From Adverse Underwriting Selection to the Risk of Portfolio Deterioration: Evidence from Group Health Insurance Using Renewal Analytics and Machine Learning Techniques

Adverse underwriting selection presents significant financial risk for health insurers by skewing the insured portfolio toward high-risk individuals and raising claim costs. Empirical evidence integrating region-wise risk analysis combining policy renewal behavior and underwriting (UW) analytics using machine learning techniques remains limited. This longitudinal study addresses the gaps by analyzing a health insurance database of 683 groups (0.40 M policies and 0.208 M claims) over five years. Influential UW and demographic characteristics with policy renewal behavior and estimation of future claim costs, combined with adverse UW selection, were identified using portfolio analysis, machine learning, and regression modeling. Results indicate that a uniform rating structure for the whole group apparently caused loss ratio escalation. Particularly, maternity-related claims were the largest portfolio segment, with 41% of claims escalating the loss ratio towards 100% and claims intensity to 101% in the northern region, indicating adverse UW. Data confirms that UW risk perils combined with demographic factors heavily influence the retention of policyholders and higher-risk groups staying within the insured portfolio for longer durations, compounding future liabilities. Hence, appropriate risk-specific UW combining region-wise claim behavior and structured demographic factors for insurers to detect new risk clusters can enhance renewal and pricing decisions to optimize portfolio sustainability.