Abstract
Aims
With increasing access to integrated administrative data, and advances in predictive analytics, it is both theoretically possible and practically feasible to use predictive risk models (PRMs) to automatically risk stratify entire birth-cohorts as to their risk of experiencing multiple adversities in childhood (Vaithianathan et al., 2013, 2018; Rouland & Vaithianathan, 2018). Such automated screening tools allow agencies to identify families at highest risk and offer them preventive services in a timely fashion. However, little is known about what protective factors might exist amongst families who are identified as high risk by PRMs. Identifying protective factors is an important step in designing preventive services for families identified by PRM tools as well as helping social workers take a strengths-based approach to these families.
Methods
We used multiple waves of the Growing Up in New Zealand (GUiNZ) study which follows a cohort of children and their families (n = 5562). Children were coded to reflect the number of adversities they experienced by 54 months based on standard measures of Adverse Childhood Experiences (ACEs) (Felitti et al., 1998). A PRM was trained to predict two or more ACEs experienced by 54 months using only administrative data available at birth and routinely held by the government, and the most at-risk children (comprising the top 20% of risk) were retained for our analysis. This study examines potential protective factors associated with having no observed ACEs despite being predicted to be at high risk of ACEs. We coded these factors from multiple waves of mother and partner surveys, with 749 factors identified as candidate protective factors. These 749 factors were coded into conceptual domains using previous literature: mother-partner, family finances, parent health and wellbeing, community or neighborhood, or parent-child. Forward, backward and multivariable regressions were utilized to identify factors with the strongest associations with having no observed ACEs despite being in the high risk GUiNZ group of children.
Results
Of the whole cohort, 790 children were identified as being at greatest risk. Of these, 164 experienced no observed ACEs. The 749 protective factors that were tested fell into the following domains: mother-partner relationship (9%), family finances (23%), parent health and wellness (14%), community or neighborhood (36%), and parent-child relationship (9%). Those that were significantly associated with high risk children with no observed ACEs were from the following domains: mother-partner relationship (40%), family finances (22%), parent health and wellness (15%), community or neighborhood (13%), and parent-child relationship (13%).
Conclusions
Our findings suggest that important protective factors exist in the domain mother-partner relationship. While many of these factors might not be mutable, these results are suggestive of a useful domain for program designers and policy-makers to consider when serving high risk families. They might also be useful factors on which to focus when approaching families for recruitment into services.
View article here.
