103440 Retaining Health Workers in Rural Areas of Timor Leste: What Can Health Worker Preferences Tell Us? June 2015 This briefing note presents the key findings and policy implications of a 2014 discrete choice experiment (DCE) conducted with 441 health workers, in all 13 districts of Timor-Leste. The key findings are:  The “probability of specialisation” was the most important factor for medical doctors followed by visits from specialists, the availability of equipment, good housing, working in higher-level facilities and an urban location  Newly graduated doctors were totally neutral toward wages  For nurses and midwives, in-service training was valued most highly, followed by transportation, equipment, a remote location and housing The following recommendations are based on the findings from this experiment:  Enhance in-house training and specialist visits for health workers based in rural areas  Increase non-financial benefits (training, supervision, housing etc.) to make rural jobs more attractive  Carry out further analysis of the DCE data to include modelling cost-effectiveness of various models Background investigated health worker preferences and choices. Retaining health workers in rural areas is one of the critical challenges for health systems in Methods most countries. In Timor-Leste a large number of health workers (particularly DCE is a study method, which uses doctors) were deployed to rural health econometric modelling based on Lancaster’s facilities in the recent past. In order to retain theory of values (Lancaster, 1966). In DCE, and keep these health workers motivated to various job attributes are investigated to serve the rural population it is important to generate different choice sets, which are understand what motivates them. It is also administered to the respondents. The important to understand what policy options answers are then analysed to identify and modalities are available to encourage priorities and trade-offs that respondents are health workers to opt for and stay in rural willing to make. areas. In February and March 2014, prior to the In 2014, a survey was conducted in Timor- main survey, qualitative research (involving Leste to better understand facility key informant interviews and focus group functionality, labour market dynamics, the discussions) was conducted in Timor-Leste to preferences of health workers, and the identify the job attributes health workers competence of doctors. This briefing note consider when deciding on rural postings. presents the findings of the discrete choice experiment (DCE) component, which Key Findings Top attributing factors The top attributing factors for the doctors, nurses and midwives are listed in order of importance below. Doctors • Wage was not significant (for young doctors) • Probability of specialisation • VIsits from specialists • Availibility of equipment The attributes that were explored during the • Good housing qualitative research included: • Working in higher level facilities • Urban location  Location (urban/ rural)  Living close to family members Nurses and midwives  Facility types • Wage was significant  Health facility equipment • In-service training  Wages • Transportation  Housing • Availibility of equipment  Motorbikes • Remote location  Training • Housing  Supervision  Workload Doctors’ overall preferences Based on the findings from the qualitative While young doctors were completely neutral study, 16 choice sets were prepared and toward wage, elder doctors did consider included in the general health worker survey wage as a significant factor. However, since questionnaire. the large majority of the sampled doctors were young, wage was not a significant factor In total, 441 health workers were surveyed on the whole. (173 doctors, 150 nurses and 118 midwives) from all 13 districts of Timor-Leste. The The largest and most positive coefficient for sampling was conducted using systematic doctors is on the higher probability of random sampling with probability proportional specialisation, which means that training is to size. Three field teams collected the the attribute with the largest range between survey data during July and August 2014. its best level and its lowest level. Beside specialisation, visits from specialists to the During the analysis, following the random doctor’s facility are a type of training that is utility theory, a model was created to analyse valued. the respondents’ choices (McFadden, 1976). The relative importance of each coefficient Good and medium-level equipment are also then generated the Marginal Rate of important to doctors, and they prefer to work Substitution (MRS) between the different in a community health centre (CHC) rather attributes. than in a health post (HP). An urban location is preferred to remote and extremely remote locations, although it seems that extremely remote locations do not carry a different 2 value from remote locations in the eyes of the doctors to work in non-urban settings. interviewed doctors. Working in a rural CHC would also result in a 51% increase in satisfaction; compared to As for personal benefits, good housing doctors in urban HPs. Good housing would comes first, while wages and having a increase satisfaction by 86%. motorbike do not seem to be important at all: for them, the coefficients are negative, but Training is a powerful job attribute that not significantly different from zero. appears to determine job choice, including location. The fact that urban health workers Nurses and midwives’ overall preferences enjoy more training adds to the attraction of urban jobs. Access to training should be In contrast with doctors’ preferences, wages determined by both training needs and the had a positive coefficient of 0.0012 per dollar, incentive it may provide in recruiting health i.e. almost the same as a remote location or workers to desired locations. Younger good equipment. For these groups, a remote doctors, female doctors and those in CHCs area is preferred to both urban and extremely and HPs tend to value training more. Keeping remote locations, but an urban location is routes to specialisation open may well be significantly preferred over an extremely highly effective in retaining doctors in rural remote one. CHCs, good housing (over poor areas. housing) and motorbike (over no motorbike) also have positive coefficients that are There are various ways that policy-makers significantly different from zero. could seek to revise current policies based on these findings. Policy Implications of the Main Findings Implications for nurses and midwives Implications for doctors Urban locations seem to be less desirable than remote locations for nurses and One of the most interesting findings of the midwives. However, urban locations are more study is the relative lack of importance appreciated than extremely remote locations. doctors placed on wages. This is in line with The policy question arises as to how to a number of recent DCEs in other countries attract nurses and midwives to extremely (Kruk, 2010; Blaauw et al., 2010). In the remote places. Timor-Leste context, doctors may see The DCE analysis revealed several options to themselves as already better-off (with a salary of US$ 610 per month) compared to encourage nurses and midwives to move other civil servants. In that case, their from urban areas to extremely remote areas. If the nurses and midwives were given preference for professional development housing and a motorbike in extremely remote opportunities rather than increased wages locations, they would prefer to work there may be reasonable. Moreover, most of the instead of urban areas (in cases where the doctors in Timor-Leste are in the early stages of their career – a stage when they may value urban job does not have housing or a career development more than financial motorbike). incentives. However, it would be difficult to incentivise health workers to move from a remote The DCE results suggest that improving location to an extremely remote location, as facilities, availability of drugs and medical the marginal disbenefit they get from moving equipment would be largely enough to compensate for an urban location with low- from a remote area to an extremely remote level equipment for doctors. As for training, area is very high. visits from specialists or a higher probability of specialisation would also incentivise 3 References Ghana: a discrete choice experiment. Bull World Health Organ., 88(5), pp. 333-341. Blaauw, D. et al. (2010) Policy interventions that attract nurses to rural areas: a Lancaster, K. J. (1966) A new approach to multicountry discrete choice experiment. Bull consumer theory. The Journal of Political World Health Organ., 88(5), pp. 350-356. Economy, 74(2), pp. 132-157. Kruk, M. et al. (2010) Rural practice McFadden, D. L. (1976) Quantal Choice preferences among medical students in Analaysis: A Survey. Annals of Economic and Social Measurement. 5(4), pp. 363-390. This project was carried out by World Bank and Oxford Policy Management (OPM) with funding from the Australian Department and European Commission. The Ministry of Health of the Government of Timor-Leste provided support at every stage of this project. Dili Institute of Technology (DIT) was the local partner. Detailed results and interpretations are available in the full report, which can be accessed at www.worldbank.org/en/country/timor-leste. If you have any queries please contact Xiaohui Hou, Senior Economist, World Bank (email: xhou@worldbank.org) or Rashid Zaman, Consultant, Health Portfolio, Oxford Policy Management (email: rashid.zaman@opml.co.uk). © World Bank and Oxford Policy Management 4