Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
190 INTEGRATING FEDERAL STATISTICS ON CHILDREN Benson, V., and M.A. Marano 1994 Current Estimates from the National Health Interview Survey. Vital Health Statis- tics, Series 10, Number 189. Hyattsville, Md.: National Center for Health Statis- tics. Edwards ,W.S., D.M. Winn, and V. Kurlantzick et al. 1994 Evaluation of National Health Interview Survey Diagnostic Reporting. Vital Health Statistics, Series 2, Number 120. National Center for Health Statistics. Fox, H.B., and M.A. McManus 1992 Medicaid Managed Care Arrangements and Their Impact on Children and Adoles- cents: A Briefing Report. Washington, D.C.: Fox Health Policy Consultants. Gaus, C., J. Scanlon, and M. Ross 1993 A New Information Framework for Health Care Reform. Presented at the Annual Meeting of National Association of Health Data Organizations, Washington, D.C., December 9. Jameson, E., and E. Wehr 1993 Drafting national health care reform legislation to protect the health interests of children. Stanford Law and Policy Review Journal 5(Fall):1. Lamberts, H., and M. Wood, eds. 1987 International Classification of Primary Care. World Health Organization of Na- tional Colleges, Academies, and Academic Associations of General Practitioners/ Family Physicians. Oxford: Oxford University Press. Leginski, W., C. Goze, F. Driggers, S. Dupman, D. Geersten, E. Kamis-Gould, M. Namerow, R. Patton, N. Wilson, and C. Worster 1989 Data Standards for Mental Health Decision Support System. DHHS Pub. No. (Adm) 89-1589, FN-10. Washington, D.C.: U.S. Department of Health and Hu- man Services. Moore, K. 1993 Children and Families: Data Needs in the Next Decade. Presented at the inter- agency Family Data Working Group meeting, Washington, D.C., May 25. National Center for Health Statistics, Centers for Disease Control and Prevention 1993 Surveys for Monitoring Health Reform. Presented at the AHSR/FHSR annual meeting, June 27-29. Newacheck, P., D.C. Hughes, J.J. Stoddard, N. Halfon 1994 Children with chronic illness and Medicaid managed care. Pediatrics 93(3):497- 500 (Commentary). Snider, S., and S. Boyce 1994 Sources of Health Insurance and Characteristics of the Uninsured: Analysis of the March 1993 Current Population Survey. EBRI Issue Brief Number 145. Washington, D.C.: Employee Benefit Research Institute. Starfield, B. 1973 Health services research: a working model. New England Journal of Medicine 289:132-136. 1992 Primary Care: Concept, Evaluation, and Policy. New York: Oxford University Press. Wood, D.L., R.B. Valdez, T. Hayashi, and A. Shen 1990a Homeless and housed families in Los Angeles: a study comparing demographic, economic and family function characteristics. American Journal of Public Health 80:1049-1053. 1990b A study comparing the health of homeless children and housed, poor children. Pediatrics 86:858-866.
MONITORING CHANGES IN HEALTH CARE 191 World Organization of National Colleges, Academies, and Academic Associations of General Practitioners/Family Physicians (WONCA) 1979 International Classification of Health Problems in Primary Care. ICHPPC-2, 2nd Ed., Oxford: Oxford University Press. Wunderlich, G., ed. 1992 Toward A National Health Care Survey. Panel on the National Health Care Sur- vey, Committee on National Statistics, National Research Council. Washington, D.C.: National Academy Press.
192 INTEGRATING FEDERAL STATISTICS ON CHILDREN Estimating the Incidence, Causes, and Consequences of Interpersonal Violence for Children and Families Colin Loftin and James A. Mercy Accurate data for estimating the incidence, causes, and consequences of violence for children and families are critical to developing effective policies and programs for prevention and control of violence. Currently, however, federal data collection activities only incidentally address key data needs, and methodology is not sufficiently consistent to provide a solid underpinning for policy and program development. In this paper, we describe the current data collection system and assess the need for im- provement. The paper is limited to data collection focused on serious assaultive violence in primary relationships, such as families, and violence involving children. As a general concept, violence is broad, including any use of force or threats of force regardless of intent or magnitude. Accordingly, it includes an argument in which threats were exchanged, spanking, and vi- cious, lethal rape. A hypothetical J-shaped distribution of the magnitude of harm from violent behavior reveals opposite problems in trying to estimate the inci- dence of violent behavior (Figure 1). On the left side of the curve, the Colin Loftin is director of the Violence Research Group, Department of Criminology and Criminal Justice, University of Maryland. James A. Mercy is with the Division of Violence Prevention, National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. 192
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 193 1/x Frequency Severity FIGURE 1 Hypothetical Relationship Between Frequency and Severity incidents are so frequent and so inconsequential that they should not, and probably cannot, be estimated. On the right side, the events are so rare that they will be missed by standard sampling schemes, yet they are so harmful that they are of major concern. Without specifying an exact cutoff, we are primarily interested in the right half of the distribution. It includes violent behavior that results in physical injury, as well as other intentional acts that pose a significant risk of injury. A wide range of extremely harmful incidents are omitted from consideration. We do not deal with negligence or failure to act that may result in injury; psychological trauma, if there is no threat of force; or self- directed violence, such as suicide or suicide attempts. We begin by describing major approaches to collecting information on serious assaultive violence and characterizing their methodology, without attempting an exhaustive description of all of the major data collection systems. We then describe what we see as some of the major problems and pressing needs. The paper ends by suggesting some feasible approaches to addressing these needs.
194 INTEGRATING FEDERAL STATISTICS ON CHILDREN EXISTING DATA COLLECTION SYSTEMS AND APPROACHES Approaches to collecting data on serious assaultive violence can be divided according to the general sampling strategies used. Household sur- veys sample dwelling units and obtain self-reports of offending and/or vic- timization from persons residing in those dwellings. Organization-based surveys sample service providers such as schools, social agencies, and emergency departments and obtain information from either self-reports or records. Com- plete enumerations or registries, such as the National Center for Health Statisticsâ vital statistics or the Federal Bureau of Investigationâs Uniform Crime Reporting system, attempt to capture all of the incidents that occur in a target population. The fourth approach combines several sampling strate- gies and is used when the incidents being studied are infrequent and no single sampling frame would yield valid estimates. The sections that follow briefly describe some examples of each approach. The characteristics of the major data sources are also described in Table 1. Household Surveys The National Crime Victimization Survey, the National Youth Survey, and the National Family Violence Surveys are based on household samples. There are, of course, major methodological differences in the these surveys, but they provide useful illustrations of this approach. The National Crime Victimization Survey The National Crime Victimization Survey (NCVS, previously known as the National Crime Panel and the National Crime Survey) is a large, con- tinuous survey of people at least 12 years old who live in U.S. residential housing units. The survey is conducted by the Bureau of the Census for the Bureau of Justice Statistics to provide national estimates of the incidence of criminal victimization (Bureau of Justice Statistics, 1992). Respondents are interviewed in person and by telephone seven times at 6-month intervals in a rotating panel design. Considerable attention is given to placing reported incidents in a specific time frame. For example, the first interview in a panel is used strictly for bounding (marking the beginning of the reference period for the second interview); thereafter, respondents are asked for the specific month of occurrence for each incident they report. NCVS data have been collected continuously since 1973 and include reports of victim- izations that may not have been reported to the police. The most distinctive feature of the NCVS is the extensive screening for crime incidents and the completion of separate incident forms that record specific information about each incident, including the exact month of oc-
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 195 currence, the victim-offender relationship, and the extent of injury and property loss. The National Youth Survey The National Youth Survey (NYS) is a longitudinal study of a national probability sample, selected in 1976, of 1,725 youths ages 11 to 17. The first interviews were conducted in 1977. By late 1993, nine waves of data had been collected (Elliott, 1994). The interview schedule contains, along with a variety of other measures, a set of self-reported delinquency items, and in some of the waves follow-up questions are designed to obtain some details about circumstances associated with the offense. Of particular inter- est are items designed to measure âserious violent offenses,â defined as aggravated assault, robbery, and rape (Elliott, 1994:3): â¢ â[Have] you attacked someone with the idea of seriously hurting or killing that person?â â¢ â[Have] you used force or strong-arm methods to get money or things from people?â â¢ â[Have] you had or tried to have sexual relations with someone against their will?â The measure of serious violent offenses is restricted to assaults, robber- ies, and rapes that involved âsome injury or a weapon.â Medically treated incidents are defined as those in which the victim was âcut or bleeding, had broken bones/jaw/nose, was unconscious, was taken to a hospital, or diedâ (Elliott, 1994:4). The survey also collects data on violent victimization of respondents. National Family Violence Surveys The National Family Violence Surveys were conducted in 1975 and 1985 (Straus et al., 1980; Gelles and Straus, 1988; Straus and Gelles, 1990). The 1975 survey was a national probability sample of 2,143 âcurrently married or cohabiting persons aged 18 through 70,â stratified by region and demographic characteristics. Response Analysis Corporation conducted the hour-long personal interviews. A random half of respondents were women and if there was more than one child, the âreferent childâ was randomly selected from the children between ages 3 and 17 who resided in the house- hold. The completion rate for the entire sample was 65 percent. The 1985 survey, conducted by Louis Harris & Associates by tele- phone, included 6,002 households consisting of two married or cohabiting adults or an adult age 18 or older who had either divorced or separated within the last two years or a single parent living with a child under age 18.
TABLE 1 Major Data Sources for Measuring Interpersonal Violence 196 Data Target Source Frequency of Violent Sponsoring Collection Population of Data Collection Behavior Organization Household Surveys National Crime Persons â¥ age 12 Mixed, personal Monthly Rape, robbery, Bureau of Justice Victimization and telephone assault Statistics (BJS) Survey (NCVS) interviews National Youth Persons ages Personal interviews 9 wave panel, Assault, robbery, Collaboration: Violence Survey (NYS) 11-17 in 1976 1977, 78, 79, 80, rape and Traumatic Stress 81, 84, 87, 89, 93 Branch, National Institute of Mental Health (NIMH) and Justice Department National Family 1975: Married and 1975: Personal 2 cross-sections, Assaults (partners Violence and Traumatic Violence Surveys cohabiting persons interviews 1975 and 1985 and children) Stress Branch, NIMH ages 18-70 1985: Adult couples, 1985: Telephone recently separated interviews persons, or single parents living in telephone households National Survey Persons â¥ age 19 Mixed, personal 2-wave panel, Hitting, shoving, Center for Population of Families and living in households interviews and 1987-1988 and throwing Research, National Households and able to be self-administered 1992-1994 things; injury in Institute for Child Health (NSFH) interviewed in questionnaire fight with partner and Human Development either English or (NICHD) Spanish INTEGRATING FEDERAL STATISTICS ON CHILDREN
National Survey of Households with at least Personal interviews Multiwave panel, Involuntary sexual Multiple funding: Children (NSC) one child (age 7-11 in 1976, 81, 87 intercourse Foundation for Child 1976) Development, Center for Population Research (NICHD) and others National Womenâs Women â¥ age 18 living Telephone interviews 3-wave panel, Assault, rape National Institute of Drug Study (NWS) in telephone households 1989, 90, 91 Abuse Youth Risk Household population Personal interviews Cross-section; Physical fights, Collaboration: Centers for Behavior ages 12-21 and self-administered follow-back of fights with injury, Disease Control and Surveillance audiocassette 1992 NHIS who fought, Prevention (CDC) and System: National questionnaire weapon carrying others Household-Based Surveys, National Health Interview Survey, Youth Risk Behavior Supplement (NHIS-YRBS) Organization-Based Surveys INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES Youth Risk Students in grades 9-12 Self-administered Cross-section; Physical fights, Center for Chronic Disease Behavior in the 50 states, the questionnaires biennial, odd-years fights with injury, Prevention and Health Surveillance District of Columbia, who fought, Promotion, CDC System: National Puerto Rico, and the weapon carrying School-Based Virgin Islands Surveys continued on next page 197
TABLE 1 Continued 198 Data Target Source Frequency of Violent Sponsoring Collection Population of Data Collection Behavior Organization Organization-Based Surveys (continued) Youth Risk Student in grades 9-12 Self-administered Cross-section; Same as above, Collaboration between Behavior in participating state or questionnaires participation varies but may be CDC and state local depts. Surveillance city with jurisdiction modified in some of health and education System: jurisdictions State and Local School-Based Surveys Monitoring the High school seniors Self-administered Annual since 1976 Assault National Institute on Drug Future questionnaires Abuse National Electronic Nonfatal injuries Medical records Monthly Intentional injuries Consumer Product Safety Injury Surveillance treated in emergency supplemented with Commission, BJS and System (NEISS) departments telephone interviews CDC with victims and witnesses National Hospital Patient visits to Medical records Began December 1991 Intentional injuries National Center for Health Ambulatory emergency and/or Statistics, CDC Medical Care outpatient departments Survey (NHAMCS) in nonfederal, short-stay hospitals INTEGRATING FEDERAL STATISTICS ON CHILDREN
Study of the Children (< age 18) Forms completed by 2 cross-sections Abuse and neglect National Center on Child National Incidence referred to community community agencies 1979-80 and 1986 Abuse and Neglect and Prevalence of professionals Child Abuse and Neglect (NIS-I and II) Complete Enumerations Uniform Crime Index crimes known to Police reports Monthly Murder b, robbery, Federal Bureau of Reporting Program: law enforcement rape, assault Investigation Return A agenciesa Uniform Crime Homicides that occurred Police reports Monthly Homicidec Federal Bureau of Reporting Program: in the U.S. Investigation Supplementary Homicide Report Uniform Crime Arrests for specified Police reports Monthly Arrests for Federal Bureau of Reporting Program: violent crimes violent crimes Investigation Age, Sex, Race, INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES and Ethnic Origin of Persons Arrested National Vital Deaths due to Death certificates Exact time of death Homicide National Center for Health Statistics Mortality intentional injuries Statistics, CDC Data continued on next page 199
TABLE 1 Continued 200 Data Target Source Frequency of Violent Sponsoring Collection Population of Data Collection Behavior Organization Combined Strategies National Incidence Children (< age 18) Multiple modes of Cross-section, 1988 Abductions by Office of Juvenile Justice Studies of Missing, data collection strangers and and Delinquency Abducted, Runaway nonfamily Prevention and Thrownaway members Children a Violent index crimes are murder and nonnegligent manslaughter, forcible rape, robbery, and aggravated assault. bIncludes nonnegligent manslaughter. cIncludes justifiable homicides as well as murders. INTEGRATING FEDERAL STATISTICS ON CHILDREN
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 201 Within the household, eligible respondents and a referent child were se- lected randomly. Black and Hispanic households were oversampled, but the entire sample is weighted to be representative of the total U.S. population. The 1985 survey had a higher completion rate (85 percent) than did the 1975 survey, and interviews were administered in less time (35 minutes) (Straus and Gelles, 1990: 529-532). Both surveys employed the conflict tactics scale (CTS), a set of ques- tions about how interpersonal conflicts are resolved that is designed to allow respondents to report violence directed at household members with less response error than would alternative approaches. The CTS is not exactly a self-report, however, because respondents provide proxy informa- tion about their behavior and the behavior of other members of the house- hold. As would be expected, the proxy responses have different character- istics than the self-reports. Other Surveys Other important household surveys on violent behaviors include the National Survey of Families and Households (Sweet et al., 1988; Brush, 1990); the National Survey of Children (Moore et al., 1989); and the Na- tional Womenâs Study (National Victim Center, 1992). Organization-Based Surveys Another strategy is to sample organizations such as schools, hospitals, police departments, and other agencies that provide services to victims and offenders. Prominent examples are school-based surveys of delinquent be- havior such as Monitoring the Future and the school-based Youth Risk Behavior Surveys and hospital-based surveys of injuries such as the Na- tional Electronic Injury Surveillance System and the National Hospital Am- bulatory Medical Care Survey (McCaig, 1994). Another important organi- zation-based study that measures violence among families and children is the Study of the National Incidence and Prevalence of Child Abuse and Neglect, also called the National Incidence Study (National Center on Child Abuse and Neglect, 1988). The main advantage of organization-based studies is that the per-case cost is lower than in household samples. This is certainly the case with school-based samples and samples of emergency departments, since one would have to screen hundreds of households to obtain a single case of assault or rape with serious physical injury.
202 INTEGRATING FEDERAL STATISTICS ON CHILDREN Monitoring the Future The Monitoring the Future studies are national multistage probability samples of senior classes in approximately 130 high schools. These studies have been conducted since about 1975 by the Survey Research Center at the University of Michigan. Among the self-report delinquency behaviors re- spondents are asked how many times during the previous 12 months they have (Osgood et al., 1989: 417): â¢ âHit an instructor or supervisorâ â¢ âTaken part in a fight where a group of your friends were against anoth- er groupâ â¢ âHurt someone badly enough to need bandages or a doctorâ â¢ âUsed a knife of gun or some other thing (like a club) to get something from a person.â National Electronic Injury Surveillance System The National Electronic Injury Surveillance System (NEISS) is designed to produce national estimates of the frequency and severity of injuries asso- ciated with specific consumer products. The system, which uses a stratified probability sample of hospital emergency departments in the United States and its territories, is conducted for the U.S. Consumer Product Safety Com- mission. Data are entered into the NEISS computer system from patient records each day; approximately 200,000 injury reports are collected through NEISS each year. Of these, about 1 percent are selected for follow-up investigation (case selection is dependent on the commissionâs product- specific priorities). Telephone interviews with victims or witnesses are then conducted to gather further information on events surrounding the in- jury incidents (U.S. Consumer Product Safety Commission, 1986). The Centers for Disease Control and Prevention and the Bureau of Justice Sta- tistics are currently exploring the use of the NEISS system to collect infor- mation on injuries due to violence. Complete Enumerations or Registries There are two major systems that attempt to completely enumerate specified classes of assaultive violence: the Federal Bureau of Investigationâs Uni- form Crime Reporting System and the homicide portion of the mortality data from the vital statistics maintained by the National Center for Health Statistics. These national enumeration systems serve two unique roles. First, they provide information on rare incidents such as homicide in specific groups (e.g., infants and spouses) (Fingerhut and Kleinman, 1989a, 1989b; Mercy
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 203 and Saltzman, 1989). Second, they provide the basis for state and local estimates. Their value is limited, however, by the fact that little informa- tion is collected and data on individuals are collected only for deaths. Also, in the case of UCR data, there has been very little research on the quality of the records, and the research that has been done suggests serious problems (Loftin, 1986; Loftin et al., 1987; Rokaw et al., 1990). Uniform Crime Reporting Program The Uniform Crime Reporting (UCR) program of the Federal Bureau of Investigation compiles information on crimes reported to or discovered by local and state police agencies in the United States. The program attempts to account, in a consistent manner, for the amount of crime known to police across the country, despite jurisdictional variation in legal definitions and practices. In concept, all crimes that come to the attention of law enforce- ment and are classifiable as UCR incidents are reported through the UCR data collection system. The main data collection instrument in the UCR is a form called âRe- turn Aâ: a police agencyâs monthly summary of recorded murders, rapes, robberies, burglaries, larcenies, auto thefts, and arsons. Other data forms submitted to the UCR program include the âSupplementary Homicide Re- portâ (SHR), the âSupplement to Return A,â âLaw Enforcement Officers Killed and Assaulted,â and âAge, Sex, Race, and Ethnic Origin of Persons Arrestedâ (Federal Bureau of Investigation, 1984). All of these except the SHR are summary reports that represent a monthly tally of various items requested by the UCR program. The SHR, in contrast, collects characteris- tics on each individual homicide incident investigated by the agency. The National Incident-Based Reporting System, which is under development, will provide incident-level information on all reportable offenses, but the system is currently implemented in only a few test jurisdictions (Poggio et al., 1985; Federal Bureau of Investigation, 1993:3). With the exception of the SHR and the systems that are under develop- ment, the only data that identify age, race, and gender of offenders or victims are the arrest data. Therefore, the UCR data are of very limited value in studying violence in specific populations, such as families and children. National Vital Statistics Mortality Data Death certificates provide the basis of the national mortality data com- piled by the Division of Vital Statistics of the National Center for Health Statistics (NCHS). Registration of births and deaths is a legal requirement in all states, and elements of the U.S. Standard Death Certificate, including
204 INTEGRATING FEDERAL STATISTICS ON CHILDREN such items as age, sex, race, place of residence, and place and cause of death, are recorded and forwarded to NCHS through state vital statistics records offices. Codes for cause of death, including those for homicide, are assigned according to definitions established by the International Classifi- cation of Diseases, Ninth Revision (World Health Organization, 1977). Combined Strategies The last type of design is actually a hybrid of approaches that draws on several sampling strategies to produce the estimates of interest. This design is motivated by the infrequency of the target incidents and the weaknesses of any single sampling frame to yield valid estimates of its incidence. The National Incidence Studies of Missing, Abducted, Runaway and Thrownaway Children (NISMART) is a good example of a combined sys- tem. NISMART collects data from six separate sources: a household sur- vey, a survey of juvenile facilities, interviews with returned runaways, po- lice records, FBI data, and a survey of community professionals (Finkelhor et al., 1992). Although the focus of NISMART is on numbers of children and not on numbers of incidents per se, the design is instructive nonethe- less. The use of multiple methods and sources of data by NISMART investi- gators was partially dictated by a wide-ranging charge to examine the âmissing children problem,â one that they concluded was a set of at least five differ- ent and distinct problems (i.e., family abductions, nonfamily abductions, runaways, thrownaways, and lost or otherwise missing children). Examin- ing even a single area, for example, nonfamily abductions, NISMART dem- onstrates the desirability of a combined approach. Although its large-scale telephone survey could generate sufficient cases to estimate the number of attempted abductions (albeit with a large standard error), it was not effec- tive in estimating completed abductions, an apparently much rarer event. For these, the NISMART police records study proved a better approach (Finkelhor et al., 1992). AREAS OF PRESSING NEED Existing data on violent behavior and its influence on children and families are fragmented, variable in quality, and do not allow for addressing many important policy issues. In this section we identify areas of the most pressing need. Obtaining Valid Responses Asking respondents to report on complex, sensitive, traumatic, and even illegal behavior committed by themselves or by members of their household
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 205 approaches the limits of what one can reasonably expect from survey meth- ods. Estimates based on different methodologies vary considerably and result in confusion and even loss of confidence in the statistical estimates of violent behavior. A couple of examples illustrate the point. A recent paper based on the National Youth Survey estimates that 36 percent of African-American males and 25 percent of non-Hispanic white males report having committed at least one aggravated assault, robbery, or rape during their 17th year, the peak year for offending (Elliott, 1994). This figure is surprisingly high and several times higher than estimates based on such sources as arrests and the reports of victims. Similarly, the National Womanâs Study estimates that, during 1990, 683,000 women age 18 or older were the victims of rape (National Victim Center, 1992: 2-3). This is more than five times the 1990 estimate based on the NCVS for people age 12 and older (Bureau of Justice Statistics, 1992: 5). Research is clearly needed in two general methodological areas: (1) eliciting candid responses about sensitive and traumatic incidents and (2) capturing the complexity of violent events. Eliciting Candid Responses Substantial research exists on enhancing the validity of responses about sensitive and illegal behaviors, such as sexual activities and drug use. Many of these existing techniques could be profitably applied to measuring vio- lence. Special problems arise, however, in household interviews in which one member of the household is reported to have abused another. Reporting violent behavior may put respondents at greater risk of victimization and thus creates serious problems for data collection. This issue is of central importance because these are exactly the situations that have high incidence rates and a substantial influence on statistical estimates. Capturing Complexity Few existing data collection mechanisms capture much of the complex- ity of violent incidents. The experience in the NCVS with series victimiza- tion is one of many examples that might be cited. Many respondents are unable to recall the details of separate incidents, even with bounded inter- views and relatively short reference periods, because there are so many similar assaults during a short period that they cannot remember them as separate incidents. These âseries incidentsâ have been characterized as incidents with long duration rather than as individual events by Biderman (1981: 795). Counting them requires special methods. Although this issue
206 INTEGRATING FEDERAL STATISTICS ON CHILDREN is an essential feature of the underlying phenomenon, NCVS is the only survey that has confronted it. Other features of violent incidents that should be captured are the se- verity of injuries and the extent of the other harmful effects of violence. This information is important not only for estimating the impact of violence on society, but also for defining incidents and comparing across data collec- tion systems. Measures of self-reported offending would profit from some of the methodology that NCVS has developed for measuring victimization. Bounding interviews, using shorter reference periods, and requiring respon- dents to provide details of specific incidents would contribute important information about the nature of violence and, other things being equal, would increase the precision of estimates. Undercoverage in Data Collection Systems on Violence Undercoverage is another problem in violence data collection systems. People who are at high risk for violence, either as offenders or as victims, are also likely to be missed in surveys of households and schools. Again, a couple of examples illustrate the point. Using evidence from police reports and an estimate of the case-fatality rate for gunshot wounds, Cook (1985) estimates that the number of nonfatal gunshot wounds is three times higher than the NCVS estimate. Although some of the difference may be due to the quality of police data, it is likely that many of the victims of gunshot wounds are not captured in household surveys. The groups that are likely to be missed in household surveysâblacks, Hispanics, and other minorities; children under the age of 10; the poor; renters; and people who move frequently (Hogan, 1993; Robinston et al., 1993)âare also those at high risk for being victims or offenders. Those who are at highest risk and who may contribute disproportionately to inci- dence estimates are the ones most likely to be missed. This problem in coverage is especially true for children and adults who are part of several households or who otherwise live in nontraditional family settings. Data on Risk Factors, Social Context, Consequences, and Sequences Existing data collection systems on violence, particularly those that focus on the more severe forms of violence, are designed primarily for estimating the incidence of violent events. They do not collect very much information on risk factors, social contexts, and other covariates of violence that might be used in evaluating causal models. Systems would be consid- erably more valuable if they measured both incidence and important covariates.
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 207 It is ironic that the data collection systems that have the best information about the most serious violent incidents (such as the vital statistics mortal- ity data and the NCVS) provide the least information about covariates, whereas those with a lot of information about covariates (such as the Na- tional Survey of Youth and the National Family Violence Surveys) provide little information about the most serious violent incidents. Data on Precursors and Long-Term Consequences of Severe Assault Violent assaults and deaths have major consequences for both primary and secondary victims (i.e., children, families, and friends of primary vic- tims), but very little is known about the magnitude and long-term pattern of those consequences. Registration systems such as vital statistics mortality data or the files of law enforcement agencies contain virtually no informa- tion about precursors or the long-term consequences of violence. Surveys provide more information, but are still quite limited. The value of existing information could be magnified severalfold by sampling cases from the registration systems and interviewing victims or next of kin to obtain infor- mation about the consequences of victimization. Topics of study would include such items as long-term medical costs, disabilities, restricted activ- ity, loss of time from work, impairment, and other circumstances surround- ing violent incidents and lifestyles. Case-control studies comparing cases in the registration systems with controls and prospective studies following a group of victims and an appropriate control group of nonvictims would be valuable. State and Local Estimates of the Incidence of Serious Assaultive Violence Many of the programs and policies designed to prevent violence or mitigate its consequences are implemented at state and local levels, and evaluating the impact of these efforts is critical. Unfortunately, few data collection systems are capable of producing small-area estimates of serious assaultive violence. Of the major data collection systems, only the vital statistics mortality data and the UCR system can produce estimates for state and local areas, and there are major limitations in the value of these sys- tems. The vital statistics mortality data provide only a limited amount of information about victims besides age, race, gender, marital status, place of residence, and cause of death; there are no data about circumstances. A revision of the U.S. Standard Death Certificate, offered for state use begin- ning in 1989, included new items on the decedentâs educational attainment and on Hispanic origin and expanded fields for indicating multiple or un-
208 INTEGRATING FEDERAL STATISTICS ON CHILDREN derlying causes of death (National Center for Health Statistics, 1993). For some states, âusual occupationâ and âkind of business or industryâ are also recorded. In addition to limited numbers of variables, confidentiality concerns affect the availability of some types of information in the public-use data. Estimates for small areas are particularly affected. Since 1982, the mortal- ity detail files have identified the decedentâs county or city of residence only if the population was 100,000 or greater. Confidentiality restrictions also prevent aggregation of daily time series after 1987 from the public microdata because exact day of death is masked. Data from the FBIâs Supplementary Homicide Report provide informa- tion about both victims and offenders for state and local areas, but there are major coverage errors (whole states are missing in some years) and content errors (Loftin, 1986; Loftin et al., 1987). STRATEGIES FOR IMPROVEMENT Coordinate Federal Focus on Data Related to Violence Chief among the strategies needed to improve data collection on vio- lence involving children and families is the need to better coordinate and integrate federal efforts to collect, analyze, and disseminate such data. There is a growing interest in the problem of violence at the federal level, and numerous departments and agencies have overlapping interests in ensuring the availability and use of high-quality data. This coordination should extend beyond intradepartmental efforts to go across departments, such as health and human services, justice, education, labor, and housing and urban development. The benefits of better coordination and integration are clear. First, pooling financial support for surveys and other data collection activities will allow more efficient use of limited resources. For example, the Na- tional Center for Injury Prevention and Control of the Centers for Disease Control and Prevention and the Bureau of Justice Statistics are currently working more or less independently to collect information on intentional injuries through the National Electronic Injury Surveillance System (de- scribed above). In addition, the National Institute for Occupational Safety and Health of the Centers for Disease Control and Prevention is interested in using these data for monitoring work-related violent injuries. This is a clear instance in which coordination and integration could improve data quality and increase efficiency in the use of federal resources. Second, more consistent estimates of the incidence and prevalence of violence and related behaviors could be produced through the use of consis- tent definitions and methodologies for data collection across federal depart-
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 209 ments. The violence field, particularly in the realm of family violence, is rife with conflicting estimates of magnitude and trends. Coordinating defi- nitions and methodologies could help improve the consistency in results across federal surveys. A mechanism is needed to achieve meaningful coordination and inte- gration. The Federal Forum on Aging Statistics, which has been instrumen- tal in this process across the National Institute on Aging, the National Cen- ter for Health Statistics, and the Bureau of the Census, is a potential model. An interagency committee or a third party (such as the Committee on Na- tional Statistics) could also be useful. Coordinated Methodological Research Program Because the methodological problems of measuring violence and its impact on families and children are interrelated and similar across data collection systems, coordinated research on basic methodology would be appropriate. A consortium of agencies or the National Science Foundation could study issues such as: â¢ the relative effectiveness of different methods of screening for and measuring the characteristics of violent behavior, especially involving chil- dren and within households, â¢ the effects of alternative collection methods that protect respondentsâ privacy in household surveys, â¢ techniques for improving coverage of persons at high risk for per- sonal violence in household surveys, and â¢ estimation procedures that incorporate information from multiple sampling frames, such as household samples and the records of service organizations. Explore More Efficient Ways of Identifying Cases Household surveys of serious assaultive violence require large, expen- sive screening operations because most of the households contacted will not have experienced a target incident. As a result, most of the expense of data collection is devoted to interviews that are not directly useful for analysis. Any procedure that increases the efficiency of screening will vastly im- prove the cost-effectiveness of surveys. Two promising approaches to reducing the costs of screening are using multipurpose screening surveys and surveys that use both household samples and samples drawn from administrative records and registration systems. With multipurpose screening, a survey designed to locate children for a study of immunization could, for example, also locate them for a study of violent behavior. The same survey could also be used to locate older re-
210 INTEGRATING FEDERAL STATISTICS ON CHILDREN spondents for a study of elder abuse and other issues concerning older citizens. Household samples and samples from administrative records are comple- mentary and used together, as in the National Incidence Studies of Missing, Abducted, Runaway, and Thrownaway Children, provide a much broader range of data than they do separately. Explore the Feasibility of Collecting Data in Key Health Care Settings Changes in the way services are delivered provide opportunities for new data sources. Two changes in the health care system have already led to more accurate and informative data collection on violence. First, the growing movement to ensure that health care providers are sensitive to evidence that their patients are being abused has led to widespread training programs for health care students and practitioners and to protocols for identifying, assessing, and referring victims of violence. This advance has the potential of greatly improving the quality and depth of data that might be collected on violence from health care providers. Second, the rising costs of health care in recent decades have spawned the development of health maintenance organizations, which, as health care reform evolves, are likely to provide primary health care to an increasingly larger proportion of the population. These organizations may be excellent sites for acquiring data on the incidence of violence, particularly family and intimate violence, in defined populations and on the health consequences of violence and its impact on child and family development. We need to explore the feasibility of developing and testing alternative methods for data collection in health maintenance organizations and other key health care settings (e.g., emergency departments, public prenatal clinics). Develop Efficient Ways for Small-Area Estimation State and local data provide a crucial underpinning for developing and evaluating local prevention policies and programs because patterns of vio- lence differ across regions and localities. For these reasons and because the great majority of violence prevention activities occur at the local level, it is important that, whenever possible, federal data collection systems allow for small-area estimation. There are several models of data collection systems that allow for esti- mating national, state, and local patterns simultaneously. For example, the Youth Risk Behavior Surveillance System has three complementary compo- nents: (1) a national school-based survey, (2) state and local surveys, and (3) a national household-based survey (Kolbe et al., 1993). These three
INTERPERSONAL VIOLENCE FOR CHILDREN AND FAMILIES 211 components provide comparable information on risk behaviors in different subpopulations of adolescents in the United States. Another example is the Fatal Accident Reporting System supported by the National Highway Traf- fic Safety Administration. This system collects detailed information on all fatal motor vehicle crashes in the United States (National Highway Traffic Safety Administration, 1993). The system also links data on these events across police records, medical examiner and coroner files, emergency medi- cal service reports, and hospital medical reports. Because this system is essentially a census, state and local data are easily obtainable. This system has provided extremely useful data for evaluating the effectiveness of state laws designed to prevent motor vehicle fatalities (e.g., safety belt laws, child safety seat laws). CONCLUSIONS Violence is a serious problem facing American families and communi- ties. Recognition of the enormous costs of violence in terms of direct medical expenses, psychological trauma, and damage to community institu- tions is growing. Given the cost of violence to communities, little has been invested in research and data collection that would provide a rational basis for prevention and control. Existing data collection systems that provide information on the incidence, patterns, and consequences of violence suffer from fragmentation, response error, undercoverage, and lack of information about important aspects of the problem. Coordination at the federal level to measure violence and its conse- quences would improve the quality, consistency, and efficiency of data col- lection efforts. Federal coordination would also encourage methodological research on such key issues as improving the validity of responses on sensi- tive issues, improving the efficiency of screening for violent incidents, cap- turing the complexity of violent events, improving coverage of persons unconventionally attached to households, and estimating the impact of vio- lence in local areas. REFERENCES Biderman, A.D. 1981 Sources of data for victimology. Journal of Criminal Law and Criminology 72: 789-817. Brush, L.D. 1990 Violent acts and injurious outcomes in married couples: Methodological issues in the National Survey of Families and Households. Gender and Society 4:56-67. Bureau of Justice Statistics 1992 Criminal Victimization in the United States, 1990. National Crime Victimization Survey Report (NCJ-134126). Washington, D.C.: U.S. Department of Justice.
212 INTEGRATING FEDERAL STATISTICS ON CHILDREN Cook, P.J. 1985 The case of the missing victims: Gunshot wounding in the National Crime Sur- vey. Journal of Quantitative Criminology 1:91-102. Elliott, D.S. 1994 Serious violent offenders: Onset, development, course, and terminationâThe American Society of Criminology 1993 Presidential Address. Criminology 32:1-21. Federal Bureau of Investigation 1993 Crime in the United States 1992. Available from the U.S. Government Printing Office. Washington, D.C.: U.S. Department of Justice. 1984 Uniform Crime Reporting Handbook. Available from the U.S. Government Print- ing Office. Washington, D.C.: U.S. Department of Justice. Fingerhut, L.A., and J.C. Kleinman 1989a Mortality among children and youth. American Journal of Public Health 79:899- 901. 1989b Trends and current status in childhood mortality: United States, 1900-85. Analyti- cal and Epidemiological Studies. Series 3, No. 26, NCHS Publication No. (PHS) 89-1410. Hyattsville, Md.: National Center for Health Statistics. Finkelhor, D., G.T. Hotaling, and A.J. Sedlak 1992 The abduction of children by strangers and nonfamily members: Estimating the incidence using multiple methods. Journal of Interpersonal Violence 7:226-243. Gelles, R.J., and M. Straus 1988 Intimate Violence. New York: Simon and Schuster. Hogan, H. 1993 The 1990 post-enumeration survey: Operations and results. Journal of the Ameri- can Statistical Association 88:1047-1060. Kolbe, L.J., L. Kann, and J.L. Collins 1993 The Youth Risk Behavior Surveillance System: Overview. Public Health Reports 108(Supp. 1):2-10. Loftin, C. 1986 The validity of robbery-murder classifications in Baltimore. Violence and Victims 1:191-204. Loftin, C., K. Kindley, S.L. Norris, and B. Wiersema 1987 An attribute approach to relationships between offenders and victims in homicide. Journal of Criminal Law and Criminology 78:259-271. Loftin, C., D. McDowall, and B. Wiersema 1992 Economic risk factors for infant homicide. Pp. 273-275 in Proceedings of the 1991 Public Health Conference on Records and Statistics. Hyattsville, Md.: Na- tional Center for Health Statistics. McCaig, L.F. 1994 National Hospital Ambulatory Medical Care Survey: 1992 Emergency Depart- ment Summary. Advanced data from Vital and Health Statistics (No. 245). Hyattsville, Md.: National Center for Health Statistics. Mercy, J.A., and L.E. Saltzman 1989 Fatal violence among spouses in the United States, 1976-1985. American Journal of Public Health 79:595-599. Moore, K.A., C.W. Nord, and J.L. Peterson 1989 Nonvoluntary sexual activity among adolescents. Family Planning Perspectives 21:110-114. National Center on Child Abuse and Neglect 1988 Study of national incidence and prevalence of child abuse and neglect, study find- ings. Washington, D.C.