REDUCING TOBACCO
USE THROUGH TAXATION
IN TRINIDAD AND
TOBAGO: MODELLING
THE LONG TERM HEALTH
AND ECONOMIC IMPACT
REDUCING
TOBACCO USE
THROUGH
TAXATION
IN TRINIDAD
AND TOBAGO:
MODELLING
THE LONG
TERM HEALTH
AND ECONOMIC
IMPACT
I
CONTENTS
Acknowledgments VII
Abstract IX
Introduction 1
Summary of Methodology 5
Methodology 5
Assumptions 5
Limitations 6
Full Methodology 9
Data Collection 9
The Microsimulation Model 11
Development of Scenarios 12
Results 17
Smoking Prevalence (Percentage) 17
Summary 18
Cases of Disease per Year 20
Cumulative Cases of Disease 20
Mortality 21
Mortality cases avoided 22
Direct Cumulative Costs Avoided 22
Discussion 25
References 28
Appendix A: Microsimulation model 35
Module One: Preductions of Smoking Prevalence Over Time 35
Multinominal Logistic Regression for Smoking Prevalence 35
Bayesian Interpretation 37
Estimation of Confidence Intervals 38
Module Two: Microsimulation Initalization – Birth, Disease, and
Death Models 38
Population Models 39
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
The Risk Factor Model 42
Relative Risks 45
Modelling Diseases 45
Methods for Approximating Missing Disease Statistics 46
Approximating attributable cases 52
Potential Years of Life Lost 53
Premature Mortality Rate 53
Costs Module 53
Premature Mortality Costs 54
Propagation of errors equation 54
Software architecture 54
Aim of the Model 54
Summary of the Architecure of the Existing Model 55
Main C++ classes used by the model 56
Tperson C++ class 56
Tdisease C++ class 57
Tscenario C++ class 60
Appendix B: Cigarette Tax Scenarios Output, 20152017, Ukraine (Results from
TaXSim Modelling) 62
References 64
LIST OF FIGURES
Figure 1: Illustration of the microsimulation model 12
Figure 2: Male and female smoking prevalence by year for each scenario 17
Figure A1: Population pyramid, 2015, Ukraine 40
Figure A3: Exsmoker relative risks as a function of time after
smoking cessation 44
Figure A2: The model structure 56
Figure A3: Multistage disease architecture 58
IV // Contents
LIST OF TABLES
Table 1: Summary of total disease cases (epidemiological) and costs
(economic) by parameter, year, and scenario, total population (values
in parentheses are uncertainty values) 3
Table 2: Never, exsmoker, and smoker prevalence (%) by age group
and sex, 2011 9
Table 3: References for disease data 10
Table 4: TaXSiM Model results 15
Table 5: Smoking prevalence by year, sex and scenario (percentage) 17
Table 6: Summary table of total disease cases (epidemiological) and
costs (economic) by parameter, year, and scenario, total population 19
Table 7: Cases per year, total population 20
Table 8: Cumulative cases for each disease by year, total population 20
Table 9: Cumulative cases avoided relative to scenario 0 for the total
Trinidad and Tobago population by 2025 and 2035 21
Table 10: Mortality cases in the total population per year 21
Table 11: Mortality Cases avoided in the total population per year 22
Table 12: Direct cumulative healthcare costs avoided (million, TT$) 22
Table A1: Summary of the parameters representing the
distribution component 40
Table A2: Parameter estimates for γ0 and η related to each disease 43
Table A3: Survival percentage for lung cancer 47
Table A4: C++ Tperson class 57
Table A5: C++ Tdisease class 59
Table A6: The C++ Tscenario class 61
V
ACKNOWLEDGMENTS
This report was prepared by a team led by Patricio V. Marquez, Lead Public Health
Specialist, Health, Nutrition and Population Global Practice, World Bank Group. Team
members include: Lise Retat, Senior Economic and Mathematical Modeller, UK Health
Forum; Abbygail Jaccard, Chief Technology Officer, UK Health Forum; Laura Webber,
Deputy CEO, UK Health Forum; Karl Theodore, Director, HEU, Centre for Health Economics,
The University of the West Indies, St. Augustine, Trinidad; Althea La Foucade, Coordinator,
HEU, Centre for Health Economics, The University of the West Indies, St. Augustine,
Trinidad; Samuel Gabriel, Researcher, HEU, Centre for Health Economics, The University of
the West Indies, St. Augustine, Trinidad; Christine Laptiste, Research Fellow, HEU, Centre for
Health Economics, The University of the West Indies, St. Augustine, Trinidad.
The comments and advice provided by the following peer reviewers were incorporated
in the final version of this assessment: Sheila Dutta, Senior Health Specialist, World Bank
Group; Santiago Herrera, Lead Economist, World Bank Group; Alberto Gonima, Consultant,
World Bank Group.
Washington, DC
August 31, 2018
VII
ABSTRACT
Background:
Tobacco is a major contributor to the rise in NonCommunicable Diseases (NCDs) and is
often linked to the increase in cardiovascular and respiratory diseases and various forms of
cancer. Trinidad and Tobago’s existing prevention and control interventions are in urgent
need of strengthening if the country is to reduce its tobacco use. Tobacco taxation has
been shown to be very effective. This study quantifies the impact of increasing tobacco
tax in Trinidad and Tobago on the future burden of smokingrelated diseases.
Methods:
The UK Health Forum microsimulation model (McPherson and others 2007) was used to
simulate a virtual ‘Trinidad and Tobago’ population and quantify the impact of different
tobacco taxation scenarios on the future burden of smokingrelated disease.
Results and conclusions:
The results showed that the higher tax increase scenario yielded the most significant
results. If tobacco tax is increased by 100% in each of the next three years it is estimated
that 2,537 new cases of smokingrelated disease will be avoided by 2035, saving
TT$254.7million to the health system. These findings support the “go big and go
fast” approach outlined in the World Bank report Tobacco Tax Reform – A Multisectoral
Perspective: At the Crossroads of Health And Development (World Bank 2018).
IX
TRINIDAD AND
TOBAGO’S EXISTI
PREVENTION
AND CONTROL
INTERVENTIONS
IN URGENT NEED
STRENGTHENING
IF THE COUNTRY
TO REDUCE ITS
TOBACCO USE.
INTRODUCTION
The rising prevalence of noncommunicable diseases (NCDs) in the Caribbean represents
one of the region’s biggest challenges. If not confronted headon, the epidemic threatens
to rapidly reverse the substantial health and economic gains that have been realized across
the Caribbean Community (CARICOM)1 in the last three decades. It is estimated that in the
Region of the Americas, which includes CARICOM, 80 percent of all deaths and 77 percent
of premature deaths among persons ages 30 to 70 can be attributed to NCDs.
Tobacco is a major contributor to this rise in NCDs, and is often linked to the increase
in cardiovascular and respiratory diseases and various forms of cancer. In the Region of
the Americas, 14 percent of deaths among adults ages 30 years and under are linked to
tobacco consumption, as are 16 percent of deaths from cardiovascular diseases, 52 percent
of deaths from chronic respiratory diseases (PAHO 2016), and 25 percent of deaths from
cancer. Associations between cigarette smoking and the risk of developing diabetes
(ŚliwińskaMossoń and Milnerowicz 2017; Will and others 2001; Rimm and others 1995) and
cerebrovascular diseases (Indira, Muralidhar, and Munisekhar 2014; Molgaard and others
1986) have also been found.
In 2011, the STEPS2 NCD Risk Factor Survey in Trinidad and Tobago indicated that
approximately 21 percent of the population smoked cigarettes, with an average daily
usage of 11.5 cigarettes (PAHO 2012). To curb tobacco use and institute penalties for
breaches of laws and regulations, tobacco control measures including the Tobacco Control
Act of 2009 and the Tobacco Control Regulations of 2013 have been implemented. Further,
the country signed the World Health Organization’s Framework Convention on Tobacco
Control (WHO FCTC) in August 2003, ratifying it one year later.
Tobacco Use and its Costs to Health and the Economy in Trinidad
and Tobago
According to the 2006 United Nations Common Country Assessment report for Trinidad
and Tobago (Government of Trinidad and Tobago, Ministry of Health 2017), tobacco
was linked to 7 percent of all NCD deaths, 9 percent of ischemic heart disease and 61
percent of lung cancer in 2004. Furthermore, in 2016, ischemic heart disease, diabetes and
cerebrovascular disease were the top three causes of premature deaths (in terms of years
of life lost) in the country (IHME 2015).
1
CARICOM members are: Antigua and Barbuda, Bahamas, Barbados, Belize, Dominica, Grenada, Guyana, Haiti, Jamaica, Montserrat,
Saint Lucia, St Kitts and Nevis, St Vincent and the Grenadines, Suriname, Trinidad and Tobago.
2
This is the World Health Organization’s STEPwise approach to Surveillance – a simple, standardized method for collecting, analysing
and disseminating data in WHO member countries.
1
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
The economic costs of NCDs are also substantial. A 2016 RTI International study (Government
of Trinidad and Tobago, Ministry of Health 2017) in Trinidad and Tobago estimated that the
economic burden of diabetes, cancer and hypertension was roughly TT$8.7 billion annually,
or approximately 5 percent of GDP. Of that total, cancer, which has tobacco use as its most
important risk factor (WHO 2018) and diabetes, cost the Government of Trinidad and Tobago
TT$2 billion and TT$3.5 billion, respectively, each year.
Taxation is Key to Strengthening Prevention and Control
Trinidad and Tobago’s existing prevention and control interventions are in urgent need
of strengthening if the country is to reduce its tobacco use. Tobacco taxation has been
shown to be most effective way to do this (Blecher and others 2014). For instance,
WHOrecommended best practices for tobacco taxation suggest that to substantially
reduce consumption, excise taxes as a percentage of the final consumer price of tobacco
products should be no less than 70 percent (WHO 2014a). However, in 2016 Trinidad and
Tobago’s excise taxes share was substantially lower, at 14.7 percent3 (WHO 2017). Evidently,
there is much work to do to achieve this goal.
This report uses selected scenarios for increasing excise taxes on tobacco products in
Trinidad and Tobago, with the aim of reducing smoking prevalence across the population.
These scenarios provide inputs for modeling the longterm health and cost benefits to
the population of proposed excise tax increases. The report provides evidence from the
modeling exercise. Impacts are calculated relative to the status quo before the tax hike
and are modeled beginning in 2017 for 2025 and 2035.
A microsimulation model was employed to simulate the longterm impact of tobacco
taxations on the future burden of a range of NCDs. Specifically, the disease outcomes
quantified were for coronary heart disease (CHD), stroke, chronic obstructive pulmonary
disease (COPD), and lung cancer. The microsimulation model has been deemed by the
Organisation for Economic Cooperation and Development (OECD) as the most relevant
method for NCD modeling based on riskfactor data (Oderkirk and others 2012). This report
complements modeling work done to estimate the fiscalrevenue impact and expected
reduction in consumption that might stem from proposed additional tobacco excise tax
increases in Trinidad and Tobago. This work has been carried out by the World Bank, using
a model based on the Tobacco Tax Simulation Model (TaXSiM) developed by WHO.
Table 1 presents a summary of total disease cases (epidemiological) and costs (economic)
avoided by parameter, year, and scenario, for the Trinidad and Tobago population.
3
For the mostsold brand.
2 // Introduction
The model estimated that by 2035, the specified tax increase would result in the
avoidance of 1,633 and 2,537 new cases of smokingrelated disease for the two scenarios
modelled. These reductions in disease will result in TT$2.09 million and TT$19.18 million
in healthcare costs avoided for the two scenarios respectively. Consequently, the results
showed that scenario two (the higher tax increase) yielded the most significant results,
supporting the “go big and go fast” approach outlined in the World Bank report Tobacco
Tax Reform – A Multisectoral Perspective: At the Crossroads of Health And Development
(World Bank 2018). This figure is conservative because: (a) only a subset of smoking
related diseases has been included (for instance, diabetes has not been included in the
model in this project); (b) indirect, social care and productivity costs have not been
estimated due to a lack of input data.
Finally, it is important to note that a nonstatistically significant impact on premature
deaths avoided was derived in the study. However, these results are based on preliminary
analysis using a subset of tobaccorelated diseases and limited availability of country
specific data inputs.
Table 1: Summary of total disease cases (epidemiological) and costs (economic) by
parameter, year, and scenario, total population (values in parentheses are uncertainty values)
EPIDEMIOLOGICAL OUTPUTS YEAR BASELINE SCENARIO 1 SCENARIO 2
2025 88,645 [±95] 88,366 [±95] 88,194 [95]
Cumulative cases
2035 210,146 [±135] 208,512 [±135] 207,609 [135]
2025 NA 279 [±134] 452 [±134]
Cumulative cases avoided
2035 NA 1,633 [±191] 2,537 [±191]
2025 10,943 [±33] 10,863 [±33] 10,823 [±33]
Cases per year
2035 14,090 [±33] 13,891 [±33] 13,785 [±33]
ECONOMIC OUTPUTS
2025 NA 15.46[±24.1] 27.81[±24.1]
Cumulative direct costs avoided
(millions, TT$)
2035 NA 155.46[±44.41] 254.73[±44.39]
3
TRINIDAD AND
TOBAGO’S EXISTI
PREVENTION
AND CONTROL
INTERVENTIONS
IN URGENT NEED
STRENGTHENING
IF THE COUNTRY
TO REDUCE ITS
TOBACCO USE.
ING SUMMARY OF METHODOLOGY
Methodology
• The model simulates a virtual population of Trinidad and Tobago, based on latest
population statistics.4
• Data on initial smoking prevalence by age and sex are extracted from the Trinidad and
Tobago Chronic NonCommunicable Disease Risk Factor Survey (Pan American STEPS)
• Scenarios take account of two different tax increases on cigarette prices, and the
impact of these tax increases on smoking prevalence and subsequent disease burden. 5
ARE
• Individual smokers included in the model have a specified smoking status, and a
probability of contracting, dying from, or surviving a disease.
• Future prevalence of smoking is calculated based on the numbers of smokers and
nonsmokers who are still alive in a particular year.
D OF
• Data for disease incidence and mortality are extracted from the Institute for Health
Metrics and Evaluation, Global Burden of Disease database.
• Relative risks of contracting diseases in smokers compared to neversmokers are
extracted from DYNAMOHIA.
• A fivemodule microsimulation model is used to predict the future health and eco
G
nomic impacts of tobacco taxes by 2025 and 2035.
• The model quantifies the future impact on health and related costs of different
levels of tax increase relative to a “no change” scenario.
Assumptions
• Smoking prevalence follows a static trend from 2011 smoking prevalence data.
Y IS
• A specified percentage of smokers who are affected by the tax increase move to the
“exsmoker” category in 2018, 2019, and 2020 in order to account for reductions in
uptake due to price increases.
• If a smoker quits as a result of the intervention, he/she becomes an exsmoker for
the rest of the simulation.
4
Ministry of Planning and Sustainable Development. 2012. Trinidad and Tobago 2011 Population and Housing Census Demographic
Report. Central Statistical Office.
5
World Bank Group. 2018. Tobacco Taxation and Impact of Policy Reforms: Trinidad and Tobago
5
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
• Time since cessation is included in the model to account for changes in disease risk
for an exsmoker.
• Smokers react quickly to tax changes so immediate effects are modelled in the year
following the year of implementation of the tax rise.
Limitations
• The model does not take account of future changes in policy or technology.
• No change in secondhand smoke exposure is modeled.
• The baseline is static over time.
• The simulation only includes four smokingrelated diseases, so results are likely to
underestimate the true effects.
• No data on nonhealthcare costs, for example lost productivity due to disease,
were available.
• No data were available to explore differences by social groups.
• The simulation did not model a possible relapse in smoking among smokers who
gave up as a result of taxinduced price increases.
• No uncertainty analysis was conducted.
6 // Summary of Methodology
7
TRINIDAD AND
TOBAGO’S EXISTI
PREVENTION
AND CONTROL
INTERVENTIONS
ARE IN URGENT
NEED OF
STRENGTHENING
THE COUNTRY IS
REDUCE ITS TOBA
ING
FULL METHODOLOGY
Data Collection
Smoking Prevalence Data
Table 2 sets out the baseline prevalence data used (never, exsmoker and smoker). The
data are from 2011.
Table 2: Never, exsmoker, and smoker prevalence (%) by age group and sex, 2011
AGE SEX YEAR SAMPLE SIZE NEVER EXSMOKER SMOKER
20–24 M 2011 116 68.5 8.6 22.9
25–29 M 2011 121.5 39.8 17.0 43.2
30–34 M 2011 121.5 39.8 17.0 43.2
35–39 M 2011 118.5 52.0 14.6 33.4
40–44 M 2011 118.5 52.0 14.6 33.4
45–49 M 2011 105.5 27.4 35.8 36.8
50–54 M 2011 105.5 27.4 35.8 36.8
55–59 M 2011 95 9.8 55.4 34.8
60–64 M 2011 95 9.8 55.4 34.8
20–24 F 2011 134 84.9 6.2 8.9
25–29 F 2011 125.5 69.2 16.5 14.3
30–34 F 2011 125.5 69.2 16.5 14.3
35–39 F 2011 150 79.2 13.5 7.3
G IF
40–44 F 2011 150 79.2 13.5 7.3
45–49 F 2011 180 78.6 13.5 7.9
50–54 F 2011 180 78.6 13.5 7.9
55–59 F 2011 165.5 74.4 19.4 6.2
60–64 F 2011 165.5 74.4 19.4 6.2
S TO
Note: Children were assumed not to smoke.
Disease Data
The following smokingrelated NCDs were modeled for this study: coronary heart disease
(CHD), stroke, lung cancer, and chronic obstructive pulmonary disease (COPD). Incidence
ACCO
9
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
and mortality data by age and sex were extracted from the Institute for Health Metrics and
Evaluation, Global Burden of Disease, and the International Agency for Research on Cancer
databases. Lung cancer data were grouped with trachea and bronchus data in the database,
so these may have been overestimated for lung cancer only (IARC 2012). No survival data
were available for these diseases in Trinidad and Tobago, therefore survival data was
calculated from incidence and mortality using WHO DISMOD II equations (WHO 2014b).
Relative risks for smokers compared to nonsmokers were extracted from DynamoHIA for
CHD (Song and others 2008; Baba and others 2006; Tolstrup and others 2014; Burns 2003;
Cronin and others 2012; U.S. Department of Health and Human Services 2014), COPD (U.S.
Department of Health and Human Services 2014; Prescott and others 1997; Johannessen
and others 2005; Terzikhan and others 2016; Thun and others 2013), lung cancer (U.S.
Department of Health and Human Services 2014; Thun and others 2013; Freedman and
others 2008; Bae and others 2007), and stroke (Mannami and others 2004; Shinton and
others 1989; Wannamethee and others 1995). As various cohort studies usually observed
participants of different age groups, their estimates were compared and combined to
cover the modeled population: Thus, relative risks for various age groups may derive from
different studies.
Exsmokers’ relative risk was assumed to decrease postcessation and was computed using
a decay function method developed by Hoogenveen and others (Hoogenveen and others
2008). This function uses current smoker relative risk for each disease as the starting point,
and then models the decline in relative risk of disease for an exsmoker over time (see
appendix A).
Table 3: References for disease data
INCIDENCE MORTALITY DIRECT HEALTHCARE COSTS
Calculations based on data from The Cost
Institute for Health Metrics
Institute for Health Metrics and of Health Services in Trinidad and Tobago.
CHD and Evaluation, Global Burden
Evaluation, Global Burden of Disease 2013 HEU, Centre for Health Economics
of Disease
(unpublished)
Calculations based on data from The Cost
Institute for Health Metrics
Institute for Health Metrics and of Health Services in Trinidad and Tobago.
Stroke and Evaluation, Global Burden
Evaluation, Global Burden of Disease 2013 HEU, Centre for Health Economics
of Disease
(unpublished)
Calculations based on data from The Cost
Institute for Health Metrics
Institute for Health Metrics and of Health Services in Trinidad and Tobago.
COPD and Evaluation, Global Burden
Evaluation, Global Burden of Disease 2013 HEU, Centre for Health Economics
of Disease
(unpublished)
Calculations based on data from The Cost
International Agency
Lung International Agency for Research on of Health Services in Trinidad and Tobago.
for Research on Cancer
cancer Cancer databases 2013 HEU, Centre for Health Economics
databases
(unpublished)
10 // Full Methodology
Health Economic Data
Calculation of Direct Healthcare Costs
Disease cost estimations were conducted using estimates of the cost of disease and health
services provided in the Cost of Health Services in Trinidad and Tobago report carried out for
the Ministry of Health in 2013 (HEU, Centre for Health Economics, 2013). This report was
used in conjunction with consultations with medical professionals in respect to the various
inputs required for treatment of the four health conditions studied. The cost per case per
year includes costs of:
• diagnostics;
• pharmaceuticals and or medical supplies;
• rehabilitation, where applicable; and
• inpatient and outpatient care.
Diagnostic costs include the average costs of laboratory tests and imaging.
Pharmaceutical costs were estimated using the most probable drugs to be prescribed
for each of the diseases, together with information on the average frequency of use,
length of use and the average price per drug. The cost of medical supplies/equipment
was included where appropriate. The cost of inpatient care was calculated based
on the average number of inpatient days per disease coupled with the average cost
per inpatient day. Outpatient costs were estimated based on the average number of
visits per disease condition multiplied by the average cost per outpatient visit. Where
applicable, the cost of rehabilitation includes physiotherapy sessions, as well as other
general rehabilitation techniques.
Population Data
To simulate the population of Trinidad and Tobago, the population by age and sex, births by
mother’s age, and total fertility rate statistics were taken from the 2011 population prospects
database. Total mortality rates were taken from the 2017 population prospects database.
The Microsimulation Model
The UK Health Forum (UKHF) microsimulation model was originally developed for the
UK government’s Foresight inquiry (McPherson and others 2007; Wang and others 2011)
and has been developed over the past decade to incorporate a number of additional
interacting risk factors, including smoking (methods are described in greater detail in (UK
Health Forum and CRUK 2016; UK Health Forum) and in appendix A). The model simulates
a virtual population that reproduces the characteristics and behavior of a large sample of
individuals (50 million). These characteristics (age, sex, smoker status) can evolve over the
11
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
life course based on known population statistics and risk factor data. Individuals can be
born and die in the model, which is modular in nature (see figure 1).
• Module 1 uses crosssectional data on the prevalence of the risk factor – cigarette
smoking in this case. For the current study, 2011 smoking prevalence data for
Trinidad and Tobago were extrapolated forward to 2035. It was assumed that the
proportions of the population within each smoking category as calculated in 2011
remained constant until 2035.
• Module 2 is a microsimulation model which uses the prevalence of the risk factor
over time, along with the specified data on the risks of developing diseases, to make
projections of future disease burden.
A wide range of different outputs is produced, including cumulative incidence. To the
authors’ knowledge, no other studies have used a microsimulation model to quantify the
future costs and health impacts of tobacco taxation policy scenarios in Trinidad and Tobago.
Figure 1: Illustration of the microsimulation model
Risk data
Population data Disease data Health economic Intervention
data scenarios
Distribution
programme
Risk
UKHF Microsimulation© programme
Input
Output data
Software
Output
Source: UK Health Forum 2017.
Development of Scenarios
An initial modeling study was carried out by the World Bank Group (Marquez and others
2018) using a version of WHO’s TaXSiM model.6 Within this model, a scenario that reflects
tobacco excise tax changes in 2017 was simulated to calculate the revenue impact as a
result of this tax increase.
6
WHO tobacco tax simulation model (TaXSiM) http://who.int/tobacco/economics/taxsim/en/.
12 // Full Methodology
The modified TaXSiM also calculated the percentage reduction in total cigarette
consumption due to the suggested tax changes. These taxation changes result in non
smokers (predominantly young people) not initiating smoking; smokers quitting, and
smokers reducing the number of cigarettes smoked.
There was one baseline and two intervention scenarios:
• Baseline: A baseline “static” trend
This assumed that smoking prevalence stays constant at 2011 rates.
• Scenario 1:
The 2017 specific excise tax rate on cigarettes is increased by 50 percent in 2018 to
TT$6.57 per 20 cigarettes; by 100 percent in 2019 (TT$13.14 per 20 cigarettes), and 100
percent in 2020 (TT$26.28 per 20 cigarettes).
The consumption as a result of the previously stated tax applied to cigarettes is
estimated to reduce by (only cessation included) a relative reduction of:
• 1.95 percent in 2018
• 5.15 percent in 2019 and
• 7.00 percent in 2020.
• Scenario 2:
The 2017 specific excise tax rate on cigarettes is increased by 150 percent in 2018 to
TT$10.95 per 20 cigarettes; by 100 percent in 2019 to TT$21.90 per 20 cigarettes; and
by 100 percent in 2020 to TT$43.80 per 20 cigarettes.
As a result of the previously stated tax applied to cigarettes, consumption is
estimated to reduce by (only cessation included) a relative reduction of:
• 5.65 percent in 2018
• 6.55 percent in 2019, and
• 7.95 percent in 2020.
Assumptions for implementation of the scenarios in the UKHF microsimulation are as follows:
• Several studies suggest that around 50 percent of the effect of price increases on
overall cigarette consumption results from participation changes (Farelly and others
2001; Centers for Disease Control 1998). Therefore, it is assumed that 50 percent
of smokers would quit and 50 percent would cut down their tobacco intake. An
estimated 50 percent reduction in cigarette consumption was used as an estimate
of the reduction in the total prevalence of smoking. While taxation that raises the
real prices of tobacco might reduce the intensity of smoking, research suggests that
people who cut down may actually inhale more, as measured by serum cotinine
13
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
levels (Fidler and others 2011). Further, the WHO target is focused on a total reduc
tion in smoking prevalence. Therefore, modeling proceeded with a focus on current
smoking prevalence, as opposed to the number of cigarettes smoked.
• Future uptake of smoking was not included in the current scenarios.
• While these average changes were not the same for each group, and usually people
under 30 years of age initiate smoking, the model did not take age differences into
account, and the relative decline in percentages of current smokers was applied to
all age groups.
• A baseline “static” trend was included. This assumed that smoking prevalence remains
constant at 2011 rates. The tax increase scenario was compared to this baseline.
• The tax increase scenarios represent the tax change adopted in 2018, 2019 and 2020.
• The scenarios are based on Monte Carlo7 simulations (individuals were sampled
from the population and simulated over time).
• The specified percentage of smokers who are affected by the tax increase move to
the exsmoker category in 2018, 2019 and 2020.
• If an individual’s smoking status is changed by the scenario, their smoking status will
remain fixed for the entire simulation.
• An immediate reduction in smoking prevalence due to the tax increases in 2018,
2019 and 2020 was assumed. We learned via personal communication with Pro
fessor Joy Townsend that there are different views on the temporal impact of a tax:
econometricians follow Becker’s model, assuming that, as tobacco is very addictive,
the reaction to price increases is slow and greater in the long run. Becker, therefore,
uses a lagged variable of y (t1) (Becker and Murphy 1998). Townsend and Atkinson
take the opposite view (Atkinson and Skegg 1973) – that smokers tend to react
quickly to a price change. A model similar to theirs was used, with an immediate
effect and then a linear trend, and in line with the modified TaXSiM model outputs
(table 4).
7
A Monte Carlo simulation is a mathematical technique which uses stochastic processes to accurately reproduce a system – in this
instance, a realistic population.
14 // Full Methodology
Table 4: TaXSiM Model results
SPECIFIC EXCISE TAX RATE 2018
INCREASED BY 100% IN 2020 TO
SPECIFIC EXCISE TAX RATE 2019
SPECIFIC EXCISE TAX RATE 2017
INCREASED BY 100% IN 2019 TO
INCREASED BY 50% TO TT$6.57
SPECIFIC TAX INCREASED TO
TT$43.80 PER 20 CIGARETTES
TT$26.28 PER 20 CIGARETTES
2018 INCREASED BY 100% TO
TT$10.95 PER 20 CIGARETTES
2019 INCREASED BY 100% TO
TT$21.90 PER 20 CIGARETTES
2017 INCREASED BY 150% TO
TT$13.14 PER 20 CIGARETTES
TT$4.38 PER 20 CIGARETTES
SPECIFIC EXCISE TAX RATE
SPECIFIC EXCISE TAX RATE
SPECIFIC EXCISE TAX RATE
BASELINE 2017: SIMPLE
PER 20 CIGARETTES
(SCENARIO 2)
(SCENARIO 2)
(SCENARIO 2)
(SCENARIO 1)
(SCENARIO 1)
(SCENARIO 1)
Total cigarettes taxed
1.16 1.12 1.00 0.86 1.03 0.90 0.76
(billion pieces)
Average cigarette price
23.45 26.48 36.33 56.49 32.82 49.73 83.37
per pack (TT$)
Average excise tax
burden (excise tax as 18.7 24.8 36.2 46.5 33.4 44 52.5
percentage of price)a
Average excise tax (per
219 328.5 657 1,314 547.5 1,095 2,190
1,000 pieces) (TT$)
Average tax burden
(total tax – import
29.9 36.0 47.4 51.7 44.6 55.2 63.7
excise and VAT as
percentage of price
Percentage change in
1.9 3.9 10.3 14.0 11.3 13.1 15.9
total cigarette taxed
Source: WBG Staff estimates.
Note: a Based on assumptions for elasticity price and elasticity income for highincome countries (Marquez and others 2018).
15
TRINIDAD AND
TOBAGO’S EXISTI
PREVENTION
AND CONTROL
INTERVENTIONS
ARE IN URGENT
NEED OF
STRENGTHENING
IF THE COUNTRY
IS TO REDUCE ITS
ING RESULTS
Smoking Prevalence (Percentage)
Table 5 shows smoking prevalence for males, females, and both males and females
combined for the baseline scenario, and scenarios 1 and 2. By 2035, smoking prevalence
decreases to 12.74 and 10.64 for scenario 1 and 2 respectively. More specifically, smoking
prevalence in men reduces to 20.80 and 17.33 for scenarios 1 and 2 respectively, with the
smoking prevalence among women falling to 4.80 and 4.05 respectively.
Table 5: Smoking prevalence by year, sex and scenario (percentage)
SCENARIO 0 (BASELINE) SCENARIO 1 SCENARIO 2
YEAR
MALE FEMALE TOTAL MALE FEMALE TOTAL MALE FEMALE TOTAL
2017 25.81 6.48 16.18 24.39 6.12 15.28 23.44 5.87 14.69
2020 26.26 6.52 16.41 23.94 5.95 14.96 22.52 5.59 14.07
2025 26.58 6.44 16.51 22.83 5.54 14.18 20.67 5.03 12.85
2030 26.90 6.33 16.57 21.71 5.13 13.39 18.88 4.50 11.66
2035 27.46 6.27 16.78 20.80 4.80 12.74 17.33 4.05 10.64
Figure 2: Male and female smoking prevalence by year for each scenario
FIGURE 2A. MALE
30.00
Male smoking prevalence (%)
25.00
G
20.00
15.00
10.00
5.00
Y
0.00
2020
2025
2030
2035
2015
2040
Scenario 0 (baseline) Scenario 1
Scenario 2
FIGURE 2B. FEMALE
7.00
Female smoking prevalence (%)
S
6.00
5.00
17
4.00
3.00
2.00
FIGURE 2A. MALE
30.00
Male smoking prevalence (%)
25.00
20.00
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
15.00
10.00
5.00
0.00
2020
2025
2030
2035
2015
2040
Scenario 0 (baseline)
Figure 2: Male and female smoking prevalence
Scenario 2
Scenario 1
by year for each scenario, Cont.
FIGURE 2B. FEMALE
7.00
Female smoking prevalence (%)
6.00
5.00
4.00
3.00
2.00
1.00
0.00
2020
2025
2030
2035
2015
2040
Scenario 0 (baseline) Scenario 1
Scenario 2
Summary
There are a number of outputs from the microsimulation.
Epidemiological Indicators
Results from the microsimulation are presented as rates per the Trinidad and Tobago
population, 2011.
Incidence
The total number of new cases of disease, divided by the total number of susceptible
people in a given year presented as a rate per population.
Cumulative incidence rate per year, per Trinidad and Tobago population
To calculate the cumulative incidence rate per year, the total number of new cases of disease
was divided by the total number of susceptible people in a given year and accumulated over
a specified period of the simulation from the year 2016. Therefore, the cumulative number of
cases represents a sum of all of the cases from the start of the simulation.
Cumulative incidence avoided per Trinidad and Tobago population over the
simulation period
The total number of cases of disease avoided or gained as compared to baseline (i.e.,
scenario 0) was estimated. A positive value represents the number of cases avoided,
whereas a negative value represents the number of cases gained.
Mortality per Trinidad and Tobago population over the simulation period
The number of deaths from a disease was estimated.
18 // Results
Mortality cases avoided per Trinidad and Tobago population over the
simulation period
The number of deaths from a disease avoided or gained as compared to baseline (i.e.,
scenario 0) was estimated.
Economic outputs
Direct costs avoided
These are cumulative direct costs across the period of the simulation. The result for 2020
represents the cumulative costs avoided for the period 2016 to 2020. These costs are
scaled to the total population of Trinidad and Tobago. Table 6 presents a summary table
of total disease cases (epidemiological) and costs (economic) by parameter, year, and
scenario as rates per Trinidad and Tobago population.
The model estimated that by 2035 the specified tax increase would result in the avoidance
of 1,633 and 2,537 new cases of smokingrelated diseases respectively for the two
scenarios modelled. These reductions in disease will result in the TT$2.09 million and
TT$19.18 million in healthcare costs respectively avoided for the two scenarios.
Table 6: Summary table of total disease cases (epidemiological) and costs (economic) by
parameter, year, and scenario, total population
SCENARIO 0–
EPIDEMIOLOGICAL OUTPUTS YEAR SCENARIO 1 SCENARIO 2
BASELINE
2025 88,645 [±95] 88,366 [±95] 88,194 [95]
Cumulative cases
2035 210,146 [±135] 208,512 [±135] 207,609 [135]
2025 NA 279 [±134] 452 [±134]
Cumulative cases avoided
2035 NA 1,633 [±191] 2,537 [±191]
2025 10,943 [±33] 10,863 [±33] 10,823 [±33]
Cases per year
2035 14,090 [±33] 13,891 [±33] 13,785 [±33]
ECONOMIC OUTPUTS
2025 NA 15.46[±24.1] 27.81[±24.1]
Cumulative direct costs avoided
(millions, TT$)
2035 NA 155.46[±44.41] 254.73[±44.39]
19
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
Cases of Disease per Year
Table 7 presents the annual cases for each disease by year.
Table 7: Cases per year, total population
LUNG
YEAR CHD COPD STROKE TOTAL
CANCER
2025 Scenario 0 5,312 [±27] 2,829 [±13] 332 [±0] 2,470 [±13] 10,943 [±33]
Scenario 1 5,299 [±27] 2,802 [±13] 332 [±0] 2,430 [±13] 10,863 [±33]
Scenario 2 5,299 [27] 2,789 [13] 332 [0] 2,404 [13] 10,823 [33]
2035 Scenario 0 7,092 [±27] 3,240 [±13] 345 [±0] 3,413 [±13] 14,090 [±33]
Scenario 1 7,065 [±27] 3,187 [±13] 332 [±0] 3,307 [±13] 13,891 [±33]
Scenario 2 7,052 [27] 3,147 [13] 332 [0] 3,254 [13] 13,785 [33]
Cumulative Cases of Disease
Table 8 presents the cumulative cases for each disease by year, and table 9 presents the
cumulative cases avoided. By 2035, the cumulative cases avoided for scenario 1 compared
to scenario 0 are 239, 452, 120,823 respectively for CHD, COPD, lung cancer and stroke.
Similarly, by 2035, the cumulative cases avoided for scenario 1 compared to scenario 0 are
359, 691, 199, and 1,288 respectively for CHD, COPD, lung cancer and stroke.
Table 8: Cumulative cases for each disease by year, total population
LUNG
YEAR CHD COPD STROKE TOTAL
CANCER
2025 Scenario 0 43,214 [±66] 23,200 [±53] 2,789 [±13] 19,442 [±40] 88,645 [±95]
Scenario 1 43,174 [±66] 23,134 [±53] 2,762 [±13] 19,296 [±40] 88,366 [±95]
Scenario 2 43,134 [66] 23,094 [53] 2,749 [13] 19,216 [40] 88,194 [95]
2035 Scenario 0 103,253 [±93] 52,709 [±66] 6,109 [±27] 48,074 [±66] 210,146 [±135]
Scenario 1 103,014 [±93] 52,258 [±66] 5,989 [±27] 47,251 [±66] 208,512 [±135]
Scenario 2 102,895 [±93] 52,019 [±66] 5,910 [±27] 46,786 [±66] 207,609 [±135]
20 // Results
Table 9: Cumulative cases avoided relative to scenario 0 for the total Trinidad and
Tobago population by 2025 and 2035
LUNG
SCENARIO 1 CHD COPD STROKE TOTAL
CANCER
2025 40 [±93] 66 [±80] 27 [±13] 146 [±53] 279 [±134]
2035 239 [±133] 452 [±93] 120 [±40] 823 [±93] 1,634 [±191]
LUNG
SCENARIO 2 CHD COPD STROKE TOTAL
CANCER
2025 80 [±93] 106 [±80] 40 [±13] 226 [±53] 452 [±134]
2035 359 [±133] 691 [±93] 199 [±40] 1,288 [±93] 2,537 [±191]
Mortality
Table 10 and table 11 present the mortality cases for scenarios 1 and 2, relative to scenario
0. Note that it was not possible to derive premature mortality costs due to the non
significance of the results (this is explained in the discussion section). Relative to scenario 0,
by 2035, 53 and 93 mortality cases are avoided for scenarios 1 and 2 respectively.
Table 10: Mortality cases in the total population per year
YEARS SCENARIOS TOTALS
2025 Scenario 0 9,894 [±53]
Scenario 1 9,867 [±53]
Scenario 2 9,880 [±53]
2035 Scenario 0 11,872 [±66]
Scenario 1 11,819 [±66]
Scenario 2 11,780 [±53]
21
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
Mortality cases avoided
Table 11: Mortality Cases avoided in the total population per year
YEARS SCENARIOS TOTALS
2025 Scenario 1 – Scenario 0 27 [±53.12]
2025 Scenario 2 – Scenario 0 13 [±53.12]
2035 Scenario 1 – Scenario 0 53 [±66.4]
2035 Scenario 2 – Scenario 0 93 [±66.4]
Direct Cumulative Costs Avoided
Table 12 presents the cumulative direct healthcare costs avoided for scenarios 1 and
2, relative to scenario 0. By 2035, scenario 1 results in costs avoided of TT$155 million
compared to scenario 0; similarly, by 2035, scenario 2 results in TT$255 million in direct
costs avoided compared to scenario 0.
Table 12: Direct cumulative healthcare costs avoided (million, TT$)
LUNG
YEAR CHD COPD STROKE TOTAL
CANCER
2025 Scenario 1 relative to Scenario 0 0.97 [±21.15] 0.78 [±7.29] 0.82 [±1.6] 14.82 [±8.83] 15.46 [±24.1]
2025 Scenario 2 relative to Scenario 0 0.2 [±21.15] 3.52 [±7.28] 2.71 [±1.59] 21.38 [±8.83] 27.81 [±24.1]
2035 Scenario 1 relative to Scenario 0 24.01 [±38.75] 25.09 [±13.52] 4.3 [±2.93] 102.07 [±16.69] 155.46 [±44.41]
8.86 254.73
2035 Scenario 2 relative to Scenario 0 49.05 [±38.75] 40.17 [±13.51] 156.65 [±16.67]
[±2.92] [±44.39]
22 // Results
23
TRINIDAD AND
TOBAGO’S EXISTI
PREVENTION
AND CONTROL
INTERVENTIONS
IN URGENT NEED
STRENGTHENING
THE COUNTRY IS
REDUCE ITS TOBA
USE. TOBACCO
ING DISCUSSION
This study explored the impact of two tobacco tax increase scenarios in Trinidad and
Tobago on the future burden of four smokingrelated diseases up to 2035. The results
showed that small changes in smoking prevalence in one year can have relatively large
impacts in terms of disease into the future. The results showed that scenario 2 (the higher
tax increase) yielded more significant results, supporting the “go big and go fast” approach
outlined in World Bank report Tobacco Tax Reform – A Multisectoral Perspective: At the
Crossroads of Health And Development (World Bank 2018). This publication suggested that
“tax strategies should focus on health gains first, then on fiscal benefits. This means going
for big tobacco excise tax rate increases starting early in the process.”
ARE
The study included just four smokingrelated diseases (CHD, COPD, stroke, lung cancer).
However, we know that smoking is responsible for many more diseases, and harms almost
every organ in the body (Centers for Disease Control 2016). Therefore, we are likely to see
much wider epidemiological benefits than those observed here. Future work could update
this study by including additional smokingrelated diseases that would also increase
D OF
the impact of the scenarios on epidemiologic outputs such as cumulative incidence,
incidence, and mortality. The relatively small number of premature deaths avoided can
be explained by a Danish study (with 15 years followup of a large cohort of smokers) that
found no evidence that heavy smokers who reduced their number of cigarettes had a
lower risk of death from CHD or from all causes (Gotfredsen and others 2003).
G IF
It was not possible to derive premature mortality costs due to the nonsignificance of the
results. There are three possible explanations for this: first, the threshold age of premature
mortality was assumed to be 65 years, which is very close to life expectancy in Trinidad
and Tobago. Second, the number of mortality cases avoided (by diseases studied) is
small (see table 10). Consequently, given that premature mortality is a subset of mortality,
S TO
premature mortality costs would be expected to be small and not significantly different.
Third, the relatively small change in smoking prevalence due to the intervention and the
relatively short time span of the simulation will play a part.
While the microsimulation method is advantageous in NCD modeling, a key disadvantage
is that the model is data intensive and required detailed and uptodate data.
ACCO
Unfortunately, it was not possible to include data on indirect costs (such as productivity
losses) because the data by disease were not available (as is the case in many countries).
However, we know from studies of total costs (Rezaei and others 2016) in Trinidad and
Tobago that the relative nonhealth cost is colossal due to premature deaths and lost
productivity. For example, the American Cancer Society’s Tobacco Atlas indicates that the
25
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
total direct and indirect cost of smoking in Trinidad and Tobago totals TT$1,858 million
(US$275million) (WHO 2002). One systematic review estimated the direct costs of smoking
to equal around 1.5–6.8 percent of national health system expenditures and 0.22–0.88
percent of GDP in the country studied (Rezaei and others 2016).
In a study in Ukraine, using the same model, the premature mortality costs avoided of
increasing tobacco tax was estimated at Hrv 16.5 billion (US$695 million) (Webber and
others 2017), and in the UK, increasing the tobacco duty escalator to 5 percent (from an
annual tobacco tax increase of 2%) would avoid £192 million in indirect costs by 2035
alone (KnuchelTakano and others 2017). Therefore, wider societal costs such as losses
in productivity are likely to be higher than direct costs, making a stronger case for the
implementation of regular tax hikes for tobacco control (Action on Smoking and Health
2015). In fact, whereas the indirect cost of tobacco use has not been assessed in Trinidad
and Tobago, in 2016 RTI International (RTI 2016) estimated annual indirect costs (in terms
of productivity losses) for smokingrelated cancers and diabetes to be TT$1.18 billion and
TT$2.32 billion respectively. According to the study, the estimated costs represent 90
percent and 58 percent, respectively of the total economic burden of cancer and diabetes.
Furthermore, it was projected that sustained prevention and control efforts can result
in significant savings in terms of productivity losses. These costs are hardly comparable
to most other studies, since they only include productivity losses while excluding other
aspects such as morbidity costs and costs of premature retirement. Nevertheless, as
noted by Rezaei et al. (Rezaei and others 2016), indirect costs exceeding direct costs is
not an uncommon occurrence. Of the 14 studies reviewed Rezaei et al., seven reported
substantially higher indirect costs, ranging from 53.3 percent to 81 percent of the total
costs of smoking. If indirect cost data by disease becomes available, then the model can
once again easily be updated in the future.
Another data limitation was the lack of trend data on smoking prevalence. Therefore, only
a static trend could be included. This may overestimate the impacts if smoking prevalence
is actually falling, or underestimate the impact if smoking prevalence is actually increasing.
We know that social groups react differently to tax increases (Krasovsky 2013). Due to small
sample sizes, it was not possible to model the longterm health impacts on different social
groups within the microsimulation.
One specific limitation of any predictive model is that it does not take account of major
future changes in circumstances, such as the behavior of the tobacco industry, or the
introduction of new drugs or technologies. In theory, their effects can be estimated by
altering parameters in the model, but these will significantly increase the degrees of
26 // Discussion
uncertainty. However, they could be simulated as different scenarios in the future relative
to a “no change” scenario.
At present, the model does not take account of multimorbidity and the joint effect of
several risk factors on disease occurrence and related mortality. However, individuals can
get more than one smokingrelated disease in their lifetime. Future work could expand
the scope of the model to take account of technological and economic changes and
their potential effects, and also to model the clustering of risk factors and diseases in the
same individuals.
The model did not take account of passive smoking/secondhand smoke. Understanding
the combined risk of smoking and passive smoking on later disease outcomes will enable
us to model the combined impact of these risk factors on later disease outcomes.
It was beyond the scope of this study, given the time constraints, to carry out an indepth
uncertainty and sensitivity analysis. We are aware that this is good practice; however,
there is a lack of validated datasets with which to compare our outputs. Furthermore,
the microsimulation is complex, relative to spreadsheet models, for example. It involves
many thousands of calculations which are completed during the simulation of 50 million
individuals. Given this complexity, local uncertainty analysis would demand many
thousands of consecutive runs and would require a supercomputer to complete the
exercise in a realistic timescale.
This study complements modelling work done by the World Bank Group and the HEU,
Centre for Health Economics of The University of the West Indies, and shows the health
and related economic benefits of increasing tobacco taxes in Trinidad and Tobago. Even
small reductions in smoking prevalence in one year will have longterm impacts on
disease incidence and subsequent health costs.
27
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
REFERENCES
Action onSmoking and Health (ASH). 2015. The Economics of Tobacco. ASH factsheet.
Washington, DC: Action on Smoking and Health.
Atkinson AB, Skegg JL. 1973. AntiSmoking Publicity and the Demand for Tobacco in the
U.K. The Manchester School. 41 (3): 265–82.
Baba S, Iso H, Mannami T, Sasaki S, Okada K, Konishi M, et al. 2006. "Cigarette Smoking
and Risk of Coronary Heart Disease Incidence Among MiddleAged Japanese Men And
Women: The JPHC Study Cohort I." European Journal of Cardiovascular Prevention and
Rehabiltation. 13 (2): 207–13.
Bae JM, Lee MS, Shin MH, Kim DH, Li ZM, Ahn YO. 2007. "Cigarette Smoking and Risk of
Lung Cancer in Korean Men: The Seoul Male Cancer Cohort Study." Journal of Korean
Medical Science. 22 (3): 508–12.
Blecher E, Drope J. 2014. "The Rewards, Risks and Challenges of Regional Tobacco Tax
Harmonisation." Tobacco Control 23: 7–11.
Burns DM. "Epidemiology of SmokingInduced Cardiovascular Disease." 2003. Progress in
Cardiovascular Diseases. 46 (1): 11–29.
Centers for Disease Control and Prevention. 1998. "Response to Increases in Cigarette Prices
By Race/Ethnicity, Income, and Age Groups – United States, 19761993." MMWR Morbidity
and Mortality Weekly Report. 47 (29): 605–9.
Centers for Disease Control and Prevention. 2016. Health Effects of Cigarette Smoking 2016
[cited 2016 09.12.2016]. https://www.cdc.gov/tobacco/data_statistics/fact_sheets/health_
effects/effects_cig_smoking/.
Cronin EM, Kearney PM, Kearney PP, Sullivan P, Perry IJ. 2012. Impact of a National Smoking
Ban on Hospital Admission for Acute Coronary Syndromes: A Longitudinal Study. Clinical
Cardiology. 35 (4): 205–9.
Farrelly MC, Bray JW, Zarkin GA, Wendling BW. 2001. "The Joint Demand for Cigarettes and
Marijuana: Evidence from the National Household Surveys on Drug Abuse." Journal of
Health Economics. 20 (1): 51–68.
Fidler JA, Stapleton JA, West R. 2011. "Variation in Saliva Cotinine as a Function of Self
Reported Attempts to Reduce Cigarette Consumption." Psychopharmacology (Berl). 217
28 // References
(4): 587–93. Becker G, Murphy KA. 1988. "Theory of Rational Addiction." Journal of Political
Economy. 96 (4): 675–700.
Freedman ND, Leitzmann MF, Hollenbeck AR, Schatzkin A, Abnet CC. 2008. "Cigarette
Smoking and Subsequent Risk of Lung Cancer In Men and Women: Analysis of a
Prospective Cohort Study." Lancet Oncology. 9 (7): 649–56.
Godtfredsen NS, Osler M, Vestbo J, Andersen I, Prescott E. 2003. "Smoking Reduction,
Smoking Cessation, and Incidence of Fatal and NonFatal Myocardial Infarction in Denmark
1976–1998: A Pooled Cohort Study." Journal of Epidemiology & Community Health. 57 (6):
412–6.
Government of Trinidad and Tobago Ministry of Health. 2017. “National Strategic Plan
for the Prevention and Control of NonCommunicable Diseases: Trinidad and Tobago
2017–2021.” Ministry of Health, Port of Spain, Trinidad and Tobago. http://www.health.gov.tt/
downloads/DownloadItem.aspx?id=385.
HEU, Centre for Health Economics, The University of the West Indies. 2013. "The Cost of
Health Services in Trinidad and Tobago." Unpublished.
Hoogenveen RT, van Baal PH, Boshuizen HC, Feenstra TL. 2008. "Dynamic Effects of
Smoking Cessation on Disease Incidence, Mortality and Quality of Life: The Role of Time
Since Cessation." Cost Effective Resource Allocation. 6:1.
Indira Y, Muralidhar MV, Munisekhar K. 2014. “Role of Smoking and Alcoholism in
Cerebrovascular Accidents.” International Journal of Physiotherapy and Research. 2 (3): 530–
66. http://www.ijmhr.org/ijpr.html.
Institute for Health Metrics and Evaluation (IHME). 2015. GBD Compare (database). Seattle,
WA: IHME, University of Washington. http://vizhub.healthdata.org/gbdcompare.
International Agency for Research on Cancer (IARC). 2012. Global Cancer Observatory
(database). Estimated Cancer Incidence, Mortality and Prevalence Worldwide in 2012. Lyon:
IARC. http://globocan.iarc.fr/Default.aspx.
Johannessen A, Omenaas E, Bakke P, Gulsvik A. 2005. "Incidence of GOLDDefined Chronic
Obstructive Pulmonary Disease in a General Adult Population." The International Journal of
Tuberculosis and Lung Disease. 9 (8): 926–32.
KnuchelTakano A, Hunt D, Jaccard A, Bhimjiyani A, Brown M, Retat L, et al. 2017. Modelling
the Implications Of Reducing Smoking Prevalence: The Benefits of Increasing the UK
Tobacco Duty Escalator to Public Health and Economic Outcomes. Tobacco Control. 27 (2):
129–135.
29
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
Krasovsky K. 2013. "Sharp Changes in Tobacco Products Affordability and the Dynamics
of Smoking Prevalence in Various Social and Income Groups in Ukraine in 2008–2012."
Tobacco Induced Diseases. 11 (1): 21.
Mannami T, Iso H, Baba S, Sasaki S, Okada K, Konishi M, et al. 2004. "Cigarette Smoking and
Risk of Stroke and its Subtypes Among MiddleAged Japanese Men and Women – The
JPHC Study Cohort I." Stroke. 35 (6): 1248–53.
Marquez PV, LaFoucade A, Theodore K, Gabriel S, Scott E, Laptiste C, et al. 2018. Trinidad
and Tobago – Tobacco taxation and impact of policy reforms (English). Global Tobacco Control
Program. Washington, DC: World Bank Group. http://documents.worldbank.org/curated/
en/910981519316913968/TrinidadandTobagoTobaccotaxationandimpactofpolicyreforms
McPherson K, Marsh T, Brown M. 2007. Tackling Obesities. Future Choices – Modelling Future
Trends in Obesity and the Impact on Health. London: UK Government Office for Sciences,
Department of Universities, Innovation and Skills.
Molgaard CA, Bartok A, Peddecord KM, J Rothrock J. 1986. “The Association between
Cerebrovascular Disease and Smoking: A CaseControl Study.” Neuroepidemiology. 5 (2):
88–94. https://doi.org/10.1159/000110818.
Oderkirk J, Sassi F, Cecchini M, Astolfi R. 2012. “Toward a New Comprehensive International
Health and Health Care Policy Decision Support Tool.” Paris: Organisation for Economic
Cooperation and Development.
Pan American Health Organization. 2016. Report on Tobacco Control for the Region of the
Americas. WHO Framework Convention on Tobacco Control: 10 Years Later. Washington DC: PAHO.
Pan American Health Organization. 2012. Panamerican Steps Chronic NonCommunicable
Disease Risk Factor Survey – Trinidad and Tobago. Washington DC: PAHO.
Prescott E, Bjerg AM, Andersen PK, Lange P, Vestbo J. 1997. "Gender Difference in Smoking
Effects on Lung Function and Risk of Hospitalization for COPD: Results from a Danish
Longitudinal Population Study." The European Respiratory Journal. 10 (4): 822–7.
Rezaei S, Sari AA, Arab M, Majdzadeh R, Poorasl AM. 2016. "Economic Burden of Smoking:
A Systematic Review of Direct and Indirect Costs." Medical Journal of the Islamic Republic of
Iran. 30: 397.
Rimm EB, Chan J, Stampfer MJ, Colditz GA, Willett WC. 1995. “Prospective Study of
Cigarette Smoking, Alcohol Use, and the Risk of Diabetes in Men.” British Medical Journal.
310 (6979): 555–59. https://doi.org/10.1136/bmj.310.6979.555.
30 // References
RTI International. 2016. Economic Burden of NonCommunicable Diseases in Trinidad and
Tobago: Preliminary Estimates. North Carolina: RTI International.
Shinton R, Beevers G. 1989. "MetaAnalysis of the Relationship Between Cigarette Smoking
and Stroke." British Medical Journal. 298 (6676): 789–94.
ŚliwińskaMossoń M, Milnerowicz H. 2017. “The Impact of Smoking on the Development
of Diabetes and its Complications.” Diabetes and Vascular Disease Research. 14 (4): 265–76.
https://doi.org/10.1177/1479164117701876.
Song YM, Cho HJ. 2008. Risk of Stroke and Myocardial Infarction after Reduction or
Cessation of Cigarette Smoking – A Cohort Study in Korean men. Stroke. 39 (9): 2432–8.
Team UNC. Common Country Assessment: Republic of Trinidad and Tobago. 2006.
Terzikhan N, Verhamme KMC, Hofman A, Stricker BH, Brusselle GG, Lahousse L. 2016.
"Prevalence and Incidence of COPD in Smokers and NonSmokers: The Rotterdam Study."
European Journal of Epidemiology. 31 (8): 785–92.
Thun MJ, Carter BD, Feskanich D, Freedman ND, Prentice R, Lopez AD, et al. 2013. "50Year
Trends in SmokingRelated Mortality in the United States." The New England Journal of
Medicine.368 (4): 351–64.
Tolstrup JS, Hvidtfeldt UA, Flachs EM, Spiegelman D, Heitmann BL, Balter K, et al. 2014.
"Smoking and Risk of Coronary Heart Disease in Younger, MiddleAged, And Older Adults."
American Journal of Public Health. 104 (1): 96–102.
UK Health Forum. Appendix B4. Detailed Methodology Technical Document. http://
econdaproject.eu/2015.
UK Health Forum and Cancer Research UK (CRUK). 2016. Aiming High: Why the UK should aim
to be tobaccofree. London: UK Health Forum and CRUK.
U.S. Department of Health and Human Services. 2014. The Health Consequences of
Smoking—50 Years of Progress: A Report of the Surgeon General. Atlanta: U.S. Department
of Health and Human Services.
Wang YC, McPherson K, Marsh T, Gortmaker SL, Brown M. 2011. "Health and Economic
Burden of the Projected Obesity Trends in the USA and the UK." Lancet. 378 (9793): 815–25.
Wannamethee SG, Shaper AG, Whincup PH, Walker M. 1995. "Smoking Cessation and the
Risk of Stroke in MiddleAged Men." Journal of the American Medical Association. 274 (2):
155–60.
31
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
Webber L, Elaine I, Andreeva T, Sotomayor N, Efren R, Jaccard A, Retat L, et al. 2017.
Modeling the LongTerm Health and Cost Impacts of Reducing Smoking Prevalence Through
Tobacco Taxation in Ukraine. Washington, DC: World Bank.
Will JC, Galuska DA, Ford ES, Mokdad A, Calle EE. 2001. “Cigarette Smoking and Diabetes
Mellitus: Evidence of a Positive Association from a Large Prospective Cohort Study.”
International Journal of Epidemiology. 30 (3): 540–46. https://doi.org/10.1093/ije/30.3.540.
World Bank. 2018. Tobacco Tax Reform – A Multisectoral Perspective: At The Crossroads Of
Health and Development. Washington, DC: World Bank.
World Health Organization. 2002. The Tobacco Atlas. Geneva: World Health Organization.
https://tobaccoatlas.org/.
World Health Organization. 2014a. Taxation. Geneva: World Health Organization.
World Health Organization. 2014b. Health Statistics and Information Systems. Software
Tools – DISMOD II. http://www.who.int/healthinfo/global_burden_disease/tools_software/en/.
World Health Organization. 2017. WHO Report on the Global Tobacco Epidemic. Country
profile for Trinidad and Tobago. Geneva: World Health Organization.
World Health Organization. 2018. Cancer. Geneva: World Health Organization.
32 // References
33
TRINIDAD AND
TOBAGO’S EXISTI
PREVENTION
AND CONTROL
INTERVENTIONS
IN URGENT NEED
STRENGTHENING
THE COUNTRY IS
REDUCE ITS TOBA
USE. TOBACCO
ING APPENDIX A:
MICROSIMULATION MODEL
The microsimulation consists of two modules. The first module calculates the predictions
Appendix A: Microsimulation model
of risk factor trends over time based on data from rolling crosssectional studies. The
Appendix A: Microsimulation model
Appendix A: Microsimulation model
second module performs the microsimulation of a virtual population, generated with
model The microsimulation consists of two modules. The first module calculates the predictions of risk
demographic characteristics matching those of the observed data. The health trajectory
factor trends over time based on data from
microsimulation
The The rolling
consists crosssectional
two
ofthe modules. The studies. The second
first module module
calculates the p
ofThe microsimulation
each consists
the of
individual from two modules.
population first module
is simulated over calculates
time allowing predictions
them to of risk
contract,
performs the microsimulation
factor trends of a virtual population,
over time studies.
based on generated
data with
from module demographic characteristics
rolling crosssectional studies. The s
factor
ules. The first module calculates the predictions trends over time based on data from
of risk rolling crosssectional The second
survive or die from a set matching those of
of diseases orthe observed
injuries
performs data.
related
the The
to health
the trajectory
analysed
microsimulation of a risk factors.
virtual of each individual
Thegenerated
population, from thewithpopulation
demograp
performs
m rolling crosssectional studies. The second module the microsimulation of a virtual population, generated with demographic characteristics
ARE
detailed description simulated
of the two over time
modules allowing
matching
is them
those
presented to
of contract,
the
below. survive
observed data.or die
The from
health a set of diseases
trajectory of or
each injuries
individual fr
matching
population, generated with demographic characteristics those of the observed data. The health trajectory of each individual from the population is
related to the analysed risk factors.
simulated over detailed them
The allowing
time description
to of the survive
contract, two modules
or die is presented
from a set of belo
dise
simulated
health trajectory of each individual from the population isover time allowing them to contract, survive or die from a set of diseases or injuries
related
related to the analysed Module One: Predictions of Smoking Prevalence Over Time
risk factors. The detailed of therisk
to the analysed
description two factors.
modules The detailed below. of the two module
description
Module
tract, survive or die from a set of diseases or injuries
One: Predictions
Appendix A: Microsimulation model of Smoking Prevalence OverisTime presented
etailed description of the two modules is presented below. For the risk factor (RF), let N be the number of categories for a given risk factor, e.g. N = 3 for
Module One: Predictions of Smoking Prevalence Over Time
Module One: Predictions of Smoking Prevalence Over Time
For the risk factor (RF), let N be
smoking. the
Let 1, 2, …, N of
! =number categories
number for a given
these categories andrisk factor,
#$ (&) e.g.
denote the prevalence of the RF tha
oking Prevalence Over Time For the risk factor (RF), let N be the number of the
For risk factor
categories for (RF), N be
let risk
a given the number
factor, e.g. N =of3categories
for for a given risk factor,
The microsimulation
Nsmoking.
= 3 for smoking. of two
consists Letcorresponds
k= modules. to
1, 2, …,The
the
N first module
category
number
smoking. ! at
these
Letcalculates
time the
t. We
categories predictions
estimate
and # of using
risk
$ (&) denotemultinomial logistic regression
Let ! = 1, 2, …, N number these categories
er of categories for a given risk factor, e.g. N = 3 for and! 1, 2,
#= (&) …,
denoteN number
the prevalence
$
these categories
of the RF and
that#$ (&) denote the preva
factor trends over time based on with rolling
data from crosssectional studies. the second
asThe module
themodel prevalence of RF category ! outcome, and time #$ (&) explanatory
t as a single variabl
D OF
categories and #$ (&) denote the prevalencethe of prevalence
corresponds
the of category
RF that to the ! at corresponds
RF that time t. We corresponds
to the
estimate #$to the
(&) category
category
using k at at time
!time
multinomial t. We
t. We
logistic estimate
estimate
regression using multinomial lo
performs the microsimulation ofFor a virtual
!<) population,
, we have generated
model with with demographic
prevalence of RF characteristics
category ! as the outcome, and time t as a single e
model
We estimate #$ (&) using multinomial logisticusing with
regression prevalence
multinomial of RF
logistic category
regression! as the outcome,
model with and time
prevalence t as a single
of RF explanatory
category k isvariable.
as the
matching those of the observed data. The health trajectory
For ! < of
) ,each
we individual from
have the population
For ! <
s the outcome, and time t as a single explanatory variable.) , we have
simulated over time
outcome, allowing
and t as a
time them to contract,
single survive or die
explanatory from a set
variable. For of ) ö =or
p (,twe
æk 0},by
1, …extrapolated
extrapolated
both
is to
usedto to crosssectional,
sex
forecast manufacture
forecast
S= the{male,
distribution
the RFtimedependent,
female}
distribution trendsand
of eachage
offor
each
group
individual
RF category
RF
discrete
A=
category
{09,
members distributions
in the 1019,
in the of
future.
6 = {#
..., For
the
future.
7079,
population.
each
For
$ (&)
80+}. ! =
The
sexandage fitted trends are
group
completion
1, … ); stratum,
& > of 0} the the
group set
simulation.
,extrapolated
is used to crosssectional,
of
stratum,manufacture
to the set
forecast ofRF
the timedependent,
crosssectional,
trends for individual
distribution of discrete
timedependent,
each members
RF distributions
category ofdiscrete
the
in the 6 = each
{#$ (
distributions
population.
future.
sexandage
&
For
)! =
each 6 = {#$ (&)! =
group stratum, the setsetof of crosssectional, timedependent, discrete distributions 6 6= {# (& )! sexandage
=
1, … );group
& > 0} stratum,
,group
is …
1, used
); the
to
& >
stratum, manufacture
0} , iscrosssectional,
used toRF trends
manufacture
the set of crosssectional,
timedependent,
for individual
RF trends members
timedependent,fordiscrete
individual ofdistributions
the
discrete population.
members of
distributions
= $
the{# $ (&)! =
population.
6 = {#$ (&)! =
1, …1, );
… ); & > & 0} > ,0}is, used
is used to tomanufacture
manufacture RF RFtrends
trends forfor individual
individual members
members of the
of the population.
population.
1, … ); & > 0}, is used to manufacture RF trends for individual members of the population.
8
The probability of the sex of a child can be made time dependent.
8
The probability of the sex of a child can be made time dependent.
8
The probability 8of the
The sex of a child
probability cansex
of the be of
made time
a child dependent.
can be made time dependent.
The
8 8
probability
The probabilityof the sexsex
of the of a
ofchild can
a child be be
can made time
made dependent.
time dependent. 33
8
The probability of the sex of a child can be made time dependent. 33
33 33
33 33
33
10
The probability of the sex of a child can be made time dependent.
41
he probability pmale=1pfemale. In the baseline model this
is taken to be the probability Nm/(Nm+Nf).
The Population editor’ menu item Population Editor\Tools\Births\show random birthList creates an
ditor\Tools\Births\show random birthList creates an
instance of the TPopulation class and uses it to generate and list a (selectable) sample of mothers
generate and list a (selectable) sample of mothersand the years in which they give birth.
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
Deaths from Modelled Diseases
The simulation models any number of specified diseases – some of which may be fatal. In the start
d diseases – some of which may be fatal. In the start year the simulation’s death model uses the diseases’ own mortality statistics to adjust the
eases’ own mortality statistics to adjust the probabilities of death by age and gender. In the start year the net effect is to maintain the same
e start year the net effect is to maintain the same probability of death by age and gender as before; in subsequent years, however, the rates at which
re; in subsequent years, however, the rates at which people die from modelled diseases will change as modelled risk factors change. The population
as modelled risk factors change. The population dynamics sketched above will be only an approximation to the simulated population’s dynamics. The
The Risk
ximation to the simulated Factor Model
population’s dynamics. The
latter will be known only on completion of the simulation.
simulation.
The distribution of risk factors (RF) in the population is estimated using regression analysis
The Risk Factor Model
stratified by both sex S = {male, The distribution
female} and of risk
agefactors
group (RF)
A= the population
in{09, is estimated
1019, ..., 7079, 80+}. using
The regression analysis stratified
ation is estimated using regression analysis stratified by both sex S = {male, female} and age group A = {09, 1019, ..., 7079, 80+}. The fitted trends are
fitted trends are extrapolated to forecast the distribution of each RF category in the future.
= {09, 1019, ..., 7079, 80+}. The fitted trends areextrapolated to forecast the distribution of each RF category in the future. For each sexandage
h RF category in the For each
future. sexandagegroup
For stratum,
each sexandage group thethe
stratum, setset
of of
crosssectional,
crosssectional, timedependent,
timedependent, discretediscrete
distributions 6 = {#$ (&)! =
dependent, discrete distributions D={p_k
distributions 6 = {#$ ((t)k=1,…N;
&)! = 1, … ); t>0}, isused
& > 0}, is used toto manufacture
manufacture RF trends
RF trends for individual
for individual members of the population.
ds for individual members
members ofpopulation.
of the the population.
We model
We model different risk factors,
different risk some of
factors, some which are
of which continuous (such
(such as
are continuous BMI) and
as BMI) some are
and some are
model different
Wecategorical
categorical risk
(smoking
(smoking
factors,
status).
status).
some of which are continuous (such as BMI) and some are
8
The probability of the sex of a child can be made time dependent.
me dependent. categorical (smoking status).
Categorical Risk Factors
Categorical Risk Factors We model different risk factors, some of which are continuous (such as BMI) and some are
Smoking
Categorical Risk Factors
Smoking is
is the
the categorical
categorical
categorical factor.
risk factor.
risk (smoking
Each individual
Each status). in
individual in the
the population
population may belong to
may belong one of
to one of the
the 33
three possible smoking categories {never smoked, exsmoker, smoker} with their probabilities {
three possible smoking categories
33 { never smoked , exsmoker , smoker } with their probabilities {p0,
p0 ,p
p11,
,
Smoking is the categorical risk factor. Each individual in the population may belong to
Categorical Risk Factors
p
p2}. These states are updated on receipt of the information that the person is either a smoker or a
2 }. These states are updated on receipt of the information that the person is either a smoker or a
one of the three Smoking
possible is the categorical risk factor.
{never Each individual
smoked, in the
exsmoker, population
smoker may belong to one of the
nonsmoker.
nonsmoker. They
They will be
will be aasmoking
neversmoker
neversmoker categories
or an exsmoker
or an exsmoker depending
depending on their
on their original
original } with
state
state their
(an
(an ex
ex
three possible smoking categories {never smoked, exsmoker, smoker} with their probabilities {p0, p1,
smoker can
smoker
probabilitiescan { p0, p1become
never
never a neversmoker).
a
, p2}. These
become states are updated on receipt of the information that the
neversmoker).
p2}. These states are updated on receipt of the information that the person is either a smoker or a
person is eitherset
complete
The complete
The
a smoker
set of
or a nonsmoker.
nonsmoker.
of longitudinal
longitudinal They will
smoking
smoking beThey
trajectories
trajectories
will
and theaprobabilities
neversmoker
andbe
a neversmoker
the or an exsmoker
probabilities or
depending
of their
their
of
an exsmoker
happening
happening onare
their original state (an ex
are
generatedon
depending
generated for their
for the smoker
the simulation
simulation can
years
original years never
state byby
(an become
allowing
exsmoker
allowing a
all neversmoker).
all possible
possible transitions
can never become
transitions between
between smoking categories:
categories:
a neversmoker).
smoking
The
{ complete
{never
never set}
smoked
smoked }of
®longitudinal
® {never
{never smokedsmoking
smoked , smoker
, smokertrajectories
}
} and the probabilities of their happening are
generated for
{exsmoker
exsmoker} the
}®®{ simulation
exsmoker,
{exsmoker years
, smokerby
smoker} } allowing all possible transitions between smoking categories:
{
{smoker
{ smoker}}®
®{ exsmoker,
{exsmoker , smoker
smoker} }
{never smoked} ® {never smoked, smoker}
When the probability of {exsmoker
When the probability of being a smoker is p the } ® {exsmoker
allowed transitions , smoker} in the state
are summarized
When the probability ofbeing
beinga smoker is pis
a smoker the
p allowed transitions
the allowed are summarized
transitions in the state
are summarized in the
update equation:
update equation: { smoker } ® { exsmoker, smoker}
state update equation:
When the probability
''
of being a smoker is p the allowed transitions are summarized in the state
0ù
é p0
update equation: é1  p 0 0 ù é p0 0ù
ê úê '' ú ê ú ê ú
ê p1
1ú = ê 0 1  p 1  p p1 (0.14)
(0.0)
(0.0)
úê 1ú
ê p2
''
ú ë p p é p0 p p
ë 2û ê ù û ú ê ú
'
é1ë 2 p
2û
0 0 ù é p0 ù
ê 'ú ê
ê p 1ú = 0 1  p 1  p ú ê p1 ú (0.0)
After
After thethe final
final simulation
simulation year the
the
yearthe smoking
smoking trajectories are
trajectories ê
completed until the ú ê person’s
person’s ú
maximum
After the final simulation year smoking trajectoriesêare
p2 ú are
ê
completed
' completed
ë p Thep
until theuntil the
person’s
p ú û ê
ë
maximum
p2 calculation
ú
û
possible age
possible age of 110 by
of 110 age supposing
by supposing that
that their smoking
their smoking ë
state û
stays fixed. life expectancy
maximum possible of 110 by supposing thatstate
their stays fixed. The
smoking life
state fixed. calculation
expectancy
stays The life
is equal
is equal to the sum
to the of the
sum of probabilities of
the probabilities of being alive in
in each
being alive possible year
each possible of life.
year of life.
After
expectancy calculation to simulation
the final
is equal the sum of year
thethe smoking trajectories
probabilities of being are completed
alive until the person’s maximum
in each
In the
In initial
the initial year
year of possible age of 110 by supposing that their smoking state stays fixed. The life expectancy calculation
the simulation,
of the simulation, a person may
a person be in
may be one of
in one of the three smoking
the three categories; after
smoking categories; N
after N
possible year of life.
updates there
updates there will
will be equal
is´
be 3
3 ´ 2N
2 to the sum
N possible
possible of the probabilities
trajectories.
trajectories. of being
trajectories
These trajectories
These alive
will
will in each
each
each have possible
have a calculated
a year of life.
calculated
probability
probability
In the occurring;
of occurring;
of
initial year of the sum of
the sum
the of these
these probabilities
probabilities is in
is 1. one of the three smoking
1.
Insimulation,
the initial a person
year of the may be
simulation, a person may be in one of the three smoking categories; after N
categories;
In each
In each year
year theN
after
the updates
updates
probability
probability there
of
of will
there
being
being a be
will be 3
smoker
a smoker3x 2Na
´or
or possible
apossible
nonsmoker
nonsmokertrajectories.
trajectories. These
depend
will depend
will These
the trajectories
ontrajectories
on the forecast
forecastwill willhave a calculated
each
smoking
smoking
scenario
each have which
scenario which
a calculatedprobability
provides
provides exactly of occurring;
exactly that
that
probability the Note
information.
information.
of occurring; sum
Note
theof these
that
that
sum these
these
ofprobabilities
states
statesprobabilities
these are is 1.
are twodimensional
is 1. and
twodimensional and
crosssectional {nonsmoking,
crosssectional { smoking},
nonsmoking, smoking }, and
and they
they are
are turned into threedimensional
turned into states {
threedimensional states {never
never
In each year the probability of being a smoker or a nonsmoker will depend on the forecast smoking
smoked,
In each year
smoked, exsmoker,
exsmoker, smoker
the probability }
smoker} ofas described
being
as describeda above.
smoker The
or
above. The time
a
time evolution
nonsmoker of
evolution of the
will three
the depend dimensional
on the
three dimensional states
states
scenario which provides exactly that information. Note that these states are twodimensional and
are
are the
the smoking
smoking trajectories
trajectories necessary
necessary for
for the
the computation
computation of
of disease
disease table
table disease
disease and
and death
death
forecast smoking scenario which provides
crosssectional exactly
{nonsmoking, that information.
smoking }, and they areNote
turnedthat
intothese states
threedimensional states {never
probabilities.
probabilities.
smoked, exsmoker, smoker } as described above. The
are twodimensional and crosssectional {nonsmoking, smoking}, and they are turned intotime evolution of the three dimensional states
Smoking
Smoking are the smoking trajectories necessary for the computation of disease table disease and death
threedimensional states {never smoked, exsmoker, smoker} as described above. The time
The microsimulation
The probabilities.
model
microsimulation model applied to
applied to smoking
smoking enables
enables usus to measure the
to measure future health
the future impact of
health impact of
changes in
changes in rates of tobacco
rates of consumption. This
tobacco consumption. includes the
This includes impact of
of giving
the impact up smoking
giving up on the
smoking on the
Smoking
following diseases:
following diseases: a) COPD; b)
a) COPD; CHD (or
b) CHD acute myocardial
(or acute infarction (AMI)
(AMI) if
myocardial infarction CHD data
if CHD are not
data are not
The microsimulation model applied to smoking enables us to measure the future health impact of
available); c)
available); c) stroke; and d)
stroke; and lung cancer.
d) lung In the
cancer. In the simulation each person
person is
simulation each categorized into
is categorized one of
into one the
of the
42 three
// Appendix changes in rates of tobacco consumption. This includes the impact of giving up smoking on the
three smoking groups:
smoking groups: smokers, exsmokers and
smokers, exsmokers people who
and people have never
never smoked.
who have Their initial
smoked. Their initial
following diseases: a) COPD; b) CHD (or acute myocardial infarction (AMI) if CHD data are not
distribution is based
distribution is based on the distribution
on the distribution of
of smokers,
smokers, exsmokers and never
exsmokers and smokers from
never smokers from published
published
available); c) stroke; and d) lung cancer. In the simulation each person is categorized into one of the
data.
data.
three smoking groups: smokers, exsmokers and people who have never smoked. Their initial
distribution is based on the distribution of smokers, exsmokers and never smokers 34 from published
evolution of the three dimensional states are the smoking trajectories necessary for the
computation of disease table disease and death probabilities.
Smoking
The microsimulation model applied to smoking enables us to measure the future health
impact of changes in rates of tobacco consumption. This includes the impact of giving up
smoking on the following diseases: a) COPD; b) CHD (or acute myocardial infarction (AMI)
if CHD data are not available); c) stroke; and d) lung cancer. In the simulation each person
is categorized into one of the three smoking groups: smokers, exsmokers and people
who have never smoked. Their initial distribution is based on the distribution of smokers,
exsmokers and never smokers from published data.
During the simulation a person may change smoking states and their relative risk will
change accordingly. Relative risks associated with smokers and people who have never
Duringhave
smoked a personfrom
been collected
the simulation published
may change data.
smoking Theand
states relative risks (RR)
their relative associated
risk will change with
accordingly.
exsmokers (RRRelative risks
) areassociated
related to thesmokers
with riskpeople
relativeand who have
of smokers never
(RRsmoker).smoked have been
The exsmoker
exsmoker
collected from published data. The relative risks (RR) associated with exsmokers (RR
exsmoker ) are
relative risks are assumed to decrease over time with the number of years since smoking
related to the relative risk of smokers (RRsmoker). The exsmoker relative risks are assumed to
cessation
decrease (T
over time
cessation
). These relative
with the numberrisks are computed
of years the model
in cessation
since smoking using
(Tcessation equations
). These 1.19
relative and
risks
are
1.20 computed in the
(Hoogenveen model
and using
others equations 1.19 and 1.20 (Hoogenveen and others 2008).
2008).
RRexsmoker ( A, S , Tcessation ) = 1 + ( RRsmoker ( A, S )  1)exp( g ( A)Tcessation ) (0.15)
(0.0)
g ( A) = g 0 exp( h A) (0.16)
(0.0)
where
where γ is
γ is regression coefficient
theregression
the coefficientof of
time dependency.
time The constants
dependency. γ0 and ηγ0
The constants are intercept
η are and
and
regression coefficient of age dependency, respectively, which are related to the specified disease
intercept and regression coefficient of age dependency, respectively, which are related to
(see Table 13).
the specified disease (see table A2).
Table 13: Parameter estimates for γ0 and η related to each disease
Disease γ0 η
AMI A2: Parameter estimates
Table for γ0 and η
0.24228 related to each disease
0.05822
Stroke 0.31947 0.01648
DISEASE γ0 η
COPD 0.20333 0.03087
AMI 0.24228 0.05822
Lung cancer 0.15637 0.02065
Stroke 0.31947 0.01648
Source: Hoogenveen and others 2008.
COPD 0.20333 0.03087
However, a minimum exists when the cessation time is equal to η1. The minimum value was
Lung cancer 0.15637 0.02065
calculated by the method detailed in equations (0.0), (0.0) and (0.0). Where time, t is equal to the
age,
Source: A of an individual.
Hoogenveen and others 2008.
r Exsmk ( t ) = 1 + ( r smk  1) f ( t ) (0.0)
f ( t ) = exp ( g 0 ( t  t0 ) exp ( ht ) )
Þ (0.0)
f ¢ ( t ) = g 0 f ( t ) e ht
( h ( t  t ) + 1)
0
43
The function f(t) has the following properties:
(see Table 13
regression ).
coefficient of age dependency, respectively, which are related to the specified disease
(see
Table Table
Table13: 13 ).
13:Parameter estimatesfor
Parameterestimates forγγ andη
00and relatedto
ηrelated eachdisease
toeach disease
Table 13: Parameter estimates for
Disease
Disease γγ γ and η related to each
000 ηηdisease
Disease
AMI
AMI γ
0.24228
0.24228
0 η
0.05822
0.05822
Reducing Tobacco Use Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
AMI
Stroke
Stroke 0.24228
0.31947
0.31947 0.05822
0.01648
0.01648
Stroke
COPD
COPD 0.31947
0.20333
0.20333 0.01648
0.03087
0.03087
COPD
Lung
Lung cancer
cancer 0.20333
0.15637
0.15637 0.03087
0.02065
0.02065
Lung cancer
Source: Hoogenveenand 0.15637
andothers
others 2008. 0.02065
Source: Hoogenveen 2008.
minimumand
Source: Hoogenveen
However, aminimum others
exists when2008. cessationtime
thecessation timeis equalto
isequal toη η Theminimum
..The minimumvalue was
valuewas
1
However, a exists when the 1
However,
calculated
calculateda minimum
by the exists
method when
detailed in the cessation
equations (0.0),time
(0.0) is
andequal
(0.0). to η 1
Where . The minimum
time, t is equal value
to the
However, by the method
a minimum detailed
exists when in
theequations
cessation(0.0),
time (0.0) andto
is equal (0.0). Where
η1. The time, t value
minimum is equal
wasto the
age,
was calculated
age, A ofan
Aof
calculatedan
by individual.
bymethod
the method
individual.
the detailed
detailed in equations
in equations (0.0), (0.0)(1.17), (1.18)
and (0.0). and (1.19).
Where time, tWhere to thet is
is equaltime,
equal
age,to the
A of anage, A of an individual.
individual.
rExsmk
r Exsmk((tt))=
=1 +((r
1+ rsmk 1
smk  )) ff ((tt))
1 (0.0)
(0.0)
r Exsmk ( t ) = 1 + ( r smk  1) f ( t ) (0.0)
(0.17)
ff ((tt))= exp((
=exp gg00((tt 00) exp ( ht ) ))
tt ) exp ( ht )
f (t ) Þ= exp ( g 0 ( t  t0 ) exp ( ht ) )
Þ (0.0)
(0.0)
ff¢¢((tt))Þ
=
= gg00ff ((tt))e
e
hhtt
((
hh((tt 00) + 1))
tt ) + 1 (0.0)
(0.18)
f ¢ ( t ) = g 0 f ( t ) e ht
( h ( t  t ) + 1)
0
The
The
The functionf(t)
function
function hasthe
f(t)has
has the
the following
following properties:
properties:
following properties:
The function f(t) has the following properties:
ff ((tt ) =1
00) = 1
(0.19)
ff¢¢((tt
00))= h
gg00e
t0
0
=1
 e
h t0
(0.0)
(0.0)
t( ) =a + h
h t0
ff(¢( t)) has
t0has ag
 minimum
0e
minimum attt =
at =tt
00 + h
11
(0.0)
ff ((¥t¥) )) =A
=
has A
a minimum at t = t0 + h 1
f (¥) = A
order
In In to
order toavoid
avoid the RR
the RR exsmoker
exsmoker
from
from increasing,
increasing, the the cessation
cessation time
time was was
set set
equal toequal to η
η1 when
1
the
when the cessation
cessation time was greater
time was
thangreater than η1 (see
η1 (see equation equation (1.20)).
(0.0)). 35
35
35
ì 1 + ( RRsmoker ( A, S )  1) exp( g ( A)Tcessation ) Tcessation < h 1
RRexsmoker ( A, S , Tcessation ) = í
î1 + ( RRsmoker ( A, S )  1) exp( g ( A)h ) Tcessation ³ h 1
1
(0.0)
(0.20)
g ( A) = g 0 exp( h A) (0.21)
(0.0)
The exsmoker relative risks as a function of time after smoking cessation were plotted in Figure A3
The exsmoker relative risks as a function of time after smoking cessation were plotted in
for AMI, stroke, COPD, and lung cancer.
Figure A3 for AMI, stroke, COPD, and lung cancer.
Figure A3: Exsmoker relative risks as a function of time after smoking cessation
1.6
Figure A3: Exsmoker relative risks as a function of time after smoking cessation
1.6 1.4
Exsmoker Relative risks
1.4
1.2
1.2
Exsmoker Relative risks
1 1
0.8
AMI
0.6 0.8
Stroke
0.4 AMI
0.6 COPD
0.2 Stroke
COPD Lung cancer
0 0.4
Lung cancer
0 5 10 15 20
0.2Time since cessation
0
0 5 10 15 20
Time since cessation
// Appendix
44Relative Risks
The reported incidence risks for any disease make no reference to any underlying risk factor. The
microsimulation requires this dependence to be made clear. The risk factor dependence of disease
incidence has to be inferred from the distribution of the risk factor in the population (here denoted
1.2 1.2
risk
risk
Exsmoker Relative risk
Exsmoker Relative ris
risks
risks
1 1.2
1 1 1
Exsmoker Relative risks
Relative
1
Relative
1
Relative
Relative
AMI AMI AMI
0.8 1 AMI AMI
0.8 0.8 0.8
0.8 0.8 Stroke Stroke
Stroke
Exsmoker
Stroke Stroke
Exsmoker
Exsmoker
0.6 0.8
Exsmoker
0.6 0.6 0.6 COPD COPD
COPD
0.6 0.6 COPD COPD
LungLungLung
cancer cancer
cancer
0.4 0.6 Lung cancer
Lung cancer
0.4 0.4 0.4
0.4 0.4
0.2 0.4
0.2 0.2 0.2
0.2 0.2
0 0.2
0 0 0
0 0 0 5 10 15 20
0 0 0 5 5 5 10 10 10 15 15 15 20 20 20
0 0 5 5 100 10 15 15 20 20 cessation
Time Time
since
Time since
since cessation
cessation
cessation Time since
Time since
Time cessation 0
since cessation 5 10 15 20
Relative Risks Time since cessation
The reported Relative Risks
incidence risks for any
Relative Risks
Relative Risks
Relative Risks disease make no reference to any underlying risk
Relative Risks
Relative Risks
The The reported incidence risks for any The reported incidence risks for any disease make no reference to any underlying risk
disease make
no no reference to any underlying risk factor. The
reported
TheThe
factor. incidence
reported
The incidence
Themicrosimulation
reported incidence
reported
risksrisks
risks
incidence
for
for for
any
requires
any
risks
any
for
disease
make
disease
this make
no
dependence
Relative Risks
disease
any make
disease noreference
make
reference
noto
reference to
be to any
toany
made
any
reference
underlying
underlying
to
underlying
clear.
any
risk
The
risk
underlying
factor.
risk factor.
risk factor
factor.
risk TheThe
The
factor. The
microsimulation
microsimulation
microsimulation requires
requires requires
this this microsimulation
dependence to be requires
made this
clear. dependence
The risk to
factor be made
dependence clear. ofThe risk factor dependen
disease
microsimulation
dependence microsimulation requires
of disease thisthis
dependence
requires dependence
incidence
dependence
hasThe
thisincidence
dependence
to
toto be
be
tomade
reported
be be
made
to
made
be
inferred
clear.
incidence clear.
clear.
madefrom
The
The The
risk
risks
clear.risk
the
factor
risk
for
The
factor
any
factor
risk
dependence
disease
distribution
dependence
dependence
factor make
dependence
of the
ofno
of disease
risk
of disease
reference
disease
of to any underlying
disease
factor
incidence
incidenceincidence
has to
has has
be to be
toinferred
be inferred
inferred
from fromfrom
the the the
distribution has to
distribution
distribution of be
the inferred
of the
of requires
the
risk riskrisk
factorfrom the
factor
in
factor the distribution
in the
population
in the population
population of
(here the
(here risk
(here
denoted factor
denoted
denoted in the population
incidence has to be inferred
incidence has to be inferred from from the microsimulation
distribution
the of
distribution the risk
of factor
the this
risk independence
the
factor population
in the to be(here
population made clear.
denoted
(here The
denoted risk factor depen
inas
the population
as p ita is
);disaggregation (here denoted
a disaggregation as p
as
process: ); it is
); is aadisaggregation
disaggregation process:
process:
asppas
);it
); p
it is
);
isit
aa is disaggregation
disaggregation process:
process:
as p); it is a disaggregation process:
process: incidence has to be inferred from the distribution of the risk factor in the populatio
as
Suppose p ); it
thatis a disaggregation
a is process:
Suppose
Suppose
Suppose
Suppose
Suppose that
that
that ais
that
a that
is a
isaa aa
risk
arisk
is
risk
is a
factor
risk
factor
risk
a isfactor
factor
factorstate
state
state ofstate
state
of of
risk
of of
risk
risk
risk
factor
risk
factorfactor
factor
factor
AAand aand
A
and
Arisk
and
denote
andfactor
denote
denote
denote
denote
by
byp
state
p by
A(d
by by
pA
A(d
of
a p
(d risk
A(d
pA¬(d,a
,a,s)
a,a,s)
,a,s)
factor
a
the,a,s)
the A
the
the
incidence
incidence and denote by pA(da,a,s) the inci
incidence
incidence
Suppose that a risk factor state of risk factor A and denote by ap A(dthe
a,a,s)incidence
the incidence
probability
probability
probability for for
the for
the the
disease disease
diseased given probability
given
dthe
d given the
riskthestate,
risk risk for
state,
state, the
,,the disease
a
, is ,athe
the
person’s d
person’sgiven
person’s
age, a,
age,the
age,
a, of
and risk
a,
gender,
and state,
and
gender,a
gender,
s. , the person’s
s. The age, a, and gender, s
relative
probability
probability forforthe
probability thefor disease
disease dd
the disease given
given d the the
given Suppose
risk risk
thestate,
risk
a
that
state,
a a
a
state,,
the the
person’s
a risk
, the factor
person’s
age,
person’s state
a,age,
and
age, a,risk
gender,
a, and
and s.The
factor
gender,
The
gender, A The
s. and
s. s.
relative
relative
denote
The
relative
The by pA(da,a,s) the
relative
risk r is defined by equation (0.0).
risk r A risk
riskis Ar
rdefined is defined
isAdefined by by by
equation equation
equation(0.0). (0.0).
(0.0). A
risk r A is
risk defined
r is by
defined equation
relative risk A defined by equation (1.22).by (0.0).
equation probability
(0.0). for the disease d given the risk state, a , the person’s age, a, and gende
risk rA is defined by equation (0.0).
A (( dAa(
p (a
a , ,s))=
ds a,,= a),r
r =)
s dr ((a
= r (Aa s(
s)
ap a ),(s )Ap
(
p (a
0 ,(
dd,,a,s)
a ,a ,,sas
, )s )
), = r Ad (a a, s ) pA ( d a 0 , a, s )
p p ,,a
d ,
A (d a
s
A) ra , a
d (a),ps
A,(
pd
)a d a
( s a
)0 s )
A
p A d a a a sd a a
d,
p A ,a s
,A d = a pA 0, d0 a 0 , a,
A
(0.0)
A d A
(0.0)
(0.0)
( )
A
r dr (ar (a
a
d,, (s)
aa0º a
s1
),0º )1,º
, s )1 º 1 r p (
a d aa ,,sa , º
s )1
= r (a a , s )
(0.22)(
(0.0)
p d a0 , a, s )
(0.0)
d (a 0(s s) º 1
A d 0
rAA r
0A
Ad
0Aad a a A Ad A
r Ad (a0 a, s ) º 1
Where
Where a isthe
0 is
is the zero
zero risk
risk state
state (for
(for example,
example, the moderate state for alcohol consumption).
Where
Where 0a
a0 is the
the zero
Where
zero risk
risk a0state
is the
state (for (for
zero example,
risk state
example, the moderate
the theexample,
(for
moderate moderate state
state thefor
for alcohol
state
moderate
alcohol consumption).
for alcohol state for
consumption). consumption).
alcohol consumption).
36 36
36 36 36
0
The incidence
incidence probabilities,
The incidence probabilities,
probabilities, as
The incidence reported,
as reported,
reported,
probabilities,can
can be be
beas expressed
expressed
reported, in
in terms
terms
can of
be expressedthe
ofterms equation,
the equation,
equation,
in terms of the equation,
The
The incidence as
probabilities, as reported, can can expressed
be expressed in terms of
in the of the equation,
p(
p
(d
daa,,ss) =åp
)= p( dda ,,a
a, s,,)s )
)p (a
pa s)
s
(,,,d
A (( påA (aAa )a , a, s )p A (a a, s )
, pA d a s= a
p d a s = å
a
p A d
a
a , a , s A
A s
(0.0)
a a
(0.0)
(0.23)
(0.0) (0.0)
=
=p
= p
pA (dd a0 ,
0 , a, s ) å
A ( d a0
A ,a ,s
a, s) å=rpA
r A
A
A (
d (a 0
dd
d
(a
aaa
a ,,
,
,as,)
s
s p
p)A
)s A
Aå(
(a
ara
a
a
A,
,ds
, s(
s )
)a a, s )p A (a a, s )
a a
a a
Combining
Combining these equations allows the conditional incidence probabilities to to in
written terms
beprobabilities of
Combining these
Combining these
these equations
Combining
equations allows
equations these
allows the
allows
equations
the conditional
conditional incidence
the conditional
allows probabilities
incidence
the conditional
incidence probabilities
incidence
probabilities to be written
to be written in
be
in terms of
written
to be
terms of in in terms of
written
known quantities
known quantities known quantities
known
terms quantities
of known quantities
p(
p
(dda a,,ss)) p ( d a, s )
p( s) ( ) p d a , s
p
(d
da ,a
a, a,,s =r
)= (a a , s
) ) ( )
p d , a , s = a , s (0.0)
(0.24)
rA p ad a
a ,,sa , s = r a a , s ( ) (0.0)
A
d
d
Ad å
åb
r
r A
A
d
d
Ad
d  a , s )
A(
( b
b 
 a
a ,
, s
s ) p
p å
A
A
A (a
ar a
a
a
Ad
,
,
, s
(
s
s )
b
 a , s ) p A (a a , s )
(0.0) (0.0)
b
b b
Previous
Previous to any
to anyseries
seriesof Monte Carlo trials, the microsimulation program preprocesses the set of
Previous
Previous to
to any
any series
series ofof
of
Previous Monte
Monte
to any
Monte Carlo
Carlo
Carlo trials,
trials,
series of the
Monte
trials, the microsimulation
the microsimulation
Carlo trials, the
microsimulation programprogram
preprocesses
microsimulation
program preprocesses
preprocesses the
program set
set of
the preprocesses
of the set of
diseases
diseases
the set of
diseases
and
and stores
diseases
and stores
stores
the
the
and
the
calibrated
calibrated
diseases and stores
stores
calibrated
incidence
incidence
the statistics
statistics
calibrated
the calibrated
incidence statistics
p
p
incidence
A (
(d
incidence
pA a
d
A(da0
a0
0
,
,a
, a
a
,
,s
, s).
s).
statistics
statistics
). pA(da0, a
pA(d0, , s).
a, .
Modelling Diseases
Modelling Diseases Modelling Diseases
Modelling Diseases
Disease modelling
Disease modelling Disease heavily
relies heavily
relies the
on the
on setsheavily
sets on the mortality,
of incidence, survival, relative risk and
Modelling
Disease Diseases
modelling modelling
relies heavily on the of incidence,
sets of
relies incidence, mortality,
sets of incidence,
mortality, survival, relative risk
relative
survival,mortality, and
survival,
and relative risk and
risk
statistics.
prevalence statistics.
prevalence prevalence statistics.
prevalence statistics.
Disease modelling
In the simulation, relies heavily
individuals on the
are assigned sets
a risk of incidence,
factor trajectory mortality,
giving their survival, relative
personal risk risk
factor
In the
In simulation, individuals
the simulation, are assigned
are
In the simulation,
individuals a
a risk
individuals
riskare
assigned factor
factor trajectory
assigned
trajectory giving
a risk their
factor
giving their personal
trajectory risk
risk their
giving
personal factor
factorpersonal risk factor
and prevalence
history
history for each
for statistics.
each year
year of for lives.
of their
their
history lives. Their
Their
each year ofprobability
probability of
of
their lives. getting
getting
Their a particular
a particular
probability risk factorrelated
risk factorrelated
of getting disease
disease
a particular risk factorrelated disease
history for each year of their lives. Their probability of getting a particular risk factorrelated disease
in a particular
in a
in a particular year
particular year
in
year
will
will
will depend on
depend
a particular
depend on
year
on
their
their
will
their
risk
risk factor
factoron
depend
risk factor
state
state
state
in that
thatfactor
in risk
their
in
year.
year. state in that year.
that year.
In the simulation, individuals are assigned a risk factor trajectory giving their personal risk
Once a person has a fatal disease (or diseases), their probability of survival will be controlled by a
Once a person has
a person a fatal
fatal
hasOnce diseasehas
(or a
diseases), probability
their(or of survival will be
Once
factor history for a
each ayear
person
disease (or
of their fatal disease
diseases),
lives. their
Their diseases),
probability
probability oftheir
ofsurvival
getting a be
probability
will controlled
of survival
controlled
particular by
will
by
risk a be controlled by a
a
factor
combination of
combination of the
the diseasesurvival
diseasesurvival
combination of thestatistics
statistics and the
and the probabilities
diseasesurvival probabilities
statistics andof
of dying
dying
the from
from other
other
probabilities ofcauses.
causes.
dying from other causes.
combination of the diseasesurvival statistics and the probabilities of dying from other causes.
related
Disease
Diseasedisease
survival
survival in a particular
statistics
statistics are
are year
modelled
modelled will
as
as depend
age
age and
and on their risk
genderdependent
genderdependent factor state in
exponential
exponential that year.
distributions.
distributions.
Disease survival statistics are modelled as age and genderdependent
Disease survival statistics are modelled as age and genderdependent exponential distributions. exponential distributions.
Methods for Approximating Missing Disease Statistics
Once a person has a fatal disease (or diseases), their probability of survival will be
Methods for Approximating Missing Disease Statistics
Methods for Approximating Missing Disease Statistics
Methods for Approximating Missing Disease Statistics
A large amount data are required for modelling these diseases. Where possible these datasets have
A large amount
A large amount
controlled by adata
data are required
are
A large amount
required
combination offor
data
for
themodelling
are
modelling these
required diseases.
for
these diseases.
diseasesurvival Where
modelling possible
these
Where
statistics possible
and these
diseases. datasets
Where
these possible
datasets
the probabilities ofhave
these datasets have
have
dying
been collected
been collected from
from published
published
been collectedsources
sources or
or analysed
analysed
from published from
from
sources either
either crosssectional
crosssectional
or analysed or
or
from either longitudinal
longitudinal
crosssectional or longitudinal
collected
beenother
from from
causes. published
Disease sources
survival or analysed
statistics from
are either
modelled crosssectional
as age and or longitudinal
genderdependent
datasets. Another
datasets. Another limitation
limitation is
is that
that often
often this
this data
data needs
needs to
to be
be in
in a
a specific
specific format.
to be in For
format. For example,
example, the For example, the
the
limitationAnother
datasets. Another datasets. limitation
is that often is that
this data often
needs to this
be indata needs format.
a specific a specific
For format.
example, the
exponential
model updates
model updatesdistributions.
individuals’
individuals’
model updates
model updates individuals’
disease
disease
disease
status
status every
every
individuals’
status every
year
year
disease so
so
year so
the
the
status relative
relative risks
risks
every year
the relative so
risks
used
used
used
in
in the
the model
model
the relative need
risks used
in the model need tothe model need to
to
needin
to
annual relative
be annual
be risks.
relative be
risks.
annual relative risks.
be annual relative risks.
This section contains the methods used in this project in cases where in for
data a particular disease
This section contains
This section the
This
contains methods
section
the used
contains
used the
methods in this
in methods
this project
used
project in
in cases where
in this
cases project
where data
data for a particular
a
cases
for where data
particular disease
for a particular disease
disease
were unavailable.
were unavailable.
were unavailable. were unavailable. 45
Terminal and nonterminal single state disease incidence from prevalence
Terminal and nonterminal single state disease incidence from prevalence
Terminal and nonterminal single state disease incidence from prevalence
Terminal and nonterminal single state disease incidence from prevalence
For terminal diseases,
For terminal
For terminal diseases, to estimate
to estimate
For terminal
diseases, to estimate
incidence
incidence
diseases, to estimate
incidence
(knowing
(knowing
(knowing
prevalence
prevalence
incidence (knowing
prevalence
and
and mortality
and mortality rates)
rates)
prevalence
mortality and
rates)
one
one
one
can
can
mortality
can rates) one can
proceed by
by finding
proceed by finding those
findingproceed incidence
incidence
those incidence probabilities
probabilities
by finding that minimize
minimize
that minimize
those incidence the
the that
probabilities distance
distance between
between
minimize the known
known
the known
the distance between the known
proceed those probabilities that the distance between the
K K
å å Ad Ad A ( A () )
b b
Previous
Previous series
to any to of Monte
any series Carlo trials,
of Monte Carlo the
trials,
microsimulation
the microsimulation program preprocesses
program the set
preprocesses of set of
the
diseases
diseases and stores the calibrated
and stores incidence
the calibrated incidence
statistics pA(da
statistics p
0A d,s
, (a a).0, a, s).
Reducing Tobacco Through Taxation in Trinidad and Tobago: Modelling the LongTerm Health and Economic Impact
UseModelling Diseases
Modelling Diseases
Disease
Disease modelling
modelling
relies heavily
relies heavily sets
on theon theofsets
incidence,
of incidence, mortality,
mortality, survival,
survival, relative
relative risk and
risk and
prevalence
prevalence statistics.
statistics.
In the simulation,
In the simulation, individuals
individuals are assigned
are assigned a risk factor
a risk factor
trajectory giving giving
trajectory their personal
their personal risk factor
risk factor
history for each
history foryear
eachof their
year oflives.
theirTheir
lives.probability of getting
Their probability of getting
a particular
a particular
risk factorrelated disease
risk factorrelated disease
in a particular
in a particular year
year will will depend
depend on theironrisk
their risk factor
factor state
state in year.
thatin that year.
Once a person
Once has a fatal
a person has a fatal disease
disease (or diseases),
(or diseases),
their probability of survival
their probability will bewill
of survival be controlled
controlled by a by a
Methods for Approximating Missing Disease Statistics
combination
combination diseasesurvival
of the of statistics
the diseasesurvival and the
statistics and
probabilities of dying
the probabilities dyingother
of from other causes.
from causes.
Disease
A large amount data survival
Disease
survival statistics
are required aremodelling
statistics
for modelled
are modelled
as age
theseasand
agegenderdependent
and genderdependent
diseases. exponential
Where possible exponential
thesedistributions.
distributions.
datasets have been collected from published sources or analysed from either cross
Methods for Approximating Missing Disease Statistics
Methods for Approximating Missing Disease Statistics
Aamount
A large large amount
sectional or longitudinal data are
data
datasets. required
are required
Another for for modelling
modelling
limitation these
is oftendiseases.
thatdiseases.
these Where
this dataWhere
possible
needs these
to bedatasets
possible these
in datasets
have have
been collected
been collected from published
from published sources
sources or analysed
or analysed
from either crosssectional
from either or longitudinal
crosssectional or longitudinal
a specific format. For example, the model updates individuals’ disease status every year so
datasets.
datasets. Another
Another limitation
limitation
is that is that this
often often
data
this
needs to be in
data needs specific
toabe in a specific
format. For example,
format. the the
For example,
the relative risks
modelused inupdates
updates
model the model
individuals’need
individuals’
diseaseto be annual
status
disease status relative
every year
every
so risks.
the
year so
relative
the relative risksin
risks used the model
used need to
in the model need to
be annual
be annual relative
relative risks. risks.
This section contains the methods used in this project in cases where data for a particular
disease wereThis This section
section
unavailable.contains the methods
contains
the methods used in this in
used project in cases
this project cases where
in where data
data for for a particular
a particular diseasedisease
were unavailable.
were unavailable.
Terminal and nonterminal single state disease incidence from prevalence
Terminal and nonterminal single state disease incidence from prevalence
Terminal and nonterminal single state disease incidence from prevalence
diseases,
For terminal For For terminal
terminal estimate
todiseases, toincidence
diseases, to estimate
estimate (knowing
incidence prevalence
(knowing
incidence (knowing
prevalence mortality
andand
prevalence rates)
mortality
and mortalityone
rates) rates)
one canone can
can proceedproceed by finding
proceed
by finding by finding
those those those incidence
incidence
incidence probabilities
probabilities
probabilities thatminimize
that that minimize
minimize the
thedistance
the distance
distance between
between the known
between the known
K K K K
the known p pre and p pre
and prevalence
computed
and computed
computed p pre
prevalence
prevalence p pre
Nonterminal
Nonterminal diseases diseases
Nonterminal
diseases
are treated intreated
are aare in away
treated
similar similar
in – way – way
although,
a similar
although, – although,
obviously,
obviously, the the mortality
obviously, the mortality
mortality probabilities
probabilities
arezero.
probabilitiesare are zero.
zero.
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics
Mortality statistics 37 37
In In
In
any
InIn any
any
year,
anyIn
In year,
year,
any
anyInyear,
In
year, in in
in
a
year,
any
in aasample
sample
sample
sample
in
in
year,
a a
sample of
asample of
of
N
sample
in aof N
people
sample
NN people
people
of
of N who
Nof
people N who
who
people
people
who have
have
have
who
who
people
have the
the
the
have
have
who
the disease,
disease,
disease,
the
the
have the
disease, a
a subset
asubset
disease,
disease,
a subset asubset
a
disease,
subset NaNN
subset
N willwill
w will
N
subsetNdie Ndie
Nfrom
die
will
will from
from
die
diethe
will the
the
from
from
die disease.
disease.
disease.
from disease.
thedisease.
the the disease.
any year,
In in
any
any a year,
year, in
in a
a of N
sample
sample people
of
of NN who
people
people have
who
who the
have
have disease,
the
the disease,
disease, a aa w will
subset
subset
subset w die from
will the
die
die
w will die
disease.
from
from
from the
thethedisease.
disease.
w w w w
w
Mortality
Mortality
Mortality statistics
statistics
Mortality
Mortality statistics
Mortality record
record
record
statistics
statistics
statisticsstatistics the the
crosssectional
the
record
record crosssectional
crosssectional
the
the
record probabilities
probabilities
crosssectional
crosssectional
the probabilities
crosssectional of of
of
probabilities
probabilities death
death
death
of
of
probabilities as
death
death
deathof as
a
as aa
result
asdeath result
result
as
as a
a of
as of
the
of
result
result
a the
disease
theof
of disease
disease
the
the – ––
disease
disease ––
Mortality
disease. Mortality
Mortality
Mortality record the
statistics
statistics
statistics crosssectional
record
record
record the
the the probabilities
crosssectional
crosssectional
crosssectional of
probabilities
probabilities of
of
probabilities a result
death
death of as
as of
aa result
deaththe
result
resultas a the
of
disease
of
of the–
the disease
disease
disease
result of – –
–
possibly
possibly
possiblystratifying
stratifying
possibly
possiblystratifying
possibly by by
age
by
stratifying
stratifying age
age
by
stratifyingbyage
age
by
possibly stratifying
possibly
possibly by age
stratifying
stratifying
the disease – possibly stratifying by age by age
by age
age
NwN
N Nww N Nw wN Nw
p
pwp
p= =
w=
ppw=
=
p
= w
= (0.0)
(0.0)
(0.0) (0.0)
(0.0)
w =
w
w
Nw N
N pw
N
w
(0.25)
(0.0) (0.0)
(0.0)
(0.0)
N N Nw
w
N
of N people who have the
Within
Within
Within
Within disease,
such
such
Within
Within such
such a such
a aa
such subset
subset
subset
subset
subset
a N
asubsetN
subsetNN
of of will
of
people
people
wof
NN die
people
people
of from
thatthat
that
people the
that
die die
die
that disease.
in indie
in
thatthat
that
die in
year
in that
year
year
thatfrom year
from
from
yearthe from
the
disease,
the
from thethe
disease,
the
disease, disease,
the
distribution
the
disease, the
distribution
distribution distribution
byby
distribution
thedistribution by yearof
yearof
yearof
yearof
byyearof
Within Within
such a such
subset aN subset wof
NN
w of people
people
of people
that that
die die
that
in in
thatdiethat
in
year year
that
from from
yearthe the
from disease,
the
disease, the
disease,
the the
distribution distribution
by by by
by yearof
yearof
w
Within
Within such
such aa
w subset
subset
w w w
w
of
w of people
people that
that die
die in
in that
that year
year from
from the
the disease,
disease, the
the distribution
distribution by yearof
yearof
d the crosssectional
by probabilities
disease
disease
disease is
disease is
not
is
disease not
not
yearofdisease of
usually
usually
is
is death
usually
not
not as a
recorded.
recorded.
is recorded.
not
usually
usually result
This
usually
recorded.
recorded. of
This the disease
distribution
distribution
This distribution
recorded.
This
This – would
would
would
This
distribution
distribution be be
be
distribution
would
would most
most
most
be
be useful.
useful.
useful.
most
most Consider
would Consider
Consider
useful.be
useful. two
most
Considertwo
two
Consider important
important
important
useful.
two
two Consider
important
important
disease is disease
not usually
disease
disease is
is not
is not
not usually
usually recorded.
recorded.
usually This distribution
recorded.
recorded. This
This distribution
This would be
distribution
distribution most be
would
would
would be most
useful.
be most
most useful. Consider
useful.
Consider
useful. Consider
Consider two
two important
two important
two important
important
e idealized,
idealized,
idealized,
idealized, special
special
special
idealized, cases.
cases.
cases.
special
special cases.
cases.
idealized,
two importantidealized,
special
idealized,
idealized, special
cases.
special
idealized,
special cases.
cases.
special cases.
cases.
NSuppose
Suppose
the
Suppose the
true
the true probabilities
probabilities ofdying
of of dying
dyingofin inthe
the the years
years afterafter
somesome
age a
age aa
are {
are {
{ { } } } }
Suppose true
Suppose
pw Suppose
Suppose
=
Suppose the
Suppose
w the
the
true
Suppose
the true
probabilities
true
true
the
the
the probabilities
probabilities
true
probabilities
true
true of
probabilities
of dying
probabilities
probabilities
probabilities of of the
in
dying
dying
of
in
of
dying
inyears
in
dying
dying
dying the
the
in
inyears
in
in
the
after
years
years
the
after
the
the some
after
after
years
(0.0)
years after
some
years
yearsafter
age
some
some
age
after
after age
0age
a
some
some
some
some 00 are
0 are
a
age
a00p
age
age
age are
{
w p
are
a
a
p
0
p,
w
0
w
0
w {
p
00,,
are
are
,w
0 0are
arep
p
p p
,ww
1w
w
p
{
0{
1
0
1
,
1,,
p
,w
pp
p
p
pp
2
w
w
w
w
,
w
w
0
w
w1
0
p
,
,
2
1
2
,
,
2,,
p
,w
pp
p
p
p3
w
w
p
w
w
,
w
1
w
w
2
1
p
,
,
3
2
3
,
,,
,p
3p
w,
p
pp
p4w
w
w
p
w
w
2
w
w
3
2
,
4
3
4 }
,
,4
,, p
pp
p
pw
w
w
w
w34
3
,}
,,
4 p
p
pw
ww4 }
4}
N 0 1 2 3 4
The The probability
probability
The probability
The
The The being
ofbeing
of of
being
probability
probability ofalive
of
probability alive
being
being
ofalive after
after
after
alive
aliveN NNyears
years
years
after
after NNis simply
issimply
simply
is
years
years is thatthat
that
issimply
simply you you
you
thatdo
that do
not
do
you
you not
die
not
do
do die
in
die
not
not in
each
in each
each
die
die inyear
in year
year
each
each year
year
TheThe probability
The
The
probability of
ofbeing
probability
probability
being of being
alive
of being
being
alive alive
after N
alive
alive
after after
years
after
after is
N yearsN
Nsimply
N years
years
years is
is simply
that
is
is simply you
simply
simply that
do
that
that
that you
not
you
you
you do
die
do do
do not
in each
not
not
not die
die
die
die in
ineach
year
in
in each
each
eachyear
year
year
year
w of people that die in that year from the disease, the distribution by yearof
would be most p
corded. This distribution
p
useful.
pp
survive (
pp
survive
survivea (
(0(a
a
+
Consider
ap
p
00N
p
survive
survive (+
+
+(a
a)
N
NN
0
)
=
two
)
0+(
+
()a
(
=
=
a
a
=
1
NN(

(
+(
)
1)1p
((
pp)(
important
1
+
+ ==
N

NN
w )
1
p
01
)w
=
=
=
w
0 survive
1
00 )(
)(
((
)(
p
1
p
1
11
(
1
w
w
1


0
p


0 )(
)(

0p
p
p
wp
p
1
p
1
1)(
w
w
1
1)(
1
)(
)()(
)(
p
1
p
)w0 (
1
11
1
w
w
1



1 )(
1)(
p

p
p
p
wp
pp
1
21)w
w
w
 2)
..
)(

2
0 )(
)()
p
)(1
..
p
1
1
1
..
..
ww
(

(



2
(
1
12)
1)
p

p

pp
..
..
ww
1
((
p
1
)(
p
N
p
1)
)w
w
 
..
..
1
N )(
..
N
p
(
1

1p
1
11
w
))
)w



N
p))
Nwp

1p
1
) ( w N ))
(0.0)
(0.0)
(0.0)
(0.26)
(0.0) (0.0)
(0.0)
(0.0)
(0.0)
(0.0)
survive survive
survive 00 w
ww001 w
ww12
1 w2
w 2 w N 1
2 w
w N
N 
1
11
Survival models
Survival models
Survival models
Survival models
Survival models
Survival models
Survival models
Survival models
Survival models
Survival models
There
There
bilities of dying in the are
years
There
There There
Thereareare
three
after
are
There
are three
some
three
three
are
are
There
three
There
There in in
use
in
in
three
three
are
in
are
areage
use
use ina
use
(they
use
in
three
three
three
(they
(they
(they
use
use
(they
0 are
are
in
in
in {
are
easily
are
are
(they
(they
use
are
use
use p easily
w
easily
0,
are
are
(they
(they
(they
extended
extended
p
easilyw
extended
easily
are ,
easily
easilypextended
wextended
easily
extended
are
are
1 easily
easily
2 , pif
extended
w ,if
the
if
extended
3 ifp
extended
the
the
extendeddata
the
w4 }data
if
if data
merit):
the
the
the
data ifdata
if
if
merit):
merit):
data
data
the
merit):
the
the merit):
merit):
merit):
data
data
data merit):
merit):
merit):
Survival model 0
alive after N years Survival model 0
Survival model 0 : a:
Survival model 0
Survival model 0 :aasingle
single
single
Survival model 0 aprobability
::a ::probability
single
single
: probability
a single ofdying
dying
of of
probability
probability dying
of
of
probability{ p{p
{
dying
dying
{
of p} {{}
w0} w
dying }
p {
pw0{}} }
0}
is simply that you do
Survival model 0
Survival model 0::not
a single
Survival model 0
Survival model 0
a die
single in aeach
a year
probability
single
single of dying
probability
probability
probability of of
of
dying
w p
0dying
dying
ww00 0 p
p
p w
w0 w0
( a0 + N ) = (1  pp p
pp )(
is is
valid
is
1p
p
00 is
valid
valid
p
00 p
valid for
is
is
p )(
for
all
for
valid
valid
is
for 1 all
all
for
all p
for
valid )years.
years.
years.
years.
all
all
for
years... (
years.
1
all pwN 1 )
years. (0.0)
wis0 valid for isall years.
valid for all years.
0 is valid for all years.
vive ww w
00 w ww w 1 w 2
w
w0
w 0
Survival model 1
Survival model 1
Survival model 1
Survival model 1
Survival model 1 : two
1: :
Survival model 1 two
:two different
different
different
two
::two : probabilities
probabilities
different
different
:: two
probabilities
differentprobabilities dying
ofdying
dying
of of
probabilities of
of
probabilities{ p{
{
of{p
dying
dyingp
, p{
0dying{,p
,w1}
pp {
{}
}
,,}
pp },} }
1}
Survival model 1
Survival model
Survival model 1
Survival model 1 two
:two different
two
different
two probabilities
different
probabilities
different of dying
probabilities
of of
dying
probabilities of
w pdying
w0 , p
dying
w w00 www
1
w0
101 p
p
p w
w
w
w
w1
0
0
,
0,
1 pp
pw
ww1
1
hey are easily extended if the data merit):
pw p
ppis is
valid
is
pw
p
00 is
valid
valid
is
00 p
valid for
is
p for
the
for
valid
valid
is
for the
first
the
for
for
valid
the first
year;
first
the
the
for year;
year;
first
first
the
year; p wp
year;
year;
the first pp thereafter.
thereafter.
thereafter.
pp
year; thereafter.
thereafter.
11 ppw
thereafter. thereafter.
0 valid
wis for isthe
valid first
for year;
0 is valid for the first
first thereafter.
year;
wyear; thereafter.
1thereafter.
0w w 1ww11 w
ngle probability of dying { pw w0}
w
w0
w 0 1 w w
w11
Survival model 2
Survival model 2 ::three
three
: three
Survival model 2 :
Survival model 2
Survival model 2 ::
Survival model 2
Survival model 2 different
three
Survival model 2
Survival model 2
different
different
three
three
: different
different
:: three
different
three
three
probabilities
probabilities
different of of
probabilities
probabilities
different
different
dying
ofdying
dying
probabilities
probabilities of{
of
probabilities
of dying
probabilities
probabilities p{p
{
dying
dying
{
of
of
of
w p
0
p,
ww{
p
0{
dying
dying
dying
0,,
p
,wp
p1
p
w
,
w
w {
w
0
p
{
1
0,
1,
p
,,
p
p
wp
p
p }
p
w
5ww
1,,5
,,
5
1
w0
}
}
,p
p}
p
ppw
w ,
5
w1,
,
5}
{}
p
pp ww0 w1 w 5 }
}
w
w 005 w
w 1
1 w
w 55
pw p
ppis is
valid
is
p
p
00 is
valid
valid
validis
is
00 ppfor for
the
for
valid
valid
is
for the
the
thefirst
for
for
is valid
valid
first
year;
first
the
the
for
first
for year;
first
first
the
year; p
year; wp
year;
the firstpp
for
year;
first
for
the
for
pp
year;
for
year;
the
second
the
for
1for
p
thepw second
second
the
theforsecond
for
second to
to to
the
second
the
the to the
fifth
the
to
second
the fifth
year;
tofifth
second the
the
to
fifth
to thep
year;
year;
fifth
fifth
the
year; pp
thereafter.
year;
year;
fifth
p
fifth
thereafter.
thereafter.
pp
year; thereafter.
thereafter.
55 ppw
thereafter.
year; thereafter.
thereafter.
{000pis valid for the firstwyear; 1for the second to the fifth w 5year; 5thereafter.
0w w w 1ww11 w w 5ww55
w 0 , pw1}
w w1 ww
different probabilities
46 w of
//0 dying
Appendix w
w
w 1 w
w11 w5
w 5
Remember
Remember
Remember that
that
Remember
Remember
Remember that
different
Remember
that different
different
that
that that
different probabilities
probabilities
probabilities
different
different will
will apply
will
probabilities
probabilities
different apply
willto
apply
will
probabilities
probabilities will apply toapply
different
to
apply
apply
will
to different
different
todifferent
to age age
and
different
to
different age
age and
and
and gender
gender
age
age
different gender
and
and
age groups.
gender
gender
and
gender groups.
groups. Typically
Typically
Typically
groups.
groups.
gender
groups. groups.the
Typically the the
the
Typically
Typically the the
Typically
the
Remember
Remember that
that different
different probabilities
probabilities will
will apply
apply to
to different
different age
age and
and gender
gender groups.
groups. Typically
Typically the
the
year; pw1 thereafter. data
data might
data might
databe
might be
might divided
divided
be divided into
divided
bedivided into 10year
10year
into 10yearage
intointo age groups.
groups.
age
10year groups.
groups.
agegroups.
datadata might
data
might
data
data be
bemight
divided
might
might be
beinto
be into
divided 10year
10year
divided
divided into
into 10year
age age
10year
10year age
groups.
age
age groups.
groups.
groups.
Calculating survival from incidence and mortality
dying { p , p , p }
Calculating survival from incidence and mortality
Calculating survival from incidence and mortality
e different probabilities of
Calculating survival from incidence and mortality
Calculating survival from incidence and mortality
psurvive (p
a
psurvive p
0(survive
a0N
+
survive (a
+ )
(Na=)(+
0 +
0 1
N
= )
()1=
N pw(01
= p)(
(w
1
 1
0)(
p1p
pw
w0 )(

w
01)(
p
1)(
1

w11)(

p p
1
w1
p

w1 )(
)(
w p
1
2 )w
1
..
2(
)1
p p
..
w2
(w
1)2p
)w..
.. (N1
p(
1

w
1 )
N
p
1w)N
pw 1)
N  )
1
(0.0)
(0.0) (0.0)
(0.0)
Survival models
Survival models
Survival models
Survival models
There
There There
are three
are
There in
three
are
are in use
use
three
three (they
inin are
(they
use
use (they
are are
easily
easily
(they are extended
easily
extended
easily extended
if the
if the
extended if if
data the
data
merit):
the merit):
data
data merit):
merit):
Survival model 0
Survival model 0
Survival model 0: a single
: a single
Survival model 0 :a: probability
a single
single of dying
probability
probability
probability {
of dying
ofofp }w{
{0p
dying
dying
w 0}{
pwp0
w 0}
}
pw 0pis is p
valid
w 0valid
pw is
0w 0
for
is all all
for
valid
valid for
years.
years.
for all
all years.
years.
Survival model 1
Survival model 1
Survival model 1
: two
: two
Survival model 1 : two
different
different
: two probabilities
different
different probabilities
of dying
probabilities
probabilities {
of dying
ofofp {0p
dying
dying
w ,wp{
0w {
1p
,p}
w
p
ww
01,},w
0p p1}1}
w
pw 0pis isp
valid
w 0 pw
valid
is
0w 0
is
for valid
thethe
for
valid first
for
first
for the year;
year;
the first
first p 1pw1 pw
wyear;
year; p
thereafter.
thereafter.
thereafter.
thereafter.
1w1
Survival model 2
Survival model 2
Survival model 2
Survival model : three
Survival model 2
2: : three
three different
:different
: different
three
three different
probabilities
different
probabilities
probabilities
of dying
of of
dying
probabilities
probabilities of{
ofp {0p
dying
dying
dying
w {
,wp
0w,{
p
1pp0
,
ww
p
w,w
10 ,
,p
5p}
p
ww
1w,1
5 }
p p5
,w w}5}
pw 0pis
is is
pw p
valid
valid
w 0 valid
is
for
is for
thethe
valid
valid first
forfor
first
the
year;
year;
the first
first p p
year;
forfor
year;
0w 0 for the first year;w1 w1 for
p
the
pw the second
for the
second
1 the
for
1wthe
to
secondto the
the
second
second fifth
to
to
to fifth
the
the
the year;
fifth
year;
fifth
fifthpyear;
w p5wthereafter.
5pw 5 pw
thereafter.
year;
year; thereafter.
thereafter.
5 thereafter.
Remember
Remember
Remember
Remember that
that
Remember
that
different
different
that
that different
differentprobabilities
probabilities
probabilities
different will
probabilities apply
apply
will
probabilities will
will
will to apply
different
to different
apply
apply toto
to ageage
and
different
different
differentand
gender
age gender
andgroups.
age
and
age andgender
groups.
gender groups.
Typically
groups.
gender thethe the
Typically
Typically
Typically
groups. the
data
data might
might
data
be be
divided
divided
might be into
divided
into
10year
into age
10year
10year
age
groups.
groups.
age
data might be divided into 10year age groups.groups.
Typically the data might be divided into 10year age groups.
Calculating survival from incidence and mortality
Calculating survival from incidence and mortality
Calculating survival from incidence and mortality
Calculating survival from incidence and mortality
Calculating
When
Whena person
When
a person
When asurvival
a (of
person from
(of
person
(of
a given
a given
(of a incidence
given
gender)
agender)
given dies
gender)
dies
gender) and
fromfrom
dies
dies mortality
disease,
a from
from
a aa disease,
they
disease, must
they
disease, must
they
theyhave
must have
contracted
contracted
must
have
contracted
have it at
contracted atit atat
it
some
it some earlier
some
earlier
some earlier
earlier
age.
age.
When For survival
age.
For
a For
survival
person modelmodel
survival 2,
(of a 2,
this
model
given is
this 2,
expressed
is expressed
this is expressed
gender) dies
age. For survival model 2, this is expressed from a disease, they must have contracted it at
some earlier age. For survival model 2, this is expressed
ˆ
p ( a()
a = p (
(a 1
 )
)1p 0 +
)( ) ( )
ˆ mortality
p
ˆ
p ˆ
p ()p
a )a =
=)+pinc
p pinc a( a
a  1
1 p
p)wp ++
ˆp
pˆ ((
aa))==pp (
(mortality
aa 11 p = 0w+
00+
mortality
mortality inc
inc w
w 0 0
( )( 0)
mortality
mortality
mortality inc
inc
inc w
ww0
p a
inc ( a  2 )(1  p
++pp ((aa22 )()(
11 
+
+
+p p
p
+
p p)
inc
inc (
)p
00inc
a(
p

a
+
2
2 )(

11 +
1
2 )(
1

1 p
pw w
wp0w
0 )0p
) p
)w
p w
wp1+
1w
1
+
+1+
inc ( )( 0 )( 1)
inc
inc
inc ww w
0 www
1
p a p p
++pp inc
inc
inc ((aa33 )()(
11 
+
+
+p+p
p
p
ww
inc
inc
w
p)(
00
0
(
(1
)(
inc
a
a(
1


a


3
3
3
pp
)(
)(
w
3)(
ww1
11
1
1
1
))p


1p
w

ww
p
pw
11
1
w
w
+p
+0w
0
)(
)( 1
1
0 )(
1


1 p
pww
wp1w
1 )12p
) p
p)ww
wp1+
1w
1
+
+1+ (0.27)
+
++ p
pinc (
(
( a 
a( a
 4 )(
4 )(
)( 1
12
4 )(

 p )( 1 p
2 pw 0 )(1  pw1 ) pw1 +
)(  )
)3
2
p ++
0 )( 1 )pw
2 2
++pp ((aa44 )()(
11 +p p
p p)(
inc
)(a
11 
 4pp 1
))
2
1pppw wp+
0
+ 1 1 pww p
1 w p
1 + (0.1)
inc
inc
inc wwinc
w
000inc ww
w 1
11 www
111
0 w 1w 1w 1
(0.1)
(0.1)
(0.1)
inc ( )( 0 )( 1)
+ p a  5 1 3 p 1  p p(0.1)
(0.1)
1+
( )( )( )
3 3
+
+ p
p
+ p ( a
a( 
a
 5
5
 )(
5 1
)(
1 
1 p
p p )( 1
)(
1 

1 p
p p )
3
)p
p p +
+ +
(( )()( )()( )) pp
33 w w w
++pp aa55 11 pp inc
inc inc11 pp w+
w 0
0w+0 w 1
w1w1 w1w1 w 1
inc
inc
inc www0
00 w
ww1
11 www1
11 38 38 38
38
inc ( )( 0 )( 1)
4
+ p
p ( a
a  6 )( 1 4 p p )( 1  p
p ) 4 p
4 4 5 +
++pp inc
inc
inc ( (
a a 6 6)()(
1 1
+
+p
 p
+
p