Background: Precise future projection of population health distribution is imperative for designing an efficient healthcare system in rapidly aging countries. Multistate-transition microsimulation models such as the US Future Elderly Model have been developed based on panel data collection, but these data may not be always available. We proposed a pseudopanel method using repeated cross-sectional representative surveys as a complementary approach, and specifically applied the model to Japan's population.

Methods: We calculated birth-cohort and sex-specific prevalence for all combinations of 14 health statuses using microdata from five waves of the Comprehensive Survey of People's Living Conditions. Combining obtained prevalences with vital statistics data, we determined transition probabilities of statuses over time using contingency tables. Assuming that state transition and mortality-exit follow the first-order Markov process, we then designed a virtual Japanese population aged older than 60 years as of 2013 and performed a microsimulation to project disease distributions to 2046 with forward, backward, and external validation tests. Following validation, we compared our projection results with those based on traditional stativ models.

Results: Our calculated morbidity and mortality rates successfully replicated governmental projections of population pyramids and matched cardiovascular and cancer incidences reported in existing epidemiological studies, supporting the validity of our estimation. Our future projection of stroke and heart disease indicated lower prevalences than expected from static models, presumably because of recent declining trends in disease incidence and fatality.

Conclusions: Our pseudopanel approach provides a valid alternative microsimulation frame for future health projection in aging societies.

Keywords: Pseudopanel approach; microsimulation; forecasting; aging; comorbidities; Japan

JEL No. C53, I1, I12, J11, J14