The need for evidence of the effectiveness of family violence interventions arises every day in a variety of settings: in courtroom deliberations, in which judges must decide appropriate punishment and treatment for offenders; in child protection investigations, in which caseworkers must consider what services to recommend for their clients; and in clinical settings, in which health professionals are expected to develop an appropriate response in examining suspicious injuries of adult and child patients. Indicators of program effectiveness are the focus of attention when program and service budgets are under review, in public hearings in which victims and advocates seek legislation to strengthen existing services or to develop new ones, and in media reports that compare the merits of customary practice with innovative approaches to addressing child maltreatment, domestic violence, and elder abuse.

Scientific research has not been able to provide clear guidance in determining the comparative effectiveness of specific efforts. Much of the literature simply describes individual programs rather than rigorously evaluating their outcomes. And most evaluations of family violence interventions simply describe the intervention that was provided (without providing much detail about the implementation process), estimate the number of clients served over selected time periods, review the types and costs of the services provided, or examine specific skills or knowledge obtained as the result of a training program. These types of studies are commonly referred to as process evaluations, because they focus on the processes of the intervention rather than its results.

An emphasis on theory and outcomes has been missing in evaluations of family violence interventions: What behaviors was the intervention seeking to

influence, and how do they relate to the overall problem of child maltreatment, domestic violence, or elder abuse? Did the child, adult, or family actually improve as a result of the program and, if so, how was this measured? Did the community experience a lower rate of family violence following the intervention? Did changes in skills or knowledge result in modifications in behavior among offenders or lead to new services in communities that reduced the violence and provided more safety or improved health? And are the results of a single program or comprehensive community intervention transferable to a larger and more diverse population of clients and multiple communities?

Both public and private agencies now seek to move beyond anecdotal reports in a search for data that can guide policy and program decisions. Yet service providers, public officials, clients, and researchers themselves are frustrated by the difficulties of measuring and assessing the impacts of treatment and prevention programs. The enormous complexity of the phenomena, the comparatively short history with service interventions, the shifting legal and social doctrines that shape public policy, the interactive nature of the problems and the services themselves, and the demographic transformations affecting American families and communities present tremendous challenges to the use of science in this field. The urgency and magnitude of these problems and the scale of resources that our society invests in treatment and prevention efforts demand better results.

Family violence victimization in the general population is widely regarded as a serious problem that affects large numbers of children and adults across the life span. Estimates of the scope of the problem vary according to the source of the information that provides the basis of the estimate and the definitions used in describing the nature of the problem. But although the exact dimensions of child maltreatment, domestic violence, and elder abuse are frequently disputed, conservative estimates suggest that these problems affect millions of children, women, and men in the United States. In 1996, a government survey (NIS-3) reported that 2.8 million cases of child maltreatment were known to local child protection agencies or community sources (including teachers, health care professionals, and other service providers) in 1993, when the data were collected (National Center on Child Abuse and Neglect, 1996b). This estimate indicates a rate of 41.9 children per 1,000, or 1 in 24 in the U.S. child population (National Center on Child Abuse and Neglect, 1996b:3-17).

No government survey has yet been published that estimates the national incidence of domestic violence in the United States. The National Crime Victimization Survey, which collects data about incidents reported as crimes, has reported that the annual rate of physical attacks by family members for women was 9.3 per 1,000 in 1992-1993 (Bachman and Saltzman, 1995). Population-based surveys suggest that the rate of adult violence involving family members may be

much higher—the National Family Violence Surveys (Straus and Gelles, 1990), for example, have reported an annual rate of 116 per 1,000 women for a violent act by an intimate partner and 34 per 1,000 for ''severe violence" by an intimate partner.

The rate of child deaths attributed to abuse and neglect remained relatively stable (around 2,000 per year) from 1979 through 1988 (McClain et al., 1993). Of the 4,869 female homicide victims reported in 1993, 31 percent (1,531) were killed by intimate partners (husbands, ex-husbands, or boyfriends) (Federal Bureau of Investigation, 1993). No reliable statistics are kept for rates of elder abuse or deaths attributed to the maltreatment of older persons. There are no self-report surveys of elder abuse, and the surveys of elder abuse reporting and recognition are incomplete.

The study of family violence consists of many separate areas of research focused on the study of child maltreatment, domestic violence, and elder abuse (Ohlin and Tonry, 1989). Each area has its own terminology, theories, experts, funding sources, data collection efforts, research instruments, and scholarly journals. Those who study domestic violence have rarely communicated with, or even read the works of, those who study child abuse. This specialization and fragmentation, combined with a lack of methodological rigor, have contributed to a research literature that has been characterized as "extensive but not definitive" (Ohlin and Tonry, 1989:5).

The absence of unifying theories, common measures, and datasets has discouraged efforts to draw inferences between studies of family violence and studies of other forms of criminal and violent behavior. Little is known, for example, about the pathways or nature of the relationship between childhood experiences with violence in the home and juvenile delinquency or adult criminal behavior, even though research has suggested that children who experience maltreatment have an increased risk both of arrest as a juvenile or adult and of committing a violent crime (Widom, 1992). Similarly, much uncertainty exists about the extent to which men who batter their intimate partners engage in other forms of violent or criminal behavior. The organization of research in this field, within both academic centers and federal research bureaucracies, often isolates different forms of family violence research and discourages efforts to integrate research on violence within the family and research on violence within the community.

The fragmentation that characterizes the research field also exists in the service delivery system. As a result of history and social forces, the law enforcement, social services, and health care efforts that are focused on different aspects of family violence must address diverse policy objectives. These goals include the punishment and rehabilitation of offenders, the protection of victims and communities, the reduction of unnecessary costs, a respect for due process and

fairness (often difficult to achieve when conflicting testimony exists or physical evidence is not available), concern for the rights and developmental needs of children and vulnerable adults, the mitigation of the adverse effects of experiences with violence, and the prevention of new incidents of violence in the home. Such services are subject to enormous social and political influences in the wake of sensational cases. The result is a patchwork of separate efforts that operate in a semiautonomous manner, although the presence or absence of one component may have direct implications for another. The basic dimensions of this loosely coupled system of treatment, prevention, and control interventions have remained ambiguous and unexamined in the research literature, and little is known about their basic character, operation, or impact.

Treatment and prevention programs are fragmented by the focus of the interventions: some efforts are designed to respond to the needs of the victims; others deal exclusively with the offenders. Some programs focus on children, others on adults. Some programs seek to serve individuals; others emphasize the importance of serving their clients in the context of their relationships as parent, child, intimate partners, or members of a family unit.

Service providers and researchers are realizing that family violence is an interactive and dynamic problem that requires approaches across multiple levels of analysis and multiple service systems (National Research Council, 1993b, 1996). This realization has profound consequences for the development of the next generation of treatment and prevention programs as well as the design of studies of their effectiveness. The emergence of a public health response and comprehensive community initiatives focused on family violence has stimulated interest in examining collaborative approaches to multiple problems within a family. These multiple problems may exist entirely within the context of family violence (such as the impact of witnessing spousal violence on the developmental outcomes of children) or they may co-exist within a family with a broad range of behaviors (such as poverty, family violence, and alcoholism or substance abuse). Artificial barriers between separate programs and institutional services still remain, but in some regions they are yielding to more integrated and inclusive approaches that focus on the need to support individuals, families, and communities as they seek to reduce the incidence of violent behavior in families. Evaluation of these comprehensive community efforts is in its infancy, challenged by the complexities noted below as well as the difficulties of evaluating programs that are not yet fully developed.

The evaluation research literature has provided limited guidance in the design of treatment and prevention interventions for family violence. Significant methodological and logistical problems include difficulties in constructing and gaining access to appropriate sample sizes, limited availability of comparison and

control groups, weak research measures and survey instruments, short time intervals for follow-up studies, and high attrition rates in both the interventions themselves and the evaluation studies. Some interventions that are in widespread practice have never been evaluated (such as the use of mandatory reporting procedures for child maltreatment cases), whereas extensive attention has focused on a few interventions in specialized settings (such as arrest policies for domestic violence, the use of home visitation services as a preventive intervention for child maltreatment, and intensive family preservation services that seek to provide family support services in crisis-oriented setting).

This imbalance between the need for knowledge in areas that are hard to measure and the availability of knowledge in discrete areas of policy and practice creates two major quandaries that this report seeks to address: (1) how to review the implications of current research findings in a form that can be useful for researchers, service providers, and policy makers in various institutional settings who are concerned with discrete and multiple aspects of family violence, (2) how to determine which interventions should be strong candidates for the next generation of evaluation studies. The problem of selecting interventions for evaluation will acquire increased significance as larger numbers of agencies in social services, law enforcement, and health care seek to address the complex dimensions of the problem of family violence within the loosely coupled but interactive system of services that has emerged at the national and local level.

In addition to the generic challenges to effective evaluations, studies in this specific area are complicated by a number of factors:

• Most family violence interventions have resulted from advocacy efforts in communities that focus on the specific needs of children, women, and the elderly. Many evaluations focus on a single program rather than an overall intervention or service strategy.
• Many interventions are not fully implemented, and the status of their implementation is often missing in evaluation reports.
• The urgent emphasis on meeting the needs of victims and taking action against offenders has caused services to be developed in the absence of a knowledge base about the causes and consequences of family violence. As a result, the theoretical frameworks that guide specific interventions are often not explicit and may vary among individual programs. In addition, limited attention has been paid, in either the research literature or the design of interventions, to the impact of unsubstantiated reports of family violence on the affected parties.
• Measures of the nature and scope of family violence often come from court or administrative records rather than health records or the observation of families and couples themselves. Administrative records often include biases that distort the ability of researchers to identify key trends or risk and protective factors that may influence patterns of aggression and the outcomes of interventions.
• When evaluations are closely coupled with program development, the

• research base may not have an opportunity or sufficient resources to develop an array of studies that can build on the strengths of qualitative and quantitative research. The programmatic emphasis on examining the results of an intervention within a short time period can inhibit efforts to learn more about the basic nature of the services that were provided and the characteristics of clients who did or did not respond to them.
• Researchers are often prevented by resource constraints as well as the nature of service systems from designing experimental studies that make use of an appropriate control group in comparing the relative effects of a selected program. Controlled studies are especially difficult to develop when services are limited by budgetary restrictions, access to services is determined by multiple agencies within a community, and service providers are not accustomed to working with researchers. Controlled studies are also not feasible in circumstances in which random assignments cannot be made for ethical or legal reasons, such as felony violence.
• Even when it is possible to conduct rigorous evaluations, program advocates have sometimes resisted scientific study because of concerns that the research will interfere with the basic service mission of the program, that the researchers do not have the capacity to measure significant interactions, or that negative findings may weaken support for existing services. Researchers, as well, have sometimes displayed a lack of appreciation for the complexities of dealing with the problems of family violence and underserved populations, focusing on factors that are relatively easy to measure, such as attendance rates or length of service, rather than addressing more difficult issues, such as the capacity to change one's own or another's behavior or consideration of the presence of multiple stressors and interactive processes that may be significant in examining the impact of policy and programs.
• The lack of opportunities for collaboration between researchers and service providers has impeded the development of appropriate measures and study designs in assessing the impact of programs. It has also discouraged research on the design and implementation of service interventions and the multiple pathways to services that address the causes and consequences of family violence.

These complications have hindered the development of a useful body of scientific knowledge about the impact of family violence interventions. In the absence of such knowledge, program decisions are influenced by anecdotal reports, marketing or lobbying efforts, and the fiscal realities of budgetary politics. Proponents of law enforcement programs may argue that stiffer penalties and mandatory sentences for offenders are effective in reducing the incidence of domestic violence and child maltreatment, but such claims lack empirical evidence. Similarly, advocates for women and children may argue that public funds should be spent on social service and public health programs designed to meet the

needs of victims of abuse and neglect, but such claims lack valid indicators of effectiveness.

The Committee on the Assessment of Family Violence Interventions was convened by the National Research Council to undertake five primary tasks:

1. Document the impact of family violence on public- and private-sector services in the United States, especially in the area of spousal abuse. This analysis should include an assessment of what is known about the scope and cost of service interventions, the adequacy and validity of current reporting and detection data that are used to estimate the incidence and prevalence of selected types of family violence, and areas of bias in the current reporting system that may exaggerate or underestimate estimates of service intervention costs.
2. Synthesize the relevant research literature and develop a conceptual framework for clarifying what is known about relationships among the risk and protective factors associated with family violence. This framework should make distinctions between areas in which evidence of harm is clear in the interaction of certain risk and protective factors and areas in which knowledge is limited or uncertain.
3. Characterize what is known about both prevention efforts and specific interventions in dealing with family violence, including an assessment of what has been learned about the strengths and limitations of each approach and factors to consider in designing program evaluations.
4. Highlight successful components and, to the extent possible, identify "best practice" models of specific intervention programs for different forms of family violence. This task includes the examination of policy and program elements and service delivery system features that appear to contribute to or inhibit the development of effective program responses to family violence.
5. Develop criteria and principles that can guide the development of future evaluations of family violence interventions, especially the identification of outcomes and methodological issues that are essential to rigorous evaluation efforts.

Early in the course of our deliberations, the committee recognized that the research literature associated with evaluations of family violence interventions was relatively immature and scattered across a wide range of research fields. Judging that the absence of an integrative review of this literature has discouraged efforts to understand the nature, objectives, and outcomes of interventions, the committee sought to develop a report that would describe the ways in which interventions and service strategies interact with families and communities. Although some types of interventions have been in place for decades (for example, child protective services), they have not been subjected to rigorous evaluations to

examine their impact on offenders and victims. Experience with other types of interventions is relatively new; many interventions now in place for domestic violence, for example, are not yet fully implemented and have not yet been evaluated. The committee therefore directed its efforts toward reviewing the knowledge base associated with the array of interventions currently in place, identifying the evaluation studies that met minimal standards of scientific quality, highlighting the major challenges to evaluations of these interventions, and examining how problems such as weak design and limited study samples could be resolved in developing the next generation of evaluation studies.

In developing the criteria and principles to guide future evaluations, the committee was not asked to address the nature of public policy, funding, and other critical factors that could influence whether or not these principles and criteria are actually used. Clearly, there is a need for further consideration of the potential roles of public agencies, both federal and state, and private foundations in improving the quality of both quantitative and qualitative evaluations for family violence interventions. The types of collaborations that might foster progress in examining the impact of the wide range of services already in place in this field also deserves attention. The committee does address the need for improved partnerships among research, practice, and policy in this report, but specific questions pertaining to the adequacy of funding for research, infrastructure, and implementation were beyond the scope of our study.

The term family violence is applied to a broad range of acts whose presence or absence results in harm to individuals who share parent-child or adult intimate relationships. These relationships are often long-term and generally involve aspects of financial or physical interdependence. ¹ The definition of family violence adopted by this committee is based on one developed by the Committee on Family Violence convened by the National Institute of Mental Health (1992:6):

Despite progress in identifying the characteristics of family violence, serious problems remain in defining the severity of the injury or threat, the amount of harm, and the extent of the consequences associated with its occurrence (Ohlin and Tonry, 1989). The question of what constitutes a family relationship is also complicated by the diversity of intimate or dependent caregiving relationships in home environments. Furthermore, uncertainty remains about the strength of the relationship among the different forms of family violence. Definitional issues are discussed further in Chapter 2.

This report examines opportunities for interactions among child maltreatment, domestic violence, and elder abuse in terms of both research and service. These relationships are important because the forms of family violence may share common risk factors or may represent a developmental continuum, such as the case of an abused child who becomes either a batterer or a victim of domestic violence. They are also important because opportunities may exist in the service delivery system to combine or integrate services and to reinforce individual program goals by adopting common strategies. The report does not address the issues of sibling violence or violence among same sex intimate adults, primarily because research in these fields is less well developed and the committee did not identify evaluations of interventions for these forms of family violence that met the selection criteria for the study.

Although the report often emphasizes the interactions among different forms of family violence, it is important to maintain diverse approaches in examining the array of programs, services, and interventions. In some situations, the origins of domestic violence may have stronger theoretical links to other forms of violence against women, such as rape and sexual harassment, than to other forms of family violence. Similarly, some forms of child maltreatment may have stronger theoretical links to interpersonal violence, such as pedophilia, assault, and homicide, than to other forms of family violence. The complexity of the field is daunting to scientific study; theory building remains in an early stage of development.

In formulating its approach, the committee developed distinctions among strategies, interventions, and programs. Although these terms are often used interchangeably in discussions of program effectiveness, they are not synonymous. By strategies, the committee refers to a general course of treatment, control, or prevention that is designed to achieve a broad social goal, such as the prevention of family violence, the identification and protection of victims, and the punishment or deterrence of offenders.

The term interventions refers to the array of services and policies used to implement strategies at the individual, family, and community levels. Examples of interventions include parenting education programs for child maltreatment cases, treatment services for batterers, and universal screening for pregnant women to enhance the access to services of victims of domestic violence. Interventions include not only organized services but also legal, judicial, and regulatory policies and procedures, such as protective orders and mandatory reporting systems, some of which have been evaluated in the research literature.

Programs are community-specific examples of particular interventions. Batterer treatment interventions, for example, may be voluntary or mandatory, depending on the setting and context of a particular program. Some interventions consist solely of a single community's program; other interventions may be composed of several programs that share a common theoretical framework but rely on different approaches in achieving their goals. These program variations often are regarded as the same intervention. The home visitation intervention, for example, consists of several models, some of which rely on public health nurses who establish contact with mothers during their pregnancy and provide frequent visits after birth; others rely on paraprofessionals who meet with the mother after her discharge from a hospital. The home visitation intervention is part of a broader child abuse prevention strategy designed to detect risk factors for child maltreatment and to strengthen parenting skills through public health and family support services.

This report focuses primarily on interventions rather than individual programs or broad strategies. Specific program examples are used when they illustrate community efforts to implement a particular intervention or when they have served as the subject of an evaluation study. Table 1-1 identifies the specific family violence interventions that are reviewed in this report. Each intervention is numbered by chapter and by type of violence (A-child, B-domestic, C-elderly). Thus, for example, intervention 4A-1 is the first child abuse intervention discussed in Chapter 4. Figure 1-1 illustrates how the interventions reviewed in this report can be arrayed in terms of the general strategies of identification, prevention, protection, treatment, legal separation, and deterrence.

The committee has identified interventions that have been evaluated in the literature, as well as key interventions worthy of study because of their perceived value to victims and communities in responding to family violence. Some evaluations that did not meet the committee's criteria for inclusion nonetheless are discussed in the report because they offer insights concerning outcome measurement and program implementation.

The literature review covers scientific journals, federal research reports, national research databases, state agency reports, and foundation studies. In our search we identified and reviewed over 2,000 studies to determine if they included evaluations that relied on a scientific design that included the use of a control or comparison group. The use of a control group is critical to good evaluation research, because it allows the researcher to compare the relative effects of the intervention under study to what might be expected in its absence (Metcalf and Thornton, 1992). In selecting evaluation studies for detailed analysis, the committee relied on the following criteria:

• The evaluation involved a program intervention that was designed to treat or prevent some aspect of child maltreatment, domestic violence, or elder abuse;
• The evaluation was conducted between 1980 and 1996 (this time period was selected to provide a contemporary history of the evaluation research literature while maintaining manageable limits on the scope of the evidence considered by the committee);
• The evaluation used an experimental or quasi-experimental design and included measurement tools and outcomes related to family violence; and
• The evaluation included a comparison group as part of the study design.

A total of 114 evaluation studies were identified that met this standard (see Table 1-2). Most of the studies reviewed by the committee have been published in the peer-reviewed literature, although some state agency reports, foundation reports, and studies in progress are included in the research base for this report. In addition, the committee relied on 35 research review papers that include detailed analyses of independent research studies, even though the individual studies within each review paper may not be consistent with the selection criteria outlined above.

The committee initiated its search with a review of the relevant on-line databases that include evaluation studies of family violence interventions.² We

also contacted several public and private national clearinghouses, which aided in identifying unpublished materials that were included in the literature review. In addition, we learned of a small number of evaluation studies that meet our criteria but have recently been completed or exist only as state agency reports and have not been included in the literature or national clearinghouse collections.

It is important to note that the range of evaluation studies presented in this report does not reflect the current range of treatment and prevention interventions. The research literature is primarily focused on small or innovative programs that provide an opportunity for research study. Science-based evaluations are rarely conducted on major existing interventions, such as foster care, domestic violence shelters, adult protective services, health care provider training programs, and court procedures, because research samples and appropriate measures are not available to assess the intervention, the intervention was already in place prior to the development of evaluation methods, the organizational setting is not

conducive to evaluation, or policy and program sponsors are not prepared to provide funds to examine the impact of the intervention.

As noted later in this report, rigorous evaluations are focused primarily on social service interventions in the area of child maltreatment and legal interventions for domestic violence. Experimental evaluations of interventions for elder abuse are almost nonexistent.

The report begins by examining what is known about the scope of family violence, risk factors, the nature and scope of interventions, and their impact and costs (Chapter 2). Next we examine the state of the art of evaluations, identify obstacles to evaluations in this field, and explore potential improvements to the use of qualitative and quantitative information (Chapter 3). Chapter 3 also highlights the importance of structuring partnerships between researchers and service providers so that relevant expertise can be applied to study designs, the selection of outcomes of interest, and the use of appropriate measures.

Chapters 4, 5, and 6 provide a comprehensive review and assessment of existing interventions in the areas of child maltreatment, domestic violence, and elder abuse. These chapters characterize the state of evaluation research and summarize what is known about treatment and preventive interventions in social service, law enforcement, and health care sectors, respectively. Interagency family violence interventions and comprehensive services or community change interventions are discussed in Chapter 7. These four chapters aim to capture the insights of the research base as well as those of service providers—a perspective generally not found in the scholarly literature.

Chapter 8 discusses a number of cross-cutting issues that arose in our assessment of treatment and prevention interventions. Chapter 9 presents the committee's recommendations.

The committee considered several different conceptual schemes in considering how to organize the material reviewed in this report. We could have presented this material in terms of the life spans of victims and offenders: child, adult, and elder. We could have sought to distinguish between victims and perpetrators and between treatment, prevention, and enforcement interventions by focusing chapters on each approach. However, we decided that an approach that divides the targets of the interventions would perpetuate the fragmentation that has characterized this area of research, policy, and practice.

The committee decided that the intended audience of this report, which includes researchers, service providers, and policy makers from distinct institutional settings, would be best served by a synthesis of research that brings the

relevant knowledge to bear on specific services and policies within their own organizational service system. At this stage in the development of family violence interventions, the institutional settings of the interventions exercise a major influence on their objectives, resources, and method of implementation.

This choice of a categorical structure for the report, however, was disturbing to several committee members, who would prefer to see a more holistic approach that builds on the needs of children, adults, and the elderly. In our analysis of cross-cutting themes in Chapter 8, we highlight key issues that cut across these institutional sectors. We also highlight, in Chapter 7, examples of interventions that experiment with comprehensive community approaches and that seek to replace bureaucratic structures with service systems focused on the developmental and multiple needs of clients and communities rather than the czonvenience of the service provider. These approaches, while not yet evaluated, represent an important dimension in the design of family violence interventions that requires further examination.

In the decades ahead, research will provide more powerful insights that can lead to improved services, policy, and practice. But the integration of research knowledge and the development of interventions will require consistent and explicit attention to the ways in which the domains of science, law, health care, and social services interact. The absence or presence of interventions in one setting—such as shelters for battered women—may directly influence the operation and effectiveness of services in another domain—such as arrest policies for batterers. This interactive process is often missed in current assessments of family violence interventions, although it is often intrinsic to the experience of service providers.

The interactions among service systems add a new level of complexity to the design of both interventions and evaluations. Addressing this complexity will require collaborative partnerships between researchers and service providers to develop approaches that are consistent with the needs of the scientific process, that can be implemented in large systems of ongoing services as well as smaller innovative programs, and that can provide guidance in determining what works, for whom, and under what conditions. An integrated research foundation, the development of reliable measures and large datasets, careful descriptions of the goals of an intervention and the characteristics of clients, and documentation of the services that were actually provided all provide important opportunities for such partnerships. Through such efforts, the institutions and agencies that have supported the development of the interventions themselves can acquire the capacity to look closely at what they have achieved and to translate this knowledge into improved efforts to address the problem of family violence.