Nonlinear modeling of health checkup factors and future claims-based healthcare costs in Japan: a retrospective cohort study

BACKGROUND: Population health checkups support disease prevention and health management. Clarifying how routinely collected checkup factors relate to subsequent healthcare costs may inform population-level risk assessment and preventive planning. We aimed to characterize nonlinear and sex-specific associations between routine checkup factors and near-term claims-based healthcare costs in Japan and to explore whether composite metabolic indices provide additional information beyond routine measures. METHODS: We linked Japan's Specific Health Checkups (FY2012-FY2022) to subsequent administrative insurance claims in Hakui City and analyzed adults aged 60-74 years. The outcome was annual claims-based healthcare costs in the third and fourth fiscal years after baseline, expressed as fee schedule points. We fitted generalized additive models with a Gamma distribution and log link in the overall sample and separately for men and women. Predictors included routine checkup measures (including glycated hemoglobin [HbA1c]) and lipid- and adiposity-related indices (atherogenic index of plasma, arteriosclerosis index, non-high-density lipoprotein cholesterol, fatty liver index, and visceral adiposity index). Model fit and the stability of the observed association patterns were assessed using 50 repeated 80/20 individual-level split validations. RESULTS: =0.06), indicating that routine health checkup variables alone explained only a small proportion of variation in subsequent claims-based healthcare costs. Age and HbA1c showed consistent associations across groups. HbA1c showed a J-shaped association with costs, with a steeper increase above approximately 6.0%. Body mass index, waist circumference, and fatty liver index showed moderate association patterns, whereas the atherogenic index of plasma, arteriosclerosis index, and visceral adiposity index provided limited additional information beyond routine measures; the fatty liver index showed a notable association pattern in men. CONCLUSIONS: Routine health checkup factors exhibited nonlinear, sex-specific associations with near-term claims-based healthcare costs in a Japanese community. These findings may inform population-level hypothesis generation and future indicator selection. However, routine checkup variables alone appear insufficient for precise individual-level prediction. External validation and integration of additional information on comorbidities, medications, healthcare utilization, and socioeconomic factors are warranted.