The Need for Better Data
Violence against women (VAW) incorporates intimate partner violence (IPV), sexual violence by any perpetrator, and other forms of violence against women (e.g., physical violence committed by acquaintances or strangers). Available data suggest that violence against women is a substantial public health problem in the United States. Police data indicate that 3,631 females died in 1996 as the result of homicide (Federal Bureau of Investigation, 1997). Thirty percent of these women were known to have been murdered by a spouse or ex-spouse. Data on nonfatal cases of assault are less easily accessible, but recent survey data suggest that approximately 1.3 million women have been physically assaulted annually and approximately 200,000 women have been raped annually by a current or former intimate partner. Data on lifetime experiences suggest that approximately 22 million women were physically assaulted and approximately 7.8 million women were raped by a current or former intimate partner (Tjaden & Thoennes, 1998). Although these and other statistics (Bachman & Saltzman, 1995; Straus & Gelles, 1990) are sufficient to suggest the magnitude of the problem, some people believe that statistics on VAW under-represent the scale of the problem, and others believe that reports of violence against women are exaggerated.
Much of the debate about the number of women affected by violence has been clouded by the lack of consensus on the scope of the term "violence against women." As indicated by the National Research Council’s report on Understanding Violence Against Women, the term has been used to describe a wide range of acts, including murder, rape and sexual assault, physical assault, emotional abuse, battering, stalking, prostitution, genital mutilation, sexual harassment, and pornography (National Research Council, 1996). Researchers have used terms related to violence against women in different ways and have used different terms to describe the same acts. Not surprisingly, these inconsistencies have contributed to varied conclusions about the incidence and prevalence of violence against women.
The lack of consistent information about the number of women affected by violence limits our ability to respond to the problem in several ways. First, it limits our ability to gauge the magnitude of violence against women in relation to other public health problems. Second, it limits our ability to identify those groups at highest risk who might benefit from focused intervention or increased services. Third, it limits our ability to monitor changes in the incidence and prevalence of violence against women over time. This, in turn, limits our ability to monitor the effectiveness of violence prevention and intervention activities. Higher quality and more timely incidence and prevalence estimates have the potential to be of use to a wide audience, including policymakers, researchers, public health practitioners, victim advocates, service providers, and media professionals.
However, obtaining accurate and reliable estimates of the number of women affected by violence is complicated by a number of factors. There is no established and ongoing mechanism for surveillance of violence against women.* Instead, people are often forced to rely on multiple data systems to obtain minimal incidence and prevalence in-formation. This can be problematic when trying to establish incidence and prevalence estimates of VAW, because these data sources are created and maintained for purposes other than monitoring the scope of the problem. For example, police collect information about violence against women for the purpose of apprehending and bringing charges against the perpetrator(s) of the violence, and thus may record few details about the victim. Hospitals collect information primarily for providing optimal patient care and for billing purposes, and thus may record few or no details about the perpetrator of the violence, even if they recognize or record the violence at all (Council on Scientific Affairs, American Medical Association, 1992). Until routine identification and documentation of VAW become part of standard patient care, hospital records may be of limited value for public health surveillance of violence against women. These and other limitations suggest that data from multiple systems are probably needed to arrive at better estimates of the number of women who are victims of violence. However, use of multiple data systems can present logistical challenges and threats to the reliability of the data because, for some incidents, information from the victim will appear in more than one data system (e.g., both police and hospital data), whereas for other incidents victim information will appear in only one data system (e.g., the victim seeks emergency department treatment but does not file a police report).
The task of obtaining surveillance information is further complicated by the repetitive nature of some types of VAW, such as intimate partner violence. As a result, it is difficult to determine if the counts obtained reflect the number of individuals affected or the number of incidents of violence. This difficulty is compounded by the necessity of relying on multiple data sources. Police may file and treat each assault separately, even if all incidents were caused by the same perpetrator, whereas hospitals may record repeated incidents in the same patient file.
In addition to these logistical challenges, there are social barriers to obtaining accurate VAW surveillance data. These barriers include the taboo nature of the topic; the guilt and shame that inhibit self-identification by victims and perpetrators; and the lack of training, fear of repercussions, and other concerns that inhibit agency personnel from recording reports of VAW in official records, even when cases are identified. Further-more, only a small fraction of all VAW victims ever seek help from either the criminal justice or the health care system.
Recognizing the need to improve the quality of the available data about violence against women, the National Center for Injury Prevention and Control (NCIPC), Centers for Disease Control and Prevention (CDC), initiated a process to begin addressing some of the conceptual and logistical difficulties inherent in the task. To narrow the scope of the task to something more manageable, CDC decided to concentrate on developing data elements for surveillance on one subset of VAW: intimate partner violence. The process involved a consultative procedure to address some of the scientific issues related to definitions and potential data elements that might be appropriate to collect as part of
*In this document, the term “surveillance” is used in the public health sense and is defined as the ongoing and systematic collection, analysis, and interpretation of health-related data.
surveillance activities. In addition, CDC funded the state health departments in Massachusetts, Michigan, and Rhode Island to pilot test methods, using the most appropriate data sources for each state, for conducting statewide surveillance of IPV among women.
The Consultative Process
The development of Intimate Partner Violence Surveillance: Uniform Definitions and Recommended Data Elements, Version 1.0 took place through a process spanning several years:
• In 1994, CDC conducted an extensive review of the literature and developed draft deﬁnitions and possible data elements to be included in an IPV surveillance system.
• These draft documents were discussed in a February 1995 exploratory meeting with consultants experienced in the areas of violence against women and data collection and measurement, and with representatives of the three funded state surveillance pilot projects (Massachusetts, Michigan, and Rhode Island).
• The documents were revised and discussed at a March 1995 meeting of the Family and Intimate Violence Prevention Subcommittee of the DHHS Advisory Committee for Injury Prevention and Control. The subcommittee was composed of researchers, practitioners, and victim advocates with expertise in the area of violence against women.
• The documents were revised and discussed at a May 1995 meeting with representatives of the three state pilot projects.
• The documents were discussed at an October 1995 workshop open to attendees at the CDC-sponsored National Violence Prevention Conference in Des Moines, Iowa.
• The documents were discussed at a January 1996 meeting with representatives of the three state pilot projects.
• Written feedback was collected from a wide variety of external reviewers who responded to CDC draft documents.
• A March 1996 meeting was held with a 12-member panel with expertise in the areas of violence against women and public health surveillance.
At the March 1996 meeting, the panel was charged with two tasks: 1) finalizing a list of data elements that were considered essential for IPV surveillance, and 2) finalizing the definitions to be used in conjunction with the data elements to ensure consistency of meaning. During the panel discussion, however, it became evident that there were no clearly identifiable criteria or procedures for determining which data elements might be most essential. The data elements presented in this report are elements on which the panel thought it would be desirable to collect information, but for which it may or may not be possible to collect information in the context of a surveillance system. The panel also developed conceptual definitions of terms to be used in conjunction with the data elements. It became evident that these definitions might need to be operationalized (i.e., made measurable) in different ways, depending on the source of the data. Given that the pilot surveillance projects in Massachusetts, Michigan, and Rhode Island would each be relying on different data sources, two documents were developed with the understanding that further revisions would be required after the pilot testing by the state projects.
CDC staff summarized the recommendations from the March 1996 meeting, incorporated changes recommended by the panel members, and produced two documents called the "Uniform Definitions Related to Intimate Partner Violence" and the "Minimum Data Set for Intimate Partner Violence Surveillance." Drafts of these documents were sent to the expert panel, the external reviewers, and others who had expressed interest. Drafts were also distributed to attendees at the 4th World Conference on Injury Prevention and Control in May 1998.
Subsequently, to conform with CDC surveillance practice (National Center for Injury Prevention and Control, 1997; Standards Committee Health Information and Surveil-lance Systems Board, 1998), CDC staff again modified the two documents. The two documents were combined, and the expert panel and representatives from the three state projects were again given an opportunity for comment. The result is this document, Intimate Partner Violence Surveillance: Uniform Definitions and Recommended Data Elements, Version 1.0.
Purpose and Scope
Intimate Partner Violence Surveillance: Uniform Definitions and Recommended Data Elements, Version 1.0, is intended for voluntary use by individuals and organizations interested in gathering surveillance data on intimate partner violence. The document is not meant as a set of mandates, but rather is intended to promote and improve consistency of IPV surveillance. If the recommended data elements can be uniformly recorded and the data made available to numerous users, then better estimates of the incidence and prevalence of IPV can be obtained and problems such as data incompatibility and high costs of collecting, linking, and using data can be substantially reduced.
Concurrent with progress toward more uniform and accessible data, existing methods of protecting the confidentiality of data must be strengthened. Guaranteeing confidentiality will serve two purposes. First, it may protect IPV victims from potential retaliation by perpetrators. Second, it may encourage victims to report IPV.
The recommended data elements are designed to collect information of value for public health surveillance of IPV and to serve as a technical reference for automation of the surveillance data. The 50 data elements are organized into 4 major sections: Identifying Information for IPV Surveillance System; Victim Demographics; Victim’s Experience of IPV; and Most Recent Violent Episode Perpetrated by Any Intimate Partner. Data elements are numbered sequentially within each section. A structured format, modified slightly from Data Elements for Emergency Department Systems (DEEDS), Release 1.0 (National Center for Injury Prevention and Control, 1997), is used to document each data element as follows: a Description/Definition of the data element; a description of its Uses; a Discussion of conceptual or operational issues; specification of the Data Type (and maximum allowed Field Length); an indication of when data element Repetition may be necessary to include all answers that may apply; Field Values/ Coding Instructions that designate recommended coding specification and valid data entries; and, where applicable, reference to one or more Data Standards or Guidelines used to define the data element and its field values, and Other References considered in developing the data element. Data types and field lengths conform to specifications in Health Level 7 (HL7), a widely used protocol for electronic data exchange (HL7, 1996), and ASTM’s
(formerly known as the American Society for Testing and Materials) E1238-94: Standard Specification for Transferring Clinical Observations Between Independent Computer Systems (ASTM, 1996).
The Technical Notes at the end of this document provide a detailed description of data types and conventions for addressing missing, unknown, and null data, as well as recommendations for dealing with data elements that are not applicable to selected groups of individuals.
Notes on the Use of Intimate Partner Violence Surveillance: Uniform Definitions and Recommended Data Elements, Version 1.0
The "Uniform Definitions" are used throughout the "Recommended Data Elements." The definitions are likely to be valuable for a wide range of policymakers, researchers, public health practitioners, victim advocates, service providers, and media professionals seeking to clarify discussions about IPV. However, most terms in the "Uniform Definitions" are defined in only a general sense, and researchers and other users may need to further refine them. Other terms, such as "cohabitation," "dating," and "psycho-logical consequences," were not defined by the expert panel and may need to be defined in subsequent versions of the "Uniform Definitions." A particular issue needing further clarification is the identification of victim and perpetrator in episodes that appear to be mutually violent.
IPV, as specified in the "Uniform Definitions" and used throughout the "Recommended Data Elements," refers to victim/perpetrator relationships among current or former intimate partners. For ease of presentation, the words "current and former" are not always used to qualify the term intimate partner violence but are always implied when the term is used. Note that the document was written to enable data collection for both female and male IPV victims, although initial pilot tests are focused on IPV against women.
As you use the "Recommended Data Elements," keep in mind the following points:
• As with all research on violence against women, ethical and safety issues are para-mount. No data should be collected or stored that would in any way jeopardize a woman’s safety. Those interested in developing a surveillance system for IPV must be particularly conscious of the need to preserve conﬁdentiality. The issue of conﬁdentiality must be balanced with the need for data linkage across multiple data sources, perhaps through mechanisms such as encryption of unique identiﬁers.
• Currently the "Recommended Data Elements" contains 50 items. Given that simplicity is an important surveillance system attribute for obtaining high quality data (Centers for Disease Control and Prevention, 1988), and given that recommendations from the three pilot projects and other locations will allow us to distinguish those data elements that might be desirable to collect from those that are possible to collect routinely, this list may eventually be shortened. Desirable data elements that are not feasible to collect as part of a surveillance system will need to be collected in other ways.
• No single agency is likely to collect all of the data elements recommended. As a con-sequence, it is likely that anyone setting up a surveillance system will need to combine data from a number of sources (e.g., health care records and police records) using a relational database (Taylor, 1995). This will allow information on data elements to be gathered from each data source used. The mechanics of how to set up relational databases are not discussed in this document, but information from the three funded state surveillance pilot projects should provide information helpful for developing such databases. A unique identiﬁer will need to be created to allow for linkage across all data sources included. This identiﬁer may or may not be identical to the data element 1.101 Case ID.
• The goals of IPV surveillance are to obtain an estimate of the number of people who are affected by intimate partner violence and to describe the characteristics of people affected, the number and types of IPV episodes, the associated injuries, and other consequences. Counting injuries as part of a surveillance system is a common proxy for estimating the number of people affected. However, the large number of cases in which multiple forms of violence co-occur and the repetitive nature of IPV mean that such a proxy may be less accurate than is desired. To obtain more accurate estimates of the number of people affected by IPV, ultimately we will need to develop some mechanism for linking data, both within and across different data sources, through the use of unique identiﬁers.
• The recommended data elements include four discrete types of violence: physical violence, sexual violence, threat of physical or sexual violence, and psychological/ emotional abuse. However, one violent episode may contain all four types of violence. A limitation of the present version of the "Recommended Data Elements" is that it will provide a count of episodes involving speciﬁc types of violence, but it cannot provide a count of the total number of discrete violent episodes, nor can it provide information about the co-occurrence of different types of violence within all episodes. However, the IPV surveillance system will allow for collection of information about the co-occurrence of different types of violence for the most recent violent episode perpetrated by any intimate partner.
• Each data element is numbered for convenience of presentation and for easy reference. The data elements are not meant to be "administered" as a survey or a questionnaire, but instead are presented as information to be gathered from appropriate data sources in the jurisdictions conducting IPV surveillance. Thus, the elements can be gathered in any order and can be obtained from one or more data sources for any given victim of intimate partner violence. Each data element includes a code set that speciﬁes recommended coding values and instructions for what to do when the data element is not applicable for a particular victim. Obviously, the accuracy and completeness of data collected on IPV victimization depend upon what is documented by the agency providing the information.
As with the other CDC guidelines for recommended data elements (National Center for Injury Prevention and Control, 1997; Standards Committee Health Information and Surveillance Systems Board, 1998), this initial release of Intimate Partner Violence Surveillance: Uniform Definitions and Recommended Data Elements, Version 1.0 is in-tended to serve as a starting point. Many data element definitions and coding specifications are new, and field testing is necessary to evaluate them. Systematic field studies are needed to gauge the usefulness of Version 1.0 for IPV surveillance, to identify optimal methods of data collection, and to specify resource requirements for implementation. Prospective users of Version 1.0 are invited to contact CDC to discuss their plans for evaluating or using some or all of the recommended data elements. Lessons learned through field use and evaluation will be a valuable source of input for subsequent revisions, but all comments and suggestions for improving this document are welcome.
As stated earlier, CDC has funded pilot tests of these data elements in Massachusetts, Michigan, and Rhode Island as part of their exploration of surveillance methods, and as a means of assessing the feasibility and utility of collecting this information on women who are IPV victims. We hope that other jurisdictions will also be able to conduct limited pilot tests. After these pilot tests are completed, the document will be revised to incorporate what has been learned. This step will enable us to refine the definitions and reduce the number of recommended data elements to make it more feasible to collect information as part of an IPV surveillance system. Eventually, we hope to develop data elements and definitions for the surveillance of family violence other than IPV (such as child abuse and elder abuse) and other forms of violence against women.
Please send questions or suggestions for improving Intimate Partner Violence Surveil-lance: Uniform Definitions and Recommended Data Elements, Version 1.0 to:
Linda E. Saltzman, PhD
Page last modified: September 25, 2008