International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
The use of specialty training to retain doctors in Malawi: a cost effectiveness analysis using stated preferences
Kate Mandeville, Godwin Ulaya, Adamson Muula, Titha Dzowela, Mylene Lagarde, Kara Hanson

Last modified: 28 March 2017



Low production and high emigration have led to few doctors in many sub-Saharan African countries, impeding the delivery of essential health services and response to new health threats. Retaining doctors in their country of training, therefore, has been the focus of recent policy efforts. Out of possible incentives, the opportunity to specialise has been found to be particularly important to doctors. Yet the specialist workforce is also small in many countries, necessitating a reliance on foreign-trained specialists or none at all. This constrains domestic training capacity, forcing countries to send doctors to other countries to specialise. The opportunity to train in more advanced health systems is often popular with doctors, but augments the risk of emigration. This presents a dilemma for policymakers: sending doctors to train in other countries is likely to increase retention of doctors in the short-term, but may produce specialists more likely to emigrate in the long-term. Stipulating a mandatory period of work before entry to specialty training would ensure better value from this investment, but delay the production of much-needed specialists. Expanding domestic training may protect against emigration of specialists, but may not be accepted by doctors. Finally, many countries will be unable to reduce their dependence on internationally trained specialists without expansion of specialty training, yet specialists are more costly to produce and employ than generalist doctors. To tease out these issues, we modelled different policy options in an example country, Malawi, using results from a discrete choice experiment in order to examine the cost-effectiveness of expanding specialty training to retain doctors in sub-Saharan Africa.


We designed a Markov model of the labour market for doctors in Malawi, incorporating data from tracing studies of Malawian graduates, doctors’ preferences for specialty training and local cost data. Expanded specialty training in Malawi or South Africa with varying mandatory service requirements were compared against baseline conditions. This is the first time to our knowledge such a model has been constructed for the medical workforce in sub-Saharan Africa. A government perspective was taken with a time horizon of 40 years. The outcome measures were cost per doctor year and cost per specialist year in the Malawian public sector.

The effectiveness of each policy option was represented by the uptake of available training places by junior doctors (doctors within seven years of graduation). These uptake rates were based on the results of a discrete choice experiment that quantified the preferences of junior doctors for different types of specialty training. This study had been undertaken on all junior doctors in Malawi that were not yet in specialty training. Latent class modelling identified four subgroups of junior doctors with distinct preferences for different types of specialty training. The cost-effectiveness analysis was run on each of these subgroups in turn in order to explore heterogeneity in the population and likely effectiveness of the policy options.


Longer periods of service before training would be more cost-effective, with five years’ mandatory service adding the most value. At the end of 40 years of expanded training in Malawi, the medical workforce would be over fifty percent larger and there would be over six times the number of specialists compared to current trends. These policies, however, would be more costly than current government spending for relatively modest gains in doctor-years. The government would need to be willing to pay at least 3.5 times more per doctor-year for a five percent minimum increase in total doctor-years over baseline and at least fifty percent more per specialist-year for a maximum six-fold increase. The most optimal option differs between subgroups of doctors, with greater increases in doctor- and specialist-years possible in those with more flexible preferences.


This is one of the first economic evaluations, to our knowledge, incorporating stated preference results to assess policy options in low- and middle-income countries. In a policy arena that is characterised by data paucity, cost-effectiveness analyses based on results of stated preference studies offer a pragmatic and timely solution to inform policy decisions. Only one other study, to our knowledge, has used the results from a discrete choice experiment to inform a cost-effectiveness analysis of health workforce policies. Methodologically, our study significantly extends this study by exploring heterogeneity in the population through the use of latent class modelling.

Our findings show that the Malawian government could obtain higher returns on their investment in medical education by expanding specialty training in Malawi. The general conclusions of this analysis are likely to be transferable to other sub-Saharan African countries seeking to maximise the value from their investment in medical undergraduate education. Training and retaining doctors consume considerable resources from limited budgets.  For example, out of its USD125 million health budget in fiscal year 2014/2015, Malawi spent around 4% on training and 36% on salaries. Decisions on health workforce policy are therefore high value, yet usually made in a low-information environment. More routine application of cost-effectiveness analyses informed by choice modelling to health workforce decisions is likely to be of considerable value.


Conference registration is required in order to view papers.