Risk factors that predict longitudinal patterns of substantiated and unsubstantiated maltreatment reports

Substantiated and unsubstantiated reports of maltreatment are associated with similar risk of emotional and behavioral problems. However, substantiation status often determines service provision.

We examined substantiated and unsubstantiated reports to identify patterns of recurrence over a five-year period and identified family risk factors that predicted recurrence patterns.

We studied a subsample (N = 246,021) of the National Child Abuse and Neglect Data System from 2011-2015.

Measures included child, caregiver, and child protective services case characteristics obtained in 2011. We used latent class analysis to identify heterogeneous classes, then entered class membership as the outcome variable in a multinomial logistic regression to identify risk factors.

Four latent classes emerged: (1) initial unsubstantiation and moderate recurrence, (2) initial unsubstantiation and low recurrence, (3) initial substantiation and moderate recurrence, and (4) initial substantiation and low recurrence. Domestic violence (relative risk ratio (RRR) = 2.56, β = 0.94, SE = .02, p < .001), caregiver substance abuse (RRR=2.23, β=0.80, SE=.02, p < .001), and Native Hawaiian/Pacific Islander race (RRR=1.67, β=0.52, SE=.11, p < .001), predicted initial substantiation status but were not meaningful predictors of long-term recurrence. Prior substantiated report and poverty predicted initial substantiation status (report RRR=1.50, β=0.41, SE=.02, p < .001; poverty RRR=1.50, β=0.41, SE=.02, p < .001) and long-term recurrence (report RRR=2.60, β=0.96, SE=.02, p < .001; poverty RRR = 1.35, β=0.30, SE=.02, p < .001). Asian American race predicted low recurrence rates (RRR=2.09, β=0.74, SE=.12, p < .001).

Similar recurrence patterns between substantiated and unsubstantiated reports emphasize the importance of providing services regardless of substantiation status. Integrating administrative databases may reveal more variables that predict long-term recurrence.