An Interpretable Machine Learning Approach to Assess Gender Equity in PM-JAY Healthcare Utilization: Evidence from Jharkhand

Abstract

Ayushman Bharat Pradhan Mantri Jan Arogya Yojana (PM-JAY) is an annual health insurance which covers about USD 7,000 per eligible family to ensure financial protection and healthcare access of the economically marginalized population in India. Despite substantial enrolment, disparities in actual service utilization persist across states, raising concerns regarding gender equity, particularly in resource-constrained regions such as Jharkhand. This paper compares gender differentials in using PM-JAY against the healthcare infrastructure at the district level and socio-demographic predictors. Method: This study utilizes the secondary data on the National Family Health Survey (NFHS-5, 201921) and official PM-JAY administrative data. Enrolment utilization gaps were measured using descriptive statistics. Ordinary Least Squares (OLS) regression was used to analyze findings based on the use of district-level determinants of PM-JAY claims. Moreover, a machine learning method that is interpretable (Decision tree) was applied to classify districts into high- and low-utilization groups through a median-based cutoff. The model establishes the hierarchical association between the literacy of women, the engagement of women in institutional healthcare, expenditure of out-of-pocket, and availability of empanelled hospitals. Results: Approximately 121 lakh Ayushman cards have been issued in Jharkhand, with substantial female enrolment. However, women account for only about 16% of authorized PM-JAY hospital admissions. Regression estimates indicate directional but statistically non-significant associations between utilization and socio-demographic predictors. The Decision Tree model achieves moderate classification accuracy (62.5%) and identifies empanelled hospital availability as the primary determinant distinguishing utilization levels across districts. Conclusion: While PM-JAY has expanded financial coverage, gender-equitable utilization remains limited. Strengthening healthcare infrastructure alongside targeted access interventions is essential to improve women’s effective participation under the scheme.

Ähnliche Arbeiten