Child Abuse and Neglect: a socio-legal Study of Mandatory Reporting in Australia

Stage 2: Numbers and outcomes of notifications of each type of child abuse and neglect: Descriptive analysis and data mining

Stage 2 of the project explored numbers and outcomes of ‘notifications’ of each type of child abuse and neglect (physical abuse, sexual abuse, psychological or emotional abuse, and neglect) made by mandated reporter groups and other reporter groups, in each jurisdiction, for each year over the decade 2003-2012. The purpose of this aspect of the project was to use data mining and summary descriptive statistics to identify the reporting practice of different reporter groups for different types of child abuse and neglect, to identify significant trends in reporting within jurisdictions over time, and to compare reporting practices with legislative requirements and changes in legislative mandatory reporting laws over the decade.

Access to data

This stage of the project relied on each jurisdiction providing unit record data regarding notifications of child abuse and neglect over the decade. Jurisdictions in Australia have differences in the process by which they receive ‘reports’ of suspected abuse and neglect (these are termed ‘child protection notifications’ for other data recording purposes eg the annual Australian Institute of Health and Welfare Child Protection Australia report, and its ‘counting rule’ for child protection notifications), and of other matters concerning children’s welfare (these are termed ‘child concern reports’ for the AIHW data recording purposes). They also have differences in their classification and treatment of these initial intakes. Because of the parameters of our research project, our focus was on reports of suspected child abuse and neglect (classed as ‘notifications’); we did not seek to explore all reports of any matters concerning children (ie those concerning children’s welfare as well as those relating to suspected abuse or neglect). Further information in this regard is depicted in Appendix 1 Table 1 (Comparative table of child protection system intake processes and our data collection approach) and Appendix 1 Figure 1 (Key points in the child protection system intake process).

Data collection

Requests to provide data were made to each State and Territory government department responsible for receiving intakes concerning suspected child abuse and neglect. Data were provided to the research team at various times over the period 21 October 2013 to 25 July 2014, often with further discussions and negotiations required to facilitate complete submission of data and to clarify other aspects of the data provided. Four jurisdictions provided data exactly as requested (Australian Capital Territory, Queensland, Victoria, Western Australia). One jurisdiction, Tasmania, provided data as requested but only for nine rather than 10 years (i.e. 2004-2012). Two jurisdictions (Northern Territory and South Australia) provided aggregate summary data for the decade rather than unit record data. One jurisdiction (New South Wales) provided summary aggregate data only for the years 2010-2012, and some other aggregate data for prior years, which enabled only partial exploration of the research questions for that jurisdiction.

Data analysis

Data mining, also termed knowledge discovery, refers to the process of analysing large datasets to isolate significant or noteworthy patterns and trends. The process includes cluster analysis (identifying groups of data records), anomaly detection (identifying unusual patterns or outlier trends), association rule learning (identifying relationships between variables), and summarisation (generating a more concise representation of the dataset including creation of reports and summaries) (Han & Kamber 2012; Stephens et al 2006).

Key steps in the data mining process adopted in this project were:

(1) selection of data and obtaining data (research project conceptualisation informing the nature of the request for data; requesting the relevant data in useable form from government child protection departments in each of the eight jurisdictions; receiving the data in excel spreadsheets containing unit record data, or for two jurisdictions, aggregate summary data, and for one jurisdiction, some aggregate tables);

(2) data pre-processing, which includes data cleaning, including removing missing and irrelevant data (this included removing data where the coded maltreatment category was ‘not stated’ or ‘other’), and collation of the data;

(3) data mining, which includes:

•