120 likes | 217 Views
Measurement Issues Regarding Estimation of Sub-National Health System Efficiency. Ajay Tandon Asian Development Bank Development Indicators and Policy Research Division Economics and Research Department September 29 th 2006. Sub-National Health System Efficiency.
E N D
Measurement Issues Regarding Estimation of Sub-National Health System Efficiency Ajay Tandon Asian Development Bank Development Indicators and Policy Research Division Economics and Research Department September 29th 2006
Sub-National Health System Efficiency • Entails measurement of health system outcomes relative to resource inputs at lower administrative tiers (e.g., at district or province level). • Arguably, of greater local relevance and often more useful than cross-country analyses for policy-makers. • Poses significant measurement challenges: relevant data often not available sub-nationally.
Sub-National Data Challenges • Sub-nationally representative surveys are expensive to conduct and – even if available – such surveys are not conducted as regularly as may be needed. • Relevant sub-national data may only be available from administrative records and this may not be reliable. • Data may not be available for pertinent variables.
Possible Solutions • Use of innovative sampling strategies to reduce cost of conducting surveys, e.g., EPI cluster surveys and Lot Quality Assurance Sampling (LQAS). • Use of econometric models and Bayesian data combinatorial techniques for estimation purposes. • Use factor analytical techniques to compute proxy indexes to capture relevant quantities of interest.
Econometric Models • Quantity of interest (Y) may be available at a national level, but its determinants (X’s) are available both at the national and sub-national level. • Y = f (X): model at national level. Predict Y at sub-national level using X. • Disadvantage: model dependence is high as estimates of Y at the sub-national level will be sensitive to choice of predictors.
Bayesian Data Combination Techniques • Derive “priors” of quantity of interest using national survey data. • Augment priors with micro-samples at the sub-national level (the “likelihood”). • Combine the priors with the likelihood to estimate “posterior” estimates at the sub-national level. • Borrow strength using prior information.
Proxy Indexes Source: Ranson, Hanson, Oliveira-Cruz, and Mills (2003)
Indonesia Sub-National Application • Output index by district: • Coverage of complete immunization. • Skilled birth attendance rate. • Iodized salt consumption. • Life expectancy. • Extent of protection from catastrophic spending. • Input index by district: • Income. • Female education. • Nurses per 100,000. • Out-of-pocket health expenditure. • Access to health facilities.