The use of age-period-cohort models in the Dutch public health foresight study

Abstract Background The Dutch Public Health Foresight Study 2024 (PHFS) provides an outlook on future trends in public health and care. However, projecting incidence, prevalence, and mortality is challenging due to uncertainties in contextual factors like societal, economic, and technological developments. It is therefore crucial to utilize all available data from recent cohorts. To account for cohort effects, the 2024 PHFS incorporated Age-Period-Cohort (APC) models. Methods Previous PHFS editions utilized Age-Period (AP) models, which projected trends for age and sex groups using logistic regression, factoring in age and time period. These projections were then linked to population forecasts from Statistics Netherlands. The inclusion of cohort effects in the recent PHFS allowed APC models to also address cohort-specific patterns. This study compared AP and APC models for lung cancer prevalence, where cohort effects are anticipated due to changing exposures to smoking habits among younger cohorts. Data from the Dutch Cancer Registry were analyzed separately for men and women. Results Lung Cancer prevalence is higher in men but the gender gap has been narrowing over time. For men, the AP and APC results were similar. In women, however, APC projections suggested a more gradual increase compared to the steep, nearly exponential rise shown by AP models. Conclusions APC models outperform AP models when strong cohort effects impact disease prevalence trends. However, APC models are sensitive to assumptions, underscoring the need for expert input and further refinement. Incorporating APC models in foresight studies provides valuable insights for public health policy-making. Key messages • Adding cohort effects improves disease prevalence modeling for future projections. • An APC model is essential for foresight studies were delayed effects are relevant to support public health policy making.