Artificial Intelligence and Machine Learning in Indian Health Insurance: A Critical Analysis and the SUTRA Framework for Risk Assessment and Moral Hazard Containment

Abstract

Purpose: Indian health insurance is expanding rapidly while remaining structurally loss-making in parts of the market, and both insurers and regulators increasingly attribute the gap to moral hazard on the demand side, supplier-induced demand on the provider side, and fraud at the interface between them. This paper critically examines what artificial intelligence (AI) and machine learning (ML) are actually delivering in the Indian health insurance market, as distinct from what is claimed for them, and develops an integrated framework through which Insurtech firms and insurers can measure risk and contain moral hazard without converting analytics into an instrument of claim denial.

Design/methodology/approach: The study adopts a pragmatist, mixed-methods design executed in two stages. Stage one is a systematic critical review of 96 sources covering the health economics of moral hazard, the AI-in-insurance literature, and Indian regulatory and industry documentation, synthesised into a gap analysis. Stage two is theory-building: the paper constructs the SUTRA framework (Sensing, Underwriting, Triage, Restraint, Assurance) and, within it, a composite Moral Hazard Exposure Index (MHEI) that decomposes exposure into demand-side, supply-side and intermediation components. A complete empirical validation protocol is specified, including construct operationalisation, sampling design, and the statistical procedures (exploratory and confirmatory factor analysis, PLS-SEM, binary logistic regression, propensity-score matching and difference-in-differences) required to test the eight hypotheses derived from the framework.

Findings: The critical analysis identifies a systematic asymmetry: AI adoption in Indian health insurance is concentrated in cost-reducing and claim-restricting applications, while applications that would reduce moral hazard at its source — prevention, care navigation, provider contracting — remain marginal. Three structural constraints explain this: fragmented and non-interoperable claims data, the absence until recently of any AI governance framework, and an incentive structure in which the returns to denying a claim are immediate and measurable whereas the returns to preventing one are deferred and diffuse. The framework responds by embedding a proportionality constraint that ties the intensity of any restraint action to the strength and auditability of the evidence supporting it.

Originality/value: The paper offers the first integrated framework that treats moral hazard containment in Indian health insurance as a measurement problem rather than a detection problem, distinguishes the three loci of hazard rather than collapsing them into “fraud”, and specifies a graduated intervention ladder that is auditable against IRDAI’s emerging AI governance expectations and the Digital Personal Data Protection Act, 2023.