Individual Risk Prediction of Adult Health Outcomes Based on the Adverse Childhood Experience (ACE) Questionnaire

• In Event: Investigating the Value of Screening for Adverse Childhood Experiences (ACEs) in Predicting Poor Health

Abstract

Introduction: Over the past 2 decades, several studies have documented associations between adult recall of adverse childhood experiences (ACEs) and risk for adult health outcomes. These studies found that participants who reported higher number of ACEs had higher risk of mental and physical health problems. Following from these epidemiological findings, there has been increasing interest in using ACEs screening to identify individuals at high risk of later disease. This screening strategy has several advantages — for example it is cheap and can engage people by acknowledging and aiming to reduce their lived experiences of adversity. This strategy has also been criticised for several reasons. Most relevant here is the concern that differences in disease risk emerging from the comparison of groups with or without high ACEs may not translate in accurate risk prediction for individuals within the groups. This is because the models developed to date assigned the same predicted probability to each individual within one the two groups (or within each group with different ACEs scores, more broadly), but the observation is typically that individuals within any of the groups have high heterogeneity in outcomes.

Hypotheses: We tested whether ACE screening can accurately classify individual risk of health outcomes. Because of the high heterogeneity observed in previous research, we hypothesise that ACE screening will show poor classification performances.

Study population: Approximatively 8,000 adults from a Health Maintenance Organisation in the United States of America.

Methods: The ACE questionnaire includes questions on adverse events recalled from childhood including psychological, physical, or sexual abuse; violence against mother; or living with household members who were substance abusers, mentally ill or suicidal, or ever imprisoned. The ACE questionnaire was mailed to adults who had previously completed a standardised medical evaluation. We adopted a commonly used coding comparing study participants with 4 or more ACEs to those with 3 or fewer ACEs.

Results: Group level predictions of adult health outcomes ranged from Odds Ratio 1.6 to 12.0, all of which were statistically significant. However, individual risk prediction was inadequate. Discrimination was overall poor — only the Area Under the Curve for suicide attempt and injected drug use was acceptable. The specificity of health outcome prediction based on ACE questionnaire was generally high, but the sensitivity very low across outcomes. Only a small proportion of study participants with high ACE score had adult health outcomes, indicating low positive predictive value. In contrast, most of study participants with low ACE score did not have adult health outcomes, indicating high negative predictive value.

Authors

• Alan Meehan, King's College London
  Presenting Author

• Stephanie Lewis, King's College London
  Non-Presenting Author

• Jessie R Baldwin, University College London (UCL)
  Non-Presenting Author

• Andrea Danese, King's College London
  Non-Presenting Author