3.1. Within Generation Results
Hypotheses one through three were addressed using both the national and regional samples. With regard to the national data, the 44 state sample included 825,763 individuals of whom 31.82% had at least one re-report over three years. In the 35 state sample, there were 505,191 unserved individuals and 162,443 served individuals. Among those served, 37.55% had at least one re-report, while among the unserved cases, 29.81% had at least one re-report (See Table 1
). Hazard curves showing re-reports over time can be found in Figure 1
(Full 44 state and 35 state samples) and Figure 2
(Served and unserved cases from 35 state sample).
There was little variability among the 44 and full 35 state samples with regard to the proportion of index reports from MHSS sources (11.16% vs. 10.32% respectively), or professional sources (52.37% and 56.29%, respectively). There was also little variance in proportions of re-reporter types between samples at 36 months post-index. The total number of unique MHSS re-report sources ranged from 7.37% (35 state unserved sample) to 9.04% (35 state served sample). The total number of unique professional re-reports ranged from 38.42% (45 state sample) to 39.64%% (35 state served sample). As can be seen from the above numbers, the percentage of unique MHSS re-reports was always lower than the percentage of MHSS index reports. The same was true for professional reports. It is important to note the small proportion of unique MHSS re-reports, which demonstrate that large SB effects on the order claimed (a 50% increase in re-reports) are mathematically impossible. Even if all unique MHSS re-reports were due to SB, the total magnitude of the SB effect would be less than ten percent at 36 months post-index for all samples.
The data did support hypothesis 2 to a limited degree. Two trends suggesting a real but small SB effect were found. First, the 35 state sample served cases had higher a higher percentage of total re-reports when compared to unserved cases (37.55% to 29.81%, respectively). This difference, however, was not mainly due to increased numbers of unique MHSS and professional re-reports. Proportions of re-reports from professionals were virtually identical between served and unserved groups (39.64% vs. 39.61%, respectively). Cases in the 35 state served sample did experience more unique MHSS re-reports than the unserved 35 state sample (9.04% vs. 7.37%, respectively). This supports a small SB effect among MHSS re-reporters but not the broader class of all professional re-reporters, as the SB hypothesis would predict.
Post-hoc analyses were performed to more precisely quantify this potential SB effect. These analyses were performed using the national data to estimate how total observed re-reports would have lessened if unique MHSS re-reports in the “served” subsample had increased only at the lesser rate seen among other reporter types. This was done mathematically by reducing the number of unique mental health social services reporters (UMHSS) re-reports among served cases to the lesser rate of re-reports among UMHSS unserved cases (See Supplementary Materials
detailed procedure and calculations).
Using this procedure, we estimate that at 3 months post-index report, SB could plausibly increase re-reports among cases served at index by +4.54%. This should be interpreted to mean that if there would have been 100 re-reports (At three months) among served index cases without SB being present, SB could inflate the number of re-reports to 104.54 re-reports.
If all index cases are considered (Not just served cases) at three months post-index report, then the increase in re-reports reduces to +1.33%. This should be interpreted to mean that if there would have been 100 re-reports (at three months) among all index cases without SB being present, SB could inflate the number of re-reports to 101.33 re-reports.
These effects degrade rapidly over time. When the same analyses are made at 36 months, the corresponding increases due to SB are only 1.84% among served index cases and +0.52% among all index cases.
Moving away from the national data, the regional study provided more specific service and family data (See Table 2
). Data in Table 2
are broken down by service type and poverty status of the family at baseline, to allow separate consideration of SB in poor and non-poor families. Among the 7185 index reports, 17.01% were from MHSS sources, with little variability by poverty status. Among re-reports, about twice as many re-reported cases had a re-report from a MHSS source (38.35%), but less than half as many had unique MHSS re-reports (only 7.66% of all re-reports). Surprisingly, this number was somewhat higher for non-poor children (10.38%) than for poor children (6.96%). With regard to hypothesis 1, we see that the percentage of UMHSS re-reports is lower than the percentage of MHSS index reports, so hypothesis 1 is not supported. Given that unique MHSS re-reports only constitute about one in fourteen re-reports, it is impossible for these re-reports to be causing a large increase in the total number of re-reports.
The regional data do provide some support for hypothesis 2, as CPS cases served by FCS or FPS (But not foster care) showed rates of UMHSS re-reports roughly twice as high (10.78% to 22.22%) as cases receiving no services (5.96% to 10.84%). This could be more evidence of a small SB effect. Contrary to expectation, foster care cases showed relatively low rates of unique MHSS re-reports.
Hypothesis 3 asserts that as service intensity increases, the proportion of unique MHSS re-reports will also increase. This hypothesis was not supported. Proportions of unique MHSS re-reports in less intensively served (FCS) and more intensively served (FPS) cases were quite similar , being, respectively, 19.47% vs. 22.22% for non-poor cases and 11.01% vs. 10.78% for served cases Foster care cases always had a relatively low proportion of unique MHSS re-reports, never exceeding 8%.
3.2. Intergenerational Results
Intergenerational issues could only be assessed using the regional sample (See Table 3
). Column 1 describes the G1s’ childhood level of child protective service involvement: Never reported (28%), reported but never provided services (27.5%), reported and provided services but were not still eligible for child welfare services when they gave birth (25.2%), reported and provided services and were still eligible for child welfare or were in foster care when they gave birth (19.2%). G1s with prior histories of services following a report did have a higher rate of re-reports for maltreating G2s than those without CPS histories (See column 3). At first glance, this might appear supportive of the idea of intergenerational SB. However, were this increase due to SB, we would expect greater proportions of professional reporters among more intensively served children. This was not the case.
Hypothesis four stated that, among subjects who became parents, those previously reported as victims of maltreatment will have higher proportions of reports from professional reporters. Hypothesis five then posited that as service intensity increases, the proportion of professional reporters would increase. There were no statistically significant differences in proportions of reports by professional reporters by prior CPS history and, in fact, the trend was in the opposite direction (See column 4). Those G1s with no prior CPS history had a higher proportion of reports made by mandated reporters (Over 57%) than any other category (Between 32% and 53%). There was a trend for increased professional reporters among those G1s who had services (53.1% and 42.8%) following a CPS report compared to those with reports but not served (32.3%) but this was not statistically significant.
Childhood CPS involvement was indeed posited as a reason for a type of intergenerational surveillance bias by Widom and colleagues [7
]. However, when looking at intergenerational surveillance, it is also possible that risky or concerning G1 behaviors or situations might place G1s in contact with professional providers separate from any child protection history. If this were to occur during pregnancy or following childbirth this could trigger concern among a professional reporter about the G2 child without the reporter having any knowledge of prior G1 maltreatment reports. A post-hoc assessment of potential surveillance by professional reporters subsequent to onset of pregnancy (And before the alleged report of G1 as perpetrator) was therefore examined by looking at service sector use related to issues that could place the G2 at risk: criminal behavior, mental illness or health care associated with domestic violence or STI. Over 35% of the G1s that were reported as alleged perpetrators were involved in systems for at least one of these issues. There was no statistically significant difference between those without CPS histories and those with CPS histories (See column 5). Although poverty is not a reason for reporting maltreatment, receipt of income assistance does mean that the G1s are in contact with another system with mandated reporters. When we added receipt of income assistance during the same time period (Not shown) the majority of all G1s were already being served by a non-CPS system prior to being reported as an alleged perpetrator, (68.7% to 90.5%) further diluting any possible “indirect SB” effect attributed to childhood CPS contact.