Humanities & Social Sciences Communications (Oct 2023)

Assessing the impact of measurement error in household consumption on estimates of catastrophic health expenditure in India

  • Sanjay K. Mohanty,
  • Suraj Maiti,
  • Santosh Kumar Sharma,
  • Laxmi Kant Dwivedi,
  • Niranjan Saggurti

DOI
https://doi.org/10.1057/s41599-023-02226-4
Journal volume & issue
Vol. 10, no. 1
pp. 1 – 12

Abstract

Read online

Abstract The National Sample Survey (NSS) collects reliable data on morbidity, health care, and health spending through its annual multi-subject and multi-round health surveys. Evidences from these surveys have been extensively used for research and policy. While these surveys collect comprehensive information on morbidity, hospitalisation, health expenditure, information on household consumption expenditure (which is used to explain the economic gradient in health outcomes) is collected through a single question. Literature suggests that having a single question on consumption expenditure results in measurement errors. In this paper, we examine the effect of measurement errors of household consumption expenditure on estimates of catastrophic health expenditure (CHE) in India using data from the 68th round of the consumption survey (2011–12) and the 71st round of the health survey (2014), carried out by the National Sample Survey (NSS). The consumption survey canvassed a detailed schedule on consumption and interviewed 101,651 households, whereas the health survey interviewed 65,932 households from across the country. Descriptive statistics, estimates of CHE, and logistic regression models were used in the analysis. We used both the budget share approach and the capacity-to-pay approach for estimating CHE. The NSS health survey was found to have underestimated monthly per capita consumption expenditure (MPCE) in India by 32%, with the level of underestimation being significant across the states. Using the budget share approach, the CHE of India in 2014–15 was estimated at 23.4% without adjusting for the underestimation of consumption and 21.1% after adjusting for it. Similarly, using the capacity-to-pay approach, CHE was estimated to be 13.4% without adjustment and 10.4% with adjustment. The estimates differed considerably across the states. In general, it was observed that the use of a single question on consumption overestimated CHE in India. The pattern was similar regarding the intensity of CHE. The predictors of CHE were similar using both the methods, but the unadjusted estimates of CHE showed significantly higher predicted probabilities of incurring CHE across household characteristics. It is recommended to include disaggregated questions on household consumption in the future rounds of the NSS-based health surveys. Researchers using NSS data need to be aware of the effect of measurement errors of consumption expenditure on estimates of catastrophic expenditure. Adjusting for the underestimation of MPCE may improve the estimation of CHE in India.