Economic evaluations of non-communicable disease interventions

BackgroundDemographic projections suggest a major increase in non-communicable disease (NCD) mortality over the next two decades in developing countries. In a climate of scarce resources, policy-makers need to know which interventions represent value for money. The prohibitive cost of performing multiple economic evaluations has generated interest in transferring the results of studies from one setting to another. This paper aims to bridge the gap in the current literature by critically evaluating the available published data on economic evaluations of NCD interventions in developing countries.MethodsWe identified and reviewed the methodological quality of 32 economic evaluations of NCD interventions in developing countries. Developing countries were defined according to the World Bank classification for low- and lower middle-income countries. We defined NCDs as the 12 categories listed in the 1993 World Bank report Investing in Health. English language literature was searched for the period January 1984 and January 2003 inclusive in Medline, Science Citation Index, HealthStar, NHS Economic Evaluation Database and Embase using medical subheading terms and free text searches. We then assessed the quality of studies according to a set of pre-defined technical criteria.ResultsWe found that the quality of studies was poor and resource allocation decisions made by local and global policy-makers on the basis of this evidence could be misleading. Furthermore we have identified some clear gaps in the literature, particularly around injuries and strategies for tackling the consequences of the emerging tobacco epidemic.ConclusionIn the face of poor evidence the role of so-called generalised cost-effectiveness analyses has an important role to play in aiding public health decision-making at the global level. Further research is needed to investigates the causes of variation among cost, effects and cost-effectiveness data within and between settings. Such analyses still need to take a broad view, present data in a transparent manner and take account of local constraints.