Working Together to Care for Kids
This study focuses on formal carers (foster and relative/kinship) who look after one or more out-of-home-care child under 18 years of age in their household. The sample of the study was drawn from carer registration databases across the states and territories (i.e., jurisdictions). This section outlines the target population on which the sample was based, sampling procedures, data collection, questionnaire design and fieldwork outcomes. It also outlines the data weighting procedures for the final sample to adjust for differential probability of sample selection.
2.1 The study population and sample
The population for this study was foster and relative/kinship carers who were registered as formal carers in state and territory departments responsible for child protection across Australia and had at least one child under 18 years of age in out-of-home care who was living with them at 31 December 2015. The population definition applied to all jurisdictions except the Northern Territory (the Northern Territory did not participate in the survey due to some legislative and administrative barriers.)
The sample selection process involved engaging with the study's Working Group representatives from each of the participating jurisdictions. Sample selection was household-based and stratified by jurisdiction and care type (foster carer households and relative/kinship carer households). The main sample process in the jurisdictions (except the Australian Capital Territory (ACT) and Tasmania) involved a randomly extracted sample from foster care and relative/kinship households at a ratio of seven in every ten of these carer households in their database. In New South Wales (NSW), the out-of-home-care system includes non-government organisations (NGOs), which drew samples using the same procedure but with the care type undefined. Given their small carer populations, all foster and relative/kinship households in the ACT and Tasmania were selected.
Each jurisdiction then sent a letter, in early to mid 2016, to the primary carers within the selected households informing them that their contact details may be divulged to a reputable research fieldwork agency for this study, and providing them with the opportunity to contact their jurisdiction within a specified time frame if they did not wish to be contacted for the study. A flyer outlining the purpose and other aspects of the study was sent with the letter. The proportion of carers who contacted their jurisdiction to advise that they did not wish to be contacted for the study was 2-4% across the four jurisdictions that provided this information.
The sampling process was implemented slightly differently in Victoria due to the administrative requirements and privacy legislation in place in this jurisdiction; the main variation was that the Department of Health and Human Services (DHHS), Victoria, directly approached the carers and this was done by the selected fieldwork company (the Social Research Centre) on behalf of DHHS.
2.2 The survey content domains
The survey went through an extensive developmental process, which involved the research team at AIFS, DSS representatives and the Working Group, as well as the fieldwork company, the Social Research Centre (SRC), at the programming stage. The final survey instrument that was programmed for the main studies was comprised of the following modules:
Introduction (confirming who the qualifying respondent carer might be):
Screening (ascertaining the number of children currently in out-of-home care in the household and/or in care within the past six months);
Section A: The child in out-of-home care and household composition;
Section B: The carer's caring experience and child wellbeing;
Section C: Services and supports;
Section D: Motivations and challenges;
Section E: Carer wellbeing and family relationships; and
Section F: Demographics.
2.3 The fieldwork
The pilot study for the WTCKS was conducted in early August 2016 to test the survey questions and length as well as the effectiveness of the survey instrument. A total of 105 pilot interviews were completed, which led to further improvement in the instrument such as the clarity of some questions and survey flow. The main fieldwork period began in early September 2016 and concluded in early November 2016.
Carers who no longer had a child currently in their care as part of the out-of-home-care system but had done so in the past six months could still be considered in-scope for the survey. Carer interviews were terminated if it was discovered the respondent had not been a formally registered primary carer of at least one child aged under 18 (in either a foster or relative/kinship care arrangement) in the past six months. This was due to the consideration of difficulties in recalling information about the child and other information such as service use if the child had left more than six months ago. If the respondent had cared for more than one child in the past six months, a random selection process was applied to identify a "study child", who would be the reference child for several questions throughout the survey. This meant that the age of each child was recorded and one was selected at random, with preference given to the child/children who were currently in care, if applicable.
The SRC made contact with 5,275 carers. The final sample size was 2,203 interviews, representing 42% of the selected sample members. The numbers of interviews across participating jurisdictions by care type are shown in Table 2.1. A relatively small proportion (n = 151 or 7%) of interviews were conducted with carers who did not currently have a child in their care. The average interview duration was 33 minutes.
Note: * The care type was unavailable for responding carers from NSW NGOs. The care type was derived based on the carer's relationship to the child in out-of-home care.
Table 2.2 shows the details of call outcomes. The number of refusals was small. The numbers not leading to an interview were typically due to non-contacts (e.g., no answer, answering machines or disconnected phone services).
The cooperation rate is another way of considering response rate. Based on the known call outcomes, it is defined as the total number of interviews as a proportion of the total number of interviews plus all refusals. The cooperation rate for the main data collection was 87%. Table 2.3 shows that the cooperation rate varied somewhat across the stratification criteria of carer type and jurisdiction, with the main difference being the higher cooperation rate for Tasmania, which is likely due to its small sample size. The cooperation rate was also slightly higher in foster carer and NGO households in comparison to relative/kinship carer households.
2.4 Data weighting
As described above, the cooperation rates varied across the participating jurisdictions and care types. A data weighting procedure was applied in an attempt to adjust for differential responses and differential coverage (e.g., oversampling of small jurisdictions). Jurisdiction, care type (foster care, relative/kinship care, care type unspecified) and SEIFA Index of Relative Socio-Economic Disadvantage (quintiles) were used in the final weighting. The results of logistic regression of having a completed interview with three variables (jurisdiction, care type and quintile of the index of relative socio-economic disadvantage) suggest that the three variables were linked with the likelihood of participation in the study. For example, carers in the top quintile of the index of relative socio-economic disadvantage were more likely than those in the bottom quintile to participate in the study.
Survey weights were computed using the ipfweight algorithm (also known as iterative proportional fitting or raking) in Stata (Mannheim).This procedure performs a stepwise adjustment of survey sampling weights to achieve known population margins. The computed weights ranged from 0.7 to 1.79.
5 The data on the extent to which carers did not wish to be part of the study for ACT, NSW and Vic. were not available. For NSW, the NGOs sent out the letters but accurate/complete information could not be traced back.
6 While other variables such as age and gender may be important for data weighting, the variables were not provided by all the participating jurisdictions.
7 See <ideas.repec.org/c/boc/bocode/s457353.html> for further information. Retrieved 9/1/2017.