This appendix contains additional detailed information regarding specific aspects of the methods used to construct the database and analyze the data.
States were considered for inclusion if they had submitted case-level data to the NCANDS for calendar years (CY) 1998–2002. CY 1998 and 1999 data submissions were made in Detailed Case Data Component (DCDC) format; CY 2000, 2001, and 2002 data submissions were made in the Child File (CF) format. Prior to use in the analysis, the DCDC data sets were converted to CF format.
Tests of unique child identification numbers (IDs) were conducted to determine whether a reasonable number of children were recorded as rereported, and the extent to which demographic data were the same from report to report. First, each State’s multiyear data set was examined to determine whether at least 2 percent of all the report-child pairs in the data set had matching unique child IDs. While no CPS system is able to prevent children from being rereported, some States do not adequately identify children who are rereported. Thus, States that do not identify the same child over repeated events will have close to 0 percent duplicate child IDs. The 2 percent threshold for matching unique IDs, while arbitrary, was chosen to exclude States that do not adequately identify rereported children and to include States with actual low rates of rereporting.
If a State’s multiyear data set met this threshold, at least 95 percent of children with the same IDs also needed to match on demographic characteristics. Additional tests were performed to assess whether the cumulative percentage of IDs that were reused from one year to the next increased systematically from year to year. Finally, all of the ID tests were repeated to assess whether the criteria for IDs were met for both the set of children who were considered victims and those children not considered victims.
Given the clear importance of history of victimization, analyses focused primarily on children with no such history of victimization. In order to identify these children, the prior victimization variable was used. Two States were excluded because their data indicated that more than 98 percent of report-child pairs had no history of prior victimization. In addition, one State was excluded due to anomalous data for prior victimization which was deemed internally inconsistent.(9) For the remaining nine States, it was possible to identify children who had not been victimized prior to their initial investigation in the data set and therefore to compare children who had a history of victimization (n=117,874) with those who did not (n=1,396,998). The data set included data from these 9 States.
A single file that included all the data submitted for the calendar years 1998–2002 was created for each State. To standardize the data files, investigations with a disposition date after December 31, 1997, but with a report date prior to January 1, 1998, were excluded.
Certain variables were derived from standard NCANDS fields to simplify and support the anticipated analyses. These recoded variables allowed for either production of single variables containing mutually exclusive categories or to combined categories which otherwise have a small number of cases. The derived variables are shown in Table A–1.
|Variables||Recoding and Derivation of Variables||Factor Categories|
|Source of Initial Report||Sources such as friends and neighbors, and victims were combined into a single category of nonprofessional||Social and Mental Health Services|
|Law Enforcement or Legal Personnel|
|Daycare and Foster Care Providers|
|Nonprofessional and Other|
|Child Age at Initial Report||Age at time of report was grouped into categories||Infants|
|Child Race and Ethnicity||Race and Hispanic ethnicity were combined into single variable with mutually exclusive categories||White only|
|American Indian and Alaskan Native Only|
|Asian and Pacific Islander Only|
|Other and multiple race, non-Hispanic|
|Unable to determine and missing|
|Maltreatment Type||A single variable with mutually exclusive categories was derived from a multiple response variable||Physical Abuse Only|
|Neglect or Medical Neglect Only|
|Sexual Abuse Only|
|Other Abuse Only|
|Multiple Abuse Types|
|Caretaker Abuse of Alcohol||Flags for the presence at the time of the report||No|
|Child With Indication of Disability||Flag for the presence of any:
|Child’s Initial Investigation Victimization Status||Flag||Nonvictim|
|Postinvestigation Services Provided||Flag, yes or no||No|
|Child Placement in Foster Care||Flag, yes or no||No|
The data from multiple reports involving the same child ID were combined to develop variables that indicated whether the child experienced one or more subsequent reports or subsequent victimizations. These variables were constructed to support the use of event history analysis procedures. This step also combined reports made within the same 24-hour period since children who were reported more than once within 24 hours were considered to be reports of the same incident. Derived variables are shown in table A–2.
|Variables||Recoding and Derivation of Variables||Factor Categories|
|Rereport Indicator||Flag for the presence of a second report for the same child||No|
|Length of Time to Rereport||Time in days between initial report date and subsequent report date was calculated|
|Revictimization Indicator||Flag for the presence of a subsequent victim disposition report for the same child victim||No|
|Length of Time to Recurrence||Time in days from initial report date to subsequent victimization was calculated|
It is important to note that there are some limitations to the analysis. First, as with any analysis utilizing administrative data, the analysis is limited by both what data are collected and the accuracy of the data. Data were tested for quality and validation procedures were performed. However, only reported child maltreatment can be analyzed; therefore, the above findings pertain only to reported child maltreatment, not to child maltreatment in general.
Further, while receipt of services was noted in the data analyzed, it was beyond the scope of this study to explore related issues such as assessment, referral and refusal of services, compliance with case plan, the length of services, worker and provider contacts, the number of services, or other factors that might aid in explaining the findings observed.
While the data analyzed represent a sizable number of cases, generalizations of the findings to the national level are not recommended. However, the degree to which the findings replicate those noted in other related research suggests that the factors identified above are important for policymakers fashioning plans for intervention and prevention of children’s reentry into the child welfare system.