Actuarial Contributions to the Study of Health Insurance

Kivonat, rövid leírás

This dissertation examines how patient characteristics and actuarial methods can be combined to design fairer, more sustainable health insurance in Hungary. I start by mapping the landscape of key private health insurance products and the pricing techniques that underpin them. Alongside classic actuarial tools—commutation functions, Markov models, and net premium calculations—I discuss the practical problem of thin or fragmented data and show where expert judgment must stand in for missing statistics. The takeaway from these product-level exercises is sobering: a “cover‑everything” all risk type contract remains prohibitively expensive for most households, which makes targeted, data‑aware design essential. The second part turns to patient behavior. We adapted and validated the 13‑item Patient Activation Measure (PAM‑13) for Hungarians aged 40+, testing distributional properties, internal consistency, test–retest reliability, factor structure, and multiple facets of construct validity in an online sample. PAM‑13 performed well (α≈0.77; ICC≈0.62), showed a single‑factor structure, correlated as expected with eHealth literacy, and—crucially—was unrelated to age, education, or income. Higher activation aligned with healthier lifestyles and more proactive information‑seeking, with most associations holding after adjustment for literacy, sociodemographics, and health status; effects were clear in chronic patients, weaker in the oldest group. These results support PAM‑13 as a robust risk indicator for our context. In the third part, I test whether incorporating patient activation can improve pricing models. I build a series of Gamma‑log GLMs for medical expenditures using the validated PAM‑13 and pre‑pandemic Hungarian data (n=779) alongside standard predictors (age, sex, education, income, region/settlement type, EQ‑5D‑5L, health and eHealth literacy). Across model variants, higher activation is associated with lower expected costs after adjustment, while health status and geography also matter; estimates are benchmarked against NEAK 2019 utilization to anchor magnitudes. The modeling section also details practicalities actuaries face—outlier handling, omnibus and Wald testing, and the trade‑off between interpretability and incremental accuracy when considering more complex learners. Policywise, the finding that activation systematically lowers spending opens a path for activation‑aware pricing or benefits (e.g., education, coaching, or incentives) that reward engagement without breaching fairness or transparency requirements. Together, the three studies—two already published, one forthcoming—show that better insurance is not only about sharper mathematics but about measuring what patients do. Validated activation metrics, coupled with transparent GLMs, can move Hungarian health insurance toward designs that are both affordable and behavior‑sensitive.