Small area estimation of health insurance coverage for Kenyan counties

Abstract

AbstractHealth insurance is important in disease management, access to quality health care and attaining Universal Health Care. National and regional data on health insurance coverage needed for policy making is mostly obtained from household surveys; however, estimates at lower administrative units like at the county level in Kenya are highly variable due to small sample sizes. Small area estimation combines survey and census data using a model to increases the effective sample size and therefore provides more precise estimates. In this study we estimate the health insurance coverage for Kenyan counties using a binary M‑quantile small area model for women $$(n=14{,}730)$$ ( n = 14,730 ) and men $$(n=12{,}007)$$ ( n = 12,007 ) aged 15 to 49 years old. This has the advantage that we avoid specifying the distribution of the random effects and distributional robustness is automatically achieved. The response variable is derived from the Kenya Demographic and Health Survey 2014 and auxiliary data from the Kenya Population and Housing Census 2009. We estimate the mean squared error using an analytical approach based on Taylor series linearization. The national direct health insurance coverage estimates are $$18\%$$ 18 % and $$21\%$$ 21 % for women and men respectively. With the current health insurance schemes, coverage remains low across the 47 counties. These county-level estimates are helpful in formulating decentralized policies and funding models.