Sensitivity analysis and methodological choices on health-related impoverishment estimates in Cambodia, 2009-17.

The measurement of health expenditure-related impoverishment as a proxy of financial risk protection is regularly used as an indicator of progress towards universal health coverage. However, the use of this indicator is greatly sensitive to analysts' choices and data sources, making comparisons across time and countries challenging. We report the results of a sensitivity analysis of critical methodological choices in estimating health-related financial impoverishment in Cambodia from 2009 to 2017. We include the following in our analysis: the construction and data sources for consumption aggregates and out-of-pocket health estimates; the use of international and national absolute and relative poverty thresholds (defined by the share of household food consumption); time and regional price adjustment methods and index sources. Marginal changes substantially affected estimates at the national and regional levels among households. In most cases, the choice of poverty thresholds and temporal and regional deflators had a significant effect. An increase of 0.01 USD in the average daily per capita poverty line resulted in relative increases in impoverished incidences of 2.90-2.62% for 2009 and 3.06-2.95% for 2014. From 2013 onwards, estimates for impoverishment in rural areas based on median food consumption were often significantly higher than estimates using official poverty lines. The high sensitivity of the impoverishment indicator cautions against its use in assessing health-related financial hardship and protection, especially with low and absolute poverty lines. In the context of low- and middle-income countries, assessing financial hardship in relative terms by using measures such as catastrophic health expenditure, complemented with research on coping strategies and their socio-economic effects on households, may be more conducive to policymaking goals and progress towards achieving universal health coverage.