Dimensions of perinatal and childhood adversities both merge and remain distinct

Perinatal and childhood periods are sensitive windows of development wherein adversity exposure can result in disadvantageous outcomes. Data-driven dimensional approaches that appreciate the co-occurrence of adversities allow for extending beyond specificity (individual adversities) and cumulative risk (non-specific summation of adversities) approaches to understand how the type and timing of adversities affect outcomes.

With evolving recommendations on what should be important in adversity research, we sought to establish a data-driven framework that accounts for both type and timing of adversity by (1) replicating dimensions of childhood adversities, (2) determining whether perinatal adversities form unique dimensions and (3) identifying whether adversities during the perinatal and childhood periods overlap or remain distinct.

Using 6815 9-10-year-olds from the baseline Adolescent Brain Cognitive Development (ABCD) study, mixed graphical models were fit independently to childhood adversities and perinatal adversities, and simultaneously to perinatal and childhood adversities, to model relationships among adversities.

Data-driven clustering approaches estimated dimensions of adversity within networks. Six dimensions of childhood adversities and five dimensions of perinatal adversities were observed. When considered simultaneously, dimensions of perinatal and childhood adversities both merged (e.g., parental circumstances during perinatal and socioeconomic status during childhood) and stayed independent (e.g., obstetric complications during perinatal and neglect during childhood) underscoring the importance of considering both type and timing when studying early life adversity.

These results highlight that it may be appropriate to study certain adversity dimensions independently, whereas for others considering the impact of timing and potential continuity in exposure is critical. Recommendations for adversity research are discussed.