Does How We Measure Matter? Evaluating Measurement Approaches to Childhood Adversity in Predicting Cognitive Aging

Background and Purpose: Early-life adversities can disrupt development and contribute to cognitive decline in later life, but it remains unclear how different measurement approaches might alter these observed effects. Life course theory and dimensional frameworks of childhood adversity highlight that adversity is not monolithic; different types – such as deprivation (e.g., neglect, poverty) and threat (e.g., abuse, violence) – may have distinct long-term consequences. Despite this, research often measures adversity as a simple cumulative count of adverse childhood experiences (ACEs), implicitly treating all events as equally harmful. This study compares three operationalizations of early adversity to assess how measurement choice impacts the prediction of later-life cognition.

Methods: We analyzed data from the Health and Retirement Study (2008 – 2020), focusing on respondents aged ≥65 at baseline (N = 11,320). Childhood adversities were retrospectively self-reported across multiple domains. Three measurement strategies were tested: (1) treating each adversity indicator as a separate predictor; (2) grouping indicators into two categories (deprivation and threat) with equality effect constraints; and (3) summing all adversities into a count score, distinguishing deprivation and threat. Cognitive function was measured at each biannual wave with the Telephone Interview for Cognitive Status (TICS-27). Latent growth curve models estimated baseline cognitive status and rate of change over time and the predictive strength of early adversity. After adjusting for relevant covariates, model fit and path coefficients were compared across approaches.

Results: All models demonstrated adequate fit, though findings varied by measurement approach. When each adversity indicator was examined separately as a predictor, several deprivation and threat exposures – such as low household income, few books in the home, sibling death, and household substance use – were significantly associated with both lower baseline cognitive function and faster cognitive decline. In the grouped adversity indicators with equality constraints model, where all deprivation and all threat items were constrained to have equal effects, neither dimension was significantly associated with baseline cognition or rate of change. In the summed score model, deprivation predicted baseline cognitive function but not cognitive decline, while threat was not associated with either outcome. Model fit indices favored the summed score model (CFI = .989, RMSEA = .029, SRMR = .018), followed by the individual indicator model and the grouped constraints model.

Conclusion and Implications: Measurement choice influenced the observed links between early adversity and cognitive aging. A simple summed score captured a general risk effect, supporting its pragmatic use, but it masked critical nuances. In contrast, examining individual items related to adversity provided a more nuanced perspective, highlighting how different types of childhood adversity uniquely contributed to later-life cognition. Measures that sum adversity into a count score risk overlooking important developmental processes by conflating deprivation, threat, or other unique forms of adversity. For social work research and practice, these results highlight the importance of dimensional adversity assessment to identify high-risk individuals and tailor early interventions. Advancing this line of research will require future work to refine measurement approaches that balance theoretical precision with feasibility in applied settings.