Contrasting Cumulative Risk and Multiple Individual Risk Models of the Relationship Between Adverse Childhood Experiences (ACEs) and Adult Health Outcomes

Abstract

Background A majority of the documented relationships between adverse childhood experiences (ACEs) and adult health outcomes are based in cross-sectional self-reported datasources such as the CDC’s Behavioral Risk Factor Surveillance System survey. Despite using the same or similar questions, there is a great deal of inconsistency in the operationalization of self-reported childhood adversity for use as a predictor variable. A cumulative risk model, resulting in a cumulative numeric ‘ACE Score’, is normative but there are differences in the way the ACE Score is calculated and used in statistical models. Alternative conceptual models are rarely used, even though predictor characterization directly impacts interpretations about the effects of adversity on outcomes. We investigated differences in model fit and performance between the cumulative ACE Score and a ‘multiple individual risk’ model that enters individual ACE events individually into prediction models. Methods We merged the 2011–2012 BRFSS data (N = 56,640) and analyzed 3 self-report outcomes. We compared model fit metrics and used Vuong’s test for model selection to arrive at a best fit model. Results The multiple individual risk model was a better fit than the categorical ACE Score for the ‘lifetime history of depression’ outcome. For the outcomes of obesity and cardiac disease, the cumulative risk and multiple individual risks models were of comparable fit, but yield very different inferences. Conclusions Additional information-rich inferences about ACE-health relationships can be obtained from including a multiple individual risk modeling strategy, either in addition to or in place of a cumulative risk ACE Score.

Adverse childhood experiences and incident coronary heart disease: a counterfactual analysis in the Whitehall II prospective cohort study

“…A major advantage of an ACE score is that it permits the examination of dose-response associations, as well as defining the total exposure to ACEs in a simple manner. On the other hand, an approach that recognises different effect sizes in a simultaneously adjusted model [ 28 , 29 ] tackles some of the challenges of ACE scores [ 27 , 30 ].…”