DIGITAL ACCESS AND HEALTHCARE UTILISATION IN URBAN INDIA: UNRAVELLING THE ROLE OF SOCIOECONOMIC INEQUALITY

Abstract

Digital health infrastructure is widely promoted as an equaliser of healthcare access in low- and middle-income countries. We ask whether, in urban India, digital access in fact narrows or widens wealth-related inequality in healthcare utilisation, and we develop the formal machinery needed to answer that question. We make three contributions. First, we treat digital access not as a binary but as a latent capability, estimating a Digital Capability Index (DCI) from six ordered indicators via a two-parameter logistic item response model (marginal reliability 0.746). Second, we derive an exact decomposition showing that the contribution of digital access to the concentration index of utilisation splits into an access channel, A = 2β_DCov(θ, r)/μ, and a returns channel, R = 2γCov(νr, r)/μ, and we prove the existence of a neutrality threshold γ* at which digital access is inequality-neutral. Third, we apply this to a micro-dataset of 14,200 urban adults calibrated to published margins of NFHS-5, the NSS 75th Round and the NSS 79th Round CAMS, identifying the effect of digital capability through a control-function instrumental-variables design built on block-level backhaul connectivity (first-stage F = 102.9). We find a sharp tension. Marginal returns to digital capability are strongly pro-poor: the effect on utilisation falls from 0.383 in the poorest decile to 0.153 in the richest. Yet digital capability is itself so pro-rich in distribution that the access channel (+0.0456) overwhelms the equalising returns channel (−0.0145), leaving a net contribution of Δ = 0.0322. Digital stratification thus accounts for 30% (95% CI 10–49%) of measured pro-rich inequality in utilisation, a Digital Inequality Multiplier of M = 1.43. Returns would need to be roughly 3.1 times more pro-poor (γ* = -0.803 against an estimated γ = -0.255) for digital access to be distributionally neutral. A conventional Wagstaff decomposition using OLS coefficients understates the digital contribution by a factor of roughly 2.9. The policy implication is uncomfortable for universal digital-health strategies: because returns are pro-poor but access is pro-rich, untargeted expansion raises average utilisation while widening the gradient, whereas expansion targeted at the bottom two quintiles improves both.