No evidence of a decrease in child maltreatment indicators in six developed countries from the late 1970s to the early 2000s

Date Published

Gilbert, R., Fluke, J., O'Donnell, M., Gonzalez-Izquierdo, A., Brownell, M., Gulliver, P., Janson, S., & Sidebotham, P. (2012). Child maltreatment: variation in trends and policies in six developed countries. The Lancet, 379(9817), 758-772

Reviewed by
Mireille De La Sablonnière-Griffin

This study aimed to determine whether policy initiatives and modern child protection systems in place since the 1970s have been successful in addressing child maltreatment. Using available data from six developed countries, the authors compared trends in child maltreatment from the late 1970s to the early 2000s. The three main indicators examined were: violent deaths, maltreatment-related injury admissions to hospital and contact with child protection agencies (i.e., reports, investigations, substantiated maltreatment, placement in out-of-home care). For this study, the authors analyzed: (1) trends within and between countries over time, and; (2) variation in maltreatment indicators between countries for the period of 2004 to 2006.

For trends within and between countries over time, the analyses were done on 12 subgroups based on two age categories (infants under 1 year and children above 1 year) and six countries (Australia [Western Australia only], Canada [Manitoba only], New Zealand, Sweden, United Kingdom and the U.S.). In terms of violent deaths and maltreatment-related injury admissions to hospital, findings indicated that these indicators remained stable or decreased after the mid-1990s in most of the countries. A decrease was noted for five age subgroups (out of 12 subgroups) for violent deaths (children above 1 year in Sweden and U.S. [since 1993], infants under 1 year in England [since 2000], and all children in Manitoba). Decreases in maltreatment syndrome or assault admissions to hospital were noted for children above 1 year in Sweden and for infants in Manitoba (for which a decrease for all maltreatment-related injury admissions to hospital was also noted); as such, a decrease in both rates (violent death and hospital admissions) was only found for Sweden and Manitoba. With regards to out-of-home care, findings demonstrated a significant increase in four age subgroups within three countries (infants in Sweden, all children in Western Australia, and children above 1 year in the U.S.). The only decrease noted was for children over 1 year in England until 2001, after which rates stabilized. Rates of reports and investigations, where available, were found to be mostly increasing, especially for infants. Officially recorded maltreatment and neglect increased in 4 age subgroups (out of 8; for at least one age group in three out of four countries). In contrast, officially recognized physical abuse only increased in Western Australia (infants) and New Zealand (children over 1 year), with the remainder being stable.

For variation in indicators between countries, the analyses were done on 3 age groups for each country (under 1 year, 1 year to 4 years, and 5 years and above). The mean annual rates in Western Australia were used as base rates, given that data for all indicators were available for this country. For violent deaths, Western Australia and Sweden had the lowest rates across age groups and the U.S. rates were more than 5 times the Australian rate. Rates for hospital admissions for maltreatment reasons were, however, very similar across countries. Regarding out-of-home care, rates were the lowest in Western Australia and Sweden, and highest in Manitoba across age groups, with the Manitoban rates being about 10 times greater than the Western Australian rates. For infants, placement rates in England, New Zealand and the U.S. were about two times higher than in Western Australia and Sweden. The rate of investigations in New Zealand and the U.S. were over 5 times greater than in Western Australia.

This study represents an important advance in understanding child maltreatment trends in that it compiled population rates using child-level data within 6 countries over a long time period. Yet, the authors clearly identify that variation or stability in indicators do not necessarily equate with change or a lack of change in child maltreatment. For example, changes in practices may provoke a decline in an indicator (recording of an event) which does not equate to a change in the occurrence of child maltreatment. Alternatively, stable rates may be an artefact of both falling rates and increased awareness in recognizing maltreatment. Thus, small and/or consistent changes must be interpreted with caution. While there is variation and little change overall between countries regarding hospital admissions and violent deaths, agency indicators were prone to change over time within a country and important differences were noted across the settings. The authors conclude that they cannot make definitive conclusions as to whether there is more child maltreatment in countries (Canada, New Zealand and the US) with higher and/or increasing rates of agency contact (investigations and out-of-home care in particular), as the rates could also indicate differences or changes in the indicators rather than in maltreatment itself.

Methodological notes

The six countries/states that were used for this study (Australia [Western Australia only], Canada [Manitoba only], New Zealand, Sweden, United Kingdom and the U.S.) were selected because of availability of data in addition to a diversity in policies and supports for child maltreatment and in social distribution of inequalities. It is important to note that while Manitoba and Western Australia are not necessarily representative of the Canadian and Australian trends, they were included because they have long records of high quality and linked data. The analyses were limited to children under 11 years of age (injuries related to physical assault or neglect in older children are less likely to be caused by parental or caregiver violence or poor supervision), and to 1979 forward (corresponding to the introduction of the International Classification of Disease [ICD], version 9). Because of the broad and changing nature of what is considered child maltreatment, the study focused on indicators of physical abuse and/or neglect (although some agency data may include sexual and emotional abuse). 

The indicator of violent deaths was defined using ICD-9 or 10 codes. The data for this indicator come from the World Health Organization and national statistical agencies in four countries – all countries have data as of 1979, until at least 2005 (U.S.) and up to 2008. The maltreatment-related injury admissions to hospital indicator was also defined using ICD-9 or 10 codes. It includes four subcategories: maltreatment syndrome, assault, undetermined cause, and adverse social circumstances. The data was drawn from local (Western Australia, Manitoba) or national (other countries) records. The years of available data varied across the countries, some with long ranges, such as Western Australia (1980-2005), Manitoba (1985-2008), and Sweden (1987-2009), while others were more limited, such as the U.S. (1997, 2000, 2003, 2006). To measure contact with child protection agencies, one to several indicators were used. For these indicators, each child with agency contact or who were placed each year were counted, except for England where the number of contacts were counted. All countries reported on out-of-home care, while four settings reported on officially recognized maltreatment, neglect and physical abuse (England, Western Australia, New Zealand and U.S.). Additional indicators include investigations (Western Australia, New Zealand and U.S.) and notifications (Western Australia and New Zealand). Again, a wide range of time intervals were observed, the longest being in Western Australia (1990-2005) and England (1988-2008 for some indicators), and the shortest being New Zealand (2004-2010) and the U.S. (2001-2007; for 20 states who participated in NCANDS only).

In term of analysis, the first component was to compute population rates using relevant population bases. These rates were calculated per 100,000 in the population because some events (such as violent deaths) are rare events. Subsequently, Poisson or negative binomial regression models were fitted for time-trend analysis of yearly incidence. These models detect any significant increase or decrease and, if relevant, the year in which the change in trend was effective. Consistency between rates within country and between countries were assessed qualitatively. In order to compare the absolute rates across countries, the mean yearly rate for 2004-06 were computed. Western Australia was used as the reference category and more than two-fold difference was considered a good indicator of difference in occurrence between the countries.