Journal Issue: Preventing Child Maltreatment Volume 19 Number 2 Fall 2009
For the purpose of developing a basic epidemiology of maltreatment, there are three primary sources of national data: the National Child Abuse and Neglect Data System (NCANDS), the National Survey of Child and Adolescent Well-Being (NSCAW), and the third National Incidence Study (NIS-3).13 Each source approaches the issue of maltreatment with a slightly different objective, and each collects data using a different method. NCANDS, described in greater detail in the accompanying box (opposite), is based on administrative data that states collect to manage their child abuse and neglect service systems. The data are tied to official reports of maltreatment, the investigation of those reports, and their disposition. Although NCANDS is comprehensive with respect to a wide range of victim, perpetrator, and service data, it is nevertheless limited in the following ways. First, NCANDS does not capture much in the way of clinical data about the family and the well-being of children, thus limiting the type of research that can be carried out with it. In addition, because NCANDS relies on official reports, state variation in reporting laws (for example, states use different definitions of abuse and neglect), evidentiary standards used by child protective services agencies to verify a report of maltreatment, and the number of investigators that a state deploys are thought to influence the process that leads to a disposition of the report.14
Certain gaps in NCANDS, such as the lack of clinical measures of child and family well-being, have been filled to a very large extent by NSCAW, which is also described in greater detail in the accompanying box (next page). NSCAW permits researchers to develop a much more comprehensive understanding of children investigated for maltreatment, from both a service and a developmental perspective. But because, like the NCANDS data, the NSCAW sample includes only children reported to public child welfare agencies, it is likely that neither source fully documents the extent of maltreatment in the United States.
The National Incidence Studies, initially mandated by Congress in 1974 and conducted periodically under the auspices of the Administration for Children and Families, are designed to provide a better estimate of the true incidence of maltreatment at a national level. The incidence studies rely on community sentinels as the reporting mechanism rather than the official data collected by state (or local) child welfare agencies. These sentinels report child maltreatment to the study team. They may also report the child to the authorities (for example, state child protective services), and child protective investigators may investigate the children thus reported. In the end, sentinel reports are compared with official reports to generate an unduplicated count of children abused during a specific time period. The third National Incidence Study, NIS-3, published in 1996, reported incidence rates that are higher than those reported with NCANDS.15 In general, findings from NIS-3 suggest that only 28 percent of the children meeting the harm standard were investigated by the child protective agencies. The under-reporting in NCANDS, judging from NIS-3, depends on the type of abuse and the report source.16 That said, I do not review the NIS findings here because the last published NIS data were collected in 1993. Maltreatment rates have dropped substantially since then, and it is simply not possible to say how findings from fifteen years ago are relevant today. As of this writing, the NIS-4 data have been collected, but the findings have not yet been released.
In addition to the three primary sources of national data, various types of self-report data address the incidence of maltreatment. The Gallup Organization, under the guidance of Murray A. Straus and colleagues, conducted perhaps the most widely cited self-report study.17 Typically self-report studies ask victims about their experiences (recollections in the case of retrospective studies). By contrast, the Gallup survey used the Parent-Child Conflict Tactics Scale, developed by Straus in the late 1970s, to ask parents about their behavior. The last Gallup survey (completed in 1995) that involved a national probability sample uncovered very high rates of maltreatment. Rates of physical abuse as reported by parents were about eleven times greater than the rate found in NCANDS and about five times greater than the rate reported with NIS.18 The Gallup survey also detected considerably more neglect.19
Research using smaller samples of self-report data has also been reported. Studies of this sort typically focus on improving estimates of the incidence of maltreatment (or understanding the difference between self-report and official report data), improving what is known about the underlying causes, or improving researchers’ understanding of how maltreatment influences child development over the long term. For example, Andrea Theodore and several colleagues sought to explain differences in officially reported abuse in North and South Carolina.20 Using the Parent-Child Conflict Tactics Scale, they found substantially higher rates of physical abuse than were officially reported. They also found that the differences between North and South Carolina using official data were larger than differences using self-report data.
Beth Molnar and several colleagues used the Conflict Tactics Scale to differentiate individual, family, and community risk factors and their influence on parent-child physical aggression.21 The findings, discussed in somewhat greater detail below, showed slightly higher rates of parent-child physical aggression than reported in other studies, including the Gallup study. The study also found that individual risk factors such as socioeconomic status, employment, and caregiver age were linked to physical aggression. Family and community protective factors, such as social support and a large social network, respectively, were associated with lower rates of physical aggression toward children.
Anne Shaffer, Lisa Huston, and Byron Egeland, in their longitudinal study of caregivers and their children, used a mix of prospective data (for example, collected from caregivers and other sources) and retrospective data (for example, self-reports of adolescents) to understand how the incidence of maltreatment was related to emotional and behavioral problems in late adolescence.22 They found that the incidence of maltreatment depends on how the data are captured. They also found a link between psychiatric disorders and how maltreatment was identified. For example, among subjects with both prospectively and retrospectively identified maltreatment, the share with any diagnosis reached nearly 75 percent. Among those children with only retrospectively identified maltreatment, the proportion with any clinical diagnosis was just under 64 percent.
Collectively, these studies illustrate how smaller, focused studies are used to clarify and otherwise sharpen researchers’ basic understanding of maltreatment: how often it happens, why it happens, and what its long-term effects are. The studies also reveal some of the fundamental problems in trying to provide reliable information for the purpose of designing preventive programs. Although maltreatment has broad implications for society as a whole, the dynamics of local communities would appear to influence parenting behavior. Studies based on national probability samples are less likely to reveal these local dynamics. By the same token, the data from smaller, focused studies are less useful when it comes to painting a national picture. Smaller studies are also expensive and are not conducted often enough to feed the continuous need for information felt by those charged with monitoring public programs. Administrative data such as NCANDS have the advantage of being routinely available. Administrative data can also be used to study maltreatment at small spatial scales.23 But, as noted, administrative sources likely under-report maltreatment, an important source of measurement error that has implications for how one uses what one learns.
In the end, the data one chooses to collect (and use) have to be matched to the question at hand. From the perspective of how one plans for and designs preventive programs, each type of data has a role. Administrative data and the data from national probability samples provide the information needed to allocate resources in relatively crude but important ways, especially if the data from smaller studies reinforce the essential findings. For example, and as discussed below, administrative data show persistently higher rates of maltreatment for young children (often under the age of one) than for older children, together with rising rates of maltreatment, particularly physical and sexual abuse, among adolescents. For the most part these same patterns are found in the small-sample studies. Administrative data also show that mothers are the most likely perpetrators and that poverty matters. Again, these findings are supported, by and large, in most if not all smaller-scale studies. What the administrative data do not provide is the detail needed to understand the mechanisms that underpin the most persistent patterns—knowledge that is essential to designing sound interventions.