Abstract

RENDON, Simón Cuartas; RAMIREZ GUEVARA, Isabel Cristina  and  CARDONA JIMENEZ, Johnatan. Evaluation of aggregation in Age-Period-Cohort (APC) models with a Bayesian approach. Ciencia en Desarrollo [online]. 2025, vol.16, n.2, pp.135-139.  Epub July 20, 2025. ISSN 0121-7488.  https://doi.org/10.19053/uptc.01217488.v16.n2.2025.17094.

Poisson Regression models of Age-Period-Cohort (APC) are employed in epidemiological studies to estimate the impact of each of these factors on the trends in incidence and mortality rates for various diseases. Typically, information on observed cases or mortality from a disease is summarized in a table with two entries: age group and calendar period in which the event of interest was recorded. One of the challenges researchers face in various contexts and scenarios requiring APC model applications is data aggregation. In most studies, the available information is aggregated either by age groups, periods, or both characteristics. In this regard, it would be relevant to assess the impact of this data aggregation on both the estimation of model effects and the predictions obtained from it. It could be inferred that this aggregation, usually done in 5-year periods or quinquennia, may significantly affect the quality of predictions, as trends or seasonal patterns may be lost in this process. Thus, this work conducts an analysis on the impact of aggregating incidence data (by periods) on future projections or predictions. In particular, a comparative study is carried out on tuberculosis mortality rates for the case of Colombia, and projections are shown for three period aggregations: five-year periods, three-year periods, and annuities.

Keywords : Age-period-cohort (APC) models; Bayesian approach; Aggregation analysis.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )