Improving the Adverse Childhood Experiences Index By Incorporating Frequency Data

Introduction

Studies have examined the associations between adverse childhood experiences (ACEs) and a variety of outcomes, finding that greater exposure to adverse experiences before the age of 18 predicts poor health outcomes. These studies typically operationalize exposure to ACEs as the number of categories of ACEs which a respondent has experienced. However, the ACE battery captures information about the frequency with which the respondent was exposed to many of the ACEs. We expect that health outcomes are related not only to the breadth of exposure to ACEs, which is measured by the traditional score, but also to the frequency of exposure. In our study, we examine whether we can improve predictions of a range of health outcomes by using information about the frequency with which respondents have experienced various ACES.

Methods

The Behavioral Risk Factor Surveillance System (BRFSS) survey administered in Washington State in 2009-2011 included an ACE module. It also included questions about their health, including questions which we categorized as relating to mental health, physical health, and risk-taking behaviors. Each category included multiple outcomes. Using this data, we calculated the traditional ACE score for each respondent, as well as a frequency weighted ACE score which weighted each ACE by the frequency with which it was experienced. All models controlled for respondent sex, race, and age, and used the sampling weights provided in the dataset. We determined which score was a better predictor of each outcome by considering the Akaike Information Criterion, an estimate of the model’s badness of fit.

Results

We found that the frequency weighted score model were better predictors of all physical and mental health outcomes. Differences were especially large for mental health outcomes, for which models using the frequency weighted definition of the ACE score had AIC values close to 20 points lower than models using the traditional definition. For reference, an AIC difference of 4 is said to strongly favor the model with the smaller value. By this criterion, the AICs strongly favored models using the frequency weighted definition of the ACE score for all mental and physical health outcomes.

In contrast, the AIC strongly favored all models using the traditional ACE score for all risk-taking behaviors outcomes.

Conclusion

We demonstrate that a frequency weighted definition of ACEs is a better predictor of certain outcomes than the traditional definition. This result is important for two reasons. First, it helps people and agencies that need to predict negative health outcomes to construct more effective models. Second, it contributes to our understanding of the mechanisms by which adverse childhood experiences lead to negative health outcomes.