Exploring the use of routinely-available, retrospective data to study the association between malaria control scale-up and micro-economic outcomes in Zambia
Article
May 15, 2017
Background Country-level evidence on the impact of malaria control on micro-economic outcomes is vital for mobilizing domestic and donor resources for malaria control. Using routinely available survey data could facilitate this investigation in a cost-efficient way.
Methods The authors used Malaria Indicator Surveys (MIS) and Living Conditions Monitoring Survey (LCMS) data from 2006 to 2010 for all 72 districts in Zambia to relate malaria control scale-up with household food spending (proxy for household well-being), educational attainment and agricultural production. The authors used two quasi-experimental designs: (1) a generalized propensity score for a continuous treatment variable (defined as coverage from owning insecticide-treated bed nets and/or receipt of indoor residual spraying); and, (2) a district fixed effects model to assess changes in the outcome relative to changes in treatment pre-post scale-up. The unit of analysis was at district level. The authors also conducted simulations post-analysis to assess statistical power.
Results Micro-economic outcomes increased (33% increase in food spending) concurrently with malaria control coverage (62% increase) from 2006 to 2010. Despite using data from all 72 districts, both analytic methods yielded wide confidence intervals that do not conclusively link outcomes and malaria control coverage increases. The authors cannot rule out positive, null or negative effects. The upper bound estimates of the results show that if malaria control coverage increases from 60 to 70%, food spending could increase up to 14%, maize production could increase up to 57%, and years of schooling could increase up to 0.5 years. Simulations indicated that the generalized propensity score model did not have good statistical power.
Conclusion While it is technically possible to use routinely available survey data to relate malaria control scale-up and micro-economic outcomes, it is not clear from this analysis that meaningful results can be obtained when survey data are highly aggregated. Researchers in similar settings should assess the feasibility of disaggregating existing survey data. Additionally, large surveys, such as LCMS and MIS, could incorporate data on both malaria coverage and household expenditures, respectively.
Methods The authors used Malaria Indicator Surveys (MIS) and Living Conditions Monitoring Survey (LCMS) data from 2006 to 2010 for all 72 districts in Zambia to relate malaria control scale-up with household food spending (proxy for household well-being), educational attainment and agricultural production. The authors used two quasi-experimental designs: (1) a generalized propensity score for a continuous treatment variable (defined as coverage from owning insecticide-treated bed nets and/or receipt of indoor residual spraying); and, (2) a district fixed effects model to assess changes in the outcome relative to changes in treatment pre-post scale-up. The unit of analysis was at district level. The authors also conducted simulations post-analysis to assess statistical power.
Results Micro-economic outcomes increased (33% increase in food spending) concurrently with malaria control coverage (62% increase) from 2006 to 2010. Despite using data from all 72 districts, both analytic methods yielded wide confidence intervals that do not conclusively link outcomes and malaria control coverage increases. The authors cannot rule out positive, null or negative effects. The upper bound estimates of the results show that if malaria control coverage increases from 60 to 70%, food spending could increase up to 14%, maize production could increase up to 57%, and years of schooling could increase up to 0.5 years. Simulations indicated that the generalized propensity score model did not have good statistical power.
Conclusion While it is technically possible to use routinely available survey data to relate malaria control scale-up and micro-economic outcomes, it is not clear from this analysis that meaningful results can be obtained when survey data are highly aggregated. Researchers in similar settings should assess the feasibility of disaggregating existing survey data. Additionally, large surveys, such as LCMS and MIS, could incorporate data on both malaria coverage and household expenditures, respectively.
Regions
Sub-Saharan Africa