Modelling Catastrophic Health Expenditures And Its Implication For Household Welfare In Malawi A Spatial Multilevel Approach

Abstract

Understanding characteristics of population groups vulnerable to catastrophic health expenditures and impoverishment due to health expenditures is important for designing financial protection programs and policies. This thesis developed a model for assessing the effect of household and neighborhood characteristics on the extent of catastrophic health expenditures and impoverishment due to health expenditures. The study used data from a cross-sectional survey conducted between April 2016 to April 2017 among 12447 households in Malawi. The outcome variables were the incidence of catastrophic health expenditures and impoverishment effects of health expenditures. Descriptive statistics such as proportions and means were used to describe characteristics of the sampled households. Moran I statistic was used to test for spatial dependence in impoverishment. Multilevel logistic model was developed to assess the effects of household and neighborhood characteristics on catastrophic health expenditures. Spatial multilevel logistic model was developed to assess the effects of household and neighborhood characteristics on impoverishment. Decomposition analysis was used to decompose socio-economic inequality in catastrophic health expenditures into its determinants. The thesis used simulation analysis to compare spatial multilevel model to multilevel and single level models in terms of overall model fit and performance of the parameter estimates. The analysis showed that 1.37% of the households incurred catastrophic expenditures. Visiting mission health facility, hospitalization, larger household size, higher socioeconomic status, living in central region and rural areas increased the odds of facing catastrophic expenditures. Majority of inequality in catastrophic expenditures is due to income, urban-rural and regional inequalities. 1.6% of Malawians were impoverished due to health expenditures. Lower socio-economic status, hospitalizations, chronic illnesses, residency in rural area increased the odds of impoverishment. There were significant spatial variations in impoverishment with higher spatial effects clustering in central region districts. Multilevel logistic model and spatial multilevel models provided the best fit to the data and unbiased estimated parameters. There is need design better prepayment mechanisms to protect vulnerable population groups and ensure progress towards universal health coverage. Policies aiming to reduce inequalities in health expenditures should simultaneously aim to reduce income, urban-rural and regional inequalities. Researchers using data from complex survey design in modelling health expenditures and its implications on household welfare should account for neighborhood and spatial dependence in the data.