Poisson regression is the best method to analyze cumulative adverse childhood experiences.

A cumulative count of adverse childhood experiences (ACEs) is associated with poor physical and mental health in adults and more recently associated with poor school performance and behavioral problems in children, although typically analyzed with binary logistic and linear regression models that may inaccurately bias the results. This study compared the results of a Poisson regression model with three binary logistic regression models of ACEs (i.e., 2-ACEs, 3-ACEs, and ≥ 4-ACEs) as well as two multiple linear regression models using ACEs as independent variables to predict children's internalizing and externalizing problem behaviors. We used 4,690 children's data from the Fragile Families and Child Wellbeing Study: a stratified, multistage sample of children born in large U.S. cities between 1998 and 2000, where births to unmarried mothers were oversampled. The children were 47.6% Black, 27.3% Latinx, and 21.1% White, and 4% were reported as other. Results showed that the Poisson regression model best fit the data compared to the logistic regression models based on comparisons of scatterplots of standardized deviance residuals. Results compared to the literature showed the Poisson and ≥ 4-ACEs model were comparable; however, the ≥4-ACEs model overpredicted negative outcomes for four or more ACEs and underpredicted negative outcomes for three or less ACEs. In addition, multiple linear regression results showed enhanced ACEs effects as suppressor variables. Poisson regression is considered the best method to analyze cumulative ACEs as the other methods yield biased results. (PsycInfo Database Record (c) 2025 APA, all rights reserved).