicable diseases such as HIV/AIDS are considered chronic because they
require ongoing management over a period of years or even decades.
9
and other NCDs. CVDs alone (including ischemic heart disease and stroke) account for 35 percent of the
country’s deaths. The three risk factors that account for the largest share of the disease burden in Sri
Lanka are high fasting plasma glucose, dietary risks, and high blood pressure (IHME, 2016).
Figure 1.2. Causes of Deaths in Sri Lanka
Cardiovascular diseases (35%)
7% Diabetes (14%)
8%
Cancers (13%)
35%
12%
Other NCDs (12%)
Injuries (12%)
12%
14% Chronic respiratory diseases (9%)
12%
Communicable, maternal, neonatal,
and nutritional diseases (7%)
Source: IHME (2017)
The increasing importance of NCDs is also evident from the major causes of disability-adjusted
life years (DALYs), a summary measure of years of life lost to death and disease. According to the Global
Burden of Disease 2016, Sri Lanka’s top three causes of DALYs are ischemic heart disease, diabetes
mellitus, and self-harm. The trend between 1990 and 2016 shows that DALYs resulting from ischemic
heart disease and diabetes mellitus are both on the rise, with diabetes exhibiting the highest increase in
DALYs. Notably, diabetes mellitus and chronic obstructive pulmonary disease are two of the three causes
that were in the 10 leading causes of DALYs in 2016 but were not on that list in 1990 (IHME, 2017).
The socioeconomic impact of rising NCDs is likely to be substantial. One of the targets of the
UN’s 2030 Agenda for Sustainable Development is to reduce premature deaths from NCDs globally by
one-third. This commitment was made partly because the rapid rise in NCDs could impede poverty
reduction initiatives in areas with limited resources. Vulnerable and socially disadvantaged people are at
a higher risk of premature mortality than the better-off due to differences in access to health services.
The death or illness of breadwinners and out-of-pocket payments for NCD-related healthcare could drain
household resources and push vulnerable households into poverty (WHO, 2017). Sri Lanka’s low poverty
rate and provision of free public care may limit the extent of these negative outcomes but unless the rising
rate of NCDs is contained, the progress made on these fronts may be under threat.
10
Given Sri Lanka’s aging population, it is important to understand the prevalence of NCDs and
related risk-factors. Aging populations, the globalization of unhealthy lifestyles, and rapid unplanned
urbanization are the driving forces behind the increase in NCDs around the world (WHO, 2017). The
current age composition of Sri Lanka’s population (as of 2015) can be characterized by what is often
referred to as a population pyramid, where the base of the pyramid consists of a large share of children
in the total population while the tip consists of a smaller share of older age groups (Figure 1.3). However,
the current age composition will be dramatically different in 2050 per projections made in 2015 by the
UN Population Division. In fact, the population of Sri Lanka is aging at a faster rate than the average for
lower-middle-income countries and the average for South Asia (Figure 1.4).
Figure 1.3. Age Composition in Sri Lanka in 2015 and 2050
Population pyramid (2015) Population pyramid (2050)
80-84 80-84
70-74 70-74
60-64 60-64
50-54 50-54
40-44 40-44
30-34 30-34
20-24 20-24
10-14 10-14
0-4 0-4
10 5 0 5 10 10 5 0 5 10
% Male % Female % Male % Female
Source: Data from UN World Population Prospectus (medium fertility variant)
Although official figures based on administrative definition of urbanization suggest that Sri
Lanka is the least urbanized country in South Asia, rapid urbanization and agglomeration has taken
place in its Western province. According to the Department of Census and Statistics estimates,
approximately 18 percent of the population in Sri Lanka resides in urban areas (Department of Census and
Statistics, 2012), which is well below the average for the South Asia region of 33 percent (Figure 1.5).
However, this official estimate of urbanization does not fully reflect the actual urbanization process that
is taking place since the official definition of “urban� is based on administrative underpinnings
(Weeraratne, 2016). A 2010 agglomeration index, which considered different features associated with
urbanization, estimated Sri Lanka’s level of agglomeration at 47 percent. When examining the country’s
urbanization from this perspective, it is evident that rapid urbanization and agglomeration has taken
11
place, particularly in the Western province around Colombo, Kandy, and Galle (including the corridors that
connect these cities) (World Bank, 2015a).4
Figure 1.4. Population Aged Over 65 Figure 1.5. Urban Population
10 50
% of population >=65
8 45
% of total population
40
6
35
4 30
2 25
20
0
15
1960
1964
1968
1972
1976
1980
1984
1988
1992
1996
2000
2004
2008
2012
10
5
Year 0
Sri Lanka Lower middle income
South Asia
Source: World Development Indicators, various years Source: World Development Indicators (2016)
Of the country’s nine provinces, the Western province is the most urbanized and is likely to be
home to the largest share of the aging population, making it ideal to study patterns of NCDs and
associated risk factors. Using the administrative definition, the Western province, with a population of
5.8 million (28 percent of Sri Lanka’s population), is the most urbanized (38 percent of its population live
in urban areas) and densely populated province in Sri Lanka (1,600 people per square kilometer). The
capital of this province, Colombo, is the most urbanized district in the country (77 percent of its residents
live in urban areas). All other districts report urbanization levels in the range of 18 percent or less
(Department of Census and Statistics, 2012). The Western province contributes 45 percent to Sri Lanka’s
gross domestic product and has the highest internal migration of all the provinces (Figure 1.6). It is
projected that another 3 million people will be added to the population of the Western province over the
next 15 to 25 years (Ministry of Megapolis and Western Development, 2016).
4
This is evident from the night-time lights in 2012 compared to those in 2002 and 1992 (World Bank, 2015a).
12
Figure 1.6. Lifetime Internal Migration by Province
500
400
Number of people ('000)
300
200
100
0
-100
-200
-300
Districts
Source: Census of Population and Housing 2012 (Department of Census and Statistics, 2012)
Increased population density accompanied by changes in the environment and lifestyles may
put residents of the Western province at an increased risk of illness. Urbanization negatively affects the
environment, particularly through increased air pollution and overcrowding, which in turn affects the
disease profile of urban centers. Nutritional outcomes may worsen because of over-dependence on
processed or fast foods and physical inactivity due to the use of vehicles and poor urban planning (such
as the absence of recreational centers). The combined effect of these factors may increase the proportion
of people who are obese, leaving much of the aging population at a risk of different circulatory system
diseases. These changes in turn may affect patterns of health service use and health care spending.
Monitoring trends and patterns of NCDs and their risk factors is key to detecting vulnerable groups
and providing early treatment. NCD management interventions will be essential to achieve the global
target of reducing premature deaths from NCDs by 2030. Managing NCDs involves detecting, screening,
and treating these diseases and providing access to palliative care for people in need. It will also be crucial
to reduce the risk factors associated with these diseases (WHO, 2017). WHO’s 2015 STEPwise approach
to Surveillance (STEPS) survey (WHO, 2015b) highlighted some challenges in Sri Lanka’s health system
performance regarding NCDs. More than one-fifth of adults with high blood pressure were not on
medication for hypertension. Moreover, 71 percent of adults had never measured their total blood
cholesterol, and about 29 percent of adults were either overweight or obese (WHO, 2015b).
1.2. Objectives
The primary objectives of this study are to examine the prevalence and distribution of NCDs and
risk factors across socioeconomic and demographic groups (in the Western province of Sri Lanka) and
assess the performance of the health system with regards to NCDs. The focus is on adult health. The study
13
also looks at patterns of health care use, the choice of public versus private care and the use of primary
level facilities. The magnitude and drivers of out-of-pocket payments and how out-of-pocket payments
relate to NCDs are also studied. The study further examines gaps in the existing health system in terms of
delivering effective care for NCDs.
The goal of this exploratory study is to contribute to policy dialogue on NCDs. The study is
uniquely placed to inform NCD-related policy dialogue because of its strategic focus on the most urbanized
Western province. As different parts of the country urbanize and their socio-economic conditions and life
styles change, they will be faced with similar challenges that the Western province is facing now. For this
reason, while the study is focused on the Western province, recommendations apply to the country as a
whole. In addition, the study presents data by socio-economic groups, which can allow for tailored
approaches to address the growing burden of NCDs. Such disaggregated data are not available in the
national health information system and most administrative data systems. This study attempts to fill those
gaps.
1.3. Data and Methods
The Data
The study mainly draws on household survey data that was collected specifically for this study
in 2015.5 The quantitative household survey consisted of 3,300 households selected using multi-stage
cluster sampling. In the first stage, all grama niladhari divisions (GND)6 in the Western province were
classified into three categories: rural, urban predominantly poor, and urban predominantly richer.7 The
classification of urban and rural areas used in the study follows the definition used by the Department of
Census. Specifically, the classification depends on whether the GND is governed by a municipal council or
an urban council. From each of these categories, 55 GNDs were selected randomly, and the survey was
conducted in 20 households in each of these GNDs.8 Overall, 3,300 households were sampled. The survey
team attempted to collect detailed health-related information related to all children under 5 years old, all
elderly people aged 60 and over, a maximum of one child between 5 and 19 years old, and a maximum of
two adults aged between 20 and 59 years old. Due to logistical constraints, it was not possible to include
all children and the elderly.9 Overall, individual-level data were collected from 10,107 people living in
these households, of whom 1,554 (15 percent) were children under 5 years of age, 2,047 (20 percent)
were children aged between 5 and 19, 4,655 (46 percent) were adults aged between 20 and 59, and 1,851
5
Additionally, the survey measured indoor air quality in 50 households (Chapter 2 for details).
6 A grama niladhari division is a subunit of a divisional secretariat.
7 This is based on the 2012 census data, which lists the percentage of households below the 40 th percentile of the national
household income distribution.
8 The survey team located the center of the GND and sampled eligible households in a particular direction till the total 20
households per GND was reached. The eligibility of these households was based on meeting five predictive variables pre-
identified through a stepwise regression model using data from the 2012/13 Household Income and Expenditure Survey.
9 A comparison of household roster and individual-level health data shows that 164 children under the age of 5 and 644 individuals
aged over 60 were not included although they were registered in the roster, suggesting that they were not available during
enumerators’ visits. We assume this exclusion is random for this study’s purpose.
14
(18 percent) were adults aged 60 and over. About 67 percent of these individuals were from urban areas
(6,733 individuals).
To understand the health-seeking behavior and the perceptions and attitudes of service users
and providers, we conducted various qualitative studies. This included 32 focus group discussions with
service users (consisting of different age groups in all three districts within the province). Each focus group
discussion involved about eight service users. These were complemented by 14 health immersion
observations in which a researcher accompanied patients to health service facilities to assess the quality
of service delivery and identify any gaps. These visits were made to primary, secondary, and tertiary-level
public and private facilities.
To assess the quality of indoor air in the province, we measured indoor PM2.5 concentration
(particulate matter smaller than 2.5 µm) using a real time continuous monitor.10 Indoor air quality was
assessed because fine particles that enter the respiratory tract originate primarily from combustion
sources and may have a wide range of health effects, especially on respiratory and cardiovascular systems.
These measurements were taken in 50 households which were deliberately selected from different
geographical locations of the Western province. In 18 households, measurements were made during meal
preparation times, spanning three hours. Another 32 households were monitored for 24 hours and their
main cooking sessions were noted.11
Methods and Definitions
This study combined three sets of data to measure the prevalence of NCDs among individuals
in the province: self-reported health status, diagnosed health status, and observed health status as
reflected in biomarkers collected directly from survey respondents. We collected data on self-reported
NCDs by asking respondents if they had an illness lasting more than six months that was not transmittable
from person to person.12 Self-reported health status, however, may not accurately capture the prevalence
and distribution of NCDs across population groups due to the impact of differential educational and
cultural backgrounds on awareness and self-reporting (Schultz and Tansel, 1997).
Our second measure of health status was based on the individuals’ diagnoses. Respondents
were presented with a list of eleven chronic conditions and were asked if they had been diagnosed with
any of them. These diseases were diabetes, hypertension, ischemic heart disease, cancer, asthma, chronic
obstructive pulmonary disease, cataracts, a cerebral vascular event, epilepsy, chronic kidney disease, and
chronic liver disease.13 Individuals had to provide a document verifying their diagnosis. We referred to
these cases as “diagnosed NCDs.� One problem with this approach, however, is that some households
10
The monitor used was the DustTrak Aerosol Monitor model 8530, TSI Inc, USA.
11 Monitors were fixed in the kitchen with the air inlet at 145 cm above the floor, 100 cm from the cook stove, and at least 150
cm away from open windows and doors. A correction factor of 1.65 was applied to all monitored data.
12 Although often considered synonymous with “chronic diseases,� NCDs are distinguished only by their non-infectious cause and
not necessarily by their duration as the term chronic implies. Some chronic diseases of long duration may be caused by infections.
These are not the focus of this study.
13 The list includes the four major types of NCDs in the world - cardiovascular diseases (like heart attacks and stroke), cancers,
chronic respiratory diseases (such as chronic obstructive pulmonary disease and asthma), and diabetes (WHO, 2017).
15
might be more systematic at keeping health records than others. Another problem is that, if certain
groups are more likely to forgo care, then relying on records to examine how NCDs are distributed across
households with different socioeconomic and demographic factors may be misleading.
The third measure was observed health status, which was not subject to these limitations. In
the survey, elevated blood pressure (hypertension) and overweight/obesity were measured using
biomarkers. These are the two leading physiological risk factors in terms of attributable deaths (WHO,
2017). For this reason, we explored them in greater detail. We compared the data on diagnoses and on
observed health status (in the case of hypertension) to understand people’s knowledge of their own
health status, the drivers of forgone care, and the ability of patients to manage chronic conditions.
The study took three readings of systolic and diastolic blood pressure from the adults in the
sample. We used the average of these three readings to classify adults as normal, pre-hypertensive, stage
1 hypertensive, or stage 2 hypertensive according to standard cut-off values presented in Table 1.1. Adults
were considered to be hypertensive if they were either stage 1 or stage 2 hypertensive, in other words,
had a systolic blood pressure of at least 140 millimeters of mercury (mmHg) or diastolic blood pressure of
90 mmHg.
Table 1.1. Cut-off Points for Classifying Hypertension Status
Systolic blood pressure (mmHg) Diastolic blood pressure (mmHg)
Normal <120 And <80
Prehypertension 120-139 Or 80-89
Stage 1 hypertension 140-159 Or 90-99
Stage 2 hypertension ≥160 Or ≥100
We examined both general (BMI-based) obesity and abdominal obesity. Of the different ways
of determining whether a body is fat or fit, the most common method is the Body Mass Index (BMI), which
is a ratio of a person’s body weight in kilograms divided by the square of their height in meters.14,15
Internationally recognized BMI cut-off points (Table 1.2) are used to identify individuals at risk of
morbidity related to general obesity (a BMI equal to or over 30). However, while a high BMI strongly
predicts the risk of chronic diseases and premature death (WHO, 2008a), the measure has some
limitations. First, it does not distinguish between body fat and lean body mass. At the same BMI, women
have more body fat than men. Second, it is not an accurate measure of body fat among the elderly.16
Therefore, we supplemented our data on BMI-based general obesity with two measures of abdominal
14
https://www.hsph.harvard.edu/obesity-prevention-source/obesity-definition/how-to-measure-body-fatness/
15
Although the risk of type 2 diabetes and cardiovascular disease among the Asian population is substantial for BMIs
lower than the existing WHO cut-off point for overweight, WHO consulted with experts and concluded that the
original cut-off points should be retained as international classifications (WHO, 2004). Asian populations also have a
higher percentage of body fat for a given BMI (Deurenberg-Yap et al, 2000).
16
Huxley et al (2010) also underscored that abdominal obesity may be better than BMI as predictor of CVD risk, but
its discriminatory capability may be higher when combined with BMI.
16
obesity - waist circumference (WC) and waist-hip ratio (WHR). The (South Asia specific) cut-off point that
we used for WC is the one recommended by the International Federation of Diabetes (IFD). In the case of
WHR, we used the WHO cut-off point based on a “substantially increased� risk of metabolic complications
(WHO, 2011). One limitation of the study design was that the survey did not include questions on physical
activity and diet. To partially fill this gap, we refer to WHO’s STEPs survey (WHO, 2015b) on selected diet
related and physical activity questions.
Table 1.2. BMI, WC and WHR classification
BMI (kg/m2)
BMI cut-off points
Underweight <18.5
Normal [18.5-25)
Overweight [25-30)
Obesity, grade I [30-35)
Obesity, grade II [35-40)
Obesity, grade III ≥40
Abdominal obesity cut-off points Men Women
Waist circumference (IFD) 90 cm 80 cm
Waist circumference (WHO) 102 cm 88 cm
Waist-hip ratio (WHO) 0.9 0.85
Note: In row 2-5, the range does not include the upper bound; The IFD cut-off points for WC are South Asia specific
The health expenditure data in this study were collected from a subset of household members
and comprises outpatient and inpatient spending on healthcare at both public and private facilities.
Individuals were first asked whether they used outpatient or inpatient care in the previous four weeks or
12 months, respectively, from either public or private facilities. If individuals reported seeking care, they
were also asked to report their OOP spending on different healthcare-related items during the relevant
reference period, including how much (if any) of the spending was reimbursed and if the patient
and/or caregiver had forgone any income. The individual’s monthly spending and forgone income was
calculated by adding one-twelfth of the amount reported for inpatient healthcare (over the previous
12 months) to the corresponding amount for outpatient healthcare (reported for the previous
month). One limitation of the expenditure data was that it was not at the household level, which
made it impossible to compute catastrophic expenditures by household and to compare our
estimates with those made in previous studies.
The study distinguishes between households belonging to the bottom 40 percent and the upper
60 percent of the national household consumption expenditure distribution. It predicted household
consumption using a small set of easily observable variables using the SWIFT17 approach (Box 1.1). The
surveyed households were then categorized into two groups based on their predicted per capita
17
The Survey of Wellbeing via Instant and Frequent Tracking (SWIFT) is an initiative by the Poverty and Equity Global
Practice of the World Bank. It estimates household income/expenditure and produces poverty and inequality
indicators based on data from household expenditure surveys and using the latest statistical methods, including
cross-validation, multiple imputation, and small area estimation.
17
consumption and data from the 2012/13 Household Income and Expenditure Survey (HIES). The first
group consisted of households that were predicted to belong to the poorest 40 percent of the national
population, while the second group consisted of households predicted to belong to the upper 60 percent
of the national population. In this report, these two groups are referred to as B40 and U60 (Yoshida et al,
2015). Twenty-one percent of the 3,300 households in this study belonged to the B40.
Box 1.1. Using the SWIFT Approach to Predict Per Capita Household Consumption
Using the Survey of Wellbeing via Instant and Frequent Tracking (SWIFT) approach, we first developed
a consumption model for the Western Province using data from the 2012/13 Household Income and
Expenditure Survey (Department of Census and Statistics, Sri Lanka, 2012). The model was designed
to predict per capita household consumption based on a small number of easily observable
independent variables, including the household’s location, demographic characteristics, housing
conditions, and durable asset ownership. Once the household survey in the Western Province had
been completed, it was possible to predict the per capita consumption of the surveyed households
using the estimated coefficients from the model and the data on the independent variables collected
by the survey.
The rest of the report is organized as follows. Chapter 2 examines the population’s health status
and the physiological risk factors for NCDs. Chapter 3 explores behavioral and environmental risk factors.
Chapter 4 reports patterns of health care use. In Chapter 5, the amount and distribution of out-of-pocket
payments is explored. This is followed by Chapter 6, which examines gaps in the existing health system in
terms of delivering effective care for NCDs. Finally, Chapter 7 presents conclusions and recommendations.
18
Chapter 2. Health Status and Physiological Risk Factors
Highlights:
• The prevalence of observed hypertension (26.1 percent) in the Western province is
similar to the country’s average while the prevalence of general obesity is 154
percent higher than the national average.
• The early onset of NCDs suggests that young adults live with ill-health for a
considerable part of their life.
• NCDs are not diseases of the rich. The prevalence of obesity is substantial even
among lower-income groups.
• The fact that so few people know their hypertensive status and that those
diagnosed with hypertension have trouble managing their condition suggest gaps
in effectiveness of preventive NCD care.
• Men and younger adults tend to forgo preventive health services.
• Obesity is a strong predictor of hypertension and diabetes in the province.
• Measuring abdominal obesity also captures general obesity whereas relying only
on measures of general obesity is likely to exclude a significant proportion of at-risk
individuals from public health initiatives.
• Measures of abdominal obesity may be better than measures of general obesity in
predicting the risk of diabetes and hypertension.
• of their life.
• NCDs are not diseases of the rich. The prevalence of obesity is substantial even
among lower-income groups.
• The fact that so few people know their hypertensive status and that those
This chapter
diagnosedexamines have of
the prevalence
with hypertension and various
NCDs managing
trouble risk
their factors and
condition their
suggest distribution
gaps
across different population groups in the Western province. The framework for our analysis of risk
in effectiveness of preventive NCD care.
factors comes from Pearson et al. (1993, pp 577-594) who distinguished three different categories of risk
• Men and younger adults tend to forgo preventive adult healthcare.
factors: (i) non-modifiable risk factors (such as age, sex); (ii) modifiable physiological risk factors (such as
• Obesity is a strong predictor of hypertension and diabetes in the province.
elevated blood pressure/ hypertension and overweight/obesity); and (iii) behavioral risk factors (such as
• Measuring abdominal obesity also captures general obesity whereas relying only
smoking tobacco and harmful use of alcohol).18 Behavioral risk factors and environmental risk factors are
on measures of general obesity is likely to exclude a significant proportion of at-risk
discussed in Chapter 3.
individuals from public health initiatives.
• Measures of abdominal obesity may be better than measures of general obesity in
predicting the risk of diabetes and hypertension.
18
Physiological risk factors are sometimes referred as metabolic risk factors and include high blood glucose levels
(hyperglycemia) and high levels of fat in the blood (hyperlipidemia). Behavioral risk factors include physical inactivity
and unhealthy diet (dietary cholesterol, saturated fat, and salt consumption). The survey used for this study did not
look at these behavioral risk factors.
19
2.1 Prevalence and Distribution of Self-reported and Diagnosed NCDs
Self-reported NCDs are concentrated among the elderly but their distribution across age groups
indicate that they have an early onset. NCDs are often associated with older age groups although young
people are also vulnerable to the risk factors (WHO, 2017). More than one-fifth (22.5 percent) of the adult
population in the Western province reported suffering from at least one NCD (Figure 2.1). Unsurprisingly,
the prevalence of self-reported NCDs is highest among the elderly (those aged 60 or over), of whom more
than half report having had an NCD (54 percent). However, almost 4 percent of adults between the ages
of 20 and 30 also reported suffering from NCDs (4 percent). The prevalence of NCDs among young adults
indicates a substantial loss of healthy life years in the province.
Figure 2.1. Prevalence of Self-reported NCDs
60% 53.9%
50%
40%
30.1%
30% All age groups
17.8%
20%
7.5%
10% 3.5%
0%
20-30 30-40 40-50 50-60 60+
Age groups
The two most widespread diagnosed chronic conditions are hypertension (elevated blood
pressure) and diabetes mellitus (high blood sugar). As many as 15 percent and 12 percent of adults were
diagnosed with hypertension and diabetes mellitus respectively (Figure 2.2). Other prevalent NCDs
include cataracts19 (6 percent), ischemic heart disease (IHD), and asthma (2 percent each). As mentioned
earlier, in Sri Lanka, diabetes is one of the three leading causes of DALYs, while hypertension is one of the
three risk factors that account for the greatest disease burden. Moreover, diabetes often coexists with
hypertension, increasing the risk for life-threating CVDs.20 The data show that one-in-ten adults in the
province have at least two of the eleven diagnosed NCDs examined, while 17% of adults have only one of
these eleven NCDs.21
19
A cataract is a clouding of the lens of the eye that is associated with age, but it can also be a result of certain
chronic diseases such as diabetes.
20
In industrial countries, the risk of coronary heart disease is two to three times higher in diabetic patients aged over
40 years old (Vaughan et al, 1993).
21
These include chronic conditions such as cataracts and hypertension (a physiological risk factor). These are
categorized as NCDs because they fulfill the definition of “a medical condition that is not transmissible.�
20
Figure 2.2. Prevalence of Diagnosed Chronic NCDs
16%
14%
12%
10%
8%
6%
4%
2%
0%
Ischaemi Cerebral Chronic Chronic
Hyperten Diabetes COPD
Cataract c heart Asthma vascular Cancer Epilepsy Kidney liver
sion Mellitus (other)
disease event Disease disease
All 14.6% 11.9% 5.8% 2.0% 2.0% 0.8% 0.4% 0.4% 0.2% 0.2% 0.0%
The prevalence of diabetes in the province (12 percent) is higher than both the national average
(8 percent) and the highest national average for other countries in the region (Figure 2.3). It is important
to understand what underlies this high prevalence as all types of diabetes mellitus (insulin-dependent,
non-insulin dependent, and malnutrition-related) are likely to lead to complications later in life and thus
to substantial economic costs. Studies show that increased food intake, obesity, and lack of exercise are
associated with non-insulin dependent diabetes mellitus (Vaughan et al, 1993, pp 561-576). The
association of diabetes with obesity and hypertension is explored in section 2.4 below.
Figure 2.3. Prevalence of Diabetes in South Asia
10
Percent of population
8
6
(20-79)
4
2
0
Banglad Afghanis
Nepal Sri Lanka Pakistan Maldives Bhutan India
esh tan
Prevalence (% population 20-79) 3.7 8 8.1 8.3 8.8 9.2 9.3 9.3
Source: World Development Indicators (2015)
Before embarking on a detailed analysis of the two most prominent physiological risk factors, in
the remainder of this section, we examine the distribution of self-reported NCDs and the five
most prevalent diagnosed NCDs/chronic conditions.
21
The prevalence of self-reported NCDs does not vary with economic status but is higher among
the less educated. The prevalence of self-reported NCDs among the poorest quintile (23 percent) is not
significantly different from the prevalence among the richest quintile (22 percent) (Figure 2.4). A
regression analysis of the probability of self-reporting NCDs confirms this finding (Annex Table A2 and A3).
With self-reporting, the less educated might under-report their actual ill-health due to lack of knowledge
about what constitutes a healthy life. Contrary to this expectation, in our survey, a substantially higher
proportion of adults who have not completed Advanced Level (A/L)22 (26 percent) self-reported suffering
from NCDs than those who have completed this level of education (14 percent). In the remainder of this
report, individuals with an education below A/L level are referred to as less educated and those who have
completed this level or higher are referred to as more educated.
Figure 2.4. Self-reported NCDs by Economic Status and Level of Education
25.0% 30.0%
24.0% 25.7%
23.0% 25.0%
22.0%
20.0%
21.0%
20.0% 14.0%
15.0%
19.0%
18.0% 10.0%
5.0%
0.0%
60 36.5% 49.9%
District of residence
Colombo 15.0% 26.8%
Gampaha 13.9% 19.2%
Kalutara 15.3% 36.8%
Note: Figures in bold show statistically significant difference in means
between categories (p-value<0.1). See Annex Table A4 for different stages
of hypertension.
In order to analytically understand the characteristics of those who are (not) aware of their
hypertensive status, we estimated a probit regression for the sample of individuals who were found to be
hypertensive. We constructed the ‘awareness’ variable using the difference between observed
hypertension and diagnosed hypertension (Annex Table A5).
Younger adults are less likely than older adults to be aware of their hypertensive status (Annex
Table A5). Among those categorized as hypertensive, we found younger adults to be less likely to be aware
of their conditions than the elderly. The elderly were about 60 percentage points more likely to be
aware of their hypertensive status than those aged 20 to 30. Higher use of care among the elderly (as
shown in Chapter 4) and the targeting of this age group for disease screening seem to have contributed
to this pattern.
29
Men are more likely to be hypertensive but less likely to be aware of it, which suggests that
they are not receiving sufficient preventive primary care. Consistent with what was pointed out earlier,
we found that men in the province were 13 percentage points less likely than women to be aware of their
hypertensive status. The focus group discussions that we held as part of the study revealed that most
adult men in the province did not seek preventive care to manage a risk factor or to delay the onset of
diseases. Primary care facilities where such screening services are most likely to be provided are mainly
used by the elderly, women, and mothers. Our regression analysis of patterns of utilization, discussed in
Chapter 3, also showed that use of outpatient care is lower among men than women.
The poorest and richest adults are more aware of their hypertensive status than those in the
middle. While there is no difference in awareness between the poorest and the richest 20 percent of
consumption expenditure, the poorest are more likely to be aware than the middle three consumption
quintiles. Moreover, those belonging to the B40 of the national consumption expenditure distribution are
8 percentage points more likely to be aware of their hypertensive status than those in the U60.
Awareness does not vary with education status and place of residence. Contrary to what might
be inferred from Table 2.1 (a comparison of diagnosed and observed hypertension), whether or not
individuals have at least an A/L level of education is not associated with awareness of their hypertensive
status. Awareness is also not significantly different between those who live in Kalutara (the district with
the highest prevalence of hypertension) and other districts. Moreover, despite the higher concentration
of health facilities in urban areas, urban residents are not more aware than rural residents about their
hypertension status.
2.3. Obesity
Obesity is another vital physiological risk factor for cardiovascular diseases. Overweight and
obesity increase the risk of high blood pressure (as shown in the previous section) and high cholesterol,
thereby increasing the likelihood of heart diseases and stroke. They also increase the risk of type II
diabetes in adults (WHO, 2008a and Shaten et al, 1993). In addition to general obesity, abdominal obesity
is a predisposing factor for CVDs, which are the main cause of obesity-related deaths (WHO, 2008a).
Abdominal obesity is related to a range of metabolic abnormalities, including several risk factors for type
II diabetes and CVDs (decreased glucose tolerance, reduced insulin sensitivity, and adverse lipid profiles).
Our data showed that the Western province has a 154 percent higher prevalence of general
obesity than the Sri Lankan average. Table 2.2 presents the estimated prevalence of obesity in the
Western province and the average for Sri Lanka based on WHO’s 2015 STEPS survey (WHO, 2015b). As
many as 15 percent of adults in the Western province were generally obese. This is mainly driven by the
high prevalence among women (17 percent, which is significantly higher than that of men (8 percent).32
The gender difference is also reflected in the averages for Sri Lanka as a whole. This difference is both
substantial and statistically significant.
32
Even after adjusting for differences in height, men have greater total lean mass and bone mineral mass and a
lower fat mass than women (WHO, 2008a).
30
Table 2.2. Prevalence of General and Abdominal Obesity
Sri Lanka Western province
t-test for
difference in
means
All Women Men All Women Men (Women=Men)
General obesity
(BMI≥30) 5.9% 8.4% 3.5% 15% 17% 8% ***
Abdominal obesity
WHR (WHO cut-off)) 75% 75% 76%
WC (IFD cut-off) 57% 65% 34% ***
WC (WHO cut-off) 29% 37% 7% ***
The high prevalence of obesity could be a result of unhealthy diet and/or physical inactivity.
The 2015 WHO STEPs survey (WHO, 2015b) found that 30 percent of Sri Lankan adults aged 18-69 have
insufficient physical activity (defined as less than 150 minutes of moderate-intensity activity per week)
(Table 2.3). The gender difference in obesity seems to be a reflection of differences in physical activity.
38 percent of women have insufficient physical activity, which is significantly higher than that of men (23
percent). Unhealthy diet appears to be another factor contributing to obesity. More than a quarter of
adults (26.6 percent) always or often eat processed foods high in salt, and about 73 percent of adults, on
average, eat less than the daily recommended amount of fruits and vegetables (i.e 5 servings of fruit
and/or vegetables per day). It is important to note that being the most urbanized and affluent province,
physical inactivity and reliance on processed foods in the Western province could be much higher than
the average for Sri Lanka.
Table 2.3. Diet and Physical Activity in Sri Lanka (Age 18-69)
All Men Women
Percentage who ate less than 5 servings of fruit and/or vegetables on
average per day 72.5 73.1 72
Percentage who always or often add salt or salty sauce to their food before
eating or as they are eating 21.8 21.8 21.8
Percentage who always or often eat processed foods high in salt 26.6 28.3 24.8
Percentage with insufficient physical activity (defined as <150 minutes of
moderate-intensity activity per week) 30.4 22.5 38.4
Median time spent in physical activity on average per day (minutes) 77.1 124.3 42.8
Percentage not engaged in vigorous activity 73.6 58.3 89.2
Source: WHO STEPs survey 2015 (WHO, 2015b)
31
The problem of abdominal obesity is also much higher in the Western province than in Sri Lanka
as a whole. The average waist circumference of men and women is 86.4cm and 84.8cm respectively.33
Both are significantly higher than the Sri Lanka average (82.3cm for men and 82.1cm for women).34 Using
the aforementioned IFD WC cut-off point for South Asia, as much as 65 percent of women and 34 percent
of men in the province have abdominal obesity. The proportion is lower using the non-ethnic specific
WHO cut-off point. However, both indicate that the potential risk of metabolic complication is very high,
especially among women for whom both abdominal and general obesity are a significant problem. The
magnitude of the problem is even higher if WHR is considered, with about three-quarters of both men
and women being at risk of abdominal obesity.
Comparing the Western province with the national average also reveals that underweight
prevalence is less widespread while obesity is more common. Nearly half of adults (47 percent) in the
Western province are either overweight or obese, while only 9 percent are underweight. For Sri Lanka, as
a whole, a much smaller proportion (30 percent) of adults aged between 18 and 69 are either overweight
or obese (with a BMI equal to or over 25), but a much higher proportion (15 percent) are underweight
(Figure 2.13). This comparison and the finding that 57 percent of adults in the Western province are
abdominally obese further strengthen the claim that obesity is a significant public health issue in the
province. Sedentary lifestyles and unhealthy diets may have contributed to this.
Figure 2.13. Comparing the Western Province with the Average for Sri Lanka
60.0% 55.4%
50.0% 44.5%
40.0% 32.0%
30.0% 23.4%
20.0% 15.3% 14.5%
9.0% 5.9%
10.0%
0.0%
Underweight Normal Overweight Obesity
Western province Sri Lanka (STEPS 2015)
A substantial proportion of adults who have a normal BMI have abdominal obesity, and this is
the case even among those who are underweight, which suggests that relying only on BMI will
undermine the effectiveness of public health initiatives on obesity. Figure 2.14 shows an interesting
distribution of abdominal obesity across levels of BMI. As expected, the prevalence of abdominal obesity
increases with increasing BMI. What is perhaps more interesting in Figure 2.14 is that 35 percent of those
who have a normal BMI (between 18.5 and 25) have abdominal obesity.35 While this is worrying enough
as it is, the WHR story is even worse. Close to three-quarters of adults within a normal BMI range have
higher WHR than the cut-off point for a substantially increased risk of metabolic complications. This
33
The average across all adults is 85.2cm.
34
The differences are statistically significant (non-overlapping confidence intervals).
35
This declines to 9 percent in the case of the standard WHO cut-off point for WC.
32
suggests that, in the case of the Western province, public health interventions should not be based
entirely on people’s BMI.
Figure 2.14. Abdominal Obesity Distribution across BMI Categories
Underweight Normal Overweight Obese (I,II,III)
120%
96%
100% 83% 82% 81% 82%
73%
80%
60% 52%
35% 41%
40%
20% 9% 4% 9%
0%
WC (IFD cut-off) WC (WHO cut-off) WHR (WHO cut-off)
Is measuring abdominal obesity better for identifying at-risk individuals (those with either
general or abdominal obesity) than just using BMI-based obesity? What proportion of adults have both
forms of obesity? Figure 2.15 splits the whole sample into four quadrants. Each dot in the figure
represents a woman or a man. Observations above the red horizontal line (BMI of 30) have general obesity
and those below do not. Similarly, observations to the right of the red vertical line have abdominal
obesity.36 The numbers in each graph show the proportion of women/men: (i) who have both general
obesity and abdominal obesity (right upper quadrant); (ii) who have neither (left lower quadrant); (iii) who
have general obesity but not abdominal obesity (left upper quadrant); and (iv) who do not have general
obesity but have abdominal obesity (right lower quadrant).
Measures of abdominal obesity may be better at identifying at-risk individuals than BMI alone.
About 16 percent of women and 7 percent of men have both forms of obesity, while 35 percent of women
and 66 percent of men have neither general nor abdominal obesity.37 The remaining individuals are obese
in either of the two forms. Interestingly, the lion’s share of these are individuals who have abdominal
obesity but not general obesity.38 For example, close to zero percent of women have a BMI above 30 and
a WC below 80cm. However, almost half of women have a BMI below 30 but a WC above 80cm. The
pattern is the same for men and holds true when using the WHR cut-off for judging abdominal obesity.
These findings reinforce our finding that measures of abdominal obesity may be better at identifying at-
risk individuals than BMI alone. It is also easier to measure WC and WHR in the field than measuring BMI
as the latter requires weighing scales.
36
Those with a WC of 90cm for men and 80cm for women according to the IFD cut-off point in the first row of graphs
and a WHR of 0.9 for men and 0.85 for women according to the WHO cut-off point in the second row of the graphs.
37
This proportion declines to 22 percent (women) and 23 percent (men) when using the WHR cut-off for abdominal
obesity.
38
We also tested whether there is a statistically significant difference in the prevalence of abdominal obesity
between those who have general obesity and those who do not. As expected, we found that the prevalence was
significantly higher among the former group.
33
Figure 2.15. General and Abdominal Obesity (% of women and men)
Female Male
10 20 30 40 50 60
10 20 30 40 50 60
0% 16% 1% 7%
BMI
BMI
35% 48% 66% 26%
50 100 150 60 80 100 120
Waist Waist
Female Male
10 20 30 40 50 60
10 20 30 40 50 60
3% 13% 1% 7%
BMI
BMI
22% 61% 23% 69%
.6 .8 1 1.2 1.4 .6 .8 1 1.2 1.4 1.6
WHR WHR
Distribution of Obesity across Groups: Beyond Gender Differences
Obesity is highly prevalent even among supposedly active age groups (Figure 2.16). To explore
the distribution of obesity (general and abdominal) and underweight across population groups other than
men and women, we carried out a series of descriptive and analytical (multivariate) analyses. We found
that as many as 39 percent of young people aged between 20 and 30 had abdominal obesity, while about
11 percent had general obesity. The prevalence of both forms of obesity increases in the subsequent age
groups, reaching a maximum level in the 40 to 50 age group and declining thereafter. With both measures
of obesity, adults under 60 are at a higher risk of diseases associated with body fat than those aged 60 or
over (Annex Table A6). As physical activity often declines after the age of 60, the explanation for higher
obesity among the under-60s may be related to diet rather than physical inactivity.
34
Figure 2.16. Obesity by Age Groups
20-30 30-40 40-50 50-60 60+
80% 66% 65%
58%
60% 54%
39%
40%
16% 20% 17%
20% 11% 8%
0%
General obesity (BMI≥30) WC > IFD cut-off
The prevalence of obesity is consistently high across economic groups. The economically worse-
off have a similar level of general obesity and a slightly lower prevalence of abdominal obesity than the
better-off (Figure 2.17). Among the poorest 20 percent, the prevalence of abdominal obesity is 52 percent,
while it is 65 percent among the richest 20 percent. However, with 52 percent of the B40 (and of the
poorest quintile) being abdominally obese, the problem of obesity is of concern even among the
economically disadvantaged. This is further corroborated by our finding that there was no meaningful
difference in the prevalence of general obesity between the B40 and the U60 or between the poorest 20
percent and the top 20 percent (Annex Table A6). However, the prevalence of underweight is largely
concentrated among the poorest quintiles (Figure 2.18).
Figure 2.17. Obesity by Economic Status
Poorest quintile 2nd poorest quintile U60 B40
Middle quintile 2nd richest quintile 70%
58%
Richest 60% 52%
50%
80% 65%
58% 40%
60% 52%54%56% 30%
20% 14% 15%
40%
14%14%13%14%16% 10%
20%
0%
0% General obesity WC > IFD cut-off
General obesity (BMI≥30) WC > IFD cut-off (BMI≥30)
35
Figure 2.18. Underweight by Economic Status
15.0% 12.9% 13.1%
9.1% 8.6%
10.0% 7.4% 7.9%
5.9%
5.0%
0.0%
Poorest 2nd Middle 2nd Richest B40 U60
quintile poorest quintile richest
quintile quintile
Some of the descriptive differences between groups may reflect differences in other
socioeconomic and demographic characteristics. Therefore, in order to identify the factors associated
with obesity and underweight, we estimated a probit regression for: (i) the probability of having general
obesity39; (ii) the probability of having abdominal obesity (according to the IFD-based WC cut-off point);
and (iii) the probability of being underweight. Annex Table A7 presents the marginal effects from a
specification akin to the one presented in Annex Table A5.
We found that women and adults aged between 30 and 60 are more likely to be obese than
men and other age groups, respectively. Even after controlling for various socioeconomic and behavioral
factors, men are 15 percentage points less likely to have general obesity than women and about 32
percentage points less likely to have abdominal obesity. The probability of being underweight is, however,
similar between the two genders. Regarding age categories, obesity appears to be more prevalent among
those aged between 30 and 60, while underweight is concentrated among the elderly and those between
20 and 30 years of age.
The richest are at a higher risk of obesity and at a lower risk of having underweight-related
health problems. The probability of being underweight is negatively associated with economic status. The
B40 have 7 percentage points higher probability of being underweight than the U60. The alternative
specification also showed that this probability decreases from lower to higher consumption expenditure
quintiles. With respect to obesity, the income gradient is not systematic, but the richest 20 percent are
much more likely to be obese than the poorest 20 percent. The B40 are 6 percentage points less likely to
be abdominally obese but are as likely to have general obesity as the U60.
Adults who have completed at least an A/L level of education are about 6 percentage points
less likely to be generally obese and 4 percentage points less likely to be underweight as those who are
less educated. Differences in education levels can lead not only to differences in economic status but also
to differences in knowledge about health. Since our analyses accounted for differences in economic
39
The regression is on the subsample of individuals who are either obese or are normal weight.
36
status, the lower probability of obesity and underweight among the educated is likely to be due to their
greater awareness about the benefits of proper diet and physical activity.40
2.4. Obesity as a Risk Factor for Diabetes and Hypertension
Adults with general or abdominal obesity have a significantly higher risk of getting diabetes
than those who are not obese. Those who have general obesity have 36 percent higher prevalence of
diagnosed diabetes (p-value<0.05) than those with a normal BMI (Figure 2.19). Depending on the measure
used, the abdominally obese also have 55 percent (p-value<0.05) to 63 percent (p-value<0.01) higher
prevalence of diagnosed diabetes than those with a normal BMI.
Figure 2.19. Obesity as a Risk Factor for Diabetes
16% 14%
Diagnosed diabetes 14%
13%
prevalence 12%
10%
9%
8%
8%
15% 6%
13% 4%
11%
2%
7% 0%
No Yes No Yes
Abdominal obesity (WC Abdominal
Normal Overweight Obese IDF cut-off) obesity (WHR
Underweight (I,II,III) WHO cut-off)
There is a strong positive association between all forms of obesity and diabetes even after
accounting for hypertension and various socioeconomic and demographic factors. Although the
association of obesity with diabetes has been long established, it is important to understand whether the
positive association remains after accounting for differences in other risk factors, including hypertension.
All three anthropometric measures of obesity used in this study have been found to be predictors of the
risk of diabetes and CVDs (Qiao and Nyamdorj, 2010 and Huxley et al, 2010). In this study, we examined
the association of diabetes with each of the three measures of obesity separately in a multivariate
regression framework (Annex Table A8). We found that adults with general obesity are 2.3 percentage
points more likely to be diagnosed with diabetes than those who are not obese. This positive association
was even stronger with measures of abdominal obesity. Those with a WC higher than the cut-off
suggested by the IFD were 5 percentage points more likely (than those with WC below the cut-off) to be
diagnosed with diabetes. Similarly, adults with WHR greater than the WHO based cut-off point were 2.7
percentage points more likely (than those with WHR less than the cut-off) to be diagnosed with diabetes.
40
In fact, the results for abdominal obesity suggest this. The significant positive association with education (column
3 of Annex Table A7) vanishes once a more refined measure of economic status is employed (consumption quintiles
in column 4). This suggests that part of the information contained in the education variable relates to economic
status. Since economic status works in the opposite direction, the actual association of obesity with awareness is
likely to be more significant than is portrayed by the results presented here.
37
Moreover, the probability of being diagnosed with diabetes was 3.5 percentage points higher among
those who were hypertensive than those who were not.
Similarly, we found that adults with general or abdominal obesity had a significantly higher
relative risk of having hypertension than those who were not obese. Figure 2.20 shows the prevalence
of hypertension across all BMI categories and by abdominal obesity status. It shows that hypertension
prevalence was 21 percent higher (p-value<0.05) among adults with general obesity than among those
with a normal BMI. Depending on the measure used, the abdominally obese have a 26 to 33 percent
higher prevalence of hypertension than those with a normal WC and WHR (p-value<0.01). Even after
accounting for other risk factors, obesity and hypertension are strongly associated (Annex Table A9).
Those with either form of obesity (general and abdominal) are 9 percentage points more likely to be
hypertensive than those who are not obese.
Figure 2.20. Obesity as a Risk Factor for Hypertension
35%
Observed hypertension prevalence 29% 28%
30%
25% 23%
28% 29% 21%
23% 24% 20%
15%
10%
5%
0%
No Yes No Yes
Normal Obese Abdominal obesity (WC IDF Abdominal obesity
Underweight Overweight (I,II,III) cut-off) (WHR WHO cut-off)
The relative magnitudes of association suggest that measures of abdominal obesity may be
better than general obesity in predicting the risk of diabetes and hypertension. The increased probability
of diabetes indicated by WC is more than twice as large as that of general obesity. Furthermore, including
all measures of obesity in the regression reveal that WC is a stronger predictor of the risk of diabetes than
WHR and general obesity (column 4 of Annex Table A8).41 Furthermore, in the case of hypertension,
including both abdominal and general obesity in the same regression suggest that WC is a stronger
predictor of hypertension than general measure of obesity (Annex Table A9). These findings further
indicate that public health interventions in the province would perform better if they rely on WC for
monitoring obesity and assessing the level of public health action needed.
Summary and implication
41
Although there is some indication that the abdominal measures are better in predicting the risk of diabetes and
CVDs (Seidell, 2010; Larsson et al., 1984; Lapidus et al., 1984) 41, a review done by ‘WHO expert consultation’
concludes that it is unclear which anthropometric measure is the most important predictor of diabetes in adults.
Neither is there consensus over which of the two measures of central obesity are better associated with CVD risk
(WHO, 2008a).
38
To conclude, physiological risk factors for CVDs are very common in the Western province, and
many adults from different socioeconomic and demographic groups are not aware of their health
status. The probability of being hypertensive is higher among men, but men are less likely to be aware of
their hypertensive status, suggesting that they are less likely to use preventive health services. Higher
awareness among women is consistent with Sri Lanka’s strong mother and child health (MCH) services.
However, even among women, gaps in awareness are evident, suggesting the need to strengthen existing
MCH services. The existence of greater health awareness among older people could be due to their
increasing contact with medical professionals as their health deteriorates with age. Men and younger
people seem to have the highest rates of forgone preventive care. While hypertension does not seem to
be associated with socioeconomic status, the poorest and richest are more likely to be aware of their
conditions than the middle 60 percent. This likely reflects utilization patterns, with the richest and poorest
having more contact with health professionals as a result of having more information and a higher
propensity to fall ill, respectively.
The prevalence of general BMI and abdominal obesity in the province poses a significant risk
for NCDs. This is especially the case among women and those between the ages of 30 and 60. Data from
secondary sources suggest that physical inactivity and unhealthy diet in the country are of concern.
Although the risk of obesity seems to be highest among the richest, it is alarmingly high even among the
poorest quintile. We found obesity in the province to be significantly associated with a higher risk of
hypertension and diabetes.
Basing public health interventions solely on BMI-based classifications risks excluding a huge
proportion of adults, especially women, who are at substantial risk of metabolic complications. In fact,
waist circumference appears to predict the risk of diabetes better than BMI-based obesity. A significant
proportion of people who would be categorized as having a normal weight are at risk of metabolic
complications related to abdominal obesity. Given that the health risk of a given amount of body fat in
the Asian population is much higher than in other regions, policymakers need to make this a high priority.
39
Chapter 3. Behavioral and Environmental Risk Factors
Highlights:
• The current smoking rate in the Western province is lower than the national average as well
as the average for countries in the region. The same is true for the proportion of current
drinkers.
• Risky behavior is predominantly practiced by elderly men in lower socioeconomic groups.
• Betel chewing in the Western province is much more common than smoking and harmful
alcohol use.
• Hypertension prevalence is higher among those who practice risky behavior.
• A huge proportion of households in the Western province use unclean sources of fuel as their
primary cooking source.
• The potential burden of indoor air pollution is the largest for the poorest and those living in
rural areas.
• Urban residents of the province have a higher risk of being exposed to outdoor sources of
pollution.
• Sanitation infrastructure in the province is widespread, but there is huge room for improving
drinking water sources for households at all economic levels.
• Unhygienic solid waste management is more prominent in rural parts of the Western
province.
3.1. Behavioral Risk Factors
Physiological risk factors such as hypertension and obesity can be indirectly modified through
• The current smoking rate in the Western province is lower than the national average as well
changes in behavioral risk factors. Monitoring these risk factors and understanding how they are
as the average for countries in the region. The same is true for the proportion of current
distributed across groups can help policymakers to design targeted interventions and predict disease
drinkers.
burdens. The World Health Organization’s 2008-2013 Action Plan for the Global Strategy for the
•
PreventionRisky
and behavior predominantly
is NCD
Control of (WHO, 2008b)practiced by elderly
identified men in lower
the monitoring socioeconomic
of commonly sharedgroups.
risk factors
• component
as a key Betel chewing in the Western
for preventing province is
and controlling much
four more
major common
NCDs: CVDs, than smoking
diabetes, and
cancer, harmful
and chronic
alcohol use.
respiratory diseases. The most common of these risk factors are tobacco use, harmful use of alcohol, an
• Hypertension
unhealthy prevalence
diet, and physical is higher
inactivity. amonglooks
This section thoseat
who
twopractice
of theserisky behavior.
behavioral risk factors (smoking
• A huge proportion of households in the Western province use unclean
and harmful alcohol use) as well as betel chewing, which is a risk factor that is sources of fuel
commonly as their in
practiced
primary cooking source.
South Asia.
• The potential burden of indoor air pollution is the heaviest for the poorest and those living in
rural areas.
• Urban residents of the province have a higher risk of being exposed to outdoor sources of
pollution.
• Sanitation infrastructure in the province is widespread, but there is huge room for improving
drinking water sources for households at all economic levels.
• Unhygienic solid waste management is40 more prominent in rural parts of the Western
province.
Smoking
The Western province has a much lower prevalence of current smokers than the national
average as well as the average for other countries in the region. Only 4 percent of adults currently smoke
tobacco products.42 This is less than one-third of the prevalence in the country as a whole (Table 3.1) and
much lower than the prevalence in other countries in the region. This low prevalence, however, masks
large differences in smoking between men and women. Hardly any women in the Western province smoke
(0.3 percent). Among men, the prevalence is around 16 percent, but this still is significantly lower than
the national prevalence of 29 percent.
Table 3.1. Current Smoking Prevalence
Sri Lanka
(STEPS
Western province 2015) Sri Lanka Pakistan India Bangladesh Nepal
All 4.4% 15
Male 15.7% 29.4 28 42 20 40 37
Female 0.3% 0.1 0 3 2 1 11
Source: Columns 3-7 are from World Development Indicators (2015)
Prevalence of smoking is highest among men in the 40 to 60 age group, but the average age at
which individuals initiated smoking has been declining over time (Figures 3.1). Among daily smokers,
younger adults started daily smoking earlier than older generations.43 Those who are currently above 60
started smoking daily at the age of 25, whereas those currently under 30 started almost a year after
turning 18. Figure 3.1 shows that this trend has persisted (p-value<0.01). Though not as stark, WHO’s 2015
STEPS survey (WHO, 2015b) also suggested this trend is also happening in Sri Lanka as a whole.
Figure 3.1. Daily Smoking Prevalence and Age of Initiation of Daily Smoking by Age Cohort
20% 25.1 30
18.9
Age of initiation of
Daily smoking
20
(percent)
daily smoking
10%
10
0% 0
Age<30 30-40 40-50 50-60 age>60
Age groups
Daily smoker (men only) Age of initiation
42
Of those who currently smoke, the majority (75 percent) are daily smokers.
43
A significant majority (93 percent) of daily smokers smoke manufactured cigarettes. Only 2.8 percent of them
smoke hand-rolled cigarettes. About 16 percent smoke cigars, cheeroots, and/or cigarillos (multiple responses were
allowed). Unfortunately, we cannot analyze the number consumed as the study did not collect such data.
41
Smoking is predominantly a behavior exhibited by men in lower socioeconomic groups (Figures
3.2).44 Daily smoking prevalence among the poorest quintile (17.3 percent) is more than double the
prevalence among the top quintile (7.2 percent). Similarly, a significant difference exists between the B40
and the U60. Moreover, the prevalence of daily smoking among those with less than an A/L level of
education (14.1 percent) is more than double the prevalence among those who have completed at least
this level of education (6.7 percent). These descriptive patterns are confirmed by a regression analysis
that accounts for differences across multiple dimensions (Annex Table A11). Men are about 12 percentage
points more likely to smoke daily than not to smoke at all. Adults who have completed at least an A/L level
of education are about 0.6 percentage points less likely to smoke daily than those who have not. Although
this is not a large difference, it may suggest that educated people have a better understanding of the
health risks of smoking. Furthermore, the B40 are 0.5 percentage points more likely to smoke daily. Using
consumption expenditure quintiles also shows that the top 60 percent are less likely to smoke daily than
the bottom 20 percent. Taken together, the findings suggest that smoking is more prevalent among those
belonging to lower socioeconomic groups.
Figure 3.2. Smoking Prevalence by Economic Status and Education Level (men only)
25.0% 21.3% 20% 17.8%
20.7%
18.7% 14.1%
20.0% 16.3% 15%
14.6%
15.0% 12.6% 9.4%
10.1% 10% 6.7%
10.0%
5%
5.0%
17.3% 15.0% 12.0% 10.1% 7.2% 17.0% 11.2%
0.0% 0%
Poorest 2nd Middle 2nd Richest B40 U60 Current Daily smoker
quintile poorest quintile richest smoker
quintile quintile
60
OOPP (among those who received care)
OOPP (among all respondents)
On average, there is no significant difference in spending between rural and urban parts of the
province (Figure 5.10). The absence of significant rural-urban difference in OOP payment in the Western
province contrasts with the national average where OOP payments are substantially higher in urban areas
than in rural areas according to the 2012/13 HIES.66 For the Western province, the HIES data also show
that OOP spending is similar in rural (LKR2,340) and urban areas (LKR2,494).67 It is important to note that
the HIES figures are at the household level (and hence much higher than our survey data at the
individual level). As mentioned earlier, the data do not allow for a valid household-level analysis
because our survey collected detailed health spending data only for a selection of individuals in the
household. Looking at a sub-sample of households (14 percent) for which the survey collected
spending data for all household members, no significant difference can be seen in the out-of-pocket
payments of rural (LKR2,889) and urban residents (LKR2,524). The lack of rural-urban difference in
OOP payments in the province may be attributable to the possibility that the rural areas in the province
are either not as rural as elsewhere in Sri Lanka or that these rural residents live close enough to Colombo
that they can use urban services and this is reflected in their spending.
65 This association could be anticipated because we identified diabetics from households’ own health records, which means that
these individuals have been diagnosed and have, therefore, used and paid for healthcare.
66 LKR2,211 (urban) vs LKR1,433 (rural). These figures are at the household level and not restricted to those who utilized care.
67 The HIES 2012/13 data show that the incidence of catastrophic expenditure is higher in rural areas (12 percent versus 9 percent
at the 10 percent threshold). This is true not only in the Western province but also nationally. Unfortunately, a valid estimate of
catastrophic expenditure cannot be computed here as the survey instrument and sample design were not designed for this task.
74
Figure 5. 10. OOP Payments by Urban or Rural Location
3000
2109 2076
LKR 2000
1000 733 713
0
OOPP (among those who received care) OOPP (among all respondents)
Rural Urban
Summary
In sum, while the amount of out-of-pocket payments paid in the Western province is modest
and is borne based on ability to pay, the increasing burden of NCDs may pose a threat to the financial
protection of households in the province in the future. In light of the increasing prevalence of NCDs and
of physiological risk factors that are associated with higher spending, containing OOP spending may
require both reducing the prevalence of these risk factors and removing the quality constraints that drive
people away from public care, especially for routine conditions.
Most of the spending in the province is incurred when seeking private care, but part of it is also
associated with the unavailability of laboratory tests and medications in public facilities, forcing patients
to use private laboratories and pharmacies. This problem will need to be addressed to ensure that Sri
Lanka can maintain its good record in financial protection.
75
Chapter 6. An Overview of the Health System of Sri Lanka
The previous four chapters examined the prevalence of NCDs and risk factors in the Western
province and the factors associated with health-seeking behavior and out-of-pocket health spending.
In part, health-seeking behavior and out-of-pocket health spending are a reflection of the country’s health
system. The lack of awareness among individuals, especially adult men, about their health conditions and
the inability to manage diagnosed chronic conditions may indicate that the health system is failing to
identify, treat, and follow up with NCD patients. This chapter highlights how Sri Lanka’s health system is
organized, with the aim of identifying: (i) better ways to address the increasing burden of NCDs and (ii)
aspects of the health system, which have been effective for maternal and child health, but are not
contributing to improvements in NCD health outcomes.
The health system of Sri Lanka consists of both a public and a private sector, with the public
sector being the primary source of care. The public health system provides services free of charge and is
funded by general tax revenues that are managed by the Ministry of Health or its nine provincial
counterparts. The public health system consists of two parallel networks providing preventive health
services and curative health services. The system provides nearly 100 percent of the country’s preventive
services, as much as 90 percent of inpatient services, and 50 percent of outpatient services. Over the last
three decades, the private sector has increased its role in providing health services, primarily curative
outpatient services in urban and suburban areas.
The private sector in Sri Lanka includes a range of providers that operate mostly in urban areas.
These include private hospitals providing outpatient and inpatient services, general practitioners,
laboratory and diagnostic facilities, physiotherapy and rehabilitation services, ambulance services, home
nursing care services, pharmacies and pharmaceutical companies. Most of the private hospital beds are
located in the Western province and a few in other major cities. The bulk of private healthcare services
are delivered by four providers based primarily in Colombo: Nawaloka, Asiri Hospital Holdings, Lanka
Hospitals, and Durdans. As of 2011, the private sector’s share in total number of hospitals and hospital
beds was 17 and 6 percent, respectively. Of the 4210 private hospital beds in the same year, 50 percent
were in Colombo (Rannan-Eliya et al. 2012).
The public sector has most of the country’s health manpower and inpatient facilities. The public
health sector employs more than 90 percent of all doctors and nurses, constituting nearly 120,000 staff
across the country (Department of Census and Statistics, 2013). There are 87 medical officers, 202 nurses,
and 42 midwives per 100,000 population (Ministry of Health, 2015) (Table 6.1). Approximately 3.8 hospital
beds are available per 1,000 population in the public sector (approximately 80,581 beds) (Figure 6.1). One
of the limitations of the public health system is that outpatient services are open only until mid-afternoon
(except emergency care, which is available 24/7). This inconvenience makes some people seek outpatient
care in private facilities despite the cost. Other than this, geographic access to public health services does
not appear to pose challenges. It is estimated that every Sri Lankan is, on average, no more than 1.4
kilometers away from a basic health clinic and 4.8 kilometers from a public health care facility.
76
Table 6.1. Key Health Personnel in Sri Lanka, 2015
Total Rate per 100,000 population
Medical Officers 18,243 87
Dental Surgeons 1,340 6.4
Assistant Medical Officers 936 4.5
Nurses 42,420 202.3
Public Health Nursing Sisters 290 1.4
Public Health Inspectors 1,604 7.7
Public Health Midwives 6,041 28.8
Hospital Midwives 2,765 13.2
Source: Annual Health Bulletin 2015, Annex Table 9 (Ministry of Health, 2015)
Figure 6.1 Hospital Beds in Sri Lanka, 2015
82000 4
Number of hospital beds
Number of hospital beds
80000
per 1000 population
3.8
78000
76000 3.6
74000
3.4
72000
70000 3.2
2011 2012 2013 2014 2015
Hospital Beds Hospital Beds per 1000 population
Source: Annual Health Bulletin 2015 (Ministry of Health, 2015)
The private sector is largely staffed by off-duty public sector doctors. Private involvement in the
country's healthcare sector began in the 1980s when government-educated and employed doctors were
also permitted to consult privately on their own time. In 2011, only 700 medical officers worked fulltime
in the private sector. Of the 16,500 medical officers in the public sector, close to thirty percent worked
part-time in private hospitals (Rannan-Eliya et al. 2012).
Preventive and curative health services in the public sector are organized through two parallel
channels: (i) community health services focused on promotive and preventive care and (ii) curative
services ranging from non-specialized care at the primary level to specialized care at hospitals. Figure 6.2
shows this clear separation between preventive and curative service provision at the local level. At a
higher level, all preventive services and primary and secondary-level curative services are the
responsibility of nine Provincial Directors of Health Services (PDHS) and 26 Regional Directors of Health
Services (RDHS), whereas tertiary hospitals are the responsibility of the central Ministry of Health.
77
Figure 6. 2. Organization of Sri Lanka’s Health System
Sri Lankan health system
Public Private
Curative
Preventive and promotive
Medical Officer of Health Primary Secondary Tertiary
areas • Primary
Medical Care • District • Teaching
Units and General Hospitals
Maternity Hospitals • Provincial
Home • Base Hospitals General
• Divisional Hospitals
Hospitals
• MCH services
• Immunization
• Family planning Only outpatient Inward facility for 4 Inward facility
• Prevention of CDs and and obstetric • specialties:
main for main
care provided at Pediatrics, General specialties, sub-
NCDs etc.
medicine, Surgery, •
this level specialties and
• Well women clinic • Obstetrics and super specialties
• Environmental health Gynecology
• Occupational health etc.
• •
•
•
78
Preventive community health services are provided through a network of 341 health unit areas
across the country (Figure 6.3). Each area is led by a Medical Officer of Health (MOH) whose field team
(which includes a public health nursing sister, a public health inspector, a supervising public health
midwife, and a second public health midwife) provides preventive and promotive services on maternal
and child health, including the implementation of an expanded program of immunization, nutrition, family
planning, and the prevention and control of communicable diseases. The key responsibilities of the MOH
also include creating awareness of healthy lifestyles and referring patients for NCD screening. While this
cadre of professionals are largely responsible for Sri Lanka’s impressive MCH outcomes, so far they have
not been able to repeat the success for NCDs. It may be necessary to reorient the focus of these
community health services to give NCDs more attention. All services provided by the MOH are free of
charge and are provided at either the patient’s house or in the field clinics by field health staff under the
direct supervision of the MOH of the area.
Figure 6.3. Numbers of Preventive and Curative Government Healthcare Facilities 2011-2015
Hospitals
Primary medical care units/central dispensaries
MOH areas
800
638 621 624 622 631
600 459 487 461 475 473
327 337 334 338 341
400
200
0
2011 2012 2013 2014 2015
Public curative services are provided by a network of 1,104 government institutions ranging
from primary medical care units to teaching hospitals (which only provide specialized care), but there
is no effective referral system. All curative services are provided at three levels: (i) the primary level (in
primary medical care units and divisional hospitals); (ii) the secondary level (in district general hospitals
and base hospitals); and (iii) the tertiary level (teaching and special hospitals and provincial general
hospitals). According to the Annual Health Bulletin 2015 (Ministry of Health, 2015), there are 631 non-
specialist and specialist public hospitals in Sri Lanka (Table 6.2). The primary-level institutions are non-
specialist hospitals, while the secondary and tertiary-level hospitals provide specialized care. The level of the
institution determines the type of service that it provides. The primary-level facilities are staffed by one or
more medical officer with a basic degree and provide only outpatient care and/or obstetric care. Secondary-
level care institutions have in-ward facilities for the four main specialties (pediatrics, general medicine,
surgery, and obstetrics and gynecology). Tertiary care institutions have in-ward facilities for most sub-
specialties and for super-specialties in addition to the main specialties. These services are accessible to the
population with no restrictions in the absence of an effective referral system.
79
Table 6.2 Number of Hospitals by Type
Number Number
Teaching Hospital 16 Divisional Hospital 482
Provincial General Hospital 3 Maternity Homes 14
District General Hospital 20 Other Hospitals 25
Base Hospital 71
Total number of Hospitals 631
The health system allows patients to bypass lower levels of care even for conditions that do not
require specialist care, so, despite long waiting times, most patients seek care directly from secondary
and tertiary hospitals. This leads to an under�use of the primary care institutions and overcrowding at
secondary and tertiary care institutions. In 2012, outpatient clinic visits were divided roughly equally
across the primary, secondary, and tertiary-level facilities in the health system, whereas, ideally, 70 to 80
percent of all care should have been delivered at the primary care level, 10 to 15 percent at the secondary
level, and the remaining (most advanced) cases treated at the tertiary level.
The curative side of the public health system is not well suited to dealing with the overwhelming
burden of non-communicable diseases. The key feature of the public health system is its commitment to
free care at all levels, and it has achieved 100 percent coverage of key MCH services such as antenatal
care, institutional deliveries, and immunizations. However, this kind of care is episodic and curative rather
than the continuous and integrated care across all levels of care that is required for chronic NCDs (Table
6.3). The primary level facilities do not routinely initiate and coordinate NCD care. In the current system,
NCD patients cannot be tracked as they receive care from many different facilities at different times.
Tracking these patients is further complicated by the fact that the health information system is
Doctor-patient familiarity is impeded by the lack of a primary care model in Sri Lanka and a
culture of self-referral. One of the reasons why people prefer private over public care is the fact that they
cannot choose between doctors in public facilities but are assigned to whomever is available when they
reach the front of the queue. Given that NCDs require long-term integrated care, a strong doctor-patient
relationship is crucial to the effectiveness of the treatment. Introducing a primary care model based on
family physicians would make it possible for such relationships to be built and would enable patients to
become partners in managing their health. It would also shift the focus of the health system from illness
episodes to the health needs of the whole family. This is particularly important in Sri Lanka where adult
men tend to not use preventive health services. If a family physician was familiar with patterns and
differences between individuals in the same family, he or she would be able to provide effective
counseling on lifestyles, which are risk factors for NCDs.
80
Table 6.3. The Ideal Approach to Chronic NCD Care
Conventional approach to ambulatory medical
Ideal approach to addressing chronic NCDs care
care
• Focus on illness and curing the sick • Focus on the overall health needs of the
• Episodic curative care/discrete interventions family
• Responsibility limited to giving advice during • Comprehensive, continuous, and person-
consultation centered care
• Users are recipients of health/ medical • Responsibility for health throughout the
interventions lifecycle
• Disjointed care provided by fragmented • Making people partners in managing their
“stand-alone� facilities and programs. health
• Integrated delivery of care with strong
communication between levels of providers.
There is a pressing need in Sri Lanka for a strong primary healthcare system focused on chronic
conditions with a service delivery approach that seeks to improve the quality of care for patients by
ensuring that services are coordinated and tailored to their needs. The current system has a strong MCH
service base but only a weak system of community-based care. The Ministry of Health recently introduced
healthy lifestyle centers in an attempt to address NCDs, but they are not yet being widely used by the
population. An individual with a medical condition needs a continuum of care and services. People with
NCDs can experience obstacles at various stages of care, including getting a diagnosis, being directed to
care, initiating and adhering to treatment. All of these stages affect the desired outcome, which is disease
control. The “cascade of care� (diagnosis, link to proper care, adherence to needed care, and the
achievement of disease control) would be improved by the development of a strong primary healthcare
system for NCDs and by improvements in the paper-based health information system to facilitate the
tracking of patients referred to higher-level facilities.
81
Chapter 7. Conclusion and Recommendations
7.1. Summary and Conclusion
This study examined the prevalence and distribution of NCDs and associated physiological and
behavioral risk factors among adults. It also examined patterns of health service use and out-of-pocket
payments and their association with NCDs. The findings show that the most common NCDs (diagnosed)
posing a threat to healthy adult life in the province include hypertension, diabetes, cataracts, ischemic
heart disease and asthma. The less educated appear to have a higher burden of NCDs, but there is no
systematic difference by economic status. More concerning is the fact that the onset of these conditions
is early, suggesting that young adults live with health conditions for a substantial part of their lives. A
special inquiry into two physiological risk factors (obesity and hypertension) for various NCDs revealed
substantial concerns.
First, more than one in four adults (26 percent) are hypertensive, but as many as 70 percent of
them are unaware of their status. This lack of awareness is common across both genders, all places of
residence, and all socioeconomic groups but is more extensive among some groups than others. Despite
having a higher probability of being hypertensive, men are less likely to be aware of their hypertensive
status, suggesting that use of preventive care differs by gender. While hypertension does not seem to be
associated with socioeconomic status, the poorest and richest quintiles are more likely to be aware of their
conditions than the middle 60 percent. While hypertension does not seem to be associated with
socioeconomic status, the poorest and richest quintiles are more likely to be aware of their conditions
than the middle 60 percent. As expected, groups that are less aware of their status are those that have
the least contact with the health system (as suggested by the low utilization of outpatient care). Men and
relatively younger people also seem to have the lowest rates of preventive care utilization.
Second, the prevalence of general obesity in the province (15 percent) is substantially higher
than the average for Sri Lanka (6 percent). Combined with the prevalence of abdominal obesity68 (57
percent), this poses a significant risk for NCDs. Obesity disproportionately affects women and adults
between 30 and 60 years of age. Although the rate of obesity is the highest among the richest quintile, it
is alarmingly high even among the poor. Consistent with evidence from other countries, obesity in the
province was found to be significantly associated with the risk of hypertension and diabetes.
Moreover, taking public health actions solely based on BMI-based risk classifications excludes
a large proportion of adults, especially women, who are at a substantial risk of metabolic complications.
Waist circumference appears to predict the risk of diabetes better than BMI-based obesity. A significant
proportion of people in the normal BMI weight category is at risk of complications related to abdominal
obesity (Figure ES3). Given that the health risk of a given level of body fat in the Asian population is much
higher than in other regions, policymakers need to give this issue a considerable amount of attention.
68
Abdominal obesity is based on waist-circumference.
82
Behavioral risk factors such as smoking, betel chewing, and harmful alcohol consumption are
generally not very widespread, but their concentration among older men and lower socioeconomic
groups should be of concern for policymakers. The clustering of these risk factors among these vulnerable
groups may exacerbate the existing economic disadvantages for such groups. Furthermore, the risk of
hypertension is higher among those who practice these habits, which suggests that the potential health
benefits of expanding preventive counseling services, especially to men, are large.69
Environmental risk factors, particularly indoor air pollution, also increase the vulnerability of
lower socioeconomic groups and those living in the rural parts of the province . About 47 percent of
households in the Western province use unclean sources of fuel (43 percent use biomass and 4 percent
use kerosene). Our readings of indoor air pollution in 50 households showed levels of pollution that were
substantially higher than WHO’s interim target-1. Indoor air pollution is a greater risk factor for the health
of poorer households, as they are more likely to depend on biomass for fuel and to have no functional
chimney. Almost one-third of households in the poorest quintile cook with biomass but have no functional
chimney, compared to only 4 percent of households in the richest quintile. Children from households that
primarily use unclean sources of fuel are about 2 percentage points more likely (than children from
households that use clean sources of fuel) to have symptoms of wheezing and whistling in the chest.
The poorer populations tend to use public facilities, while private facilities are largely used by
the better-off, mainly for reasons of convenience and shorter waiting times . Among those who used
some outpatient or inpatient care, the probability of choosing private care rather than public care is higher
for the better-off, the better educated, and the elderly. Analysis of health-seeking behavior suggests that
much of this preference is driven by differences in waiting times and the soft skills of health providers
between public and private providers of care rather than by any major differences in infrastructure,
amenities, or perceived clinical quality. Quantitative data, however, also show that individuals prefer to
use public facilities for more serious healthcare concerns.
The data also revealed that a high proportion of people seeking care bypass primary care
facilities in favor of higher-level services. This is particularly true for adult preventive services but is rarer
for preventive MCH services. Consumers also often bypass primary care facilities for curative needs that
could be met at the primary level. For example, 63 percent of respondents reported that they would go
to a secondary facility for curative care for chronic conditions, while only 23 percent responded that they
would seek care from primary facilities. Furthermore, a significant proportion of households did not know
where they could seek care for adult preventive health services, such as counseling on nutrition. This an
issue on which policymakers need to focus.
Out-of-pocket (OOP) payments are modest given Sri Lanka’s per capita income, but ensuring
this does not escalate will require public health actions to control NCDs. Most out-of-pocket spending in
the province is for private care, but part of it is likely due to the unavailability of laboratory tests and
medications in public facilities, forcing patients to use private laboratories and pharmacies. Considering
the increasing prevalence of NCDs and physiological risk factors that are associated with higher healthcare
use and spending, containing OOP spending will require reducing the prevalence of these risk factors and
69
Partly because age and these lifestyles are correlated, this association vanishes in a multivariate regression.
83
removing the quality constraints that drive people away from public care, especially for routine
conditions. Early diagnoses and management of risk factors can reduce an otherwise costly treatment of
NCDs (both for individuals and the government).
The curative side of the public health system is not well suited to deal with the overwhelming
burden of non-communicable diseases. Primary level facilities provide facility based episodic NCD care,
and they do not routinely initiate or coordinate such care. A culture of self-referral and lack of an effective
gatekeeping mechanism produce discontinuity in client information between providers and constrain the
doctor-patient relationship. The ability to choose doctors appears to be an important factor driving
patients to utilize private facilities. Given that NCDs require long-term integrated care, however, a strong
doctor-patient relationship is crucial for the effectiveness of treatment. In the current system, NCD patients
cannot be tracked as they receive care from different facilities at different times, oftentimes resulting in lack
of continuity of care. Tracking these patients is further complicated by the absence of an electronic health
information system.
7.2. Recommendations
While the analyses were based on the Western province, recommendations are made for the
country as a whole. There are two reasons for this. First, the focus on the Western province is a strategic
one. Given the rate of urbanization and the relative homogeneity and size of the country, it can be argued
that other provinces will follow the trends exhibited by the Western province. As other provinces urbanize
and their socio-economic conditions and life styles change, they will be faced with similar challenges that
the Western province is facing now. Therefore, reorienting the health system in anticipation of these
challenges could lead to prevention of risk factors and early detection, which would result in significant
health gains nationwide. Second, several of the recommendations are systemic and institutional, and
therefore would apply to the whole country.
A multi-pronged approach, consisting of multi-sectoral preventive interventions, health system
reorientation and strengthening, and a targeted approach aimed at those most vulnerable to NCDs and
NCD risk-factors, is required to address the challenges posed by the behavioral, physiological, and
environmental risk factors for NCDs, as identified in this study. Vulnerable populations include men,
people with multiple risk factors and the poor (who suffer more from smoking, betel chewing, indoors
pollution, etc.).
Even though the health sector bears most of the burden of prevention and treatment of NCDs,
most interventions that could create health promoting environments lie outside the health sector.
Acknowledging this, Sri Lanka has recently approved a National Multi-sectoral Action Plan for the
prevention and control of NCDs (2016-2020) focusing on the following four strategic areas: i) leadership,
advocacy, and partnership; ii) health promotion and risk reduction; iii) reorientation of the health system
for early detection and management of NCDs and risk factors; and iv) surveillance, monitoring and
evaluation, and research. The recommendations listed below are consistent with this Action Plan and are
based on the findings presented in this study.
84
A. Interventions to control risk factors and prevent the onset of NCDs
I. Introducing and expanding population-based interventions: Population-based interventions
such as community-wide campaigns and national NCD literacy campaigns together with
regulations and corporate social responsibility can effectively reduce the trend in unhealthy aging
population. Such interventions are key as primary prevention of NCDs and are affordable even in
low income settings. They do not require health system strengthening and have low cost of
implementation. These interventions address not only those already suffering from NCDs but also
those who are the most exposed to NCD risk factors. It should be noted that most of the
recommended policies require a multi-sectoral approach. The Lancet NCD Action Group and the
NCD Alliance propose the delivery of five priority interventions based on their health effects, cost
effectiveness, low cost of implementation and political and financial feasibility (Beaglehole et al.
2011). Among these five set of interventions, four are population-based, namely:
a) Reduce unhealthy diet and promote physical activity: through mass-media campaigns,
fiscal measures (food taxes and subsidies), food labelling and marketing restrictions to
reduce unhealthy diet (such as saturated and trans-fat, and sugar in sweetened drinks).
Fiscal measures could be used not only to discourage the consumption of unhealthy diet
but also to promote the consumption of fruits and vegetables. Global experience has
shown that fiscal policies can be an important instrument to curb consumption of
unhealthy foods (Table 7.1). The empirical evidence suggests that fiscal policy
interventions are most effective for sugar-sweetened beverage taxes, reducing
consumption by 20-50 percent, and fruit and vegetable subsidies, increasing consumption
by 10-30 percent. Combining the two policies could also potentially reduce substitution
with unhealthy foods. Prior to implementing such policies, however, several important
questions need to be answered regarding the type and structure of taxes to use, which
products to tax, implications for revenue generation, and the distributional consequences
of such policies (particularly for the poor and other vulnerable groups). In addition, such
policies are often met with strong opposition from the private sector, and the government
should be prepared to address this. In many countries, civil society involvement has been
critical in the adoption and implementation of necessary legislation. It should be noted
that Sri Lanka has already taken some steps to address these issues.
85
Table 7.1. Effectiveness of fiscal policies on diet
Food/beverage taxes Nutrient-focused Subsidies
taxes
Effect on Strongest evidence for Reduce consumption Subsidies increase
consumption SSB taxes – reduce of target but may healthy food intake.
consumption by same increase consumption Strongest evidence for
percentage as tax rate. of non-target fruit and vegetable
nutrients; may apply subsidies.
to core foods; better if
paired with subsidy.
Effects on Substitution will affect Disease outcome Subsidies may also
body total calorie intake. affected by increase total calorie
weight/disease Most effective to substitution – nutrient intake and body weight.
outcomes target sugar- profile taxes less likely Very likely to reduce
sweetened beverages. to have unintended dietary NCD risk factors.
Limited evidence for effects than single
disease outcomes. nutrient-based taxes.
Differential May be most effective May be more likely to Mixed socioeconomic
effects for low-income have regressive effects status effects for
populations; may have as more likely to apply population subsidies,
greater effect on those to core foods. may benefit wealthy.
who consume most. Targeted low-income
subsidies effective.
Source: World Health Organization (2016). Fiscal Policies for Diet and Prevention of Non-Communicable Diseases
Countries that have already introduced diet-related fiscal policies include Ecuador,
Mexico, Thailand, and Hungary. Others, such as Colombia and the Philippines, are in the
process of adopting such policies. In Hungary, the introduction of a public health product
tax (PHPT) resulted in 26-32% of consumers decreasing the intake of products subject to
PHPT. While price increases were the primary reason for behavior change, a significant
share of consumers (26-32 percent depending on food categories) indicated that they
reduced consumption as a result of higher health consciousness. Importantly, the policy
was particularly effective among individuals most at risk, as consumers with poor self-
reported health status were twice as likely to decrease consumption of foods subject to
PHPT than those of good health. In addition, the policy had an impact on the food
industry. Almost 40 percent of manufactures reformulated their products, 30 percent
completely removed unfavorable ingredients, and 70 percent reduced the number of
unfavorable components in the product.
86
Table 7.2. Examples of taxes on drinks and foods in other countries
Country Year Foods Taxed Tax Rate
Denmark 2011 Products with more than Kr16/kg (£1.76; €2.15; $2.84) of
2.3% of saturated fat: meat, saturated fat
dairy products, animal fats,
and oils
Fiji 2006 Soft drinks 5% on imported drinks
Finland 2011 Soft drinks and Soft drinks €0.075/L
confectionery (£0.06; $0.10); confectionery
€0.75/kg
France* 2012 SSBs and artificially €0.11/1.5L
sweetened beverages
French 2002 Sweetened drinks, 60 francs/L
Polynesia* confectionery, and ice- (£0.41; €0.55; $0.66) for
cream imported drinks
Hungary* 2011 Foods high in sugar, fat or 10 forint (£0.03; €0.04; $0.05)
salt, and sugary drinks per item
Mexico 2014 Non-dairy SSBs and high SSBs: 1 peso/L
energy dense foods (EDF) EDFs: ad valorem tax of 8% for a
defined list of non-essential
foods containing ≥275
calories/100g
Nauru 2007 Sugar, confectionery, 30% import levy
carbonated drinks, cordial,
and flavored milks
Norway 1981 Sugar, chocolate, and Variable
sugary drinks
Samoa 1984 Soft drinks 0.40 tala/L
(£0.11; €0.14; $0.18)
South Africa Proposed for SSBs (exemptions: pure fruit Each gram above a threshold of
implementation juices and milk-based 4g/100ml is taxed at
in 2017 drinks) R0.021/gram
United Proposed for SSBs (exemptions: pure fruit 24p/L if sugar content
Kingdom* implementation juices and milk-based >8gr/100ml; 18p/L if sugar
in 2018 drinks) content of 5-8 gr/100ml
United States Various Sugar-sweetened soft drinks 1-8%
in 23 states (SSSDs and
other foods in 35 states)
Source: Landon (2012) and Hagenaars et al (2017).
*At least a portion of revenues is earmarked for health.
A number of countries have already introduced mandatory front-of-package labels,
including the United Kingdom, Ecuador, Chile, Mexico. In 2012, Chile passed legislation
(Law 20.606 on Food Nutritional Composition and Food Marketing) aimed to reduce
consumption of unhealthy foods in order to address the growing obesity epidemic in the
87
country. Specifically, the law restricted food marketing to children under 14 years of age,
regulated school food environment, and required front of package warning labels (Figure
7.1). The Ministry of Health established thresholds for calories, saturated fat, sodium, and
sugar (decreasing them at 24 months and 36 months after implementation). Foods that
exceeded these cutoff values had to be labeled accordingly. Focus groups were conducted
to determine which labels would be most effective and recognizable, particularly for
children. Unlike the commonly used traffic light front-of-pack labeling, Chile chose a black
sticker to convey the information. While front-of-package (FOP) labels have been found
to be more effective than mandated nutritional information (Becker et al., 2015), the
design of the labels also appears to be important. A recent study found that the black
warning labels can be more effective in influencing children’s choice as compared to the
traffic light warnings (Arrua et al., 2017).
Figure 7.1. Chile’s Front of Package Warning Food Labeling
Source: Ministry of Health Chile (2017). Translation (from left to right): 1) high in calories, 2) high in saturated fat, 3) high in
sodium, 4) high in sugar.
b) Reduce consumption of dietary salt: through mass-media campaigns and voluntary
action by food industry to reduce salt content of processed foods. Sri Lanka could use
innovative public health campaigns to inform households about the health risks of dietary
salt they put in the food they prepare at home. Also, the government can nudge the
private sector to change industry norms with regards to salt and fat content of processed
foods as it is inevitable that consumption of processed foods will increase with
urbanization. A number of countries have already introduced policies to reduce salt
consumption (Table 7.3). Recognizing the difficulty of establishing strict regulation on salt
use and high costs associated with enforcement, in 2002 the United Kingdom decided to
implement voluntary targets of salt reduction for a range of processed foods. The
government collaborated with industry representatives to establish a set of guidelines
and specific targets for salt reduction. As a result, salt intake fell from 9.5 grams in 2005
to 8.1 grams per day in 2009. The success of the program has been largely achieved as a
result of an effective partnership approach with the industry, with the government
showing a keen interest in understanding the technical barriers that producers face (He
et al., 2014). Similarly, Kuwait also introduced voluntary targets by targeting one of the
main sources of sodium in the country – bread. The Ministry of Health was able to
88
negotiate with the local producer, responsible for more than 80% of bread production in
Kuwait, to reduce the amount of salt by 10% (WHO, 2016c). Working together with
universities, research institutions, and associations representing the baking and food
industries, the Federal Ministry of Health in Argentina introduced nutritional guidelines for
salt consumption, a coordinated national plan for salt reduction, and a law reducing access
to salt shakers in restaurants (Meiro-Lorenzo et al., 2011). Understanding the potential
constraints, South Africa’s government gave the industry three years to make the necessary
changes in order to meet the newly stipulated legislation regarding salt levels.
c) In addition to legislation, communication campaigns could also be effective at reducing
sodium consumption. The communication for behavioral impact (COMBI) approach
promoted by the WHO uses multiple communication channels to encourage schools,
communities, health service providers, and local authorities to take action to reduce salt
consumption. China’s Shandong Ministry of Health Action on Salt Reduction and
Hypertension (or SMASH) initiative is a good example of this multifaceted approach.
Working together with restaurants, SMASH developed sodium standards for Shandong
cuisines, produced lower salt menus, and conducted communication activities to raise
awareness about recommended salt levels. A mid-term evaluation showed a decline in
salt consumption from 12.5 grams per day in 2011 to 11.6 grams per day in 2013 among
adults 18-69 (WHO, 2016c).
Table 7.3. Examples of countries with legislation on salt reduction
Mandatory salt targets Argentina (most foods), Belgium (bread), Bulgaria (bread,
milk products, meat products and lutenica), Greece
(bread, tomato products), Hungary (bread), Netherlands
(bread), Paraguay (bread), Portugal (bread), South Africa
(most foods)
Taxation on high salt foods Fiji (tax on MSG), Hungary, Portugal
Regulation on Front of Pack Labeling Chile, Ecuador, Finland, Indonesia, Korea (on children’s
foods), Mexico, Portugal, Thailand (on 5 snack food
categories)
Standards for salt as part of procurement Argentina, Brazil, Bulgaria, Cook Islands, Costa Rica,
policies in public institution settings Estonia, Finland, France, Greece, Hungary, Israel, South
Korea, Kuwait, Latvia, Lithuania, Malaysia, Mexico,
Romania, Slovenia, Spain, Sweden, USA, and UK
Source: Trieu et al. (2015)
d) Control tobacco use through accelerated implementation of the WHO Framework
Convention on Tobacco Control. It is to be noted that Sri Lanka already has a robust
tobacco taxation policy but it could benefit from other complimentary interventions such
as prohibiting illicit trade of tobacco products and banning point of sale display and all
other forms of advertising. Regulating the content and emissions of tobacco products
could further reduce the health damage caused by the use of tobacco. In April 2017, Sri
Lanka was selected to receive technical support under the WHO FCTC 2030 project aimed
89
at strengthening the Framework Convention on Tobacco Control (FCTC) measures. While
Sri Lanka has already made significant progress in implementing tobacco control
measures, additional steps could be taken to strengthen FCTC implementation.
Specifically, increasing the excise tax rate from the existing 63 percent to the WHO-
recommended level of 75 percent could further reduce the prevalence of smoking and
deter the young population from initiating smoking. While countries often fail to raise
tobacco taxes due to their perceived regressivity, a number of recent studies have found
that tobacco taxes can indeed be pro-poor policies (Verguet et al., 2015, Postolovska et
al., 2017, World Bank, 2017). Not only do they avert premature mortality, but they also
improve financial risk protection by reducing the burden of tobacco-related diseases and
associated out-of-pocket expenditures. Prohibiting smoking in restaurants and other
public spaces is also important and will require not only regulation but also significant
enforcement to be effective. In addition, the availability of cessation services and help-
lines could assist current smokers to stop smoking. The introduction of plain packaging as
has been done in Australia could also significantly reduce the attractiveness and appeal
of tobacco products. Studies have also shown that plain packaging increases the salience
of health warnings on cigarette packs and leads to further reductions in the prevalence of
smoking than large health warnings alone (Borland et al., 2013).
Figure 7.2. Example of plain packaging for tobacco products
Source: WHO (2016). Plain packaging of tobacco products
90
e) Reduce harmful alcohol consumption: through tax increase, banning advertisements and
restricting access. Enforcement of the National Alcohol Policy and establishment of
mechanisms to reduce the production and sale of illicit alcohol would also help.
II. Targeted campaigns promoting healthy behavior: To maximize impact, customize campaign
messages for different target groups and use tailored platform to communicate messages.
a) Campaigns on behavioral risk factors: The relatively high prevalence of smoking,
excessive alcohol use, and betel chewing among the socio-economically disadvantaged,
men and the elderly suggest that these groups of people may not fully appreciate the
health risks of such lifestyle choices. Campaign message with hard-hitting evidence on
health effects of these lifestyle related risk factors such as smoking could be designed
such that they are appealing to these groups of population. The platform used for such
campaigns could also be tailored to these population groups. In addition, to deter early
initiation of unhealthy behavior, Sri Lanka could design and implement school health
programs. There are a few notable examples of such campaigns in other countries. In
2008, Australia launched the Measure-Up campaign aimed at reducing the prevalence of
non-communicable disease by raising awareness between waist measurement, physical
activity, healthy eating, and obesity risk among adults. The focus was on waist
circumference as an indicator of an unhealthy lifestyle. Qualitative research prior to the
introduction of the campaign found that this was a compelling, credible and easy to
understand message. The campaign resulted in high unprompted and prompted
awareness (38 percent and 89 percent, respectively), but there were no significant
changes in reported consumption of fruits and vegetables or physical activity (King et al.,
2013). Brazil was able to achieve higher levels of physical activity through a community
based program - the Academia da Cidade program (ACP). The ACP provides free
supervised leisure-time physical activity sessions, nutrition education, and health
monitoring (including blood pressure measurements, anthropometric and nutritional
assessments) in Recife, Brazil. The city identifies public spaces, such as parks, beaches,
and recreation centers, and conducts the necessary renovations to ensure that the space
can be used for ACP. Physical education teachers are paid for the by the city government
(Simoes et al., 2013).
b) Campaigns on utilization of preventive check-ups and counseling services: Campaigns
that motivate people to have regular preventive check-ups and counseling services should
especially target young adult men who are found to forgo such health services. This would
help to delay the onset of NCDs and provide early treatment. It is, however, important to
ensure that the elderly are not left behind, as they continue to be the most vulnerable
due to their age and high prevalence of NCDs and behavioral risk factors. This is
particularly important considering Sri Lanka’s rapid aging of the population and the
implications this will have on future healthcare costs. The private sector could also play
an important role in raising awareness about healthy behaviors and increasing uptake of
91
services. In the United Arab Emirates, for example, the Ministry of Health formed a public
private partnership with Bin Sina Pharmacy in an effort to combat the rising rates of
obesity and diabetes. Bin Sina provides health examinations, assistance, and advice on
cholesterol, blood pressures, diabetes, and obesity to everyone who visits their outlets at
a subsidized price. Almost 28,000 people participated in the first year of the program, of
whom 27% were diagnosed with high levels of cholesterol (Meiro-Lorenzo et al., 2011).
c) Campaigns on healthy weight: Campaigns on the health benefits of maintaining a healthy
weight and how to achieve it should target women and younger adults who are at a higher
risk of being overweight. This effort should raise awareness not only about the BMI-based
risk of body fat but also about the health risks of abdominal obesity. These campaigns
could involve messages on healthy foods and food based dietary guidelines, unhealthy
diet (both for food prepared at home and purchased processed foods), and the health
benefits of physical activity. The “Agita Sao Paulo� program in Brazil encourages citizens
to adopt an active lifestyle by doing at least 30 minutes of moderate physical activity per
day. It targets three main groups: students, workers, and the elderly. The program is
known for its multi-sectoral approach, broad use of partnerships, and simple messaging,
and has been replicated in other parts of Brazil and the region (Meiro-Lorenza et al.,
2011). Worksite interventions to address nutrition and physical activity have also been
found to be moderately effective (Anderson et al., 2010). Table 7.4 presents a list of
strategies that have been found to be effective in changing individual behaviors.
B. Health system reorientation and strengthening
I. Introducing integrated and continuous care with primary care as default first contact: The
study’s findings of widespread risk factors, low awareness of health conditions and ineffective
management of diagnosed conditions suggest that Sri Lanka’s health system is not as effective
at dealing with NCDs as it has been for maternal and child health. The system provides facility
based episodic care, but there is no routine initiation and coordination of NCD care at the
primary level. A new NCD case is typically diagnosed within an outpatient department or in
the hospital during inpatient admission, and its management is usually centered around a
single disease by a specialist rather than person-centered care that primary care providers
could provide. Effective management of NCDs is also constrained by a culture of self-referral,
which limits doctor-patient familiarity. As such, establishing an integrated NCD care system
that goes beyond a facility based episodic care to reach and screen those who forgo
preventive care and ensure necessary follow-up is essential. By institutionalizing primary care
as the first point of contact, a more productive doctor-patient relationship can be established.
The following actions are needed for the introduction of an effective and integrated chronic
care model primary care service delivery:
a) Constitute primary care teams to enable provision of comprehensive NCD care. This
requires appropriately training providers to meet complex NCD needs, including facility
based health promotion and behavior change services.
92
b) Regularly assess the capacity of the health system and the pillars of health service
delivery (human resource, facilities and drugs) to provide high quality integrated primary
NCD care services. Evaluating existing public facilities in terms of availability of essential
NCD drugs and diagnostic facilities and investing to fill these gaps would help contain OOP
expenses related to NCD care.
c) Develop referral chains after careful geographic mapping of appropriate facilities to
complete the feedback loop and ensure the continuum of care.
d) Invest in an electronic information system to enable the transfer of patient information
between providers in the integrated delivery of care.
II. Institutionalize primary care level opportunistic NCD screening and counseling. Develop and
implement basic health services such as screening services for blood pressure, cholesterol and
diabetes, and interpersonal communication program (for improved diet and life style).
Provision of life-style counseling to care seekers would enhance self-regulatory behavior.
Opportunistic screening is a cost-effective way of reaching those who are less likely to seek
preventive and counseling services (e.g. men).
III. Improve soft skills of providers: Improve communication skills of public healthcare providers
through appropriate training and performance based rewards. One of the major reasons
reported for choosing private over public care is the ‘hostile attitude’ of staff in public
facilities. Unlike acute episodic care, NCD care requires continuous interaction with health
care providers for which patient comfort and trust are essential. Providers could receive
additional training on the importance of good communication with patients. The Institute of
Medicine (IOM) identifies patient-centered care as one of the six main elements of high-
quality care. Specifically, patient-centered care is defined as “respecting and responding to
patients’ wants, needs and preferences, so that they can make choices in their care that best
fit their individual circumstances� (IOM, 2001). A recent review found that patient-centered
care was positively associated with clinical effectiveness and safety consistently across a
range of disease areas, study designs and settings (Doyle et al., 2013). Patient experience or
responsiveness, however, is not commonly used to assess the performance of individual
health providers, but several countries have included patient experience indicators in pay-for-
performance schemes. The UK, for example, includes an indicator to track the length of a GP
consultation (at least 10 minutes) in its Quality and Outcomes Framework (QOF) program
(OECD, 2014). Brazil’s Social Organizations in Health (OSS) performance scheme for hospitals
includes two performance indicators on patient satisfaction: percentage of patient complaints
addressed and completion of patient satisfaction surveys. Meanwhile, Turkey’s performance
based contracting scheme in family medicine includes an administrative system comprising
35 indicators, among which are abiding with working hours and duties, maintenance and
security of health records, and ensuring patient confidentiality (World Bank, 2013). Box 7.1
provides an example of an effective integrated primary care model in Costa Rica.
93
Box 7.1. Costa Rica’s EBAIS Primary Care Model
Costa Rica’s Equipo Basico de Atencion Integral de Salud (EBAIS; or basic integrated health team) model
is another prime example of an integrated primary care approach. Initiated in 1995, the program aims to
ensure first contact access, comprehensiveness, continuity, and coordination of care. Each team consists
of a physician, nurse, technical assistant (similar to a community health worker), a medical clerk, and a
pharmacist. All providers are trained to provide all primary care (from prenatal to geriatric care), and each
team member has a clearly defined role. The physician is responsible for the provision of curative and
preventive care, while the nurse performs basic clinical tasks and provides health counseling. The
technical assistant is responsible for health promotion activities, disease prevention, epidemiological data
collection, basic sanitation services, identification of disease risk factors, and referrals. Technical
assistants conduct home and community visits (e.g. churches, schools) and can also follow up with
patients who miss their appointments. The medical clerk conducts patient registration and data
collection, while the pharmacist is responsible for dispensing prescribed medicines. Individuals are
assigned to an EBAIS team based on where they live, with the goal of 4500 patients per team. Maintaining
a reasonable patient to clinician ratio, geographic empanelment ensures that individuals have access to
providers and results in relatively low wait times. Data collection is a critical element, as all information
(collected from both home and clinical visits) is sent to the health area administration and subsequently
to the Social Security Agency (Caja Costarricense de Seguro Social or CCSS). The CCSS in turn uses the
information to revise targets in the management contracts for each area. If an area fails to meet the
targets, together with the CCSS it develops an action plan to improve performance. In 2014, EBAIS teams
conducted 75% of all medical consultations in Costa Rica.
Source: Pesec et al. (2017)
Implementing these recommendations will have economic benefits both at the micro and
macro levels. At the micro level, the prevention and early management of NCDs protects households from
loss of productivity and financial risk due to high OOP payment for medical care. At the macro level, the
fiscal implications of rising NCDs could be substantial. While the aging of the population will result in
higher costs for public provision of health care, NCDs could reduce the tax base of the economy by
affecting productivity and labor supply. Given the large societal costs of premature mortality and
morbidity due to NCDs, there is a strong case for investment in prevention and management of NCDs. Yet,
the question of fiscal sustainability of these reforms remains to be explored, especially if these
interventions are to be financed from the existing health budget. Efficiency gains from the gate-keeper
system could be one source of fiscal space to strengthen the health system and launch population-wide
interventions. Notwithstanding this, proper examination of the fiscal implications of these interventions
requires scrutiny. It is also worth noting that given the developmental threats that NCDs pose in aging
populations, there is a case for the Ministry of Finance to allocate more resources to the health sector.
94
Table 7.4: Interventions on Diet and Physical Activity: Summary Results from a Systematic Review
Settings Impacts Examples
Policy and Effective
environment interventions
• Government regulatory policies to support a healthier composition of staple foods (e.g. replacing palm
with soya oil reduces the saturated fatty acid content of the oil).
• Environmental interventions targeting the built environment, policies that reduce barriers to physical
activity, transport policies and policies to increase space for recreational activity.
• Point-of-decision prompts to encourage using the stairs (e.g. information on the benefits of physical activity
beside elevators and stairs)
Moderately
effective
interventions • Pricing strategies (fiscal policies) and point-of-purchase prompts in grocery stores, vending machines,
cafeterias and restaurants to support healthier choices
• Multi-targeted approaches to encourage walking and cycling to school, healthier commuting and leisure
activities
Mass media Effective Mass media campaigns promoting physical activity: with community-based, supportive activities such as
interventions programs in schools and local communities; or associated with policies to address local environmental
barriers to participation
Moderately
effective
interventions
• Intensive mass media campaigns using one simple message, e.g. increasing consumption of low-fat milk,
or fruit and vegetables
• National "health brand" or logos to assist consumers to make healthy food choices
• Long-term, intensive mass media campaigns promoting healthy diets
95
Settings Impacts Examples
School Effective High-intensity school-based interventions that focus on diet and/or physical activity, are comprehensive, multi-
settings interventions component and include:
- curriculum on diet and/or physical activity taught by trained teachers
- supportive school environment/policies
- a physical activity program
- a parental/family component
- healthy food options available through school food services: cafeteria, vending machines, etc.
Moderately
effective
interventions
• A focused approach, for example programs aimed at reducing sedentary behavior and increasing
participation in physical activity, accompanied by supportive activities within the curriculum
• A formative assessment that addresses the needs of the school and cultural contexts
Workplace Effective Multi-component programs promoting healthy dietary habits and/or physical activity, that:
interventions
- provide healthy food and beverages at the workplace facilities, e.g. in the cafeteria or vending
machines
- provide space for fitness or signs to encourage the use of stairs
- involve workers in program planning and implementation
- involve the family in interventions through self-learn programs, newsletters, festivals, etc. or
- provide individual behavior change strategies and self-monitoring
Community Effective
interventions
• Diet education programs that: target high-risk groups (e.g. menopausal, pre-diabetic women); and are
multi-component;
• Community development campaigns with intersectoral cooperation and/or focused on a common goal
(e.g. reduction in cardiovascular disease risk)
• Group-based physical activity programs or classes for a homogenous group of individuals
96
Settings Impacts Examples
Moderately Interventions that use an existing phone-in service to provide dietary advice
effective
interventions
- Community-wide interventions conducted as part of a national or global campaign (e.g. healthy
lifestyles strategy or “Healthy Village�) in a homogenous community
- Programs that target low-income/low literacy populations and include diet education in the standard
program
- Computer/web-based interventions with interactive personalized feedback, targeting high-risk groups
Supermarket tours and on-site educational programs to support the purchase of healthier foods
- Walking school bus
Primary care Effective Interventions targeting chronic NCD risk groups that:
interventions
- include persons who are inactive, consume less than five servings of fruits and vegetables daily, consume
a lot of dietary fat, are overweight, or have a family history of obesity, heart disease, cancer and/or type 2
diabetes and
- include at least one session (health risk appraisal) with a health-care professional, with brief negotiation or
discussion to decide on reasonable, attainable goals, and a follow-up consultation with trained personnel
who are supported by targeted information and are linked and/or coordinated with other stakeholders
such as community sports organizations or ongoing mass media physical activity campaigns
Moderately • Cholesterol screening programs that provide clients with their results and follow-up
effective education, ideally in person
interventions • Weight loss programs using health professionals with:
- personal or telephone/Internet consultations over a period of at least four
weeks, and
- a self-help program that includes self-monitoring.
Source: WHO (2009)
Note: Evidence from a systematic review. Interventions were labelled effective if the study had a robust experimental design, sufficient sample size, and
significant effects on specified outcome variables. They were also determined to be generalizable to other settings. Moderate interventions lacked one or more
elements but were determined to be sufficiently robust in certain settings.
97
References
Amarasekera, N., N. Gunawardena, N. De Silva, & A. Weerasinghe, 2010. "Prevalence of childhood
atopic diseases in the Western Province of Sri Lanka", Ceylon Medical Journal, 55(1), pp. 5-
8.Anderson, L. M., Quinn, T. A., Glanz, K., Ramirez, G., Kahwati, L. C., Johnson, D. B., ... & Katz, D. L.
(2009). The effectiveness of worksite nutrition and physical activity interventions for controlling
employee overweight and obesity: a systematic review. American journal of preventive
medicine, 37(4), 340-357.
Arrúa, A., Curutchet, M. R., Rey, N., Barreto, P., Golovchenko, N., Sellanes, A., ... & Ares, G., 2017. Impact
of front-of-pack nutrition information and label design on children's choice of two snack foods:
Comparison of warnings and the traffic-light system. Appetite, 116, 139-146.
Beaglehole, R., R. Bonita, R. Horton, C. Adams, G. Alleyne, and P. Asaria. 2011. "Priority Actions for the
Non-communicable Disease Crisis." The Lancet 377 (9775): 1438-47
Becker, M. W., Bello, N. M., Sundar, R. P., Peltier, C., & Bix, L., 2015. Front of pack labels enhance
attention to nutrition information in novel and commercial brands. Food policy, 56, 76-86.
Bonilla-ChacÃn, M. E., Iglesias, R., Suaya, A., Trezza, C., & MacÃas, C., 2016. Learning From The Mexican
Experience With Taxes On Sugar-Sweetened Beverages And Energy-Dense Foods Of Low
Nutritional Value. Health, Nutrition and Population Discussion Paper.
Borland, R., Savvas, S., Sharkie, F., & Moore, K., 2013. The impact of structural packaging design on
young adult smokers9 perceptions of tobacco products. Tobacco Control, 22(2), 97-102.
Chartier, R., M. Phillips, P. Mosquin, M. Elledge, K.Bronstein, S. Nandasena, V. Thornburg, J. Thornburg,
C. Rodes. 2016. "A comparative study of human exposures to household air pollution from
commonly used cookstoves in Sri Lanka," International Journal of Indoor Environment and Health
27(1): 147-159
Chu, N.-S., 2001."Effects of Betel Chewing on the Central and Autonomic Nervous Systems," Journal of
Biomedical Science, 8(3): 229-236
Courtemanche, C., R. Tchernis, and B. Ukert, 2016. "The effect of smoking on obesity: evidence from a
randomized trail," NBER. Working paper 21937
Dalpatadu, S., P. Perera, R. Wickramasinghe, & R. Rannan-Eliya, n.d. Public Hospital Governance in Sri
Lanka: A Case Study on Processes and Performance, s.l.: s.n.
Department of Census and Statistics, 2009. Sri Lanka Demographic and Health Survey, 2006-2007,
Colombo: Ministry of Healthcare and Nutrition.
Department of Census and Statistics, 2012. Census of Population and Housing , 2012, Colombo: Ministry
of Healthcare and Nutrition.
98
Department of Census and Statistics, 2013. Annual Health Bulletin, Colombo: Department of Census and
Statistics .
Department of Census and Statistics. 2015. Household Income and Expenditure Survey data 2012/13
Deurenberg-Yap, M., G. Schmidt, W.A. van Staveren, P. Deurenberg, 2000. "The paradox of low body
mass index and high body fat percentage among Chinese, Malays and Indians in Singapore,"
International Journal of Obesity and Related Metabolic Disorders, 24(8): 1011-1017
Doyle, C., Lennox, L., & Bell, D., 2013. A systematic review of evidence on the links between patient
experience and clinical safety and effectiveness. BMJ open, 3(1), e001570.
Epstein, F. H. 1989. "The relationship of lifestyle to international trends in CHD," International Journal of
Epidemiology 18 (supplement 1): S203-S209
Hagenaars, L. L., Jeurissen, P. P. T., & Klazinga, N. S., 2017. The taxation of unhealthy energy-dense foods
(EDFs) and sugar-sweetened beverages (SSBs): An overview of patterns observed in the policy
content and policy context of 13 case studies. Health Policy, 121(8), 887-894.
He, F. J., Brinsden, H. C., & MacGregor, G. A., 2014. Salt reduction in the United Kingdom: a successful
experiment in public health. Journal of human hypertension, 28(6), 345.
Huxley R, S. Mendis, E. Zheleznyakov, S. Reddy, J. Chan, 2010. "Body mass index, waist circumference
and waist: hip ratio as predictors of cardiovascular risk – a review of the literature," European
Journal of Clinical Nutrition, 64(1):16�22.
IHME, 2017. Global Burden of Disease Profile Sri Lanka. Institute for Health Metrics and Evaluation,
Seattle, USA
Institute of Medicine, 2001. Crossing the Quality Chasm: A New Health System for the 21st Century.
Washington, DC: The National Academies Press.
Jha, P., R. Nugent, S. Verguet, D. Bloom, and R. Hum, 2013. Disease Control Priorities in Developing
Countries. 3rd edition, Working Paper 2.
Jones, A., 2011. "Models for health care" in: D. Hendy and M. Clements (eds.), Oxford Handbook of
Economic Forecasting. Oxford: Oxford University Press.
King, E. L., Grunseit, A. C., O’Hara, B. J., & Bauman, A. E., 2013. Evaluating the effectiveness of an
Australian obesity mass-media campaign: how did the ‘Measure-Up’campaign measure up in New
South Wales?. Health education research, 28(6), 1029-1039.
Ko, Y-C., Y-L. Huang, C-H. Lee, MJ. Chen, LM. Lin, CC. Tsai, 1995. "Betel quid chewing, cigarette smoking
and alcohol consumption related to oral cancer in Taiwan." Journal of Oral Pathology and
Medicine 24(10): 450-453
Kotelawala, H. “Sri Lanka Death Toll rises in Garbage Dump Collapse� The New York Times. 17 April 2017.
Google News Web. Accessed on 20 May 2017
99
Kumara, A. & R. Samaratunga, 2016. "Patterns and determinants of out-of-pocket health care
expenditure in Sri Lanka: evidence from household surveys," Health Policy and Planning, 31(8):
970-983.
Landon J, and H. Graff. 2013., “What is the Role of Health-related Food Duties? A Report of a National
Heart Forum Meeting held 29th June 2012�, London: National Heart Forum; 2012
http://nhfshare.heartforum.org.uk/RMAssets/NHFMediaReleases/2012/Health-
related%20food%20duties%20meeting%20report%20FINAL.pdf, accessed 4 June 2015.
Lapidus L, C. Bengtsson, B. Larsson, K. Pennert, E. Rybo, L.Sjostrom, 1984. "Distribution of adipose tissue
and risk of cardiovascular disease and death: a 12 year follow up of participants in the population
study of women in Gothenburg, Sweden," British Medical Journal, 289 (6454): 1257-1261.
Larsson, B., K. Svardsudd, L. Welin, L. Wilhelmsen, P.Bjorntorp, G. Tibblin. 1984. "Abdominal adipose
tissue distribution, obesity, and risk of cardiovascular disease and death: 13 year follow up of
participants in the study of men born in 1913," British Medical Journal, 288 (6428): 1401-1404
Lin, W., F.X. Pi-sunyer, C. Liu, TC. Li, CI. Li, CY. Huang, CC, Lin., 2009. "Betel nut chewing is strongly
associated with general and central obesity in Chinese male middle-aged adults," Obesity (Silver
Spring) 17(6): 1247-1254
Meiro-Lorenzo, M., Villafana, T., & Harrit, M. (2011). Effective Responses to Non-communicable
Diseases. Health, Nutrition, and Population Discussion Paper, World Bank, Washington, DC.
Ministry of Health, 2015. “Annual Health Bulletin�. Colombo.
http://www.health.gov.lk/moh_final/english/public/elfinder/files/publications/AHB/2017/AHB%2
02015.pdf
Ministry of Health, 2016. "Sri Lanka National Health Accounts 2013". The Ministry of Health, Sri Lanka
Colombo
Ministry of Megapolis and Western Development, 2016. Western Region Megapolis Master Plan - From
Island to Continent, Colombo: Ministry of Megapolis and Western Development.
Nandasena, S., A. Wickremasinghe, & N. Sathiakumar, 2010. "Air Pollution and Public Health in
Developing Countries: Is Sri Lanka Different?" Journal of the College of Community Physicians of Sri
Lanka, 17(1): 15-20.
O’Donnell, O., E. van Doorslaer, RP. Rannan-Eliya, A. Somanathan, SR. Adhikari, B. Akkazieva, D.
Harbianto, CC. Garg, P. Hanvoravongchai, AN. Herrin, 2008. Who pays for healthcare in Asia?
Journal of Health Economics, Volume 27, pp. 460-75.
OECD., 2014. Paying for Performance in Health Care Implications for Health System Performance and
Accountability: Implications for Health System Performance and Accountability. OECD Publishing.
Pearson, T.A, D.T. Jamison, J. Trejo-Gutierrez, 1993. "Cardiovascular disease" in D.T Jamison, W.H.
Mosley, A.R. Measham, and J.L.Bobadilla, JL. (eds). DCP in Developing Countries, Oxford Medical
Publication, pp 577-592
100
Pesec, M., Ratcliffe, H. L., Karlage, A., Hirschhorn, L. R., Gawande, A., & Bitton, A., 2017. Primary Health
Care That Works: The Costa Rican Experience. Health Affairs, 36(3), 531-538.
Pinkowish, M. 1999. "Hand in glove: smoking cessation and weight gain" Patient Care 33(2): 134
Postolovska, I., Lavado, R. F., Tarr, G., & Verguet, S., 2017. Estimating the Distributional Impact of
Increasing Taxes on Tobacco Products in Armenia (No. 26386). The World Bank.
Qiao, Q. and R. Nyamdorj, 2010. "The optimal cut-off values and the performance of waist
circumference and waist-hip ratio for diagnosing type II diabetes," European Journal of Clinical
Nutrition. 64(1): 23-29
Rannan-Eliya, R. P. & L. Sikurajapathy, 2009. "Sri Lanka: Good Practice In Expanding Health Care
Coverage, " Colombo: Institute of Health Policy.
Rannan-Eliya RP., I.K. Liyanage, J. Jayanthan, N. Wijemanne, S. de Alwis, S. Amarasinghe, I. Siriwardana,
S. Dalpatadu and P. Cooray. 2012. Study of Quality of Care in Public and Private Sectors. Institute
for Health Policy, Sri Lanka.
Rimm, E.B., P. Williams, K. Fosher, M. Criqui, and M.J. Stampfer, 1999. "Moderate alcohol intake and
lower risk of coronary heart disease: meta-analysis of effects on lipids and haemostatic factors,"
BMJ 319:1523–8
Roth, G.A., M.H. Forouzanfar, A.E. Moran, AE., R. Barber, G. Nguyen, VL. Feigin, M. Naghavi, GA.
Mensah, C JL. Murray, 2015. “Demographic and Epidemiologic Drivers of Global Cardiovascular
Mortality,� The New England Journal of Medicine, 342:14
Russell, S., 2004. “The economic burden of illness for households in developing countries: a review of
studies focusing on malaria, tuberculosis, and human immunodeficiency virus/acquired
immunodeficiency syndrome,� American Journal of Tropical Medicine and Hygiene, Volume 71
(Supp.2), pp. 147-155.
Santos Silva, J. & S. Tenreyro, 2006. "The log gravity," Review of Economics and Statistics, 4(88), pp. 641-
658.
Schultz, T.P. and A. Tansel, 1997. "Wage and labor supply effects of illness in Cote d'Ivoire and Ghana:
instrumental variable estimates for days disabled," Journal of Development Economics 53(2): 251-
286
Seidell, JC., 2010. "Waist circumference and waist-hip ratio in relation to all-cause mortality, cancer and
sleep apnea," European Journal of Clinical Nutrition, 64(1): 35-41
Shaten, B.J., G.D. Smith, L.H. Kuller, e, JD. Neaton, 1993. "Risk factors for the development of type II
diabetes among men enrolled in the usual care group of the Multiple Risk Factor Intervention
Trial," Diabetes Care 16(10): 1331-1339
Simoes, E. J., Hallal, P., Pratt, M., Ramos, L., Munk, M., Damascena, W., ... & Brownson, R. C., 2009.
Effects of a community-based, professionally supervised intervention on physical activity levels
among residents of Recife, Brazil. American Journal of Public Health, 99(1), 68-75.
101
Sumpter, C. C. D., 2013. "Systematic review of meta-analysis of the associations between indoor air
pollution and tuberculosis," Tropical Medicine and International Health, 18(1), pp. 101-108.
Trieu, K., Neal, B., Hawkes, C., Dunford, E., Campbell, N., Rodriguez-Fernandez, R., ... & Webster, J.,
2015. Salt reduction initiatives around the world–a systematic review of progress towards the
global target. PloS one, 10(7), e0130247.
United Nations, Department of Economic and Social Affairs, Population Division, 2015. World Population
Prospects: The 2015 Revision, DVD Edition. Data retrieved on May 16, 2017
Vaughan, J.P., L. Gilson, and A. Mils, 1993. ‘Diabetes’ in D.T. Jamison, W.H. Mosley, A.R. Measham, and
J.L. Bobadilla, JL. (eds). DCP in Developing Countries, Oxford Medical Publication, pp 577-592
Verguet, S., Gauvreau, C. L., Mishra, S., MacLennan, M., Murphy, S. M., Brouwer, E. D., ... & Jamison,
D. T., 2015. The consequences of tobacco tax on household health and finances in rich and poor
smokers in China: an extended cost-effectiveness analysis. The Lancet Global Health, 3(4), e206-
e216.
Wagstaff, A., 2007. " The economic consequences of health shocks: evidence from Vietnam." Journal of
Health Economics 26(1): 82-100
Wakefield, M., Germain, D., Durkin, S., Hammond, D., Goldberg, M., & Borland, R. (2012). Do larger
pictorial health warnings diminish the need for plain packaging of cigarettes?. Addiction, 107(6),
1159-1167.
Wakefield, M. A., Loken, B., & Hornik, R. C. (2010). Use of mass media campaigns to change health
behaviour. The Lancet, 376(9748), 1261-1271.
Weeraratne, B., 2016. Can we produce better estimates of urbanization in Sri Lanka? Talking Economics,
the Blog of IPS, Sri Lanka
World Bank. 2013. Turkey - Performance based contracting scheme in family medicine : design and
achievements. Washington DC ; World Bank.
http://documents.worldbank.org/curated/en/880861468338993763/Turkey-Performance-based-
contracting-scheme-in-family-medicine-design-and-achievements
World Bank, 2015a. Sri Lanka- Ending Poverty and Promoting Shared Prosperity: A Systematic Country
Diagnostic, Washington DC: The World Bank.
World Bank, 2015b. Sri Lanka Country Snapshot, Colombo: The World Bank Group.
World Bank, 2017. Tobacco Tax Reform: At the crossroads of health and development. World Bank,
Washington, DC.
http://documents.worldbank.org/curated/en/491661505803109617/pdf/119792-REVISED-oct3-
pm-v2-PUBLIC-FINALWBGTobaccoTaxReformFullReportweb.pdf
World Development Indicators, 2014. Data retrieved May 1 2017 from World Development Indicators
(WDI) Online database, World Bank
World Development Indicators, 2015. Data retrieved May 1 2017 from World Development Indicators
(WDI) Online database, World Bank
102
WHO, 1985. Diabetes Mellitus. Report of a WHO Study Group. Technical Report Series 727, World Health
Organization, Geneva.
WHO, 2004. Appropriate body mass index for Asian populations and its implications for policy and
intervention strategies. Lancet, 363(9403): 157-163
WHO, 2005. WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur
dioxide: Global update 2005: Summary of risk assessment. World Health Organization, Geneva.
WHO, 2008a. Waist Circumference and Waist-Hip Ratio: Report of a WHO Expert Consultation. World
Health Organization, Geneva.
WHO, 2008b. 2008-2013 Action plan for the global strategy for the prevention and control of NCD, 2008-
13. Geneva, World Health Organization, Geneva.
WHO, 2009. Interventions on Diet and Physical Activity: What works. Summary Report. World Health
Organization, Geneva.
WHO, 2010. The World Health Report - Health Systems Financing: the Path to Universal Coverage. World
Health Organization, Geneva.
WHO, 2011. "Waist Circumference and Waist-Hip Ratio: Report of a WHO Expert Consultation." World
Health Organization, Geneva.
WHO, 2013. Population-based approaches to childhood obesity prevention. World Health Organization.
Geneva. http://apps.who.int/iris/bitstream/10665/80149/1/9789241504782_eng.pdf?ua=1
WHO, 2014a. Non-communicable diseases (NCD) Country profiles, 2014. World Health Organization,
Geneva.
WHO, 2014b. WHO Guidelines for Indoor Air Quality- Household Fuel Combustion, World Health
Organization, Geneva.
WHO, 2015a. Sri Lanka: WHO statistical profile. World Health Organization, Geneva.
WHO, 2015b. Non-communicable disease risk factor survey Sri Lanka: STEPS Survey Sri Lanka 2015.
World Health Organization, Geneva.
WHO, 2016a. Diabetes country profiles, 2016. World Health Organization, Geneva.
WHO, 2016b. Plain packaging of tobacco products. Evidence, design, and implementation. World Health
Organization, Geneva.
WHO, 2016c. SHAKE the Salt Habit. The SHAKE Technical Package for Salt Reduction. World Health
Organization: Geneva. http://apps.who.int/iris/bitstream/10665/250135/1/9789241511346-
eng.pdf?ua=1
WHO, 2017. Non-communicable diseases, fact sheet, accessed on May 17, 2017, available at
http://www.who.int/mediacentre/factsheets/fs355/en/. World Health Organization, Geneva.
Yilma, Z., A. Mebratie, R. Sparrow, M. Dekker, G. Alemu, AS. Bedi, 2014. “Channels of impoverishment
due to ill-health in rural Ethiopia.�. Institute of Social Studies Working Paper No. 592.
103
Yoshida, N., R. Munoz, A. Skiner, CK. Lee, M. Brataj, W. Durbin, D. Sharma, 2015. SWIFT Data Collection
Guidelines, Version 2, Washington DC: The World Bank Group.
104
Annex 1: Statistical Tables
Table A1. Distribution of Most Prevalent NCDs Across Socioeconomic Groups
Hyperten
observati
reported
Diabetes
Ischemic
Cataract
Mellitus
Asthma
disease
heart
NCD
Self-
sion
ons
Sex Male 23.5% 14.4% 13.8% 3.1% 1.9% 6.6% 1974
Female 22.2% 10.9% 14.9% 1.6% 2.0% 5.6% 4532
Residence Rural 20.2% 10.9% 14.4% 1.8% 1.9% 5.8% 2148
Urban 25.9% 13.2% 15.0% 2.3% 2.0% 5.8% 4358
Economic U60 22.4% 12.3% 14.4% 1.9% 2.0% 6.0% 5095
status B40 23.0% 10.2% 15.6% 2.4% 2.0% 5.0% 1411
Consumption Poorest quintile 23.0% 9.9% 15.5% 2.2% 1.9% 4.9% 1518
quintiles 2nd poorest 1355
quintile 24.1% 12.7% 15.1% 1.7% 1.5% 6.6%
Middle quintile 20.7% 11.9% 12.4% 1.7% 2.2% 5.6% 1157
2nd richest 1240
quintile 22.6% 13.1% 14.7% 1.7% 2.3% 7.5%
Richest 22.1% 12.1% 15.3% 2.5% 2.0% 4.5% 1236
Education Low educated 25.7% 12.9% 17.1% 2.2% 2.0% 6.8% 4942
Highly educated 14.0% 9.2% 8.1% 1.4% 1.8% 3.4% 1564
Age (10 year) Age 20-30 3.5% 0.9% 1.9% 0.2% 1.7% 0.4% 924
Age 30-40 7.5% 4.3% 3.8% 0.3% 1.1% 0.6% 1617
Age 40-50 17.8% 9.6% 9.5% 1.2% 1.4% 4.2% 1246
Age 50-60 30.1% 21.3% 22.4% 2.2% 2.5% 8.0% 868
Age>60 53.9% 24.7% 36.5% 5.9% 3.4% 16.2% 1851
Age (elderly) Age <60 13.8% 8.3% 8.5% 0.9% 1.6% 2.9% 4655
Age>60 53.9% 24.7% 36.5% 5.9% 3.4% 16.2% 1851
District Colombo 26.2% 13.0% 15.0% 2.7% 2.3% 5.7% 3915
Gampaha 21.2% 10.8% 13.9% 1.3% 1.8% 3.4% 1585
Kalutara 17.4% 11.5% 15.3% 1.9% 1.6% 10.5% 1006
All 22.5% 11.9% 14.6% 2.0% 2.0% 5.8% 6506
Note: The statistically significant difference in means (p-value<0.1) between socioeconomic groups are in bold.
105
Table A2. Probit (Marginal Effects) for the Probability of Self-reported and Diagnosed NCDs
(1) (2) (3) (4) (5) (6)
Self- Ischemic
reported heart
VARIABLES NCD Diabetes Hypertension disease Asthma Cataract
Male -0.0315** -0.00143 -0.0428*** 0.00658** -0.00262 -0.00986**
(0.0134) (0.00941) (0.00909) (0.00333) (0.00399) (0.00449)
3060 0.626*** 0.420*** 0.473*** 0.0753*** 0.0203** 0.247***
(0.0227) (0.0337) (0.0295) (0.0208) (0.00813) (0.0387)
Urban -0.000193 0.00442 -0.00654 -0.00186 -0.00908* 0.00885
(0.0168) (0.0116) (0.0124) (0.00376) (0.00546) (0.00641)
Education (at least
completed A/L) -0.0339** 0.00721 -0.0228** -0.000415 -0.00284 -0.0116**
(0.0143) (0.0102) (0.0105) (0.00336) (0.00425) (0.00513)
Unemployed 0.0246* 0.00472 0.00758 0.00135 -0.000876 -0.00304
(0.0140) (0.00928) (0.0100) (0.00292) (0.00428) (0.00510)
Gampaha -0.0522*** -0.00966 -0.0114 -0.00703** -0.00861** -0.00372
(0.0161) (0.0119) (0.0123) (0.00315) (0.00427) (0.00715)
Kalutara -0.117*** -0.00172 -0.00840 -0.00481 -0.00859** 0.0500***
(0.0157) (0.0131) (0.0139) (0.00295) (0.00394) (0.0127)
B40 -0.00330 -0.0178* 0.00765 -0.00241 0.000774 -0.00675
(0.0147) (0.00944) (0.0110) (0.00266) (0.00418) (0.00518)
Observations 6,499 6,499 6,499 6,499 6,499 6,499
Note: Robust standard errors in parentheses. *** p<0.01. ** p<0.05. * p<0.1. The reference group is women
aged 20 to 30 living in rural areas of Colombo, who have not completed an A/L level of education, are employed,
and belong to the B40 households in the national consumption expenditure. Regression also controls for ethnic
dummies.
106
Table A3. Probit (Marginal Effects) for the Probability of Self-reported and Diagnosed NCDs
(1) (2) (3) (4) (5) (6)
Self- Ischemic
reported heart
NCD Diabetes Hypertension disease Asthma Cataract
Male -0.0316** -0.00159 -0.0424*** 0.00648** -0.00264 -0.00993**
(0.0134) (0.00939) (0.00908) (0.00329) (0.00398) (0.00446)
3060 0.625*** 0.420*** 0.471*** 0.0736*** 0.0202** 0.247***
(0.0227) (0.0337) (0.0295) (0.0205) (0.00802) (0.0389)
Urban -0.000973 0.00417 -0.00941 -0.00231 -0.00924* 0.00959
(0.0169) (0.0117) (0.0126) (0.00374) (0.00544) (0.00638)
Education (at least
completed A/L) -0.0355** 0.00646 -0.0262** -0.00133 -0.00348 -0.0105**
(0.0145) (0.0105) (0.0108) (0.00322) (0.00425) (0.00532)
Unemployed 0.0248* 0.00470 0.00877 0.00131 -0.000686 -0.00333
(0.0141) (0.00928) (0.0100) (0.00286) (0.00426) (0.00507)
2nd poorest quintile 0.00680 0.0253* -0.00517 0.00454 -0.00375 0.0152*
(0.0180) (0.0133) (0.0126) (0.00420) (0.00455) (0.00802)
Middle quintile 0.00762 0.0222 -0.0179 0.00513 0.00358 0.00603
(0.0189) (0.0138) (0.0126) (0.00459) (0.00585) (0.00793)
2nd richest quintile 0.00709 0.0259* -0.00578 -8.91e-05 -0.00173 0.0143*
(0.0189) (0.0143) (0.0136) (0.00379) (0.00503) (0.00855)
Richest 0.0112 0.0231 0.00949 0.00873* 0.00168 0.00212
(0.0199) (0.0146) (0.0147) (0.00505) (0.00580) (0.00759)
Gampaha -0.0530*** -0.0100 -0.0134 -0.00731** -0.00895** -0.00295
(0.0162) (0.0120) (0.0123) (0.00305) (0.00405) (0.00721)
Kalutara -0.118*** -0.00198 -0.00894 -0.00459 -0.00868** 0.0501***
(0.0157) (0.0131) (0.0138) (0.00289) (0.00390) (0.0127)
Observations 6,499 6,499 6,499 6,499 6,499 6,499
Note: Robust standard errors in parentheses. *** p<0.01. ** p<0.05. * p<0.1. The reference group is women
aged 20 to 30 living in rural areas of Colombo, who have not completed an A/L level of education, are employed,
and belong to the B40 households in the national consumption expenditure. Regression also controls for ethnic
dummies.
107
Table A4. Distribution of Observed Hypertension
Hypertension* Observed hypertension status
Pre- Stage 1 Stage 2
Diagnosed Observed Normal hypertension hypertension hypertension
All 14.6% 26.1% 39.7% 34.2% 17.1% 9.0%
Male 13.8% 34.2% 24.9% 40.8% 23.2% 11.0%
Female 14.9% 23.3% 44.7% 32.0% 15.0% 8.3%
Rural 14.4% 25.5% 39.4% 35.1% 16.6% 8.9%
Urban 15.0% 26.9% 40.1% 33.0% 17.7% 9.2%
U60 14.4% 26.6% 38.7% 34.8% 17.5% 9.0%
B40 15.6% 24.2% 43.5% 32.3% 15.3% 8.9%
Poorest
quintile 15.5% 24.2% 44.0% 31.8% 15.4% 8.8%
2nd poorest
quintile 15.1% 25.8% 39.1% 35.1% 15.7% 10.1%
Middle quintile 12.4% 27.2% 40.2% 32.6% 19.2% 8.0%
2nd richest
quintile 14.7% 27.4% 34.3% 38.4% 18.9% 8.5%
Richest 15.3% 26.0% 40.2% 33.8% 16.4% 9.6%
Low educated 17.1% 28.5% 37.1% 34.4% 18.3% 10.1%
Highly
educated 8.1% 19.7% 46.6% 33.8% 13.7% 5.9%
Age <60 8.5% 19.4% 46.3% 34.2% 13.7% 5.8%
Age>60 36.5% 49.9% 15.7% 34.3% 29.3% 20.6%
Colombo 15.0% 26.8% 39.5% 33.7% 17.9% 8.8%
Gampaha 13.9% 19.2% 46.3% 34.5% 12.8% 6.4%
Kalutara 15.3% 36.8% 28.3% 34.8% 23.0% 13.9%
Note: *Test of difference in means is reported only for the first two columns. For these columns, the figures in bold
show statistically significant differences in means between categories (p-value<0.1).
108
Table A5. Probit for Observed Hypertension and Awareness of Being Hypertensive
(1) (2) (5) (6)
Hypertensive Hypertensive Aware being Aware being
(observed) (observed) hypertensive hypertensive
Obese (BMI>=30) 0.0882*** 0.0883***
(0.0191) (0.0191)
Daily smoker -0.0243 -0.0252
(0.0313) (0.0312)
Daily betel chewer 0.00792 0.00804
(0.0246) (0.0246)
Excess alcohol 0.0387 0.0394
(0.0427) (0.0427)
Male 0.0456*** 0.0457*** -0.128*** -0.127***
(0.0175) (0.0175) (0.0248) (0.0248)
3060 0.566*** 0.567*** 0.601*** 0.597***
(0.0248) (0.0249) (0.0988) (0.0989)
Urban 0.0307* 0.0317* -0.0320 -0.0500
(0.0179) (0.0179) (0.0336) (0.0343)
Education (at least
completed A/L) -0.0124 -0.0111 -0.0188 -0.0370
(0.0155) (0.0159) (0.0311) (0.0320)
Unemployed 0.0134 0.0127 0.00170 0.00534
(0.0150) (0.0150) (0.0273) (0.0272)
2nd poorest quintile 0.00653 -0.0666**
(0.0193) (0.0331)
Middle quintile 0.0174 -0.118***
(0.0206) (0.0318)
2nd richest quintile -0.00779 -0.0905***
(0.0195) (0.0344)
Richest 0.00584 0.00251
(0.0210) (0.0385)
Gampaha -0.0587*** -0.0585*** -0.0329 -0.0437
(0.0180) (0.0181) (0.0348) (0.0347)
Kalutara 0.134*** 0.135*** -0.0381 -0.0399
(0.0238) (0.0238) (0.0347) (0.0346)
B40 -0.00508 0.0816**
109
(0.0157) (0.0317)
Observations 5,796 5,796 1,782 1,782
Note: Robust standard errors in parentheses. *** p<0.01. ** p<0.05. * p<0.1.
The reference group is women who do not practice daily smoking, -daily betel chewing, or excess
alcohol consumption, are aged 20 to 30, live in rural areas of Colombo, have not completed an
A/L level of education, are employed, and belonging to either the B40 of households in the
national consumption expenditure or the poorest quintile of households in the sample
(depending on the specification). Regression also controls for ethnic dummies.
110
Table A6. Distribution of Obesity and Underweight Across Population Groups
Abdominal
obesity BMI-based classification
WC (IDF Obese Obese Obese
cut-off) Obesity Underweight Normal Overweight I II III
All 57% 14.5% 9.0% 44.5% 32.0% 10.9% 2.7% 1.0%
Male 34% 7.9% 9.2% 54.9% 28.0% 6.6% 1.1% 0.2%
Female 65% 16.8% 8.9% 41.0% 33.3% 12.3% 3.2% 1.3%
Rural 56% 12.9% 9.7% 45.5% 31.9% 10.0% 2.1% 0.8%
Urban 58% 16.9% 7.9% 43.1% 32.1% 12.1% 3.5% 1.3%
U60 58% 14.4% 7.9% 45.0% 32.8% 11.1% 2.4% 0.9%
B40 52% 15.1% 13.1% 42.8% 29.0% 9.9% 3.9% 1.3%
Poorest
quintile 52% 14.4% 12.9% 43.5% 29.2% 9.5% 3.6% 1.2%
2nd poorest
quintile 54% 14.3% 9.1% 47.5% 29.1% 10.7% 2.7% 1.0%
Middle
quintile 56% 13.3% 8.6% 45.6% 32.5% 10.3% 2.5% 0.5%
2nd richest
quintile 58% 14.3% 7.4% 43.2% 35.1% 11.6% 1.9% 0.8%
Richest 65% 16.5% 5.9% 42.8% 34.8% 12.5% 2.4% 1.5%
Low
educated 56% 15.2% 9.4% 44.0% 31.5% 11.2% 3.0% 1.0%
Highly
educated 59% 12.9% 8.0% 45.8% 33.4% 10.0% 1.9% 0.9%
Age <60 58% 16.3% 8.0% 41.7% 34.0% 12.1% 3.1% 1.1%
Age>60 54% 8.1% 12.5% 54.6% 24.7% 6.5% 1.1% 0.6%
Colombo 56% 16.0% 8.4% 43.4% 32.2% 11.6% 3.4% 1.0%
Gampaha 59% 14.3% 8.7% 42.3% 34.7% 10.4% 2.7% 1.2%
Kalutara 54% 11.9% 10.8% 50.8% 26.6% 10.0% 1.2% 0.6%
Sri Lanka
average
(STEPS 2015) N/A 5.9% 15.3% 55.4% 23.4% N/A N/A N/A
Note: *Test of difference in means is reported only for the first three columns. The statistically significant
differences in means between socioeconomic groups (p-value<0.1) in these three columns are reported in bold.
111
Table A7. Probit for Obesity and Underweight
General obesity (BMI≥30) Abdominal obesity (WC) Underweight (BMI<18.5)
(1) (2) (3) (4) (5) (6)
Daily smoker -0.0996** -0.0969** -0.116*** -0.106*** 0.162*** 0.159***
(0.0387) (0.0391) (0.0385) (0.0386) (0.0423) (0.0420)
Daily betel chewer -0.106*** -0.100*** -0.0737** -0.0620** 0.00345 0.000670
(0.0273) (0.0279) (0.0295) (0.0297) (0.0246) (0.0243)
Excess alcohol 0.0599 0.0656 -0.0272 -0.0298 0.0556 0.0559
(0.0684) (0.0690) (0.0488) (0.0487) (0.0468) (0.0469)
Male -0.151*** -0.154*** -0.316*** -0.321*** -0.0270 -0.0264
(0.0177) (0.0175) (0.0168) (0.0169) (0.0169) (0.0170)
3060 -0.0170 -0.0218 0.227*** 0.221*** -0.0270 -0.0255
(0.0256) (0.0255) (0.0202) (0.0203) (0.0188) (0.0189)
Urban 0.0133 0.00565 0.0187 0.00327 -0.0135 -0.00899
(0.0210) (0.0212) (0.0186) (0.0187) (0.0180) (0.0180)
Education (at least
completed A/L) -0.0464*** -0.0614*** 0.0413** 0.0159 -0.0458*** -0.0408**
(0.0175) (0.0176) (0.0166) (0.0171) (0.0156) (0.0161)
Unemployed -0.00786 -0.00454 -0.0135 -0.00655 0.0228 0.0206
(0.0182) (0.0182) (0.0166) (0.0167) (0.0160) (0.0161)
2nd poorest quintile -0.00973 0.0124 -0.0546***
(0.0227) (0.0208) (0.0160)
Middle quintile 0.00705 0.0332 -0.0591***
(0.0252) (0.0220) (0.0167)
2nd richest quintile 0.0393 0.0820*** -0.0594***
(0.0264) (0.0213) (0.0175)
Richest 0.0693** 0.145*** -0.0804***
(0.0276) (0.0213) (0.0170)
Gampaha -0.00663 -0.0132 0.0377* 0.0250 0.0151 0.0190
(0.0224) (0.0223) (0.0195) (0.0197) (0.0196) (0.0197)
Kalutara -0.0883*** -0.0920*** -0.000925 -0.00713 -0.000502 0.000944
(0.0210) (0.0207) (0.0220) (0.0220) (0.0209) (0.0209)
B40 -0.0121 -0.0637*** 0.0733***
(0.0191) (0.0175) (0.0177)
112
Observations 3,499 3,499 6,168 6,168 3,153 3,153
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Regression also controls for ethnic
dummies.
113
Table A8. Correlates of Diagnosed Diabetes
(1) (2) (3) (4)
Diagnosed Diagnosed Diagnosed Diagnosed
VARIABLES diabetes diabetes diabetes diabetes
Hypertensive 0.0358*** 0.0344*** 0.0383*** 0.0340***
(0.00940) (0.00900) (0.00915) (0.00928)
General obesity (BMI>=30) 0.0227* 0.00616
(0.0123) (0.0117)
Abdominal obesity (WC IFD cut-off) 0.0493*** 0.0453***
(0.00784) (0.00911)
Abdominal obesity (WHR WHO cut-off) 0.0267*** 0.0103
(0.00840) (0.00990)
B40 -0.0169* -0.0161* -0.0185** -0.0152
(0.00974) (0.00938) (0.00937) (0.00971)
Male 0.00755 0.0200* 0.000838 0.0241**
(0.0101) (0.0106) (0.00966) (0.0111)
3060 0.416*** 0.387*** 0.395*** 0.390***
(0.0377) (0.0361) (0.0363) (0.0377)
Urban 0.00493 0.00490 0.00615 0.00410
(0.0118) (0.0114) (0.0114) (0.0117)
Education (at least completed A/L) 0.00869 0.00663 0.00843 0.00595
(0.0106) (0.0102) (0.0103) (0.0104)
Unemployed 0.00353 0.00662 0.00522 0.00521
(0.00970) (0.00931) (0.00944) (0.00960)
Gampaha -0.000442 -0.00255 -0.00132 -0.00173
(0.0127) (0.0122) (0.0123) (0.0125)
Kalutara 0.000530 -0.00316 -0.00343 0.000913
(0.0135) (0.0126) (0.0128) (0.0133)
Observations 5,796 6,141 6,141 5,772
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Regression also controls
for ethnic dummies.
114
Table A9. Correlates of Observed Hypertension (probit)
(1) (2) (3)
VARIABLES Hypertensive Hypertensive Hypertensive
Obesity (BMI≥30) 0.0882*** 0.0548***
(0.0191) (0.0192)
Abdominal obesity
(WC IFD cut-off) 0.0899*** 0.0779***
(0.0123) (0.0133)
Smoker (daily) -0.0243 -0.0259 -0.0197
(0.0313) (0.0305) (0.0319)
Chew betel daily 0.00792 0.00643 0.0121
(0.0246) (0.0238) (0.0249)
Alcohol excessive 0.0387 0.0285 0.0412
(0.0427) (0.0410) (0.0429)
Male 0.0456*** 0.0800*** 0.0697***
(0.0175) (0.0180) (0.0184)
3060 0.566*** 0.540*** 0.552***
(0.0248) (0.0248) (0.0254)
Urban 0.0307* 0.0278 0.0303*
(0.0179) (0.0172) (0.0179)
Education (at least
completed A/L) -0.0124 -0.0179 -0.0150
(0.0155) (0.0150) (0.0155)
Unemployed 0.0134 0.0130 0.0153
(0.0150) (0.0147) (0.0151)
B40 -0.00508 0.000405 -0.000340
(0.0157) (0.0152) (0.0159)
Gampaha -0.0587*** -0.0652*** -0.0612***
(0.0180) (0.0172) (0.0180)
Kalutara 0.134*** 0.124*** 0.132***
(0.0238) (0.0228) (0.0239)
Observations 5,796 6,141 5,772
Note: Robust standard errors in parentheses.
*** p<0.01, ** p<0.05, * p<0.1. Regression also
controls for ethnic dummies.
115
Table A10. Distribution of Current Smoking Among Groups (both sex)
Current
smokers Daily smokers Age of initiation of daily smoking
All 4.4% 3.4% 22.0
Sex
Male 15.7% 12.3% 22.1
Female 0.3% 0.1% 18.0
Sector
Rural 4.5% 3.3% 22.1
Urban 4.4% 3.5% 21.7
Economic status
U60 4.1% 3.2% 21.7
B40 5.6% 4.1% 22.7
Poorest quintile 5.7% 4.1% 22.5
2nd poorest quintile 4.9% 3.9% 22.0
Middle quintile 4.4% 3.3% 21.2
2nd richest quintile 3.8% 3.0% 22.0
Richest 3.1% 2.2% 21.5
Educational status
Low educated 5.2% 4.0% 21.9
Highly educated 2.3% 1.7% 22.3
Age groups
Age <60 4.2% 3.1% 21.1
Age>60 5.3% 4.2% 25.1
District
Colombo 5.4% 4.2% 22.3
Gampaha 2.9% 2.0% 20.5
Kalutara 5.3% 4.2% 22.5
Note: The statistically significant differences in means between socioeconomic groups (p-value<0.1) are
reported in bold. The number of observations for the last column is 281.
116
Table A11. Correlates of Daily Smoking
(1) (2) (3) (4)
Smoke daily vs Smoke daily vs Smoke daily vs Smoke daily vs
VARIABLES not not not smoker not smoker
Male 0.118*** 0.116*** 0.123*** 0.120***
(0.0114) (0.0111) (0.0118) (0.0116)
3060 0.00502 0.00518 0.00528 0.00543
(0.00355) (0.00346) (0.00356) (0.00346)
Urban -0.00182 -0.00102 -0.00183 -0.000978
(0.00219) (0.00205) (0.00219) (0.00204)
Education (at least
completed A/L) -0.00570*** -0.00443*** -0.00567*** -0.00438***
(0.00168) (0.00166) (0.00166) (0.00163)
Unemployed -0.000466 -0.000908 -0.000539 -0.00102
(0.00193) (0.00184) (0.00191) (0.00182)
2nd poorest quintile -0.000842 -0.000940
(0.00184) (0.00179)
Middle quintile -0.00314** -0.00314**
(0.00158) (0.00155)
2nd richest quintile -0.00573*** -0.00578***
(0.00168) (0.00166)
Richest -0.00596*** -0.00597***
(0.00173) (0.00171)
Gampaha -0.00754*** -0.00710*** -0.00746*** -0.00698***
(0.00198) (0.00190) (0.00197) (0.00188)
Kalutara -0.000920 -0.000478 -0.000827 -0.000341
(0.00212) (0.00209) (0.00212) (0.00210)
B40 0.00504** 0.00517**
(0.00233) (0.00234)
Pseudo R2 0.2864 0.2929 0.2945 0.3015
Observations 6,499 6,499 6,420 6,420
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Regression also controls for
ethnic dummies.
117
Table A12. Socioeconomic and Demographic Correlates of Excessive Drinking
(1) (2)
Excess drinker Excess drinker
Male 0.0227*** 0.0227***
(0.00463) (0.00462)
3060 0.0299*** 0.0299***
(0.00976) (0.00968)
Urban -0.00758** -0.00736**
(0.00369) (0.00373)
Education (at least completed
A/L) -0.00217 -0.00168
(0.00277) (0.00294)
Unemployed -0.00604** -0.00608**
(0.00262) (0.00261)
2nd poorest quintile -0.00425
(0.00264)
Middle quintile -0.00743***
(0.00228)
2nd richest quintile -0.00616**
(0.00248)
Richest -0.00769***
(0.00245)
Gampaha -0.00103 -0.000846
(0.00311) (0.00313)
Kalutara 0.00310 0.00296
(0.00395) (0.00392)
B40 0.0101***
(0.00372)
Pseudo R2 0.1252 0.1254
Observations 6,499 6,499
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Regression also controls for ethnic dummies.
118
Table A 13. Distribution of Betel Chewing
Ever chew betel Daily chew betel
All 21.3% 6.5%
Female 15.5% 2.7%
Male 37.4% 17.0%
Rural 25.7% 8.3%
Urban 15.0% 4.0%
U60 20.4% 5.5%
B40 24.7% 10.3%
Poorest quintile 24.8% 9.9%
2nd poorest quintile 22.0% 7.2%
Middle quintile 21.1% 6.6%
2nd richest quintile 21.8% 5.4%
Richest 15.8% 2.7%
Low educated 23.5% 8.2%
Highly educated 15.4% 2.1%
Age <60 17.9% 4.6%
Age>60 33.6% 13.3%
Colombo 21.7% 6.5%
Gampaha 17.4% 6.3%
Kalutara 27.8% 7.1%
Note: The statistically significant differences in means between
socioeconomic groups (p-value<0.1) are reported in bold.
119
Table A14. Socioeconomic and Demographic Correlates of Betel Chewing
(1) (2) (3) (4)
Chew daily vs Chew daily vs never Chew daily vs
not Chew daily vs not tried never tried
Male 0.0759*** 0.0739*** 0.0888*** 0.0863***
(0.00880) (0.00866) (0.0102) (0.0101)
3060 0.0772*** 0.0760*** 0.0944*** 0.0918***
(0.0160) (0.0158) (0.0186) (0.0182)
Urban -0.0444*** -0.0388*** -0.0615*** -0.0539***
(0.00811) (0.00783) (0.0104) (0.00997)
Education (at least
completed A/L) -0.0298*** -0.0245*** -0.0342*** -0.0286***
(0.00406) (0.00415) (0.00433) (0.00435)
Unemployed -0.00298 -0.00418 -0.00649 -0.00798
(0.00509) (0.00493) (0.00562) (0.00546)
2nd poorest quintile -0.0129*** -0.0135***
(0.00469) (0.00515)
Middle quintile -0.0191*** -0.0214***
(0.00424) (0.00455)
2nd richest quintile -0.0224*** -0.0247***
(0.00422) (0.00454)
Richest -0.0367*** -0.0391***
(0.00388) (0.00429)
Gampaha -0.0175*** -0.0153*** -0.0246*** -0.0219***
(0.00500) (0.00500) (0.00530) (0.00527)
Kalutara -0.0111* -0.0102* -0.0114* -0.00985
(0.00585) (0.00574) (0.00651) (0.00645)
B40 0.0354*** 0.0381***
(0.00719) (0.00799)
Pseudo R2 0.1684 0.177 0.2041 0.2137
Observations 6,499 6,499 5,648 5,648
Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Regression also
controls for ethnic dummies.
120
Table A15. Households’ Primary Source of Fuel
Unclean (biomass
Electricity LPG Kerosene Biomass plus kerosene)
All 1% 52% 4% 43% 47%
Rural 1% 38% 2% 58% 60%
Urban 1% 73% 7% 19% 26%
Poorest quintile 1% 28% 7% 65% 72%
2nd poorest
quintile 1% 46% 5% 47% 53%
Middle quintile 2% 50% 3% 45% 48%
2nd richest
quintile 1% 59% 4% 35% 39%
Richest 2% 75% 1% 22% 23%
U60 1% 57% 4% 38% 42%
B40 1% 28% 7% 65% 71%
Colombo 1% 69% 6% 25% 31%
Gampaha 2% 45% 4% 49% 53%
Kalutara 1% 32% 2% 66% 67%
Note: The statistically significant differences in means between socioeconomic groups (p-
value<0.1) are reported in bold.
121
Table A16. Risky Cooking Practices
Traditional Cooking usually done Availability of functional
stove use in sleeping/living chimney (Who cook in the
(Biomass users) space (Biomass users) building using biomass)
All 58% 7% 74%
Sector of
residence
Rural 58% 7% 76%
Urban 58% 5% 65%
Economic status
Poorest quintile 60% 8% 62%
2nd poorest
quintile 60% 5% 71%
Middle quintile 54% 7% 79%
2nd richest
quintile 56% 7% 84%
Richest 61% 5% 86%
U60 57% 6% 79%
B40 61% 9% 60%
Poorest quintile 60% 8% 62%
2nd poorest
quintile 60% 5% 71%
Middle quintile 54% 7% 79%
2nd richest
quintile 56% 7% 84%
Richest 61% 5% 86%
Colombo 50% 6% 66%
Gampaha 67% 2% 79%
Kalutara 52% 15% 72%
Note: The statistically significant differences in means between socioeconomic groups (p-value<0.1)
are reported in bold.
122
Table A17. Hygiene, Sanitation, and Working Conditions (percentage of households)
Someone
Drinking works in
water source: Flush hazardous
piped in to toilet/pour Share work
dwelling/yard flush toilet toilet places
All 49% 92% 10% 17%
Rural 43% 90% 10% 16%
Urban 60% 96% 10% 18%
U60 49% 93% 9% 15%
B40 52% 87% 15% 27%
Poorest quintile 52% 87% 15% 27%
2nd poorest
quintile 48% 93% 11% 20%
Middle quintile 49% 90% 10% 16%
2nd richest
quintile 52% 95% 6% 13%
Richest 47% 95% 8% 8%
Colombo 66% 94% 14% 19%
Gampaha 42% 90% 8% 11%
Kalutara 32% 94% 6% 25%
Note: The statistically significant differences in means between socioeconomic
groups (p-value<0.1) are reported in bold.
123
Table A18. Correlates of Outpatient Healthcare Use (by source of care)
(1) (2) (3) (4)
Outpatient Outpatient Outpatient Outpatient
VARIABLES public care public care private care private care
2nd poorest quintile 0.0195* 0.0329**
(0.0111) (0.0158)
Middle quintile -0.00927 0.0507***
(0.0109) (0.0169)
2nd richest quintile -0.0156 0.0798***
(0.0110) (0.0178)
Richest -0.0410*** 0.0740***
(0.0112) (0.0190)
Head has >=A/L education -0.0381*** -0.0301*** 0.0378*** 0.0308**
(0.00882) (0.00924) (0.0126) (0.0127)
Poor living conditions 0.0153 0.0109 -0.0302** -0.0273**
(0.0100) (0.00988) (0.0134) (0.0135)
Unclean source of fuel 0.0376*** 0.0294*** -0.0271** -0.0220**
(0.00868) (0.00869) (0.0110) (0.0112)
Male -0.0136** -0.0122* -0.0162* -0.0174**
(0.00634) (0.00631) (0.00843) (0.00840)
- -
Age 0.00158*** 0.00142*** -0.00761*** -0.00773***
(0.000470) (0.000467) (0.000597) (0.000597)
4.75e- 4.59e-
Age square 05*** 05*** 0.000107*** 0.000108***
(5.82e-06) (5.78e-06) (7.78e-06) (7.79e-06)
Work hazard -0.00360 -0.00637 0.0395*** 0.0423***
(0.00919) (0.00907) (0.0130) (0.0131)
Unsanitary garbage disposal 0.00686 0.00621 0.00298 0.00232
(0.0138) (0.0136) (0.0179) (0.0178)
Urban 0.0233** 0.0260*** -0.00497 -0.00758
(0.00970) (0.00967) (0.0143) (0.0144)
Gampaha -0.00235 0.00238 0.0690*** 0.0662***
(0.0105) (0.0107) (0.0158) (0.0159)
Kalutara -0.0326*** -0.0289*** 0.0332* 0.0294*
(0.0106) (0.0106) (0.0172) (0.0171)
B40 0.00436 -0.0542***
(0.00930) (0.0118)
Observations 10,106 10,106 10,107 10,107
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Regression
also controls for ethnic dummies.
124
Table A19. Correlates of Healthcare Use
(1) (2) (3) (4)
VARIABLES Outpatient care Outpatient care Inpatient care Inpatient care
B40 -0.0455*** 0.00172
(0.0141) (0.00596)
Head has >=A/L education 0.00417 0.00550 -0.00454 -0.00440
(0.0140) (0.0143) (0.00555) (0.00572)
Poor living conditions -0.00710 -0.00847 -0.00517 -0.00541
(0.0154) (0.0154) (0.00613) (0.00611)
Unclean source of fuel 0.00129 -0.000176 0.00410 0.00405
(0.0128) (0.0130) (0.00523) (0.00534)
Male -0.0316*** -0.0315*** 0.00784* 0.00797*
(0.00965) (0.00965) (0.00450) (0.00451)
Age -0.00930*** -0.00926*** -0.000954*** -0.000949***
(0.000728) (0.000728) (0.000293) (0.000293)
Age square 0.000157*** 0.000156*** 1.76e-05*** 1.75e-05***
(9.73e-06) (9.72e-06) (3.64e-06) (3.64e-06)
Work hazard 0.0276* 0.0270* 0.0103* 0.0101
(0.0145) (0.0146) (0.00614) (0.00617)
Unsanitary garbage disposal 0.000626 -0.000379 -0.00843 -0.00835
(0.0206) (0.0205) (0.00711) (0.00710)
Urban 0.0161 0.0159 0.00186 0.00167
(0.0161) (0.0162) (0.00632) (0.00637)
Gampaha 0.0535*** 0.0545*** -0.00485 -0.00504
(0.0173) (0.0174) (0.00636) (0.00635)
Kalutara -0.0155 -0.0150 0.0263*** 0.0265***
(0.0181) (0.0181) (0.00933) (0.00934)
2nd poorest quintile 0.0502*** 0.00244
(0.0173) (0.00689)
Middle quintile 0.0336* -0.00388
(0.0183) (0.00708)
2nd richest quintile 0.0516*** -0.00411
(0.0189) (0.00723)
Richest 0.0335* 0.000392
(0.0203) (0.00805)
Observations 10,107 10,107 10,107 10,107
Note: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1. Regression also controls
for ethnic dummies.
125
Table A20. Correlates of Healthcare Use (adult sample)
(1) (2) (3) (4)
Outpatient Outpatient Inpatient Inpatient
care care care care
B40 -0.0408** -0.00680
(0.0173) (0.00711)
Head has >=A/L education -0.0224 -0.0212 -0.00969 -0.00944
(0.0171) (0.0175) (0.00662) (0.00680)
Poor living conditions -0.0364* -0.0380** -0.000975 -0.00150
(0.0191) (0.0192) (0.00832) (0.00816)
Unclean source of fuel -0.0108 -0.0128 0.00448 0.00439
(0.0158) (0.0160) (0.00645) (0.00658)
Male -0.0510*** -0.0510*** 0.0122* 0.0125**
(0.0134) (0.0134) (0.00630) (0.00629)
Age 0.00669*** 0.00667*** 0.000205 0.000210
(0.00247) (0.00247) (0.000992) (0.000987)
Age square -3.11e-06 -3.07e-06 4.14e-06 4.09e-06
(2.38e-05) (2.38e-05) (9.26e-06) (9.21e-06)
Underweight -0.0177 -0.0181 -0.00295 -0.00278
(0.0244) (0.0244) (0.00951) (0.00952)
Overweight 0.0625*** 0.0626*** -0.00877 -0.00825
(0.0154) (0.0154) (0.00617) (0.00615)
Obesity 0.0669*** 0.0668*** 0.00584 0.00608
(0.0204) (0.0203) (0.00927) (0.00922)
Hypertensive 0.0424*** 0.0427*** -0.00124 -0.00127
(0.0150) (0.0150) (0.00651) (0.00650)
Diabetic 0.273*** 0.273*** 0.0438*** 0.0435***
(0.0211) (0.0211) (0.0103) (0.0103)
Work hazard 0.0322* 0.0313* 0.0101 0.00996
(0.0186) (0.0186) (0.00784) (0.00783)
Unsanitary garbage disposal 0.0204 0.0196 0.000542 0.000369
(0.0269) (0.0269) (0.00949) (0.00941)
Urban 0.0439** 0.0438** -0.00363 -0.00417
(0.0195) (0.0197) (0.00808) (0.00813)
Gampaha 0.0469** 0.0480** -0.0197*** -0.0200***
(0.0210) (0.0211) (0.00708) (0.00701)
Kalutara -0.00147 -0.000811 0.0281** 0.0284**
(0.0228) (0.0229) (0.0116) (0.0116)
2nd poorest quintile 0.0435** 0.0147
(0.0213) (0.0100)
Middle quintile 0.0265 0.00297
126
(0.0225) (0.00963)
2nd richest quintile 0.0475** 0.00464
(0.0233) (0.00982)
Richest 0.0271 0.00970
(0.0241) (0.0110)
Observations 5,803 5,803 5,803 5,803
Note: Robust standard errors in parentheses;
*** p<0.01, ** p<0.05, * p<0.1. Regression also controls for ethnic dummies.
127
Table A21. Correlates of Outpatient Healthcare Use by Source of Care (adult sample)
(1) (2) (3) (4)
Outpatient Outpatient
Outpatient Outpatient private private
VARIABLES public care public care care care
2nd poorest quintile 0.0284** 0.00447
(0.0144) (0.0184)
Middle quintile -0.00789 0.0365*
(0.0139) (0.0197)
2nd richest quintile -0.00973 0.0678***
(0.0144) (0.0210)
Richest -0.0493*** 0.0694***
(0.0132) (0.0219)
Head has >=A/L
education -0.0483*** -0.0389*** 0.0274* 0.0173
(0.0104) (0.0110) (0.0148) (0.0147)
Poor living conditions -0.00573 -0.0109 -0.0306* -0.0260
(0.0124) (0.0120) (0.0157) (0.0159)
Unclean source of fuel 0.0417*** 0.0309*** -0.0391*** -0.0322**
(0.0111) (0.0109) (0.0129) (0.0133)
Male -0.0231*** -0.0205** -0.0176 -0.0207*
(0.00857) (0.00856) (0.0112) (0.0111)
Age 0.0112*** 0.0113*** -0.00182 -0.00184
(0.00177) (0.00178) (0.00194) (0.00193)
-6.86e- -6.97e-
Age square 05*** 05*** 4.31e-05** 4.31e-05**
(1.63e-05) (1.63e-05) (1.87e-05) (1.86e-05)
Underweight -0.0147 -0.0165 -0.0126 -0.0108
(0.0155) (0.0152) (0.0197) (0.0198)
Overweight 0.0313*** 0.0324*** 0.0255** 0.0240*
(0.0110) (0.0110) (0.0127) (0.0127)
Obesity 0.0389*** 0.0397*** 0.0327* 0.0302*
(0.0149) (0.0149) (0.0174) (0.0172)
Hypertensive -0.00286 -0.00263 0.0444*** 0.0446***
(0.00960) (0.00950) (0.0129) (0.0129)
Diabetic 0.143*** 0.142*** 0.105*** 0.105***
(0.0170) (0.0169) (0.0184) (0.0183)
Work hazard 0.00390 -0.000106 0.0345** 0.0389**
(0.0123) (0.0120) (0.0159) (0.0161)
Unsanitary garbage
disposal 0.0101 0.00943 0.0173 0.0169
(0.0178) (0.0175) (0.0224) (0.0223)
Urban 0.0337*** 0.0370*** 0.0165 0.0133
(0.0116) (0.0114) (0.0164) (0.0166)
Gampaha -0.00758 -0.00195 0.0712*** 0.0676***
(0.0127) (0.0129) (0.0187) (0.0188)
Kalutara -0.0232* -0.0189 0.0474** 0.0427**
(0.0137) (0.0137) (0.0204) (0.0203)
128
B40 0.00134 -0.0414***
(0.0118) (0.0142)
Observations 5,802 5,802 5,803 5,803
Note: Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1. Regression also controls for ethnic dummies.
129
Table A22. Probability of Using Private Care Among Those who Used Care during the Reference Period
(1) (2)
VARIABLES Private outpatient care Private outpatient care
2nd poorest quintile -0.00634
(0.0301)
Middle quintile 0.0677**
(0.0308)
2nd richest quintile 0.125***
(0.0290)
Richest 0.157***
(0.0304)
Head has >=A/L education 0.124*** 0.0975***
(0.0235) (0.0246)
Poor living conditions -0.0967*** -0.0839***
(0.0291) (0.0292)
Unclean source of fuel -0.113*** -0.0874***
(0.0236) (0.0242)
Male 0.00298 -0.00593
(0.0182) (0.0182)
Age -0.00599*** -0.00659***
(0.00125) (0.00125)
Age squared 3.23e-05** 3.76e-05**
(1.58e-05) (1.58e-05)
Work hazard 0.0626*** 0.0702***
(0.0241) (0.0242)
Unsanitary garbage disposal 0.0149 0.0166
(0.0367) (0.0362)
Urban -0.0459 -0.0528*
(0.0292) (0.0295)
Gampaha 0.0967*** 0.0893***
(0.0281) (0.0285)
Kalutara 0.127*** 0.119***
(0.0306) (0.0308)
B40 -0.0735***
(0.0279)
Observations 3,117 3,117
Note: Robust standard errors in parentheses;
*** p<0.01, ** p<0.05, * p<0.1. Regression also controls for ethnic dummies.
130
Table A23. Costs of Healthcare
Those Who Used Care All Individuals
% %
Total Share of incurred incurred
health forgone Total health Share of OOP OOP OOP OOP
cost income expenditure reimbursement payment payment payment payment
Male 2484 6.1% 2298 1.6% 98% 2209 34% 774
Female 2222 3.0% 2108 2.7% 98% 2038 34% 700
All 2312 4.1% 2173 2.3% 98% 2096 34% 725
Note: Costs and expenditure are reported for one month (in Sri Lankan rupees). Reimbursement was from an
insurance provider, government or someone else. The figures in bold show statistically significant differences across
gender.
131
Table A24. Distribution of OOP Spending
Those Who Used Care All Individuals
% who % who
incurred OOP OOP incurred OOP OOP
payment payment payment payment
All 98% 2096 34% 725
Male 98% 2209 34% 774
Female 98% 2038 34% 700
Rural 97% 2109 34% 733
Urban 98% 2076 34% 713
U60 98% 2237 35% 802
B40 97% 1526 29% 462
Poorest quintile 97% 1577 30% 492
2nd poorest
quintile 97% 1329 35% 485
Middle quintile 98% 2227 33% 747
2nd richest
quintile 98% 2574 37% 961
Richest 98% 3027 35% 1071
Low educated 98% 1931 34% 667
Highly educated 98% 2726 34% 945
Under 5 99% 1389 45% 625
5 to 19 97% 1408 31% 451
20 to 60 98% 2305 28% 665
Age>60 97% 2810 50% 1449
Colombo 97% 2210 32% 733
Gampaha 98% 1771 36% 653
Kalutara 97% 2549 32% 846
Adult (sample)
Not obese 97% 2367 33% 803
Obese 97% 3108 36% 1138
Not hypertensive 97% 2478 29% 751
Hypertensive 98% 2553 44% 1154
Not diabetic 97% 2350 29% 696
Diabetic 98% 2889 64% 1877
Note: Figures in bold show statistically significant differences across corresponding groups.
132
Table A25. Predictors of OOP Spending
(1) (2) (3) (4)
Log OOP
spending if
Probit (marginal OOP
VARIABLES GPML estimates Poisson GLM effects) spending>0
B40 -0.422*** -0.351*** -0.0473*** -0.345***
(0.0985) (0.113) (0.0142) (0.0711)
Head has >=A/L
education 0.352*** 0.318*** 0.00343 0.337***
(0.0885) (0.0816) (0.0141) (0.0626)
Poor living conditions -0.172* -0.246** -0.00362 -0.234***
(0.0967) (0.108) (0.0153) (0.0737)
Unclean source of
fuel -0.287*** -0.354*** 0.00318 -0.361***
(0.0796) (0.0775) (0.0129) (0.0568)
Male 0.00110 0.0830 -0.0253** 0.0521
(0.0669) (0.0716) (0.00995) (0.0479)
Age -0.0164*** -0.00748* -0.00977*** -0.00232
(0.00458) (0.00389) (0.000746) (0.00316)
Age squared 0.000436*** 0.000306*** 0.000164*** 0.000111***
(6.27e-05) (4.75e-05) (1.01e-05) (4.10e-05)
Work hazard 0.202** 0.204** 0.0254* 0.0568
(0.0943) (0.0943) (0.0146) (0.0662)
Unsanitary garbage
disposal 0.132 0.145 -0.000428 0.0217
(0.151) (0.158) (0.0205) (0.101)
Urban -0.120 -0.117 0.0190 -0.0178
(0.106) (0.105) (0.0161) (0.0693)
Gampaha 0.00681 -0.0776 0.0434** 0.0622
(0.109) (0.107) (0.0172) (0.0697)
Kalutara 0.182 0.117 0.00516 0.258***
(0.118) (0.129) (0.0186) (0.0859)
Constant 6.489*** 6.436*** 6.793***
(0.127) (0.122) (0.0917)
Observations 10,094 10,094 10,094 3,518
R-squared 0.089
Note: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1. Regression also controls
for ethnic dummies.
133
Table A26. Predictors of OOP Spending (adult sample)
(1) (2) (3) (4)
Log OOP
Probit spending if
GPML Poisson (marginal OOP
VARIABLES estimates GLM effects) spending>0
B40 -0.406*** -0.284* -0.0492*** -0.318***
(0.140) (0.154) (0.0173) (0.0936)
Head has >=A/L
education 0.179 0.228** -0.0191 0.308***
(0.116) (0.105) (0.0173) (0.0806)
Poor living conditions -0.269** -0.330** -0.0259 -0.373***
(0.136) (0.151) (0.0192) (0.1000)
Unclean source of
fuel -0.303*** -0.404*** -0.00289 -0.404***
(0.108) (0.104) (0.0159) (0.0760)
Male 0.140 0.234** -0.0439*** 0.201***
(0.104) (0.0990) (0.0139) (0.0697)
Age 0.00817 0.0104 0.00559** -0.0362***
(0.0176) (0.0186) (0.00254) (0.0120)
Age squared 0.000166 0.000137 9.10e-06 0.000392***
(0.000161) (0.000157) (2.46e-05) (0.000110)
Underweight 0.0210 -0.122 -0.0226 0.0582
(0.162) (0.168) (0.0248) (0.126)
Overweight 0.206* 0.103 0.0569*** -0.00448
(0.106) (0.104) (0.0156) (0.0720)
Obesity 0.425*** 0.404*** 0.0761*** 0.0811
(0.140) (0.137) (0.0207) (0.100)
Hypertensive 0.177* 0.0695 0.0427*** 0.0673
(0.0994) (0.108) (0.0155) (0.0713)
Diabetic 0.702*** 0.600*** 0.291*** 0.162**
(0.104) (0.103) (0.0210) (0.0738)
Work hazard 0.378*** 0.274** 0.0248 0.0668
(0.139) (0.121) (0.0187) (0.0903)
Unsanitary garbage
disposal 0.404* 0.271 0.0252 0.0800
(0.224) (0.202) (0.0262) (0.130)
Urban -0.176 -0.207 0.0448** 0.0612
(0.146) (0.133) (0.0197) (0.0900)
Gampaha -0.147 -0.201 0.0267 0.0961
(0.146) (0.137) (0.0211) (0.0899)
Kalutara 0.149 0.0934 0.0218 0.258**
134
(0.162) (0.166) (0.0234) (0.105)
Constant 5.710*** 5.834*** 7.589***
(0.467) (0.500) (0.320)
Observations 5,792 5,792 5,792 2,044
R-squared 0.096
Note: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1.
Regression also controls for ethnic dummies.
135