• No se han encontrado resultados

CAPÍTULO 3. IMPLEMENTACIÓN Y PRUEBA

3.4 Pruebas

The NSCAW II is the second in a series of national studies collecting data from families reported to CPS for child maltreatment. The total sample consists of a national probability sample of 5872 children (0 to 17.5 years at time of sampling) and their families, who were investigated for child maltreatment between February 2008 and April 2009. Baseline data on these families were collected between April 2008 and December 2009. Released in the summer of 2012, the second wave of data (18 month follow-up) allow for the examination of

longitudinal indicators of child harm. Follow-up data were collected between October 1, 2009 and January 8, 2011 for all children participating in the baseline study regardless of age at follow-up. In NSCAW II, data were collected from interviews with children, caregivers, CPS caseworkers, and teachers (see Appendices 1-3 for measures table). Data collected from

children, caregivers, and CPS caseworkers has been utilized in this dissertation. Only one child per family was included in the NSCAW II. Therefore, children were not nested within

caregivers.

The total sample contained both families where the children remained in the home and families where the children were removed from the home following the index, or baseline, investigation. Primary caregiver interviews were conducted only in families where the children remained in the home at baseline making it possible to use self-reported variables in this sub-group including self-reported indicators of caregiver problematic substance use, emotional maltreatment, harsh discipline, caregiver depression, domestic violence, and criminal involvement. In order to utilize the caregiver-reported variables, two sets of analyses were

41

conducted. With separate analyses, the dissertation examined data only available for families in which the child remained in the home following the baseline investigation–the majority of CPS-involved children (87%; Dolan, Smith, Casanueva, & Ringeisen, 2011)–which is not available when children were placed in out-of-home care at baseline. For Aim 1, the samples for each hypothesis contain the entire sample: both families where the children remained in the home and families where the children were removed from the home following the investigation. For Aim 2, the samples for each hypothesis contain only families in which the children remained in the home after the baseline investigation.

Sampling Strategy

In order to understand the sampling strategy of NSCAW II, it is important to understand the sampling process in

NSCAW I (NDACAN, 2007).

A two-stage stratified

sampling design was utilized for both NSCAW I and NSCAW II (see Figure 5).

The first stage involved dividing the United States into nine sampling strata

(NDACAN, 2007). In NSCAW I, eight of the nine

strata correspond to the eight states with the largest child welfare caseloads in the nation. The ninth stratum contains the remaining 38 states and the District of Columbia. Four states were

Figure 5: Sampling Strategy

42

excluded from the study because CPS caseworkers were required to make contact with the family to discuss the study before the NSCAW research team. The second stage involved forming primary sampling units (PSUs) at the county level within the strata and randomly selecting the ones which were included in the study. PSUs were formed for all counties in the United States which investigate at least 60 to 67 cases of maltreatment per year. The random selection of PSUs involved using a probability-proportionate-to-size (PPS) procedure which gave a higher likelihood of selection to areas with larger caseloads. From the nine strata, 100 PSUs were randomly selected. Seven of the 100 PSUs were determined to be very small so they were combined with adjacent counties in order to make one PSU for the study. Six of these initial 100 PSUs refused and were replaced by PSUs of approximately the same size. Eight of the 100 PSUs were determined to be ineligible for the study because they were in states in which the local child protection agency had to make the first contact with the family before the

NSCAW I team could contact the family. This resulted in a total of 92 PSUs containing 97 counties in the NSCAW I sample. The same number of children was chosen from each of the 92 PSUs. The sample design required oversampling of infants in order to make sure there would be enough children to follow through permanency planning, oversampling of children who were sexually abused in order to allow for analysis of this type of abuse alone, and oversampling of children whose case was receiving ongoing services following the investigation so that there would be enough power to examine this process.

In NSCAW II, there are eight sampling strata rather than nine (NDACAN, 2011). In the time frame between the NSCAW I data collection and NSCAW II data collection, one of the eight states with the largest child welfare caseloads in the nation had a policy change requiring CPS workers to make the first contact. This state had to be excluded from NSCAW II resulting in

43

eight strata with seven strata corresponding to the largest child welfare caseloads in the nation and the eighth stratum containing the remaining eligible states. For NSCAW II, the same counties that participated in NSCAW I were asked to participate in the study. Seventy six percent of these counties participated in NSCAW II (NDACAN, 2011). Eight states were excluded from the NSCAW II study because state law required a CPS worker to make the first contact with the family rather than the NSCAW II study team. Affecting 9 PSUs, this was the most common reason that counties included in NSCAW I did not participate in NSCAW II (NDACAN, 2011). NSCAW II contains 81 PSUs of which 71 were in the NSCAW I sample. A complex weighting strategy accounting for stratification, clustering, weighting, and

oversampling of some subgroups was developed by the original study team to create a national probability sample.

Selection of Dataset

The NSCAW II dataset was selected for this dissertation over other datasets for several reasons. First, NSCAW II contains a wealth of information not available in most child welfare datasets, including many indicators of child harm and child well-being. Second, the dataset includes caregiver self-report data on problematic substance use. NSCAW II contains a large sample of children, necessary for path analysis, and it is a longitudinal sample with which it is possible to make national estimates. Finally, NSCAW II contains recently collected data.

Documento similar