An application of these analytical choices can be found in this issue, as performed by Sisitsky et al. (
2023). The authors utilized data from a population-based birth cohort of mostly racial and ethnic minority youth, generally exposed to heightened levels of ELA, who were born between 1998 and 2000 across 20 large cities in the United States (the Future of Families and Child Wellbeing Study, N = 2,483, 51.6% male). ELA was measured using a range of variables when children were 3 years old, including parent-rated reports as well as community-level statistics, and subsequently combined into a broader multi-adversity model. Interestingly, confirmatory factor analyses showed that a unidimensional model (similar to a
cumulative model) fit the data poorly, while a two-dimensional model of ELA (similar to the threat-deprivation
dimensional model) showed a marginal fit. The optimal solution consisted of a four-dimensional model, differentiating between home threat, community threat, neglect, and lack of stimulation. This indicates that, in some cases, an approach that retains more specificity in adversities can better capture variance within ELA, as compared to higher-order solutions. However, a key difference with the traditional
specificity model – which typically focuses on one single adversity – is that here multiple adversities were still examined in a comprehensive manner, allowing to model the co-occurrence of ELAs. As a next step, a
person-centered approach was used to examine associations between the four ELA-dimensions and child biopsychosocial outcomes at age 9. These analyses identified 8 distinct subgroups based on unique patterns of exposure to home threat, community threat, neglect, and lack of stimulation. While 5 of the subgroups were characterized by the levels of a single ELA dimension (e.g., community threat), the other 3 subgroups (collectively representing over half of the sample) showed varying levels across multiple ELA dimensions. This suggests that the specificity model may work best for some individuals, but misses the more complex patterns of ELA experienced by others. In turn, these subgroups were found to be differentially associated with internalizing and externalizing behaviors (but not telomere length), indicating that, as the authors put it, “
it is not just the amount of ELA, but the combination of exposures that predict child mental health outcomes”. Notably, the authors also ran associations using
variable-centered analyses for comparative purposes. These generally produced consistent results in terms of identifying unique associations between specific adversities and outcomes, however were less informative regarding the impact of heightened exposure to multiple adversities.
Overall, the study of Sisitsky et al. (
2023) provides an important example of how different analytical approaches can be integrated to better understand the complexity of ELA and its associations with child mental health, while still producing findings that have the potential to inform public health policies and intervention strategies. A key message that emerges from this work is that one size
does not fit all – in many cases, population-level screening tools that focus on exposure to single adversities may suffice to identify at-risk individuals; however, these strategies should be complemented with a more personalized approach to capture those with complex ELA profiles, in order to improve risk prediction and offer more tailored support. Nonetheless, caution should be exercised when interpreting the identified subgroups and their outcomes, as the subgroups were derived from a specific cohort and may not generalize to other populations, and as such await replication. Further, exposures and outcomes were measured at a single time point, which may obscure important developmental dynamics in the relationship between ELA and mental health. Indeed, timing and chronicity of exposure to adversity are important, but understudied factors in ELA-outcome associations, due to challenges in measuring these characteristics reliably and the limited availability of cohorts with repeated ELA data. In this context, it is noteworthy that childhood adversities are often preceded by prenatal adversities (e.g., maternal exposure to stressful life events or psychopathology), as evidenced by the known stability of risk factors across these developmental periods. This means that observed effects of childhood adversity on outcomes may be partly due to exposures occurring before birth, and conversely, that prenatal effects on outcomes may be partly mediated by adversities during childhood. Despite this, information on prenatal and postnatal ELA are rarely studied simultaneously, pointing to an important avenue for future research.