ABSTRACT Title of dissertation: A Comparative Analysis of Multiple Level Risk Factors Between Child Homicide and Child Abuse and Neglect. Debra L. Stanley, Doctor of Philosophy, 1995 Dissertation directed by: Charles F. Wellford Prof , essor Department of Criminal Justi d C . . ce an r1m1nology. The primary purpose of this study was to explore and compare the risk factors between two major categories of child homicide and child abuse and neglect. The two child homicide categories are intrafamilial for all children murdered by a caretaker , and extrafamilial for all other homicides involving noncaretakers. Using State of Maryland Child Fatality Review data and Baltimore city Child Abuse and Neglect data, for the period between January 1993 and June 1994, multiple level risk factors are compared . The three levels of risk include individual, family, and community factors. The first phase of the analysis found that Baltimore city and all other Maryland city child homicide data are somewhat similar when examining each level of risk. The second phase of the analysis compares risk factors between each child homicide category. The typical chi ld homicide victim was found to be a black male, with most intra- fami lial victims under 10 years of age , and most extra- familial victims between 10 and 17 years of age. The intrafamilial suspects were primarily the biological father between 26 and 48 years of age , while the typical extrafamilial suspect characteristics mirrored that of their victims. The third phase of the analysis compares both categories of child homicide and child abuse and neglect incidents . The vict ims' age , gender , and birth order position appear to differ when comparing child homicide and child abuse and neglect data. The suspect profiles appear to be similar for intrafamilial homicide and child abuse and neglect. Most victims ' are living with a single parent and have experienced prior abuse or neglect. Also , most child homicide and child abuse and neglect victims have similar community level c haracteristics . The final phase of the ana l ysis examin es the specific causes and circumstances of death and injury . Intra- familial homicide and child abuse and neglect incidents have similar characteristics with regard to causes and circumstances of death or injury. A COMPARATIVE ANALYSIS OF MULTIPLE LEVEL RISK FACTORS BETWEEN CHILD HOMICIDE AND CHILD ABUSE AND NEGLECT By Debra L . Stanley Dissertation submitted to the Faculty of the Graduate School of The Univers i ty of Maryland in partial fulfillment of the requirements for the degree of Doctor of Philosophy 1995 I' ' I , , f I Advisory conuni ttee: Professor Charles F . Wellford , Chairman/Advisor Professor Colin Loftin Associate Professor David McDowall Pro fessor Robert Hardy Chief , State Medical Examiner John Smialek , M.D. I /J )·f '. I C Copyright By Debra L. Stanley 1995 Dedication To all the childr en , past , present , a nd future ... ii Acknowledgement I wou l d like to thank my committee chair , Ch arles Wellford , for his support and guidance . Working with him has been an inva luable experience . I also thank my oth er committee members for their suggest i ons and comments , Colin Loftin , David McDowall, and Robert Hardy. I especially acknow l edge John Smi a l e k, Chief Medical Examiner, for his support and the resources necessary for pursuing thi s area of research. I wish to thank the members of the Maryland Child Fatality Review Team, for sharing their experience and expertise . I am especially indebted to Carrie Fowl er , Ann Dixon , Dan Timmel, Dan Wilson, Bob Meney, Ursu l a Cain- Jordan , and Car l a Simon. Wi thout your deep commitment the proper revi e w of chi ldhood deaths and this research , would not be possible. I a l so thank Diane Gordy , Vicky Young, and the staff at the Department of Human Resources , and the staff at the Ba l timore City Department of Socia l Services Administration , for their assistance and the use of their data. I am especially indebted to Frank Chachulski and James McLaughlin , without their s upport I would stil l be collecting the data . iii I wish to thank Mary Ann Hawkes and Nicky Rafter who have both been an inspiration as a teacher , scholar, and friend. Mary Ann ' s encouragement led me to my graduate studies and to Nicky. Nicky was instrumental in stimulating my interest in child homicide , and guiding me through the early stages of research. I am deeply indebted to April Pattavina , who has been my main source of strength throughout graduate school. The unsparing encouragement and friendship from these three people has greatly enriched my life . A Special thanks to Susan Pease and Karen Beyard, at Central Connecticut State University, for the opportunity to work with the best! Susan , thanks for your steady support . I am so grateful to my family. Their incredible s upport and l ove provide me with a steady foundation. A very special thanks mom and dad! The solid presence and friendship of my brother Eric, and my sisters , especially Janis for feeding me so well. Finally, I wish to thank all of my friends and family members for their perseverance, support, and love. I l ook forward to getting reacquainted after this long journey ... iv Table of contents Section 1. 2. 3 • 4 . 5. Introducti o n 1 .1 Statement of the Problem 1.2 Status of Current Literature .. 1.3 Child Fatality Review Teams 1.4 Purpose o f Study. 1.4.1 Objectives of Study 1.4. 2 Research Questions . 1.5 Limitations of Study. Literature Review. 2 .1 Section I . 2 . 1.1 Definitions. 2 . 1.2 Intrafamilial Child Homi c ide 2 .1. 3 Extrafamilial Child Homicide 2 . 2 Sect i on II . 2 . 2 .1 Ava ilability of Data. 2 . 2 .2 Limited Sampl ing Proc e dures 2 . 2 . 3 Wea k Methodologica l Designs 2 . 2 .4 Measurement Limitations . Summary of Literature Review Child Fatality Review Teams. Mary l a nd Child Fatality Review Team Research Methods 4 .1 Data Sources 4.1.1 Child Homicide Data 4 . 1 . 2 Child Abuse and Neglect Data 4.1 . 3 Census Data. 4.2 Methodology. 4. 2 .1 Homicide Risk Fa c t or Patterns . 4. 2 . 2 Risk Factor Differe nce s - Child and Child Abus e and Neg l ect Ris k Factor Prediction . 1 3 5 7 9 11 13 1 5 1 8 18 18 23 32 36 37 39 41 43 45 47 53 60 60 61 63 65 67 70 Homi cide 79 81 4 . 2 . 3 4.2.4 Cau ses and Circumstances of Death and Injury 84 87 87 9 1 Results Missing Data 5 .1 State and City Child Homicide Data. V 6. 5.2 5.3 5.4 5 .1.1 5 .1. 2 5 .1. 3 Individual L 1 v· eve Factors ictim Profiles Suspect Profil Famil L es · · Y evel Factors ~ommunity Level Fact~rs . ummary. · Intrafamilial ve · · · · . . Homicide rsus Extrafamilial Child 5.2.1 Individu;1 L. · evel Factors R~se~rch Question Ia. 5 . 2.2 5.2.3 Victim Profiles . Suspect Profiles . Family Level Fact~rs. Research Question Ib .. Community Level Factors. Research Question Ic. Summary . Child Homicide versus Chiid Abuse Neglect 5.3 .1 Individual Level Factors Research Question IIa. 5 . 3.2 5.3 . 3 Victim Profiles . Suspect Profiles . . Family Level Factors. Research Question IIb .. Community Level Factors. 91 91 95 99 104 104 105 .106 106 106 110 115 115 121 121 122 . 123 124 124 124 128 .132 132 139 Research Question I]§. Causes and Circumstances of Death and Injury 5.4.1 Causes of Death and Injury. Research Question IIIa. 143 144 144 144 147 150 150 150 156 159 5.4 . 2 Child Homicide ... . Child Abuse and Neglect ... . Circumstances of Death and Injury Research Question IIIb. Child Homicide. Child Abuse and Neglect. Summary. Summary and Recommendations 162 162 163 and 170 176 177 Comparison of State and City Child Homicide. Comparison of Child Homicide Categories . Comparison of Child Homicide and Child Abuse Neglect Causes and Circumstances of Death and Injury Future Research and Policy Recommendations Appendix A - Logistic Regression Analysis . References 18 2 200 vi LIST OF TABLES Number Pages 1 Chi l d Homicides by Victim's Place of Residence 62 2 Child Homicide by Category and Place of Victim's Residence 63 3 Balt imore City Chi ld Abuse and Neglect Cases 65 4 Research Design and Methodology 69 5 Distributions for Items Excluded due to Missing Data 88 6 Child Homicide Victim Profiles 92 7 Child Homicide Suspect Profiles 96 8 Child Homicide Victim Family Characteristics 101 9 Child Homicide Victims' and Siblings Prior Abuse and Neglect 103 10 Child Homicide Categories and Victim Profiles 107 11 Chi - Square Statistics - Child Homicide Type by Victim Profile Items 109 12 Child Homicide Categories and Suspect Profiles 111 13 Chi-Square Statistics - Child Homicide Type by Suspect Profile Items 115 14 Child Homicide Categories and Victim Family Characteristics 117 1 5 Child Homicide categories and Victims' and Siblings Prior Abuse and Neglect 118 16 Chi-Square Statistics - Ch ild Homicide Type by Suspect Profile Items 121 17 Child Homicide and child Abuse and Neglect - Victim Profiles 125 vii j Number Pages 18 Chi-Square Statistics - Child Homicide and Child Abuse and Neglect by Victim Profile Items 127 19 Child Homicide and Child Abuse and Neglect - Suspect Profiles 129 20 Chi - Square Statistics - Child Homicide and Child Abuse and Neglect by Suspect Profile Items 132 21 Child Homicide and Child Abuse and Neglect - Victim Family Characteristics 134 22 Child Homicide and Child Abuse and Neglect Prior Abuse and Neglect of Victims and Siblings 135 23 Chi-Square Statistics - Child Homicide and Child Abuse and Neglect by Victim's Family Profile Items 24 Child Homicide and Child Abuse and Neglect - Victim ' s Community Level Risk Factors 25 Chi - Square Statistics - Child Homicide and Child Abuse and Neglect by Community Level Risk Factors 26 27 Child Homicide Categories by Causes of Death Child Homicide Victim ' s Age by Causes of Death 28 Child Homicide Victim's Gender by Causes of Death 29 Child Abuse and Neglect Victim's Age by Causes of Death 30 Child Abuse and Neglect Victim's Gender by Causes of Death 31 32 Child Homicide Categories by the Circumstances of Death Age of Child Homicide Victim by Circumstances of Death viii 138 140 142 145 146 147 148 149 151 154 Number 33 Age of Child Abuse and Neglect Victim by Circumstances of Injury 34 Multiple Logistic Regression Results for Intrafamilial Child Homicide by Victim ' s Profile Items 35 Multiple Logistic Regression Results for Intrafamilial Child Homicide by Suspect Profile Items 36 Multiple Logistic Regress i on Results for Intrafamilial Child Homicide by Victim ' s Family Level Risk Factors 37 Multiple Logistic Regression Results for Child Abuse and Neglect and Intrafamilial Homicide by Victim ' s Family Profile Items 38 Multiple Logistic Regression Results for Child Abuse and Neglect and Extrafamilial Homicide by Victim's Family Profile Items 39 Mu l tiple Logistic Regression Results for Child Abuse and Neglect and Intrafamilial Homicide by suspect ' s Profile Items 40 Multiple Logistic Regression Results for Child Abuse and Neglect and Extrafamilial Homicide by suspect Profile Items 41 Multiple Logistic Regression Results for Pages 157 184 186 188 Child 189 Child 191 Child 192 Child 193 Child Abuse and Neglect and Intrafamilial child Homicide by Family Level Risk Factors 195 42 Mu l tiple Logistic Regression Results for Child Abuse and Neglect and Extrafamilial Child Homicide by Family Level Risk Factors 1 96 43 Multiple Logistic Regression Results for Child Abuse and Neglect and Intrafamilial child Homicide by Community Level Risk Factors 197 ix Number 44 Multiple Logistic Regression Results for Child Abuse and Neglect and Extrafamil ia l Pages Child Homicide by Communit y Level Risk Factors 198 X CHAPTER 1. INTRODUCTION 1.1 Statement of the Problem Evidence of child abuse and child homicide date back to the earliest days of civilization when ancient laws and attitudes overlooked the welfare of children. Histor icall y , violence aga inst chi l dren has always existed, however, national statistics show that the rate and seriousness have steadi ly increased throughout the 190 0s (Jason, 1 983 ; Berg, 1987; Christoffe l, 1990). Prior to the 1900s, poor medica l treatment and c hildhood infectious diseases had a major impac t on high child mortality rates. As medical t reatment improved and infectious diseases were better controlled, public attention shifted away from rapidly declining child mortality rates . However, by the mid 1950s a deceleration in the rate of decline of childhood deaths began to occur. Whil e natural death rates continue d t o decline, the unnatural or all non disease related deaths of children began to increase (Vital Statistics, 1989). In the ear l y 1900s external causes (i.e., accidents, homi c ide, suicide , and undetermined causes) accounted for less than 10 % of all childhood deaths. By mid-century, external causes accounted for over 35 % o f all childhood d e aths (Fingerhut 1 and Kleinman , 1989). Recent national health statistics (1991) show that over 64 % of all deaths of children are due to external causes , with over 18 % the result of homicide (NCHS , 1994). Today , homicide is one of the leading causes of death for children under 18 years of age. Although homicide in general is widely researched , few studies focus on child homicide . Empirical evidence supporting causal relationships found among adult homicide risk factors is not generalizable to child homicide. Both parental offenders and characteristics such as cause of death , age of child , and t i me and place of injury, suggest risk factors vary from those found in adult homicides (Kaplun and Reich , 1976 ; Christoffel et al. , 1983 ; Jason, 1983 ; Muscat , 1988; Goetting , 1990) . Risk factors occur before the incidence of violence and are associated with an increased level of risk . Knowledge about the risks and the unique circumstances surrounding child homicides is limited. Research data invariably show two major categories of child homicide. Separate categories help to explain the victim and offender relationships common among most child homicides (Jason , 1983) . The first category , intrafamilial homicide , usually involves younger children (under 10 2 years of age) who are murdered by a caretaker. The caretaker , most commonly a parent or s ubsti tute parent , is responsible f or the wel l-b e ing of the victim a t the time of death . A c hild ' s death is often the r esu lt o f , or r elated to , abusive or neglect ful behavior (Kaplun and Reich , 197 6 ; Fe in , 1979 ; Jason , 1983 ; Anderson et al ., 1983 ; Krugman , 1985 ; Ho llander , 1 986 ; Plass , 1993) In less typical intrafamilial homi cide incidents the perpetrator may be a re l ative other than a parent , an acquaintance to the family, or a hired da y care person . Researchers have speculated that ch i ld abuse and neglect may be maj o r contributing factors in the high proportion of chi l d homicides (Fein , 1979 ; Mil l er and Block , 198 2 ; Mccurdy and Daro , 1 993) . National data also s uggests that a strong link between chi ld abuse a nd neglect and chi ld h omicide exis t s . Statistics f or 45 s tates repor t ed that over 10 % of the total child abuse cases in 1991 were fatal (NCAND, 1 993). The second category of chi l d homicide , extrafamilial homicide, t ypica lly involve o l de r c hildren between the ages of 11 and 17 years . The victim and offender relationship often involves adolescent peers , or other acquaintances , and more rarely , strangers (Jason and Andereck , 1983 ) . The death of a teenager i s oft en the 3 result of criminal , illegal drug , and/or gang related activities or conflicts. Although intrafamilial child homicides have historically , occurred more frequently , extrafamilial homicides are a growing problem among adolescents involved in illegal drug or street crime activities (Fein , 1979; Miller and Block , 1982 ; Mccurdy and Daro , 1993) . Since the 1980s , national health statistics show extrafamilial homicide to be the leading cause of death for black , males , between 15 and 19 years of age (Christoffe l , 1990) In summary , violence against children has always been a societal problem, although , in recent years , the level of violence has become more serious (Berg , 1987). Of the two categories of child homicide , intrafamilial homicides have typically been more common ; however , with the rise in urban street violence , extrafamilial homicide rates are rapidly increasing . Although , the full extent of the violence problem is not clearly reflected in current child homicide research. As homicide rates for children continue to rise, empirical research remains limited. Insufficient empirical evidence restricts our knowledge to mere speculation , and limits the development of effective prevention . The current study attempts to present a detailed empirical analysis of both categories of child 4 homicide using child fatality review team data. An extensive group of var i ables is examined to validate support for specific r i sk factors of child homicide . 1.2 Status of Current Literature The literature has only recently begun to address chi ld homicide . Most of the research i s generated by the medical and psychological disciplines . Little attention i s given to the social and crimino l ogical aspects of child homicide. various definitions are used to exp l ain the unique victim and offender relationships , and the age specific victimization of chi l dren. Child h omicide research typically focuses on either intra- or extrafamilial homic ide, with the majority focusing on intrafamilial. Researchers have only specul ated that risk factor differences exist between each category of child homicide ; there is no empirical evidence to support factors that place children at greater risk for either category of homicide. Little i s known about the c ircumstances leading to a child ' s death . Most research does not attempt to de scribe the c irc ums tances or define actual or potential risk factor patterns of ch i ld homicide . There is no s upportive evidence to explain risk factor relationships. 5 The underlying assumption is that any one of these factors can increase a child ' s fatality risk. Most child homicide studies rely on case study descriptions of basic offender patterns. Of mor e than 50 published research studies on child homicide , all but seven are strictly descriptive case studies . Qualitative research cannot s upport any specific hypothe ses by estab l ishing association or making causal inferences about selected variables. Prior research ha s help e d establish a strong foundati on for studying child homicide . However, there are specif ic research questions that need addressing that prior research has ignored. Because of the prior qualitative research, we have a better understanding of what direction to take more scientific approaches for future research . Future research questions should address specific individual, family , and environmental fa ctors to learn about which high risk predictors are mo s t preva l e nt to child h omicides . Questions should also address the specific circumstances unique to child homicide incidents , and establish causal relationships between risk factor variables . Most child homicide studies that employ any form of scientific procedure appear to suffer from methodol og i ca l limitations . Some methodological limitations include , the 6 inability to define and separate out intra - versus extrafamilial homicide incidents in a data sample. Inadequate data sources , poor sampling procedures , and weak research designs limit the quality of research. As a result , most current child homicide research is limited in i t ' s empirical analyses and methods of design . (A complete review of the literature and a discussion of the methodological limitations is discussed in chapter 2). 1 . 3 Child Fatali ty Review Teams This study uses Child Fatality Review Team data that includes multiple agency data sources . The comprehensive data collection procedures used by a child fatality review team greatly enhances the quality of data. The review team data facilitates the exploration of causal relationships among risk factors common to child homicide incidents. Multiple agency Child Fata l ity Review Teams have slowly emerged in response to the increasing awareness of severe violence against children in the United States. Since 1978 , when the first team originated in Los Angeles, California , over 44 state level Child Fatality Review Teams have been established across the nation (Durfee et al . , 1992). Team membership is dependent on multiple 7 agency participation. Team members from various public institutions - social, medical, legal , forensic, educational, and research agencies, have joined together to establish Child Fatality Review Teams. The initial mission of most review teams has been to develop open communication among public agencies concerned with the welfare of children. If agencies share case information, deaths may be avoided or at least more accurately classified. Moreover, opening communication lines among various agencies facilitates identifying and protecting the deceased child ' s siblings and other family members, who may be at risk. The development of Child Fatality Review Teams, is also in response to the vast problems with intra- and interagency recordkeeping and inadequate databases. A long term goal of Child Fatality Review Teams is to develop a database that would provide more accurate information to all agencies involved in the welfare of children. A central database that provides sophisticated knowledge on the patterns and trends from past incidents would be developed. The data could then be used to evaluate and identify problem areas in a community , families at risk, and to help prevent the misclassification of deaths. More accurate 8 classifications of childhood deaths may enhance the identification of intentional injuries and neglectful harm to children , which may prevent future deaths, injuries and neglect. Multipl e agency data collected by the Child Fatality Review Teams will improve future child homicide data sources , and as a result , improve child homicide research, and prevention. An ultimate goal of review teams is to develop intervention strategi es and prevention measures for children at risk. The State of Maryland established a review team and began reviewing childhood deaths in 19 91. However , the Maryland team has just begun developing a database with multiple level data. (A complete discussion of the Mary land Child Fatality Review Team is presented in c hapter 3). 1.4 Purpose of Study The purpose of this study is to explore and compare the relationship of risk factors between child homicide and child abuse and neglect. The association of multipl e l eve l risk factors , such as individual, family, and socio- cultural , need to be compared between child homicide and child abuse a nd neglect (Biller and Solomon , 1986) . Thi s study explores multipl e levels of ris k factors, which 9 include individual , family , and community level factors. (Specific risk factors are discussed in chapter 4 - Research Design). This study examines empirical ev idence of risk factor relationships associated with each category of child homicide . Risk factors most prevalent in chi ld homicide are compared with risk factors common to child abuse and neglect . Also, risk factors are structured into multidimensional models to establish any causal implications for both child homicide categories and child abuse and neglect. This study departs from other child homicide studies in several ways. First, this study uses a comparison group of child abuse and neglect data. The comparison group is a representative sample of typical victims of abuse or neglect residing in the City of Baltimore , Maryland. None of the previous child homicide research has employed a child abuse and neglect comparison group. A second departure of this study is that it employs the total population of child homicide victims in the State of Maryland between January 1 993 and June 1994 . All of the child homicide cases are reviewed and/or autopsied by the State of Maryland Medical Examiner's office. Most prior research employ small samples of child homicide cases that usually target a particular age group . 10 Finally , this study improve s upon previous child homicide studies by employing multiple agency child homicide data. Multip l e agency Child Fatality Review Team data incorporates all forensic , medical , l ega l, social , and education data re l ating to a specific homicide. Child Fatality Review Team members review all homicides to ensure a full invest igation and that all details are documented. No existing research has ana l yzed child homicide incidents using Child Fatality Review Team data. 1.4.1 Objectives of study First - Thi s study identifies differences in risk factor patterns between two categories of child homicide, intrafamilial and extrafamilial homicides. Risk factors are categorized into three broad categories , 1) Individual factors: age , race, gender (victim and offender profiles) , victims ' relationship with offenders, vict im and offender drug use, and birth order position of victim; 2) Victims ' Family factors - marita l status of parents , socioeconomic status , family size (number of chi l dren) , and prior history of abuse or neg l ect of victim and their sib lings , and 3) Community Factors - place of victims ' residence , economi c status of community , popu l ation under 18 years of 11 age , racial distribution of corrununity , and population living in single parent households. Second - Using the same risk factors li sted above, this study identifies differences in risk factor patterns between child homicide and child abuse and neglect incidents. The major assumption is that certain patterns of risk factors will be more prevalent in child homicide incidents rather than in chi ld abuse and neglect. Third - based on descriptive statistics for risk factors, further examination of the data is performed to show and explain the association of statistical l y significant risk factors for child homicide. To determine whether any causal relationships exist among risk factors, multidimensional prediction models are used . Fourth - This study enhances our knowledge about the circumstances leading to child homicide and child abuse and neglect. Facts about specific causes of death and injury (i.e., gunshot wound versus suffocation; and use of blunt instrument versu s malnourishment), and details of the circumstances of death are compared with the circumstances of chi ld abuse and neglect incidents. Variables compared include - time and place of death or injury, drug related injury, type of weapon used , and number of victims . This study also classifies child 12 homicide data by cause of death , and child abuse and neglect data by type of injury. Fifth - based on the findings, this study develops valid conclusions for both theoretica l and policy building strategies . By identifying statistically significant risk factors of child homicide a framework is developed , in the way of prevention strategies and much needed public policy development. 1 .4 .2 Research questions 1. Are there different risk factor patterns between intrafamilial and extrafamilial child homicides? a . What specific individual level risk factors will be different? i.e . , age , gender , race , (victim and offender profiles , victim and offender relationship , birth order of victim, and offender drug use . b. What specific family level risk factors will be different? i . e ., marital status , socioeconomic status, family size , and prior history of abuse. c . What specific community leve l risk factors will be different? i.e ., place of victims ' residence , percent of the population living in single parent households , percent of the popu l ation under 18 13 years of age, percent of the population who are nonwhite , percent of the population who are living below poverty level. 2 . Are there different risk factor patterns between child abuse and neglect and chi ld homicide incidents? a. Are individual level risk factors different between child abuse and neglect and child homicide? b. Are family level risk factors different between child abuse and neglect and child homicide? c. Are community level risk factors different between child abuse and neglect and child homicide? 3. What are the significant risk factors for predicting child homicide and child abuse and neglect? 4. What are the unique causes and circumstances of death or types of injuries? a. Type of injury, i.e., gunshot wounds, beatings , fire , strangulation, stab wounds, and malnourishment. b. Circumstances , i.e ., time, place, drug related , type of weapon used, and number of victims. In summary , the purpose of this study is to explore and explain the relationship of risk factors between child 14 homicide and child abu se and neglect. A risk focused, approach seeks to eliminate or reduce the effects of identified precursors of violence. The risk focus of this study is a prospective approach that provides knowledge necessary for preventing future violence. 1.5 Limitations of Study The initial intention of this study was to employ a comparison group consisting of State level child abuse and neglect data. However, the State of Maryland, Department of Social Services , currently does not have a statewide computer network system for maintaining c lient records. The only county or city within the Sta t e of Maryland operating on a Child Protective Services computer network system is the City of Baltimore. The other twenty-three (23) counties maintain traditional recordkeeping procedures that require using manual file numbering systems. Every county operates independently from the other counties , there is no standard record filing or numbering system. Al~o, all county level records are stored at multiple satellite offices throughout each county. Files are in the custodial care of the social worker assigned to the case. Therefore, it was not possible to compile a statewide data sample that would 15 identify each social worker , office locat i on , and county . For these r easons , it wa s not possible to develop a statewide accountabili t y of all child abuse a nd neglect cases by county file numbers. As a result, a statewide compar i son group could not be constructed for this study . The comparison group f or this study includes 210 chi l d abuse and neglect cases randonly selected from Balt i more City , Child Protective Service (CPS) records. Thi s study includes the f o llowing chapters : Chapter 2 - Revi ew of Previous Research This chapter discus ses t ie definitions and historical context of child homicide inc idents. A review of the studies that focus on child h omi cide is i ncluded in this chapter. A s ummary of the me : hodo l ogical limi tations of the current literature is also discussed. Ch apter 3 - Child Fatalit y Revi ew Teams A thorough hi s tor i cal and methodologi ca l explanat i on o f Child Fatality Review Teams es tab lished across the country in over 44 states is discussed. Specific referen ce is made to the State of Maryland Chi ld Fatal ity Review Team hi story, goal s , a nd data collect i o n and case review procedures . 1 6 Chapter 4 - Research Design A detailed discussion of the research plan and the methods used to analyze the data is provided. Included in this chapter is a discussion about the three data sources, 1) child homicide data, 2) child abuse and neglect data, 3) census data. A discussion detailing the variables and statistical procedures used to measure the variables is also included in this chapter. The dependent variables include a binary variable for both types of child homicide , intrafamilial and extrafamilial , and a variable for child abuse and neglect. Also, the development of multiple logistic regression models is explained in this chapter. Chapter 5 - Results The results of the study are presented and described in relation to each research question addressed in this study. Chapter 6 - summary and recommendations A discussion of the findings in relation to existing theories and policies are included in this chapter. Also a brief summary of the study, and recommendations for future research and policy are discussed. 17 CHAPTER 2 . LITERATURE REVIEW This chapter reviews current child homicide literature in a historical context and discusses methodological limitations. The literature review is divided into two sections. The first section has three subsect i ons , 1) defini tions u sed in describing child homicide ; 2) historical and background literature on intrafamilial homicides ; and 3) historical and background literature on extrafamilial homicides. The second section addresses methodological issues relevant to current child homicide literature . Based on methodological limitations of the current research, the second section has four s ubsections, 1) avai l ability of data sources ; 2) sampling procedures ; 3) research design ; and 4) measurement issues . 2.1 Section I 2.1.1 Definitions The term " child homicide " includes a ll intentional or neglect provoking deaths of children l ess than 18 years of age caused by another person . Throughout history, many terms have been used to define or describe chi ld homicide and their unique victim- offender relationships. Definitions have been used to identify specific 18 circumstances, explain age specific incidents, or to describe the unique victim and offender relationships. A review of the most widely used terms is discussed below. One of the first terms used in research to describe child homicides was "infanticide". The term infanticide, was used to describe the death of an infant; a chi ld killed in the first year of life. The act of infanticide has often been a solution to poor economic conditions , poverty, and/or starvation. It was an accepted practice in many cultures and societies around the world before the 1800s. Some of the earliest recorded incidents of infanticide date back to the 7th Century B . C. when the Chinese experienced extreme impoverishment (Breiner , 1990). In most early civilizations , infanticide was not considered a crime (Kukull, 1977; Parker and Good, 1981). While intentional death was quite common for infants and not treated as a crime, the murder of an older child was a serious crime and punishable by death of the offender . Early research indicates there were few recorded incidents of child killings outside the category of infanticide (Breiner, 1990). The research on infant i cide dominates the literature on chi ld homicide right up to the 1960s. A second common term linked with the c hi ld homicide literature is "neonaticide". Neonaticide is the 19 in tention~! de a th of a child l ess than 24 hours after birth (M yers , 1970) . Most neonaticides h ave been committed by the birth mother and have not been considered a cr ime . The p r actice of neon aticide has been an accepted form of i n fan t i c ide and was not considered a crime until the 20th c entury . Historically , neonaticide was used as a means t o contro l the population ; a pract ice widely supported in countries l i ke Greece and China (Breiner , 1 990) . Recent incidents of neonaticides have been the r esult of p oor e conomic and social circumstances (Resnick , 1970 ; Chri stoffe l e t al. , 1983 ). Re search is limited in this area o f c hild homicide becau se most data sources do not separate out infant victims less t h an 24 hours of age . For exam~l e , the Fe d e ral Bureau o f I nvestigation (FBI) , Uni form Crime Reports (UCR) only di s tingui s h infan ts b y l ess t ha n a wee k , or more than a week but less than one year (FBI-UCR , 1992 ). A tr: ird term common t o the child homi c ide l i t e ra t u re is " filic ide ". The term filicide is used to c larify the victim ard offend e r r e l a t ionships common t o c h ild h omi c ide incident ~. Fi licide i s a t erm u sed to describe t h e homicide o f c hildren by a parent (Re sni c k , 1969; Mye rs , 197 o) _ r.rhe r e are a t l e as t five cate gories o f f il i c ide that a tt empt to expl a in p a r e n ta l mur d e ring of c h i ldre n (at 2 0 least 24 hours of age). The categories are 1) Altruistic Filicide - to relieve a child from the suffering of an illness; 2) Acutely Psychotic Filicide - a result of a parent who is under the influence of hallucinogen i cs or delirious; or experiences epileptic or schizophrenic psychosis, 3 ) Unwanted Filicide - illegitimate children who are no longer wanted; 4) Accidenta l Filicide - a result of child abuse or neglect; and 5) Spousal Revenge Filicide - this i s a deliberate attempt to make the other spouse s uffer (Resnick, 1969). Parental homicide o f chi ldren is the most widely studied v i ct im and offender relationship in both historic and current literature . Additional terms have been used to describe the ci rcumstances s urrounding the cause of a child's death. Terms such as " subtle fatal child abuse" describe abuse or neglect that h as left no anatomical evidence indica tive of the true cause o f death (Zumwalt and Hirsch, 1980; Christoffe l et al., 1981; Krugman and Peterson , 1985) Fatal child abuse describes physical or sexual assault; and fatal child neglect describes a lack of proper care or s upervision that may cause the death of a c hild (Nixon et al., 1 981). current literature has shifted away from using s u c h terms as infanticide, n eonaticide, and f ilicide . As 2 1 discussed earlier , there are generally two categories or terms u sed to describe child h omi cide incidents today . The first , intrafamilial child homic i de, is usually the result of c hild abuse or neglect and is perpetrated by t he victims' caretaker . Victims are often less t han 10 years Of age; howe ver , the maj ority tend to be les s than five years of age (Christoffel , 1990) . The second category, extrafamil ia l c hild homi c ide , i s generally associated with street crime o r illegal drug activities and is most common among adolescent peers. Victims and offenders are most often o lder than 10 years of age, with the majority between 14 and 17 years (Jason, 198 3) . Some of the literature uses the term "Gang Homicides " to refe r to extrafamilia l child homicides (Curry and Spergel , 1 988 ) A third type of child homicide are those that involve strangers who murder children. Homicides i nvolving s trangers are the most publicized by t h e media , a nd the most fear e d by parents . Howe ver , child homicides involving strangers are t ypically very rare incidents. Therefore , this study will not examine the stranger c hild homicide category as a separate category, stranger homicides are included within the extrafamilial child h omi c ide category . Also included in the extrafamilial h omicide ca t egory 22 :I I are " justifiable homicides " by the po l ice. With the increase in adolescent street crime and i l legal drug activities there is also an increase in po l ice s hoot i ngs involvin g adolescents . J u s ti f i abl e homici des generally involve the victim fleeing from the scene of a crime when shot by t h e po l ice . The v i c tim is cons i dered a s u spect of criminal activity which ini ti a tes t he po l ice pursu it. Because the victim- offender re l ationsh ip does not involve a caretaker , such i n cidents are p l aced i n t h e category for extrafamilial chi ld homi cide , and will be treated t he same as other extrafamilial incidents. 2 . 1 . 2 Intrafamilial Homicide Research Whil e studying the severity of child abuse and neglect in the 1950s , Dr . Henry C. Kempe (a pediatrici an) and his associates comp l eted one of the first studies that recognized child homicide as a major issue . The authors discovered hidden pathological information relating injury incidents closer to child abuse and negl ect. They found that a large portion of severe child abuse and neglect incidents l ead to chi ld fatalities. This wa s the first public acknowledgement that fatal child abuse and neglect are a growing problem. 23 Kempe and h is colleagues (19 62) identified several major factors evident in most fatal chi ld abuse or n eglect cases. For example , they found that most victims are less than three years of age , and most perpetrators are the caretakers - parent or guardian , of the victim. Also, mos t injuries are internal and difficult t o detect during an external anatomical examination. They also discove r ed t h at fatal abuse and n eglect are not restricted to the lower classes. Kempe.and his associates were surprised to discover that fa tal child abuse and neglect are widespread throughout the general population . Kemp e and his associates a l so found a h igh rate of fa iled marriages among child abusing familie s . Although the stu dy did not provide empirical evidence or support for the detection of potential victims , their work i s important for practical purposes. Kempe and his col leagues are responsible for modeling the c hild abuse law that h as graduall y been adopted by e v e ry state in the nation 1 • Their work a l so affects medical reporting strategies and proposes a medical model for treatment of child abusers. The resu lts of their work capture the attention of Ameri cans and influence further investigation and interest in c hild 1prior to 19 64 , the r e were no effective child abuse reporting laws . 24 I I I I I I ' homicide research . What followed was the p ub l ication of over forty-five qua l i t ative c h i l d homicide studies generated by the medical sciences . These stu d i es foc u s on the medical detection and appr opr i ate identification of speci fic injuries , and a l so forensic and pathological indi cat i ons for classifying a death . The medi cal studies lead t h e way in building a foundation for child homi cide research. Medical researchers speculate about the relat i o n shi p between chi ld homicide and child abuse and neglect , characteristics of v ict i ms and offenders , t ypes of fatal injuries , and environmental factors (Myers , 1970; Chris t offel et a l., 1 981; Blaser , 1983 ; Jason , 1984; Paulson and Rushforth , 1986 ; Zumwa l t and Hirsch , 1987; Winpisinge r , et al. , 1991). Although it seems plausible that most intrafamilial homicides are linke d with abusive families , there are no empiricall y tested re l ationships of the specific variables linked with child abuse and neglect. Lester Adelson ' s 1961 stu dy is one of the first to explore child homicide data. He too discovers that hidden e lements of fata l child abuse go unnoticed wi thout the assistance of an autopsy. Ade l son emphasizes how long term abuse may not be readily apparent until the point of an autops y. He a l so stresses h ow a young c hild is unable 25 f1 IQ p I to reveal the details of the incident(s) leading to their abuse and ultimate death. This study, as well as other early studies have led to national policy changes requiring full autopsy investigations for all deaths involving children. The requirement of medi cal examiner investigations in the death of a l l children has enhanced the quality of knowledge, upgraded the classification of homicides , and increased the quality and availability of data. Adelson (1961) also emphasizes that the c ircumstances , motive , and intent of death involving children vary from adult homicide incidents. Most adult homicide literature is not gene ralizable to children , also, most homicide research does not include child vict ims. For example, a major study involving race , socioeconomi c status and homicide only analyzes cases involving victims 16 years and older (Centerwall , 198 4). Certain child homicide sociodemographics (Straus , 1987 ), environmental factors (Adelson , 1961 ), geographic distributions and trends (Goetting, 1990), may be differe nt from adult homicides. Some studies make assumptions about risk factors that appear to be common among child homicide victims . For example , distinctive social profiles of child homicide appear to b e dependent on the age of the child . Age is a 26 factor that appears to be close l y associated with intrafamilial homic i des. For example , the younger the chi ld the more likely the perpetrator will be a parent . Mu ch of the research shows that chi l dren l ess than fo u r years of age are overr epresented among intrafamili a l homicide victims a • I (Zumwalt and Hirsch, 1980 ; Jason et 1 1983; Jason and Andereck , 1983; Krugman , 1 985 ; Schloesser et al ., 1992 ; Plass , 1993). Perhaps because younger children are physical l y more vulnerable and social l y iso lated, they are at greater risk of intrafamilial homicide. Biller and Solomon (19 86) suggest that t he younger a child the l ess likely pre vious abuse or negl ect will be reported or that the victim will retaliate. Most s tudies focus on a particular age group when exami ning intrafamilial child homi c ides . For instance , a study using Australian death records sampled only victims less than five years of age and omitted all of the o lder d eceased children (Nixon , et al. , 1981) · Althou gh thi s study provided useful and adequate information regarding the younger c h ildre n , it did not provide any information r e garding o lder c hild homicid e victims . Fe w studies l ook at al l age groups a nd·make comparison s regarding the c irc ums tances leading to d eath. In fact , mos t studies vary wide ly when sampling on age ; s u gges ting there are 2 7 ' major differences between age groups and risk of homicide. Limiting data by age group may bias the results and prohi bit making genera l izations . Th e gender of the victim only seems to be related to extrafamilial child homi c ide inc idents. Most of t he research suggests that males are the most vulnerab le victim, as well as t h e most common perpetrator in extrafamilial homicide (Curry and Spergel , 198 8 , Goetting , 1990) . However , gender does not appear to be a factor in intrafamilial child homicide incidents . Most res earch s u ggests that both male and female victims are at t h e same level of risk (Goetting , 19 89 , Plass , 1993 ) . Anothe r factor identified i n recent research focuses on the birth order position of the victim . Some studies report that the victim is most likely to be the o n l y child in the family (Smith , 1989), especially when examining ext r a famili a l child homi c ide vic tims . There are discrepanci e s in t h e research in that oth er studies support that vict im ' s are more typically the child of a l arge family (Curry and Spergel , 1988). The intrafamilial homicide r esearch s uggests t h at the victim is most ofte n the last born child in the fami ly (Schloesser et a l., 1992 ) . These discrepancies in birth order position of t h e victim warrant further r e search . 28 Winpisinger and her colleagues (1991) examined famil y factor s that link mothers' characteristics with child homicide. The risk factors examined were race, marital status and education of the mother at the time of victim' s birth , mo t her' s age , and birth defects. The authors were interested in the association between the birth of a child to a n unwed mother and the risk of being killed. Although they do not examine causal relationships the data show a strong association between the birth of a child to an unwed mother and the risk of homicide. Most intrafami lial child homicide r esearch is directed toward victims ' mothers, as both the primary caretaker and the offender . The victim and offender relationship may be a key factor in determining the level of risk of a potential victim . Historically, children have been at the greates t risk in their own home, especially when the child is less than one year of age . Only one child homicide study examines male caretakers as the perpetrator. The study examines fathers or substitute fathers who are the offenders and specifically se l ects only thos e cases involving ma l e offe nders (Scot t, 197 3 ) . Howe ver , the 29 cases examined cannot be generalized to al l child h omicides . Male parent/guardian offende r s are a major group overlooked by ch ild homicide research . Cu rrent 29 I I .J research can only s pec ulate that mothers are the most common offenders because research only examines female offenders (Myers , 1967 ; D'Orban, 1979; Weisheit , 1986 , Silverman and Kennedy, 1988) . Substitute fathers are often targeted as offenders in child abuse and neglect cases (Arrunerman and Hersen , 1990) , suggesting that male caretakers may be a factor in child homicide research. Several studies address social structural factors and their relationship to intrafami l ial child homicide. The current research has shown a strong relationship between child homic i de and measures of poverty (Boone , 1982; Jason et al. , 198 3 ; McDowall , 1986). Although child homicides occur at all levels of the socioeconomic spectrum (Kempe et al., 19 62 ), research findings suggest that they are mor e like ly to invo lve l owe r socioeconomic families . Seve ral studies show that the rates of child h omicide occur more frequently in areas characterized by poverty, racia l minori t i es , a nd u r b aniz ation (Abel , 1986; Christoffel et al. , 1983). These studies suggest that certain social structural factors increa se the ris k of child h omicide and t hat the rates of child homicide are not randomly distributed. Fi a l a a n d LaFree (1 988 ) examine social structural f actors i n a cross - n a tiona l study of child h omicide using 30 1974 World Health Organization (WHO) data. The social structural factors examined were economi c stress , social disorganization, culture of violence, and social isolation. Comparing 22 l ess developed nations with 18 more developed nations , they found that none of the social structural factors predicted child homicide in less developed nations . However, low levels of government spending on social programs , high proportions of women in the work force , and low proportions of women in college and professional occupations were associated with high child homicide rates in the more d e veloped nations. To further support these findings , Gartner (1990) expanded the WHO data to include 1 9 years (1965- 1984) of data and found similar patterns. Gartner points out that welfare spending and females in the labor force link child homicide to economic stress , social i solation, and lack of socia l support . Gartner (1991) expands the research further whe n she examines family characteristics, welfare spending , and child homicide. In this study, family characteristics are d efin e d as one of the following : proportion of births to unmarried mothers , proportion of births to teenage mothe rs, number of children under five per 100 wome n aged 15-45 , and t h e crude divorce rate . Using 1 5 years of WHO 31 ' I' data analyzed in five year intervals, Gartner aggregated measures of family structure conditioned by the level of government spending on social programs (i.e. social security expenditures). She found statistical significance for each of the family characteristics. study shows that where government spending on social This programs is low, chi ld homicide rates increase based on the prevalence of mothers who are single , teens , divorced, employed or who have many young children . Cross-national studies have been successful at measuring ecological relationships with child homicide. What these studies suggest is that risk factors should be measured at the individual level to find out if they will show similar patterns. They clearly express the need for detailed victim and offender data for use in future child homicide research. 2.1.3 Extrafamilial Homicide Research Historically, extrafamilial child homicides have been rare occurrenc es in most societies. In the Unite d States e vide nce of a major extrafamilial homicide probl e m e rupt e d in the 1980s. Prior to the 1980s most victims were the result of intrafamilial child homicide. Since the 1980s ' the rates of t een d eaths , or gang homicides r e l a ting t o 32 4 s treet crime and illega l drug activities, continue to be a growing phenomenon in this country. Today there i s a new level of ri sk for ado lescent children out in the streets. However , e xtrafami lia l child homicide research is stil l in the preliminary stages of development. The bulk of the literature focuses o n the offender rather than the victim . Wolfgang and Fen:·acuti ' s " subculture of violence " theory (19 67) attempt s to explain s ubcultural values and norms , and enviro nme ntal factors . They identified a subculture of v i o l enc2 in certain areas of the nation that support n orms separat2 from those of the dominant parent culture . Th e authors tested this theory while examining homicide pattern s in Philadelphia. They found that c hildren exposed to a v iolent subculture come to accept violence as a norma l response to interpersonal and soci al conflicts 2 • Some critics suggest that the levels of violence accepted within a culture are reflected in the levels of violence directed at ~hildren. The violent subculture in 2 " Lower c l ass boys , for example , appear more likely to be orient e d t o ward di r e ct expression of aggres s ion than are middle c l as s b oys . Th2 type of punishment meted out by parents t o mi sbeh aving children is related t o thi s c lass orientation t oward aggression. Lower-class mothers report that they or the i r husbands are likely to strike the ir children o r threaten them ... " (Wolfgang and Fe rrac uti , 1967, p. 1 54) . 33 which the child is raised increases their risk for future victimization (Wolfgang and Ferracuti, 1967). The theory suggests that violent teens will attack others similar in age, race, and economic status who share their cultural values. Erlanger (1979) found some correlation between family social class and physical child discipline, but added that it was not strong enough to be of great theoretical significance. current research has not examined the relationship between sociocultural factors and extrafamilial child homicide in comparison with child abuse and neglect (Biller and Solomon, 1986). Wolfgang and Ferracuti's research (1967) found that violence rates are highest in urban areas that support teenage gangs whose members engage in violence. However, there is no extrafamilial homicide research that addresses whether homicides occur in neighborhoods with high rates of reported child abuse and neglect. However , based on the subculture of violence theory multiple levels of risk factors need to be examined when studying child homicide. Although there is some disagreement (Hirschi, 1969; Erlanger, 1979), it seems plausible that there is some association with the circumstances of extrafamilial child homicides and the subculture of violence theory. Wheth er abuse and neglect are direct or indirect factors 34 associated with extrafami lial child homicides are an issue never addressed in research. Child abuse as a social indicator for the quality of life of families in a particular neighborhood is a matter of interest both theoretically and practically (Garbarino, 1981). Three common elements of extrafamilial child homicides are that victims and offenders are of similar age, gender , and race. Much of the research shows that black, male, urban, adolescents are overrepresented among drug related street murders (Sorenson et al., 199 3 ; Plass, 1993). Victims and offenders are most often acquaintances through their illegal activities (Goetting, 1989; Harries, 1 993). Studies measuring race consistently show a disproportionately higher number of minority homicide victims (Silverman et al., 1990; Plass, 1993; Sorenson et al., 1993). The National Child Mortality statistics suggests that homicide is a leading cause of death for black children between 15 and 19 years (Fingerhut and Kleinman, 1989). The only empirical analysis on race is a cross-national study that compares two jurisdictions (Chicago and Ontario) for patterns of racial differences in murdered children (Silverman et al., 1990). Their findings are the same as those from earlier studies , 35 blacks died at a much higher rate (almost nine times) that for whites in Chicago. Older black children were most often killed by a nonfamily member. These deaths appear to be closely linked with gang activity and perhaps drugs, however, neither of these elements are controlled for in this study. This macro-level study could not respond to specific family factors , social inequality or the general effects of a subculture of vio l ence. Individual level analyses have not examined the circumstances that surround child homicide incidents (Silverman et al., 1990). Research has not included the examination of family factors, sociodemographic or environmental factors and their connection to victim and offender relationships. A category of extrafamilial child homicide overlooked by researchers involves sibling homicides. With the growing number of adolescents involved in homicide, there is concern that sibling homicides are increasing in rate (Wilson, 1993). currently there are only two exploratory studies that address this issue (Rowley et al ., 1987). 2 . 2 Section II Methodological Issues Although several ·attempts have been made to identify relevant risk factors for child homicide victims, most have not been very successful. Most child homicide 36 studies rely on case study descriptions of basic victimization patterns. Out of more than 53 published academic studies on child homicide all but seven are strictly qualitative case studies. Although some current studies employ more empirically based methods, each of those studies suffer methodologically. Four major areas of methodological limitation include: 1) availability of data sources; 2) sampling procedures; 3) research designs; and 4) measurement problems. Each methodological limitation is discussed in the following four subsections. 2 .2.1 Availability of Data Sources Access to reliable data sources has restricted the quality of child homicide research. Similar to child abuse and neglect research , in the past many believed that inquiries involving the death of children should be restricted to social and medical practitioners to protect the victim. As a result , most data regarding the Victimization of children remains buried within bureaucratic systems. Public agencies have traditionally operated under policies that restrict access to any family or child oriented issues. These policies date back to a time when society believed family issues should not be resolved by government , but more privately through medical 37 and social service agents. Such practices have been resistant to change and slow to update, and have greatly limited data resources for studying child homicide and child abuse and neglect. Intrafamilial homicides are not identified through traditional data systems. National level statistics do not distinguish victim and offender relationships common to intrafamilial homicides. The national level arrest data collected by the Federal Bureau of Investigation-Uniform Crime Reporting system, do not include information specific to the victim and offender relationship (FBI-UCR ' 1992) . The Uniform Crime Reports do not separate out the family issues surrounding the death of children. Data must be collected at the state or local level to examine these relationships. Most studies rely on public agencies for documentation of child homicide incidents. Data collected from public records are often incomplete, subjective, and Death classification problems have also been inaccurate. of major concern for several decades. Agencies may underrecord fatal child abuse cases or intentional injury cases due to reporting categories of accidental and unknown deaths, and the misclassification of the cause of death (Jason et al., 1983; Straus, 1987; Durfee et al., 38 1992) . For example , a fatal child abuse incident may be missed because it is classified as an accidental death. Historically, classification procedures have not been consistent across agencies , and jurisdictions. Outdated statistics is another major problem with child homicide data. Mortality data are not co llected on a regular basis and are generally several years o ld. Also , data sources often invo l ve a small number of cases , which do not include a complete homicide population. Working from incomplete data sources e liminates the possibility of randomizing and may create an element of bias. 2.2.2 Limited Sampling Procedures A common problem among chi l d homicide studies is the inadequate methods used for selecting cases to be analyzed. Most c urrent studies use incomplete data sources , which are not randomized, and not representative samp l es of the tota l population . Although random selection is not imperative , randomizing makes it possible to generalize the findings to a larger population . The one study that looks at place of residence is not generali zable to the larger population. The subjects are drawn from an urban, predominately black location 39 (Detroit), with a high homicide rate (Goetting , 1990) Another problem with most child homicide data is the selection of spec i fic victim age groups. For instance ' most victim-offender studies suffer from sampling bias because they only examine cases that involve low socioeconomic status mothers (Weisheit , 1986 ; Goetting , 1988). A random selection is never employed to examine cases from all social classes. As a result , an understanding of the role of parental characteristics is not available (Ammerman and Hersen , 1990). Research that has examined family sociodemographics as predictors of child homicide (Resnick , 1969 ; Scott , 1973; Jason et al., 1983) often represent small samples not generalizable to the l arger population. Descriptive studies offer weak explanations because of limited sample size and weak empirical techniques (Widom, 1989). Percentages measuring characteristic variables are not necessarily generalizable when using small nonrandomized samples. For example, in one study offender characteristics were profiled using only twenty-nine (nonrandomized) cases in which male guardians were charged with killing their children (Scott, 1973). The child victim and caretaker offender relationship is one of the most studied areas of intrafamilial child 40 homicides. Both the relationship and offender characteristics have attracted a wide interest among researchers. Although most of the research suggests that the majority of offenders are the mothers of the victims , discriminating sampling methods prevent generalizing any of the results (Resnick , 1 969 ; Kaplun and Reich , 1976 ; Goetting, 1986 ; Silverman and Kennedy , 1988) . Also, few studies use comparison groups. Weak sampling techniques and a lack of comparison groups question the significance of the results. Comparison groups would strengthen the empirical findings of studies that use a small number of cases. Comparison groups help to establish causal inference between independent and dependent variables . Testing a proposed causal relationship involves determin i ng whether confounding variables exist . A comparison group also allows the experimenter to identify disturbance variables. 2 .2 . 3 Weak Methodological Designs out of more than 53 published studies involving child homicide most suffer from weak methodological designs. Most of the existing studies rely on case study description of victimization patterns. Although case t enhanced our knowledge about the s udy discussions have 41 seriousness of violence against children , such studies are not scientifically significant. Most data in descriptive studies involve small data sets that are not adequate for any adept statistical analyses (Scott, 1973; Krugman, 1985; Goetting, 1988; Muscat, 1988;). One study found a direct correlation between homicide rates and the degree of urbanization and poverty level of the community of residence. However, only eight child homicide cases were available in the data source (Scott, 1973). The study did use a comparison population needed for comparing results and usually involve a small number of cases. The relationships between and among specific variables have not been empirically tested by current studies. It is difficult to use current chi ld homicide research as a basis for comparing age because each study uses different age cutoffs to identify adolescents or youths. some studies cutoff the age at 15 years (Goetting, 1989), Zimring (1984) compared victims less than 16 years, and Rowley et al. ( 198 7) examined al 1 victims under 18 years. Also , most research that examines socioeconomic conditions does not employ multivariate analysis and do not control for other factors. Thes e studies do not provide strong foundations for building future research strategies. 42 2.2.4 Measurement Limitations Most of the current child homicide research employs aggregate level data. If the researcher ana l yzes group level data and attempt s to focus on individuals, then the research findings will be inaccurately interpreted. Garbarino (1976) measures New York county homicide rates by county level socioeconomic characteri st i cs . These characteristics are us ed to determine the l eve l of murders committed by mothers. Thi s does not, however, measure individual levels of abuse and socioeconomic charac - teristics. Group level SES data used to determine who commits murder, is an overgeneralization that can not be applied to individuals (Muscat, 1988) . Census data may identify poor neighborhoods, but this is macro level data that overlooks specifi c individual characteristics (Christoffel et al ., 1983). The analysis of national homicide rates by gross national product (measures economic conditions) is not plausible. If the research agenda is to determine who or what is responsible for the violence, individual level data i s necessary (Fiala & LaFree, 1988). In on e study state level homicide rates were inaccurately used to det · · d' 'd al level characteristics (Straus , 1987) ermine in ivi u Also , Justice & Duncan (1976) never identify the dependent 43 variable, and they do not explain how violence is determined. Therefore, there is no basis for their conclusions that life changes increase the amount of violence against children. In some research, measurement models are flawed and information is not properly documented (Straus, 1987). Lacking are adequate indicators for categories of social factors , and the authors do not control for any disturbance variables· (Straus, 1987). It is difficult to identify what the author is measuring and the results are often unclear when stepwise regression is used. Winpisinger and her col l eagues (1991) were unclear when explaining the results of their analysis. Information is missing regarding the variables and the methods used to measure the variables, and the relationship between the variables is questionable. In an early study of violence against children , the correlations of five indices to abuse were stated as statistically significant , but they were not confirmed in the results (Garbarino, 1976) Kaplun & Reich (1976) briefly mention a confidence coeffic i ent formula for a binomial distribution of a 90 % probability the percentage would fall into a specific theY never explain what was mea s ured with the 90 % probability. category . However, or how they came up 44 Summary of Literature Review A review of the literature indicates that previous child homicide research is almost nonexistent, and few studies demonstrate any systematic empirical investigation. Selective and discriminating sampling, in most studies , prevents generalizing to the larger population. Qualitative research cannot support any specific hypotheses by establishing association or making causal inferences about selected variables. However , previous research provides specific research questions and helps establish a strong foundation for studying child homicide. Based on the knowledge discovered through qualitative research we know what direction a more scientific approach should take. Current research consistently includes many assumptions about predictors of child homicide, but actual empirical analyses of the data are rare. One recent book (1993) written on victim-offender homicide relationships barely acknowledges the childhood victim (Wilson , 1993). The bulk of early child homicide literature concentrates on psychologically based factors (Resnick , 1969) over- looking the socially based family and community issues. We know very little about the family structure, or community factors of the victims and their famili es . 45 Based on the sub:ultura l vio l ence theory we need to look at multiple level factors . Attributes based on multip l e indicators (Wiersema and Pattavina, 1993) , would provide a multidimensional approach f or ide ntifying c hild homicide risk factors . To clarify what the relationship is between child h omi:ide a nd chi ld abuse and neglect , we need to l ook at three levels of risk factors: socio- demographic , fami ly, and community factors . The literature includes many assumpti ons about the ri s k factors of c hil d homi:ide , but actual empirical evidence is ei ther unavailable or it suffers from methodological errors . The unde rlyi~g assumption is that any o n e of these factors can incLease a child's fatality risk. The majority of literature focuses on descriptive patterns of victims , but li ttle i s known about the nature and causes of child homi c ide . 46 CHAPTER 3 . CHILD FATALITY REVIEW TEAMS This chapter provides historical and methodological explanations of Chi l d Fatality Revi ew Teams (CFRT) established across the country . The child homicide data used in this study was obtained from the State of Maryland and Baltimore city Child Fatality Review Team Records . Therefore , a detailed account of the State of Maryland CFRT process is also provided, to demonstrate the advantages of using the data for this study. Multi agency Child Fatality Review Teams have slowly emerged in response to the i ncreasing awareness of severe violence against children in the United States . Since 1 978 , when the first CFRT was established in Los Angeles , Cali fornia , over 44 Child Fatality Review Teams have developed across the nation (Durfee et a l., 1992) . The American Bar Association, Center on Children and the Law (1 991) defines "Child Fatality Revi e w Teams " as a group of people who meet to review child deaths , ideally all deaths of children below the age of 18 years. The CFRT is an active s urve illance of childhood d eaths . Su ch a team can be internal or external. An internal chi ld review team reviews child deaths related to a particul ar agency . For exampl e , the Child Protective Service Agency would only 47 11 ·1 ,,, ,, ,,, ,, i' ( I' " " 1-:: ,,. 11 ·· i!. fi r /,:. review cases involving their former cl ients, the death of children who had contact with their agency before their death. An external review team does no t limi t its wo rk to any one agency, but, considers the activities of all agencies. An external team membership generally consists of multidi sciplinary and multiagency representatives . Memberships have included representatives from med ical examiner/coroner offices , pediatric physicians, child protective service agencies , law enforcement , prosecuting attorneys , public health and menta l health , education , Sudde n Infant Death (SID) programs , domestic violence programs, and social scientists. Teams have been established at two different geographical levels. Generally, there is a state leve l team, whi c h h as traditionally b een r espon s ibl e for establishing protocol for the review process to be used in that particular state. The State team mandates change and policy for the statewide local child review system . Local teams are established for the purpose of r e viewing c ases in a s maller area . For example , some sta t es have established local teams at the county or city level. Local teams are required t o foll ow the state established policies and procedures f or reviewing cases . Typically, the medical examiner provid es the 48 ,. I· ,. , '' 'I .; : multiagency team with an inclusive monthly listing of all child deaths. Members of the team check the names with their own records for prior agency contacts with the child br family. Monthly meetings are held to review this multiagency data and discuss details of suspicious deaths. Cases are investigated completely within the membership . Reviews may lead to criminal charges and prosecution , changes in classification of deaths , and intra-agency support for the surviving family members. The primary justification for the development of a systematic review of child fatalities was based on a number of problems associated with chi ldhood deaths. Several of these problems have addressed some very basic issues. For instance , prior to reviews we did not know the numbe r of annual deaths of children or the accurate causes of their deaths (Durfee, 1989). We were unable to account for all the fatal abuse and neglect incidents on a national leve l, many states did not keep track of t hi s data. The development of CFRTs is also in response to the vast problems with intra- and inter-agency communication , r ecordkeeping and inadequate databases . Death certificates have often reflected different causes of death from what is indicated by police or child protective service records. The value of multiagency inquiries of childhood 49 1 ·•1• ,. ;,:: '. 'I ,·: '' 1" ;·: , .. :;; ·' ., deaths is the identification of child abuse and neglect as a cause of death. Many so called accidental deaths are found to be linked with child abuse and neglect or the result of abuse or neglect. Without t he review process many such deaths have gone unnoticed as anything more than an accident. To remedy these problems , Child Fatality Review Teams have been developed across the United States. Review teams around the country have targeted numerous goals and objectives. An initial goal for most review teams has been to develop open communication between public agencies concerned with the welfare of children. The theory is that if agencies share case information, deaths may be avoided or at least more accurately classified. Also , opening communication lines among various agencies facilitat es identifying and protecting the deceased child ' s siblings who are at risk. Other objectives have included improving the response and investigation of child deaths by all agencies involved in the welfare of children, educating the community in prevention, establ ishing protocol for health and social service agencies for reporting suspicious deaths, identifying the full extent of child homicides, and Properly classifying the death of children. A long term goal of Child Fatality Review Teams is to 50 l •" ; ·' ' :, ..- ......... ._........ -- · .. ~- develop an accurate child fatality database that would provide information to all agencies involved in the welfare of children . . A central database that will provide sophisticated knowledge on the patterns and trends from past incide nts would be developed. The data could be used to evaluate and identify problem areas in a community , families at risk , and help prevent the misclassi fication of deaths. More accurate classifications of childhood deaths may increase awareness about intentional injuries and neglectful harm inflicted on children, which may prevent future deaths . When teams begin to systematically examine child fatalities, more homicides will be discovered (Durfee, 1989). Multiagency data collected by the Child Fatality Review Teams will improve future child homicide data sources , and as a result, improve child homicide research , and prevention. Multiagency databases will more adequately portray the dimensions of childh ood deaths (Kaplan , 1992). An ultimate goal of review teams is to develop intervention strategies and early prevention measures for children at risk. Although team guidelines and procedures vary from team to team, many improvements have been made nationally, regarding the investigation, prevention, and information of chi ld d eaths. Review teams have recognized timely 51 I ,- ,, ;a ,, ·., Maryland Child Fatality Review Team (CFRT) In 1988 the Office of the Chief Medical Examiner (OCME) , together with the Departments of Human Resources (DHR) and Health and Mental Hygiene (DHMH) established a CFRT for the State of Maryland. In i tially the multiagency team met to discuss social and legal issues surrounding the death of children. They he l d several workshops and discussed strategies for investigating child fatalities . The CFRT outlined fatality case investigation procedures and guidelines for the police , child protective services , and the state medical examiner ' s office . Since 1991, both state and city teams have been involved in actual review and investigation of all deaths involving children less than 15 years of age. Agency representation at meetings always includes ocME , DHR , a pediatrician , and a SIDS specialist. At most meetings there is a police representative (usually a Homicide Unit detective) , a representative from DHMH, and two researchers , one a trained public health and injury prevention epidemiologist , the other a criminologist. The CFRT reviews deaths of children below the age of 15 years. Cases that involve children between 15 and 18 years are not reviewed unless the cause of death is suicide . However , CFRT data collection forms are 53 completed on all children under the age of 18 years. Although the team has been meeting on a regular basis since 1991, data has only systematically been collected since 1993. The team has just begun to develop a statewide child fatality database. Today, the Maryland CFRT consists of a multi- disciplinary and multiagency membership. The Maryland CFRT has several major goals. The team investigates and establi shes a cause of death in cases where the cause of death was unknown. It identifies possible child abuse and neglect related fatalities not previously reported to child protective services . The team recommends further police or child protective service investigations in cases suspected of abuse or neglect, and educates the public about prevention. The team also establishes policies and procedures for investigating child deaths, and keeps committee members informed of all current interagency policies and procedures. There are two active teams, one is the State of Maryland CFRT , and the other is the City of Baltimore CFRT. The state team is responsible for developing the main mission and goals of all the Maryland based CFRTs , and setting the standards for reviewing and investigating all child deaths. The state team is also responsible for 54 establishing access to data sources within each of the agencies ; developing and maintaining data co llection instruments; overseeing the review and continued investigation of chi ld fatality cases in all geographical areas except Baltimore City; and establishing strategies to enhance the investigation of child fatalities. An ultimate goal of the state team is to develop strategies and interventions for preventing future suspicious deaths. Also , to properly recognize deaths caused from abuse or neg l ect not otherwise.identified. The Baltimore City CFRT is a local membership established for the specific purpose of investigating Baltimore City child fatalities. The Baltimore CFRT is responsible for reviewing all child fatalities that occur in the city of Baltimore on a monthly basis. Baltimore CFRT members represent local agencies who generally are aware of any contact with the child or the family prior to death. MultiagencY history involving deceased children are discussed at each review. CFRT recommendations are made to help in reducing mistakes in ascertaining the cause and manner of death. A recent recommendation (September 1993) was helpful in the revision of police procedures for SIDS death scene investigations. The death scene investigation provides an accurate documentation of 55 the scene in terms of environmental risk factors and risk factors associated with sleeping conditions. The new revisions require that police provide as much detail about the environment and specific location of a SIDS death. Full documentation , including a sketch and a written exp lanation, is forwarded to the medical examiner , and documented in the CFRT database. The medi ca l examiner now has more detailed information for distinguishing SIDS cases from abuse and other medical conditions that may have caused the death. With close invest igation and sophisticated pathological testings many infant deaths are found not to be accidental. In cases that involve (subtle fatal c hild abuse) fatal h ead injuries, chemical assaults (poisoning), asphyxiation , drowning , and cardiac arrhythmia ' s, the death may not always be classified as a homicide . If the pathological examination is limited (nonanatomic evidence ) then many times a child homicide is never uncovered (Zumwalt and Hirsch , 1980) · The state of Maryland CFRT plans to expand it s membership to include a representative from the State Attorney General's Office, and also o ther health , education , and social service representatives. The sta t e team will eventually move away f rom the twice monthly task 56 of reviewing cases, as additional local teams are established at the county level. The state team members are prepared to review the more difficult or pending cases once the local team completes their own investigation. The team is currently developing a single child fatality database. The State CFRT has conducted several workshops with community members to review current data collection forms , determine the priorities for multiagency data purposes, and to discuss availability of resources for structuring a database system. The CFRT has also discussed expanding the database by eventually matching birth and death certificates together. Birth certificates include additional sociodemographic data not included on death certificates and often incomplete in both the social service and police records. The current Maryland Child Fatality Review Team database is a comprehensive multiagency resource. The CFRT database consists of consolidated information extracted from each agency involved with the review process. Data is obtained from OCME records, local police department records , child protective service records, and hospital records (when applicable). A Data collection form has been developed by the State CFRT , to consolidate 57 multiagency information . A complete exp lanati on of each data source included in the Maryland CFRT data follows. 1) Medical Examiner (OCME) Records Data is extracted from Medical Examiner records . Each file contains an Autopsy report (external inspection and/or i nternal autopsy) , medical facil i ty report (when applicable), 24 hour police report (when applicable) , OCME investigation report and notes , and a death certificate. 2) Law Enforcement Records Police investigation reports that follow the initial 24 -hour police investigation. Any information regarding arrest , or suspect(s) , and interviews of family or witnesses , may be provided . 3) Chi l d Protective Services (CPS) Records Data is extracted from Department of Social Services I Child Protective service records. When a death has been investigated or a member of a victim ' s family (to include the victim) has been investigated, CPS places a copy of the report in the victim ' s records. All previous contacts with CPS are filed in the victim's record. Rating forms 58 that identify risk level of the victim(s) are provided following each caseworker contact visit . 4) Vi tal Statistics Data Deaths of children under 19 years of age in Maryland during 1993 , are verified through Vital Statistics , Department of Health. 5) Child Fatality Review Team Data Besides the data collected from the above sources , the CFRT also provides meeting notes concerning the specific cases discussed . Additional information is often gathered from members who have had further contact with the victim ' s family , or follow-up investigation of the death . If a change of manner or cause of death occurs , because of the CFRT investigation , the new manner or cause is corrected in the OCME records and the State Vital Statis t ics office is notified . 59 CHAPTER 4. RESEARCH METHODS This chapter describes the research design and methods of analysis used in the study. There are two sections , the first section describes the three data sources used in the study, and the second section states the research questions , describes the risk factor variables , and explains the methods used to measure each of the targeted research questions. 4.1 Data Sources This study employs three major data sources: 1) 1993- 1994 Child Homicide Data - State of Maryland Child Fatality Review Team homicide cases; 2) 1993-1994 Child Abuse and Neglect Data - City of Baltimore, Maryland, Department of Human Services , Child Protective Service substantiated child abuse and neglect cases ; and 3) 1990 U.S. Bu reau of census Data - City of Baltimore, Maryland census tracts. A complete description of each data sample follows. 60 4.1.1 Child Homicide Data This study departs from previous child homicide studies by employing multiple agency child homicide data , collected by the State of Maryland Child Fatality Review Team. As discussed previously in chapter 3 , Child Fatality Review Team data incorporates all incident based data from forensic, medical, legal , and social service records in the state under study. All death records are located in the State Medical Examiner ' s Office. Homicide cases are reviewed by members of the Child Fatality Review Team to ensure full investigation of the circumstances leading to a death , as well as detailed documentation. No existing research comparatively analyzes child homicide incidents using multiple agency Child Fatality Review Team data. Also, most prior research examines only a sample of child homicide cases, usually targeting a particular age group. This study examines all child homicides in the state of Maryland that occurred over an eighteen month period, from January 1993 through June 1994. For this time period, there were 82 child homicide victims , less than 18 years of age, who died and resided in the State of Maryland. All unnatural child death incidents are reviewed by the state of Maryland , Medical Examiner ' s 61 Table 1 Child Homicides by Victim's Place of Residence Victim's Place of Residence Baltimore City Anne Arundel County Baltimore County Cecil County Frederick County Harford County Howard County Montgomery County Prince George ' s County Wicomico County Total n % 55 67 2 2 1 1 3 5 1 1 1 1 2 2 2 2 14 18 1 1 82 100 % office for either , or both an external or internal anatomical examination to determine the cause of death. Therefore, it can be assumed that all child homicides in the State of Maryland for this time period are included in this data set. Table 1 presents the distribution of homicide cases , based on the residence of the victim, across the state of Maryland . As shown in table 1 , 67 % of the victims resided in Baltimore city at the time of death, and 33 % of the victims were dispersed among nine other counties throughout the State of Maryland . Table 2 presents the distributions for both state and city homicide cases . The data are arranged into subsamples based on the two child homicide categories , 62 Table 2 Child Homicide Type by Place of Victim's Residence Intrafamilial Extrafamilial Total Baltimore City 22 %(12)* 78 % (43) 100 %(55) Other Maryland Cities 56 % (14) 44 lb (13) 100 %(27) Total 3Y2i (26) 67 % (56) 100 %(82) *Number of cases in parenthesis 1) intrafamilial homicide , and 2) extrafamilial homicide. J Intrafamilial and extrafamilial homicides define the two dependent variables used in this study. As shown in table 2 , the majority of child homicide cases are defined as extrafamilial homicide. Sixty-seven percent of the state homicides and 78 % of the city are extrafamilial homicides. 4.1.2 Child Abuse and Neglect data This study also includes a comparison group of Child Protective service substantiated (or confirmed) child abuse and neglect cases . Restricting the selection to JThree cases in this data set involved victims of justifiable homicide. They were killed by a police officer during the pursuit of criminal su~p~cts. E~c~ of the three cases were placed in the extrafamilial homicide category. 63 confirmed child abuse and neglect cases increases the reliability of the data. Reports of child abuse and neglect are not always confirmed f ollowing a Child Protective Service ' s investigation. As no ted in Ch apter 1 , the int e ntion of this study is to employ a comparison group consisting of State leve l child abuse and neglect data. Th e State of Maryland, Department of Social Services, do not have any means for collecting or accounting for statewide case files. Therefore, the study design for the comparison group had to be changed to city level child abuse and neglect data. The data sample of 210 child abuse and neglect cases were systematically sel e ct e d fr om Baltimore City , Child Protective Service records, for the same e ighteen mon t h time period as the homicide incidents; January 1993 and June 1994. The abuse and neglect cases were randomized using a sampling interval of every nineteenth case . 4 The statistically determined samp l e size was based on a 95 % confi dence interval to control for any error in estimating to the g e neral population. These data are a repre s entative 4Cases could not be selected by the type of abuse or neglect involved. Physical , sexual , and neglect victims were all included in the sample. 64 Table 3 Baltimore City Child Abuse and Neglect Cases Case Type n Q 0 -- - - - -------- -- Physical Abuse 61 29 Sexual Abuse 39 18 Neglect 110 53 Total 211 100 % sample of typical child abuse and neglect victims residing in Baltimore , Maryland . Table 3 shows the distributions of child abuse and neglect cases by type. There are 210 victims included in the comparison group of child abuse and neglect cases. As shown in table 3 , there are two categories of child abuse 1) physical abuse and 2) sexual abuse . The comparison group is primarily made up of neglect victims with 53 %. Also , 29 % are physical abuse victims , and 18 % sexual abuse victims. 4. 1. 3 Census data This study employs 1990 United States Bureau of Census data to map out high risk communities , and to develop risk patterns specific to communities experiencing 65 high rates of child homicide and child abuse and neglect. The U.S. Census data are a descriptive sampling of a wide variety of characteristics of the U.S. population aggregated by geographic areas; such as states, counties, and cities. The most recent U.S. Bureau of Census survey data, collected in 1990, was used for this study. The census data sample consists of census tracts for the City of Baltimore, Maryland. Census tracts were identified for each of the 55 Baltimore City homicide cases and the 210 child abuse and neglect cases. The census data variables ~ere merged into the Baltimore City child homicide and child abuse and neglect data sets. Census data were used to examine community level risk factors for both child homicide and child abuse and neglect victims 5 • The variables used to describe community level risk factors include, the percentage of the population living below the poverty level, the Percentage of the population under 18 years of age, the Percentage of the population who are nonwhite, and the Percentage of the population living in single parent sBal timore City census data is .agyreg~ted into c ensus tracts that define and outline specific neighborhoods. However, census tracts are not availabl e for all Ma r yl a nd communities. 66 households. Communi ty characteri stics were identified for each victim' s census trac t , based on their add ress at the time of d eath or injury. The community level risk factors are explained fur ther in section 4. 2 below. 4.2 Methodology The follow i ng section describes each r esearch questio n, the variab l e s u sed to describe the risk factors , and the methodol ogy u sed to analyze the research questions. The purpose of this study is t o explore the ris k factor re l ationshir s betwee n child h omi cide and child abuse and neg l ect . The study examines the differences in risk facto r patterns between intrafamilial and extra- familial c hild homicide , and child abu se and neglect. There are three l e ~e l s of risk factors that are examined in this study , individual , family , and community. Each o f the risk factor l eve l s i nvolve a gro up o f items that d escribe a particul ar risk factor l evel . Individual leve l risk factors d escribe the profile of the victim and suspec t for each h omi cide and abu se or neglect case . Fami ly l evel risk factors describe t h e c h arac teristi c s o f t h e vict im ' s family a s well as provide t he histo ry of previ o u s e xper i ences o f child abus e o r neglect invo lving the victim and any o f t heir siblings . Communi ty l e v e l 67 risk factors describe the victim's community of residence with regard to income status , number of children , race, and marital status. The comparison group variables are designed to match those collected in the homicide data sample, allowing the comparison of key variables from each risk factor level. The specific variables used to describe the risk factors are defined in section 4.2.1 of this chapter. The examination of the data addresses three research questions. Table 4 identifies the data and the methodology used to analyze each of the research questions . The first phase of the analysis compares Baltimore city child homicide cases with all other State of Maryland cities, to determine if there are any significant differences between s tate and city leve l homicide incidents. If no significant differences are found, then the use of city homicide data in later phases of the analysis will be justified. The second phase of the analysis compares statewide intra- and extrafamilial homicide data. Comparisons are made to determine whether or not there are any differences among individual , family , and community leve l factors. The third phase of analysis examines the causes and circumstances surrounding statewide child homicides and child abuse and neglect incidents. The 68 r O'\ \.0 Table 4 Research Design and Methodology Questions 1. Risk Factor Patterns: A. Intrafamilial Homicides 1 . Individual Factors 2. Family Factors 3 . Community Factors B. Extrafamil i al Homicides 1. Individual Factors 2 . Family Factors 3 . Community Factors 2. Risk Factor Dif ferences: Homicide vs . Child Abuse and Neglect 1 . Indi vidual Factors 2 . Family Factors 3 . Community Factors 3 . Causes / Circumstances of homicid e vs . A. Type o f I nj u ries B. Type o f Weapon C. Ti me of I n j ury D. Place of Death E . Drug Related Death F. Number o f Victims Data Set (s ) State Level Homicide Data City Census Dat a State Lev el Homicide Data City Census Data Baltimore Ci ty - Homicide Da ta Child Abuse Data Census Data abuse and negl ect St ate Level Homicide Data Child Abuse and Neglect Da ta Methods Frequencies Chi-Squares Regression Frequencies Chi - Squares Regression Frequencies Chi-Squares Regression Fr equencies final phase of analysis compares Ba l timore city child homicide and child abuse and neglect incidents. If earlier results find risk factor differences between t h e two types of child homicide then each type of homicide will be compared with chi l d abu se and neg l ec t data. 4.2.1 Homicide Risk Factor Patterns Are there different risk factor pat t erns between intrafamilial and extrafami l ial ch i ld homicides? The first research question examines the nature of the association of specific risk factors between intrafami l ial and extrafamilial homicides . One of the major concerns with prior child homicide research is the lack of comparative risk factor analyses between each type of child homicide . This study is designed to examine the effects of multiple leve l risk factors between intra - and extrafamilial child homicide . For purposes of this study , intrafamilial and extrafamilial child homicides are the dependent variables . The three levels of risk factor variab l es are described be l ow. a. What specific indi vidual leve l risk factors will be different? The variables used to measure , individual level risk factors include ag e , 70 gender, race , (both victim and offender profiles) , victim-offender relationship , and birthorder of the victim (i . e. , f i rst born, middle child , last born , or only children) Prior research suggests there are patterns of victimization for each type of child homicide . For instance , several studies show males , under four years of age, overrepresented among fatal abuse victims (Schloesser et al., 1992 ; Plass , 1993). However , for extrafamilial Victims most prior research identifies black , male , urban , adolescents as overrepresented among drug related street murders (Goetting , 1989) . Victim profiles appear to be dependent on the age of the child . Victims less than 10 years generally fall into the area of fatal child abuse or intrafamilial homicide research . Child homicides involving victims between the age of 10 and 19 years , are generally defined as street murders , or extrafamilial homicides. Separate age divisions suggest there may be different risk factors for intra - and extrafamilial child homicide incidents. Victim and offender relationships may differ according to age of the victim. Typically, children less than ten years of age are generally limited in their exposure to people outside the family (Jason, 1983) . Suggesting that younger 71 children would generally be victimized by a nonstranger. Current intrafamilial homicide research specifies that the child's caretaker is characteristically the perpetrator (Miller and Block , 1982; Christoffel , 1990) . The most common extrafamilial homicide victim has a similar characteristic profile as that of their offender (Rowley et l a . ' 1987 ; Harries 1993). Extrafamilial homicide research specifies that adolescents are more often victimized by another peer member or an adult acquaintance known outside the family residence , and occasionally a stranger is responsible . Previous research also suggests an apparent connection between adolescent street murders and illegal drug activities (Goetting, 1990). The rarest victim and offender relationship found in child homicides involve strangers (Jason and Andereck , 1983). With the exception of justifiable police homicides, extrafamilial homicide victim and offender relationships are typically nonstrangers. Victims and offenders are also expected to have similar individual Profiles. However , extrafamilial individual profiles will typically vary greatly from intrafamilial victim and offender profiles. Some research suggests that the birth order position of a victim is an important risk factor of child homicide. 72 For example, two separate studies found that most victims were either the only child or youngest sibling (Mitchell, 1989; Ammerman, 1991). However , each of these studies is limited by the age of the victims in their data samples. This study examines the birth order position of the homicide victim to determine if there are differences between the two types of child homicide victims . Birth order is expected to show entirely opposite conclusions for each type of homicide. Intrafamilial homicide should indicate more last born victims , and extrafamilial homicide is expected to show more first born victims. Also, intrafamilial homicide victims are typically younger, and more often the last born child, suggesting that they would have fewer chances of experiencing prior violence (Jason, 1983i. A complete explanation of how each of the individual level variables are coded follows. Gender is represented by a binary variable coded one for males and zero for females. Age of the victim is a binary variable coded one for less than 10 years of age and zero for ten and older. Age of the suspect is coded as ab' . ble with one for over 25 years of age and inary varia zero for 14 through 25 years of age. categorical variable coded one for black , two for white, Race is a three for Asian, and four for Hispanic. Birth order o f 73 the victim is represented as a categorical variable, coded one for first born, two for middle child, three for last born , and 4 for only children . The victim-offender relationship is a categorical variable coded one for natural parent , two for foster or step parent , three for Parent ' s paramour, four for other relative, five for babysitter, six for acquaintance, seven for stranger , and eight for police officer. For purposes of the multiple logistic regression analysis the victim- offender relationship variable is recoded as a binary variable. The new variable is coded as one for caretakers and zero for non-caretakers. b . What specific (victim) family level risk factors will be different? The items used to measure (victims') family level risk factors include parents' marital status , socioeconomic status , family size , and previous history of abuse. Very little is known about family level risk factors anct child homicide incidents. Previous research focusing on offenders of intrafamilial child homicides finds that most victims come from single parent families (Goetting, 1989; Plass, 1993 ). Family demographic variables such as , l status, single parent, teen parent, ow socioeconomic 74 nonintact families with many children (Gartner, 1991), are assumed to aff e ct child abuse, however, they have not been examined regarding child homicide. Based on prior research one may assume that whatever factors affect child abuse, may also affect intrafamilial child homicides. This study examines specific family level risk factor patterns between both intra- and extrafamilial homicides. Socio-economic status has not been analyzed using individual level data, and aggregate level data has l imited the prior research from generalizing to individual victim and offender characteristics . Prior history of child abuse or neglect of victims and their siblings are expected to be higher among extrafamilial homicide victims, than for intrafamilial homicide victims. These conclusions are based on the age of the child at the time of death, in that most extrafamilial homicide victims are older than ten years of age. Also, most intrafamilial research sugges t s that victims are the youngest child in t he fami ly , s u ggesting t hat they may not h ave a pr i or history o f b e i n g abused or ne g l ected (Schloe sser et al ., 1993 ). Explanations as to hOW each of the fami l y level variables are code d are descr ibed below. Of the victim's parents is repre s ent e d Marital status by a vari'able coded one for married, two for categorica l 7 5 single , three for widowed. Socio- economi c status i s me asured using Public Welfare Assistance status . Publi c ass i stance s tatus is represented by four binary variables, coded one for partial public a ssistance , (food stamps, partial income or s ubsistence), two for Medicaid (medical benefits) , three f or no public assis tance , four for unknown public assistance status , and five f or full (AFDC ) p ubli c assistance . Persons with no known i ncome are represented in the suppressed categor i es of each of these . Family size is represented by the number of children in the victim ' s family. Family size is a discrete var iable that indicate s the actual number of children (th e maximum family s ize is 8 additional chi ldren) . Prior h i story of chi ld abus e or neglect o f both the v i ct im and t heir s iblings is represented by a binary variabl e coded one for prior history , zero for no histo ry . The actual number of times a r eport for prior abuse o r neglect was substantiated is repres ented as binary variable coded one for three or more r eports and zero for n o reports . For purposes of t he multiple l ogistic regression ana l ysis both marital status and family size are recoded as b inary variab l es. Marital statu s i s coded as one for singl e and ze r o for married . Family size is coded one for s iblings and zero for no s iblings . 76 c. What specific community level risk factors will be different? The variables used to measure community level risk factors include p l ace of vict im's residence (census tract), percentage of the population living below the poverty level, percentage of the population under 1 8 years of age, percentage of the nonwhite population, and percentage of single parent househo lds. Specific community characteristics may be identified as key risk factors to incidents of child homicide and/or child abuse and neglect . Currently, there are no extrafamilial homicide studies that focus on community level risk factors. Thi s study will analyze Baltimore City level data to compare community level risk factors between intra- and extrafamilial child homicides . The victims ' addresses are matched to a specific census tract identified from Baltimore City Census Bureau data. The tract identifiers are then merged into the homicide and child abuse and neglect data sets. Census tracts are discrete variables indicating the actua l number of the tract. The community level variables are coded as follows. The percentage of the population under the Poverty level is us e d as a measure of the socioeconomic 77 ·status of each community. The percentage distributions for poverty and nonwhite are spread between O and 100 %. Therefore, the item for the percentage of the population liv· ing below the pove~ty leve l is coded as a binary ariable using one for more than 50 % of the community V . members living below the poverty level and zero for less than 50 % of the community living below the poverty level. The item measuring the percentage of the population who are nonwhite is coded as a binary variable using one for more than 50 % of the community members who are nonwhite and zero for less than 50 % of the community members who are nonwhite. The percentage distributions for population under 18 years of age and members living in single parent households are between o and 50 %. The item for percentage of the population und~r 18 years of age is coded as a binary variable with one for more than 25 % of the community members under 18 years of age and zero for less than 25% o f the population under 18 years. The item for Percentage of population living in single parent households is coded one for more than 25 % of the Population living in single parent households and zero for .less than 25 % of the population living in single parent households. 78 The data used to measure child homicide risk factor patterns · l d inc u e the U.S. Census data for Baltimore City, a nd State of Ma ryland child homicide data. The 82 child e incide nts are categorized into one of the two homicid · · . categories of c hild homicide , intrafamilial and extrafamilial. 4 · 2 · 2 Risk Factor Differences - Child Homicide and Child Abuse and Neglect Are the r e di ffe r e nt risk factor patte rns between c hild hom· · i c ide a nd c hild abus e a nd n e glect incide nts ? This n e xt res e arch question addresses whether there are diffe r e nt r isk factor patterns between child homicide and child a buse and n e g lect incidents . The limi tat ions o f mos t prior c hild homicide research demonstrates the n eed f o r comprehe nsively examining the var ious risk f actors of b o th homicide and child abuse and neglect incidents . Onc e establishing specific patterns for both homi c i de and abuse and neglec t , p at terns of risk should be c ompared b etwe en h omicid e a nd abuse and neglect . some r isk fa c t ors inc luded in pre vious child abuse and negl ect r e search are age , g e nder and r ace of both vi c tim and of f e nder , s ing l e Pa r e nting , socioeconomi c status , p r i or his t o r y of abuse , 79 I I I I I I I I I I I - --... ~ ---~ ----- illegal drug use , and poverty of neighborhood. It is a comb· ination of these models that target high risk abusive families used in this study t o examine risk factors for child homicide. A major assumption of prior research is that c ertain patterns of risk factors would be more evident in child homicide incidents than in child abuse a nd neglect. This study examines risk factor patterns Using e a c h o f the three levels of risk describe d in section 4.2.1. Each o f the f ollowing research questions are addressed in this phase of the analysis . a. Are there different individua l level risk factors between child homicide , and child abuse and n e gl ect inc idents? b. Are the r e different family level risk f actors between child homi c ide , a nd child a buse and negl ec t incidents? c . Are t h e r e di fferent c ommuni t y l e v e l ri s k f ac t o r s b etween child homicid e , a nd c h ild a buse and n egl ect inc idents ? 80 I . ~- --~---- '::- ·=-- ·-·· Although one might assume there is a link between intrafamilial homicide characteristics and abusive fami l ies , there are no d' t stu i es o suppor t these rela t ionships. Child abuse and neglect may be major contributing factors in the high proportion of homicides (Mccurdy and Daro , 1993) . This study examines pr i or h' lstory of child abuse and neglect of homicide and abuse and neglect victims a~d their s iblings. 4 · 2 . 3 Ri sk Factor Predicti on ~tare the significant risk factors for predicting child ~icide and child abuse and neglec t? Based on the feasibili ty of the data, t h e association of significant risk factors for child homicide and child abuse and neglect are examined . Multiple logis t i c regression statistics6 are used to measure t h e b i nary dependent variable of·intrafamilial and extrafamilial h omici d e and a dependent variable measuring child abuse and neglect incidents . Multi-dimensional prediction Gw·th trol study such as t hi s , the samp l e is l a case con . Selected on the outcome of t h e dependen~ var~able of homici' d L . t · egression can obtain adJusted odds e . ogis ic r ff· · t (H ratios from the e stimated slope coe icien s osmer and Lemeshow, 1989). 81 ·------- - ---· -~===--==-·---=-- -- ase on e in ings o earlier risk models are developed b d th f ' d' f factor analyses . The methods used to analyze the risk factor items and the homicide and chi ld abuse and neglect data include cross - tabulations of frequency distributions and percentages, chi -square values , and multiple logistic regression models. cross-tabulations will provide descriptive statistics describing the characteristics of child homicide and child abuse and neglect data. Risk factor patterns will be identified between the categories of homicide and child abuse and neglect . The three l evels of risk factors will be interpreted as separate groups of items. All of the individual victim profile items will be examined as a group, the individual suspect profile items, the vict im' s family factors, and community level risk factors will be examined separately across both homicide categories and c hild abuse and neglect incidents. Chi- square values are used to test for significant comparative differences in the cross- tabulated data . Risk factors showing statistical significance, are further analyzed, us ing multiple logistic regression models. Multiple l ogistic regression models are used to determine the l evel of prediction for each risk factor item . The dependent 82 .. ---;---~--::--.:-- - ·~-- variable is binary measuring both intrafamilial and extrafamilial homicides . The first phase of analysis compares state and city l eve l homi c i de data by each risk fact or level. This comparison will determine if any risk factor differences exis t between stat e and city cases. If there are no differences at the state and city level s , later s tages of the analysis will use onl y c ity level homicide data. Onl y Baltimore City homi c ide data will be used to examine risk factor dif ferences be tween homicide and c hild abuse a nd neglect incidents . The second phase of analysis compares state leve l intrafamilial and extrafamilia l child homicides by each risk factor level . This comparison wil l determine if any risk facto r differences exist between each category of child homicide . Based on findings between intrafamilial and extrafamilial homicide, t h e third phase of analys i s will compare homic ide and abuse and neglect risk factor patterns. If t here are no differences between each category of homic ide , then in the third phase of analysis chi ld homicide incidents wil l be not be broken down into separate homi cide categor ies. However , if risk factor di fferences are found between intrafamilial and extrafamilial homicide incidents, then separate analyses will be completed for each category of 83 homicide. Each category of homicide will remain separate so that risk factor d i fferences between intrafamilial and child abuse and neglect , and extra- fami lial homicide and child abuse and neglect incidents are examined . 4 . 2.4 Causes and Circumstances of Death or Injury ~tare the unique causes and circumstances of child ~icide death and child abuse or neglect injury? This study examines the unique circumstances leading to death, and the caus es of death and types of injuries in both child homicide and child abuse and neglect incident s . a. Cause of death or Type of Injury . The variable categories for cause of death or type of injury are gunshot wounds , physical abuse , fire , strangulation, stab wounds , and malnouri shment . The cause of death or injury variable is coded as , 0 ne for gun shot wound, two for physical beating, three for fire or scalding, four for strangulation, hanging, or asphyxia, five for stab wounds , and six for malnutrition, dehydration, or child neglect . Facts about the specifi c causes of death or injury were examined to determine the frequency distributions of each subcategory of this 84 I I I I V . ariable and to compare the characteristics with child abuse and neglect incidents . b. Circumstances of the incident . The variables used to measure the circumstances of death or injury include time of death/injury, place of death/injury, drug related injury, types of weapons used, and number of offenders. The characteristics of circumstances leading to the homicide of children appear to differ depending on the age of the child. For example , a child less than four years of age is more physically vulnerable to violence than an adolescent. An adolescent , however , is exposed to v · lolence through peer association of other adolescents Participating in street crime. Such age distinctions, require specific age groupings be established before examining any causal relationships with child homicide. State level homicide data will b e used to examine the circumstances and causes of death , and types of injuries resulting in the death of a child . Also , Baltimore City child abuse and neglect data will be used to examine the circumstances and caus es of inj ury fo und in child abus e anct neglect inc idents. The c irc umstances of death or injury variable is categorica l coded as one fo r argument preceding death , t wo 85 .:, • I ' 1: ., ~- -·-:~---- -·- :-~ .... . ---- The for abuse or neglect, three for street , drug, or gang s hooting , four police shooting, and five for arson. time of injury is a categorical variab l e coded one for OOOl-1000 hours, two for 1001 - 1800 hours, and three for The place of injury variable is coded lBOl-2459 hours . one for victim's residence, two for another person ' s residence , three for daycare center , four for public roadway or alley, six for other (i . e. , vacant parking lot , field, or inside parked vehicle). The item used to measure the type of weapon is coded as, one for hands or feet ' two for knife or sharp object , three for fire or hot liquid, four for strangulating device, five for neglect (i . e ., no sustenance , no food or water) , six for unknown type of firearm , seven for shotgun , eight for automatic rifle , and nine for handgun. The number of additional victims is a discrete variabl e measuring the actual number of additional victims. Frequency distributions and Percentages were examined for each subcategory of item of cause and circumstance. Distributions were compared between each type of homicide and child abuse and neglect. 86 , I i ! I '' ""'\----=.:--=------- -- CHAPTER 5. RESULTS This chapter presents the results of the study in four separate sections. The first section presents the results of the comparative analysis between Baltimore city and a ll other Maryland city child homicide data . The second section includes the results of the comparat ive analysis between intrafamilial and extrafamil ial homicide sing state level child homi c i de data. The third section u . presents the results of the comparison between child h omicide and child abuse and neglect incidents using Baltimore city leve l data. Finally, t h e fourth section presents the results compar i ng child homicide and child abuse and neglect incidents with regard to the causes and circumstances of death and injury. The results of the study are described in relation to existing theories and policies. Before presenting the findings the problem of missing data is briefly discussed . Missing Data Several chi ld homicide and child abuse and neglect cases are missing data for spec i fic risk factor items . Table 5 shows that this was especially true for two risk f t d t the chi l d homicide data . ac or items with regar 0 87 11 'I I I i ., Table 5 Di stributi ons f or Items Excluded due to Mi ssing Data Item other Cities (n=27) Suspect History n Mental He alth 4 Perpetrated D.V./CAN* 6 Victim of D.V./CAN Alcohol/Drug Abuse 4 Prior Crime Convictions 6 No History 4 Missing data 3 Socioeconomic Status Partial AFDC 2 Working 1 Full AFDC Missing data 24 % 17 25.5 17 24 . 5 17 67 33 Baltimore City (n=55) n 1 5 5 5 8 31 6 6 15 28 4 21 21 21 33 22 22 56 *D.V. means Domestic Vio l ence and CAN means Child Abuse and Neglect Thirty-four homicide cases are missing data for the item measuring the suspect ' s prior criminal , alcohol , drug abuse , and victimization history. The suspect ' s prior history of crime , and drug and alcohol abuse is not consistently documented in the Child Fatality Review Team data. Also , documentation confirming that a suspect has been a victim of either domestic violence or child abuse and neglect is not consistent. The second item, measuring the socioeconomic status of t h e victim ' s family is missing data for more than half of the state and city homicide 88 cases. 7 Victims ' Although the available data suggest that most families across the state are receiving full PUblic financial assistance , more than half of the cases are missing data for this item . Because of the large Volume of missing data for both of these risk factor items , each are dropped from the suspect profile analyses in this study . Also , with regard to homicide suspect data , four cases involve victims who were shot by the police while fleeing from the scene of a crime , and for eight cases a suspect was not identified by authorities at the time of data collection. Although , for all 82 homicide cases Child Fatality Review Team records identify whether the caretaker is the suspect , a specific suspect is not necessarily identified. As a result, when examining s uspect profile items , these twelve cases are eliminated from the analyses and identified as missing in both the tables and discussion. In terms of the prior abuse or neglect of victims and Siblings , data are missing for homicide victims residing outside the city of Baltimore. The prior abuse and 7Alth h th' i'tem is included in the documents . oug is . . th' ~aintained by the Child Fatality Review Team~ is l~formation is seldom completed by the agencies at the time of the child ' s death. 89 neglect of victims and their siblings is not consistently documented in the Chi ld Fatality Review Team records , when a homicide victim was not a resident of Baltimore city. State l evel chi ld abuse and neglect data are not available from the Department of Chi ld Protective services . Therefore , twenty homicide cases are missing data on victims ' prior experience of abuse or neglect, and nine homicide cases are missing data on sibling ' s prior abuse or neglect . Cases with missing prior abuse and neglect data are e liminated from the family risk factor analyses and identified as missing in the results tab l es and discussion. Items that have small numbers of missing data are included in the analyses , however, any case with missing data on a particular item, i s dropped from that portion of the analysis. Wh enever an item is mi ssing data , it is clear l y identified in the spec i fic table(s) and mentioned in the discussion . As noted in chapter 1 , the U.S . Cens u s data and the Child Abuse and Neglect data are for the city of Baltimore. Therefore , no interpretation of community level factors is made regarding homicide cases outside Baltimore city . Also , Ba l timore city child abuse and neglect data can only be compared with Baltimore city homicide data. 90 ~ection 5 . 1 - State and City Child Homi cide Data Results for the first phase of the child homicide analysis are based on a comparison between Baltimore city and all other State of Maryland city level homicide cases . The purpose of the analysis is to determine whether the characteristics of Baltimore city child homicides are different from child homicides in other cities in the State of Maryland. Data are examined with respect to two of the three risk factor levels . The victim and offender Profiles, and the victim ' s family characteristics are compared in this phase of the analysis . 8 Findings that show no major differences , provi de justification for using only the Baltimore city homicide cases in other phases of analyses for this study . S . 1 . 1 Individual Level Factors Victim Profi les The first group of items examined are the individual level factors that describe the victims ' profile . The 8Community level factors are €ntire state therefore , they are I ' final phase of the analysis . 91 not available for the only examined in the Table 6 Child Homicide Victim Profiles Item Other Cities Baltimore City (n==27) (n==55) ---- ------------ ----------- ----- - - ----- ~tim ' s Age n % n % < 10 l0 - 17 years 14 51 1 4 25 years 13 49 41 75 Gender --=-:::.=.. Male 16 60 46 84 Female 11 40 9 16 Race2 ------------=- Black 17 63 50 91 White 10 37 5 9 Asian Hispanic B' ~horder Position First Born 1 5 12 26 Middle Child 2 9 14 30 Last Born 2 9 2 4 Only Child 17 77 19 40 Missing data 5 8 items used to describe the victims' profile include age 10 , gende r , race , and birth order position. According to the Wh 9Although additional categories of race were included bl:n collecting the data, only two ~at~gories o~ race, h ~k and white , were found among victims of child omicide. i 1°Consistent with prior research, in that most antrafamilial victims are younger than ten years of age , t~~ ~ost extrafamilial victims a~e older than ten ye~rs , th item measuring age is coded 1nto two age categories Vi~~~ghout the study . Victims older than ten years, and ims less than ten years. 92 results presented in table 6, approximately 15 % of the Maryland homicide cases are missing data on the item measuring victim's birth order position. In comparing Baltimore city child homicide cases with 0ther Maryland child homicides, the results show a variation in the age of victims. Seventy-five percent of th e child homicide victims in Baltimore are between 10 and 17 years of age, the other 25 % of Baltimore victims are younger than 10 years of age. While the distributions for all other Maryland child homicide victims are split in half with regard to age; 51 % are younger than 10 years and 49 % are between 10 and 17 years. In terms of victims ' gender, males are predominately more victimized showing that 84 % of the Baltimore citY victims, and 60 % of all other Maryland victims are male. Racial percentages show that most victims are black, with a higher percentage for Baltimore victims at 91 %, while 63 % of other Maryland Nine percent of the remaining v · ictims are black. Balt' h't while 37 % of the other imore victims are w 1 e, Maryland victims are white. The last victim profile item in table the birth order position of the 6, measures homicide victim. The data shoW that there are some d'f and other Maryland child 1 ferences between Baltimore ho · · micide victims. The largest category statewide is for 93 Victims who are the only child. About 40 % of all Baltimore ci·ty , and 77 90· f 11 th · t· o a o er vic ims are only children. The smallest category for Baltimore victims is for last born children , at 4%, whereas the smallest category for other Maryland victims is for first born children at 5 %. Twenty- six percent of the Baltimore Victims are first born children , and 9 % of the other Maryland victims are last born children. Also, the category for middle children is somewhat different with 3 0 % of the Baltimore city victims , and 9 % of all other Victims. In summary, victim profiles appear to be similar on all items except age of the victim when comparing Baltimore city and all other State of Maryland victims . The majority of all victims are black, males, who are the 0 nly children in their family . However , most Baltimore Victims are older than 10 years of age , while all other Maryland victims are split equally with half who are between 10 and 17 years , and half who are younger than 10 Years of age. As noted in chapter 4 , table 2, 78 % of the Baltimore homicide cases involve extrafamilial homicides. The Baltimore victim profile is most typical of extra- familial homicide victims described in prior research. Also , all the other Maryland homicide cases are split 94 equally between both intrafamilial and extrafamilial homicides . In conformance with prior research , the distribu tions are typica l of intra - and ext r afami l ial homicide victim profi l es . Al though the victim profiles are similar across the State of Maryl and , there are fewer cases outside the city of Ba l t i more . Because 67 % of the Child homicide victims are in Baltimore city, the Percentage distributions across each risk factor item are much smaller for the victims located outs i de Ba l timore . Suspect Profiles The second group of items compared between Baltimore and other state of Maryland child homicide data are the items describing the suspects ' profile . The suspect Profi l e items include age , gender , race , and type of relationship between the suspect and victim . As noted earlier , the item measuring the suspect ' s criminal and Victimization history is dropped from the analysis because of missing data . Al so noted earl ier , twelve extrafamilial homic i de cases have missing data , because suspects have either not been identified by the police , or the case was a justifiable homicide . The results for this phase of the analysis are presented in table 7 · 95 Table 7 Child Homicide Suspect Profi l es * Item - - - - Age ----=--- 14 - 25 yrs 26 - 48 yrs Missing Gender --=-=-Male data Female Missing data Race -----=--Black White Asian Hispanic Missing data ~elation to Victim Natural Parent Foster/Step Parent Parent ' s Paramour Other Relative Babysitter Acqua i ntance Strangers Police Officer Missing data Other Cities (n=27) n % 15 60 10 40 2 21 4 2 19 6 2 7 1 4 2 9 2 1 1 85 15 76 24 27 4 15 8 34 8 4 Baltimore City (N=55) n 33 12 10 38 7 10 41 3 1 10 7 4 1 1 26 7 3 7 % 73 27 84 16 91 7 2 14 8 2 2 54 14 6 ~retaker to Victim No 13 52 43 78 Yes 14 48 12 22 *Missing Data _ 4 police shootings - no suspect data is needed also 7 Baltimore city and 1 other city cases do ' ' . . not have a specific suspect identified . 96 Overall , statewide child homicide suspects have similar profiles . In terms of age, most suspects st atewide are between 14 and 25 years of age . Seventy- three percent of the Baltimore suspects , and 60 % of the other Maryland suspects are under the age of 26 years . Only 27 % of the Baltimore suspects , and 40 % of the other Maryland suspects are older, specifically, between 26 and 48 Years of age. The percentages for gender are the same, With 84 % of the Baltimore suspects , and 85 % of the other Maryland suspects who are male . Racial distributions are also somewhat similar in that most homicide offenders are black. However, the Baltimore percentage is larger with 9 1 %, while only 76 % of the other Maryland suspects are black. Baltimore had fewer white suspects at only 7 %, While 24 % of the other Maryland suspects were white. Also , one Baltimore homicide case involved an Asian suspect. Distributions for each category describing the suspect ' s relationship with the victim were somewhat different . The friends and acquaintances category was the largest category with 54 % of the Baltimore, and 34 % of the other Maryland suspects. The second largest relationship categ h 27 g f the other Maryland suspects and 14 % ory, as o o of the Baltimore suspects who are natural parents to their 97 Victims. Al t ' d so , one case ou si e of Baltimore involved a foster parent , and one Balt i more case involved a relative other than a parent . More of the suspects outside Baltimore are the parent ' s paramour , only 8 % of the Baltimore , while 15 % of the other Maryland suspects are a Parent's paramour. The reverse is true with regard to st rangers , with 14 % of the Baltimore , and only 8 % of the other Maryland suspects who are a stranger to their Victim . In addition , four cases involved a babysitter, one in Baltimore , and.three outside of Baltimore . Four cases involved the police, three Baltimore and one case outside of Baltimore involved the police shooting the Victim who was fleeing from the scene of a crime . The last suspect profile item in table 7, identifies whether the suspect was the victim ' s caretaker at the time of death. Forty- eight percent of the suspects outside of Baltimore city, and 22 % of the Baltimore city suspects Were the victims ' caretaker at the time of death. In summary , the comparison of suspect profile items indicate that the majority of suspects statewide are black, males , and you~ger than 26 years of age . However , the victim and offender re l ationship varies when comparing b t d other Maryland city suspects. The e ween Baltimore an ma . suspects are acquainted or friends Jority of Baltimore 98 I with their victims, while the other Maryland suspects are equally the caretaker to the victim or a friend or acquaintance to the victim at the time of death. These findings suggest that the cases outside of Baltimore are equally distributed between intra - and extrafamilial child homicide. While the majori ty of Baltimore city child homicides are extrafamilial child homicides . Consistent with most extrafamilial homicide research , the findings show that most Baltimore suspects are black, males , between 14 and 25 years of age , and although acquainted with their victim, they are not typically the victim ' s caretaker. Also , in support of prior research , extra- familial homicide victims and suspects have similar individual level characteristics. 5 . 1 . 2 Family Level Factors The next group of items compared between Baltimore and other State of Maryland child homicide cases measure the victim ' s family characteristics . The following items describe the victims ' family , marital status of parents, number of siblings , victims ' prior child abuse or neglect, siblings ' prior child abuse or neglect, and the number of reported prior abuse and neglect incidents involving the 99 -·---- - - - --·- - -- - Victims or their siblings. I tems for prior child abuse or neglect of either the victim or their siblings measure all incidents prior to the death of the victim . As noted earlier , the item for-soc i oeconomic status is dropped from the analysis because of the large quantity of missing data . Al so , 18 cases outside of Baltimore city are missing data on prior abuse or neglect of victims , ands cases outside of Baltimore city are missing data on the Prior abuse or neglect of siblings . The results comparing fami l y characteristics are presented in tables 8 and 9. The results in table 8 show that patterns for the marital status of the victims ' parents are similar for all Maryland victims , with 92 % of the Baltimore city, and 77 % of the other Maryland victims living with a single parent at the time of death . 11 Twenty- three percent of the Victims outside Baltimore and 6% of the Baltimore city Victims ' parents are married; one Baltimore victim has a Widowed parent. The next item measures the size of the victims' family based on the number of children in addition to the 11Al though , for each case it was clearly. sp~cified in the Child Fatality Review Tea~ data that a victim was living with a single parent, it was uncle~r as to whether the parent was in fact a single parent , divorced , or legally separated from their spouse. 100 Table 8 Child Homicide Victim Family Characteristics Item --- - -- - - --- ~rital Sta~~~ - - -- -- - Married Single Widowed Missing data F . ~ily Size Victim Only 1 sibling 2 siblings 3+ siblings Missing data Other Cities (n=27) n % 6 23 20 77 1 17 3 1 1 5 77 13 5 5 Baltimore City (n - 55) n % 3 6 4 6 92 1 2 5 19 7 7 14 8 40 15 15 30 Victim. Table 8 shows that 40 % of the Baltimore victims f and 77 % of the other Maryland victims are the only Children in the family. About 13 % of the Maryland victims outside of Baltimore , and 15 % of the Baltimore victims have one sibling. Fifteen percent of the Baltimore and 5 % of the other Maryland victims have two siblings . Also , 3 0 % of the Baltimore victims have at least three or more Siblings , while only 5 % of the other victims have three or more children. In summary, although the percentage distributions are somewhat higher for Baltimore victims, most child homicide 101 victims were living with a single parent at the time of death. Also, the percentages for family size are somewhat different for Maryland victims when comparing with ore vic ims ou si e o Baltimore Baltimore ci· ty vi· cti· ms. M · t · t · d f city are the only children in the family , while more Baltimore city victims have three or more siblings. The results comparing Baltimore city and other state of Maryland victims with regard to prior child abuse or neglect of victims and siblings are presented in table 9 . As shown , the findings are very different for Baltimore city homicide victims versus other Maryland homicide victims when comparing the distributions of prior abuse or neglect of homicide victims and their siblings. None of the child homicide victims or their siblings , outside of Baltimore city, are shown to have experienced prior abuse or neglect. While 40 % of the Baltimore homicide victims, and 49 % of their siblings experienced prior abuse or neglect. The majority of all victims and their siblings who experienced prior abuse or neglect had three or more reported incidents. sixty-four percent of the Baltimore homicide victims experienced three or more incident s of abuse or neglect. seventy-eight percent of the Baltimore homicide siblings experienced three or more incidents of abuse or neglect. suggesting that when a child homicide 102 I I I I I I I I I I Table 9 Chi l d Homicide Victims ' and Siblings Prior Abuse and Neglect Item - - - --- - - ------- ---- - - Victim ' · . - s Prior History .£!._Abuse or Neglect Yes No Missing data ~er of Times Prior ~use\Neglect Reported 1 report 2 reports 3 or more Sibl' - ings Prior Abuse No Siblings Yes No Missing data ~ber of Times Siblings p . ~or Abuse 1 report 2 reports 3 or more Other Cities (n=27) n % 9 1 00 18 17 2 8 100 Baltimore City (n- 55) n 22 33 6 2 14 17 18 19 4 14 % 40 60 27 9 64 49 51 22 78 Victim or their sibling did experience prior abuse or neglect the violence was typically an ongoing problem in the household . In summary , the majority of child homicide victims in the State of Maryland were living with a single parent at the time of their death . However , when comparing between 103 I I I Baltimore . t d M l d . . . Cly an ary an victims outside of Baltimore , all other family characteristics are different . The majority of Baltimore city victims have one or more Siblings. Also , the major i ty of both Baltimore homicide Victims and their sibl ings experienced prior abuse or neglect , with most experiencing three or more incidents of Prior abuse or neglect. None of the homicide victims or their siblings outside of Baltimore experienced prior abuse or neglect , however , as noted earlier much of the data is missing with regard to prior abuse and neglect . 5 - 1 . 3 Community Level Factors State and city homicide data cou l d not be compared on community level factors. Items measuring community level factors were extracted from the U.S. Bureau of Census , and were only available for Baltimore city . Community level factors are analyzed in the third phase of analysis of this study . Summary . In summary, when comparing Baltimore city child homicide cases and all other state of Maryland child homicide victims by victim and suspect profil es , and Victims ' family characteristics , the patterns are somewhat 104 similar . However , because there are fewer cases outside the city of Baltimore , the percentages are much smaller than the Baltimore percentages. The results indicate that most child homicide incidents are extrafamilial homicides. The victim profile patterns are similar on all items except for age. Most Baltimore v i ctims are between 1 0 and 17 years , while all other victims are equally distributed between those under 10 years of age and those between 10 and 17 years. The patterns for suspect profile items are similar for all child homicide suspects in Maryland. The patterns for victims ' family items are with regard to the marital status of the victims ' parents . Although most victims are only children , the distribution for victims ' outside Baltimore are much larger . Also , the items measuring prior abuse or neglect of victims or their siblings are different. None of the victims or their siblings outside of Baltimore experienced abuse or neglect prior to the victim ' s death. Secti on 5.2 Intrafarnilial versus Extrafarnilial Child Homicide The second phase of this study involves a comparative analysis between the two categories of child homicide. 105 Intrafamilial and extrafamilial homicide are compared by each level of risk , individual , family, and community . The purpose of this analysis is to determine whether risk factor patterns are different for each type of homicide . State of Maryland child homicide data are u sed to compare intrafamilial and extrafamilial child homicide categories . The results are presented in tables 1 0 through 16 . The results show that there are 26 i ntrafamilial and 56 extrafamilial State of Maryland ch i ld homicide cases . 5 . 2 . 1 Individual level factors Research Question Ia . What specific individual level risk factors will be different between intrafarnilial and extrafamilial chi ld homicide incidents? Victim Profiles The first group of items compared between each category of homicide inc l ude the victim profile items . The results presented in table 10, show very different distributions in terms of the victim ' s age. Ninety- six percent of the intrafamilial victims are less than 10 years of age , whereas , 95 % of the extrafamilial victims are between 10 and 17 years of age . Also , when comparing gender and race the patterns are slightly different for 106 I I I I I I I I I I I I I I Table 10 Child Homicide Categories and Victim Profi l es Type of Homicide Item Intrafamilial Extrafamilial --- (n==26) (n==56) Victim ' s ; --- -------- ------------- ----- ------ ~ n % n ~l < 10 years 0 l 0 - 17 25 96 3 5 years 1 4 53 95 Gender ~ Male 17 65 46 82 Female 9 35 10 1 8 Race -----=- Black 16 62 51 91 Whit e 10 38 5 9 Asian Hispanic Other ~thorder Position First Born 3 12 10 23 Middle Child 8 30 8 1 9 Last Born 2 8 2 5 Only Chi ld 1 3 50 23 53 Missing data 13 each h . . omi cide category . Although black male victims represent the majority of victims in both homicide categories , the distributions vary. Males represent 82 % anct blacks 91 %, of the extrafamilial homi c ide victims , in comparison to 65 % male and 62 % black intrafamilial h omicide victims . Whi te victims are the only other racial racia l group represented in both homi c ide catego r i es , with 107 a higher percentage for intrafamilial at 38 %, while only 9 % of the extrafamilial victims are white . The final item used to describe the victims ' profile is birth order position of the victim . Previous studies suggest that birth order position is an important risk factor of child homicide victims. Prior research finds that intrafamilial victims are typically the last born children, and extrafamilial victims are typically the first born children (Smith , 1989) . The current study disputes most prior child homicide research. The results show that birth order position is similar across both homicide categories , with more than 50 % of the victims of both intra- and extrafamilial homicide the only children in the family. The category for middle children is the second largest for both types of homicide , JQ iJ; of the intrafamilial , and 19 % of the extrafamilial victims are middle children. The last born children make up the smallest birth order category, with only 8 % of the intrafamilial , and 5 % of the extrafamilial victims. Also , 23 % of the extrafamilial homicide victims are the first born children , and 12 % of the intrafamilial victims. In surrunary , the findings show two distinct patterns when comparing the type of child homicide and the age of the victim . Intrafamilial homicide victims are typically 108 less than 10 years of age, and extrafamilial victims are typically older than 10 years of age. However, when comparing all other victim profile items, the patterns are very similar for both types of homicides. The majority of all child homicide victims are black, males, and almost half are the only child in their family. To examine the relationship between intrafamilial and extrafamil ial chi ld homicide and the victim profile items, chi - square values are used to determine statistical significance. 12 As presented in table 11, critical chi - square values, and the necessary degrees of freedom are Table 11 Chi-Square Statistics - Child Homicide Type by Victim Profile It ems Item Age'' Gender Raced Birthorder Position 61.12*** 2.16 8.48** 2 .51 df 1 1 1 3 *p<.05 **p< .01 ***p< .001 levels of statistical significance dYates Corrected value ]? -When a crosstab cell has less than 5 frequencies, the chi-square value may be distorted (Levin and Fox, 1988). To correct for possible misleading results, a Yate's corrected formula is used for all 2 x 2 tables that contain any cell s with less than 5 frequencies. 109 reached at the . 0001 level of significance for age , and at the . 01 l eve l of significance for race . The items measuring gender and birth order position are not statistically significant . Based on the findings we can ass ume that there is a relationship between the items measuring age and race of the victim and each type o f child homicide . The results suggest that the proportion of victims unde r 10 years of age is greater among intrafamilial h omicide. Also , the proportion of vi c tims between 10 and 17 years of age is greater among extrafamil ial homicide . The proportion of black victims , rather than whit e victims i s greater for both intra- and extrafamilial h omicide. To examine these relationships furth er mul tipl e logi st i c regression models are analyzed (S ee Appendix A). Suspect Profi l es The next group o f items compared between intra- familial and e xtrafamil ia l h omi cide measure the s u s p ects profiles . As d iscu s sed earlier in thi s chapte r , twe lve c ases are missing suspect data, four involved vi c tims shot by t h e police , and e ight h ad no s uspect identified. The s e t we lve c a ses are dropped f r om thi s phase o f the a nalys i s . 110 Table 12 Child Homicide Categories and Suspect Profiles Item Caretaker to Victim Yes No Suspects ' Age 14 - 25 yrs 26-48 yrs Missing data Gender Male Femal e Mis s ing data Race Black White Asian Hispanic Missing data Relation to Victim Natural Parent Step/Foster Parent Parent' s Paramour Other Relative Babysitter Acquaintance St rangers Police Officer Missing data Type of Homicide Intrafamilial Extrafamilial n 26 8 18 18 8 17 9 n 15 1 6 1 3 111 (n=26) (n=56) 100 31 69 69 31 65 35 % 58 4 23 4 11 n 56 35 9 12 41 3 12 43 1 12 n 36 9 4 7 % 100 80 20 9 3 7 98 2 74 18 8 % As shown in table 12, a ll twelve cases with missing data are extrafamilial homicide cases , therefore results are based on 44 extrafamilial homicide cases, and 26 intrafamilial homicides. The results clearly establish the definitional boundaries between intrafamilial and extrafamilial homicides. All of the intrafamilial suspects are the victims' caretaker at the time of death, but none of the extrafamilial suspects are in a caretaking role. Most prior intrafamilial homicide research profiles the suspect as young , female , single parent, with more than one child to care for, and living below the poverty level (Jason and Andereck , 1983; schloesser et al., 1992). Whereas, most prior extrafamilial homicide research profiles the suspect as a black, male, adolescent , living below the poverty level (Goetting, 1990; Plass, 1993; Harries, 1993). In terms of comparing the age of the homicide suspect , the results in table 12 show that most of the intrafamilial suspects are older than extrafamilial suspects. sixty-nine percent of the intrafamilial homicide offenders , are between 26 and 48 years of age , whereas, SO % of the extrafamilial homicide offenders are between 14 and 25 years of age. The majority of both intrafamilial and extrafamilial homicide s uspects are 112 overrepresented by males , at 69 % for intrafamilial and 93 % for extrafamilial suspects. The findings also show that both types of homicide suspects are predominately black. The distribution for extrafamilial offenders is higher at 98 %, whereas , 65 % of the intrafamilial suspects are b l ack . An interesting pattern was found for regarding white suspects . None of the extrafamilial suspects are white , while 35 % of the intrafamilial suspects are white. Al so , only one suspect , an extrafamilial homicide suspect is Asian. The last suspect profile item describes the type of relationship between the suspect and the victim. The results in table 12 s how very different patterns for each type of homicide . More than half (58 %) of the intrafamilial homicide suspects are the victims ' natural parents , all but three of the natural parents are the father of the victim. Also , 4% (one) intrafamilial suspect is a step father , 23 % are the parent ' s paramour , 4 % or one suspect is a grandmother , and 11 % are the victims ' babysitters at the time of death. In comparison , 74 % of extrafamilial suspects are friends or acquaintances to their victim. Another 18 % are strangers and 8 % are po l ice officers who shot the victim (typically the victim was fleeing from a crime scene). Of the three incidents 113 of extrafami l ial homicide that involved more than one suspect , all are friends or acquaintances to the vict i m. In summary , there are two d i stinct suspect profiles for both intrafamilial and extrafamil ial homicide suspects . The current data indicate that intrafamilial homicide suspects are typically the natural father of the victim, between 26 and 48 years of age , and black. Whil e the majority of extrafamilial homicide suspects have a similar profile as their victim. The extrafamilial suspects are typically black , male , adolescents , between 14 and 25 years o f age , and are an acquaintance or friend of the victim. The items measuring the suspect ' s profile are examined further using chi-square statistics. Table 13 shows that the critical values of chi-square statistics , and the degrees of freedom were reached at both the p< . 001 levels of significance for four of the five suspect profile items. The item measuring gender reached the .05 l evel of significance ~ Based on these findings we can assume there is an association between both types of child homicide and each suspect profile item . The results suggest that the proportions of black , mal es who are older than 26 years of age , and a caretaker to the homicide victim, are greater among intrafami l ial homi cides . The 114 Table 13 Chi-Square Statistics - Child Homicide Type by Suspect Profil e Items Item x2 df -------------------- -------- Agea 14.41*** 1 Genderd 5.38* 1 Race 17.82*** 2 Relation to Victim 71.14*** 7 Care takerd 77.45*** 1 *p< . 05 **p< .01 ***p<.001 l eve l s of statistical s ignifican ce aYates Corrected valu e proportions of black , males, between 14 and 26 years of age , who are friends or an acquaintance t o the h omicide victim, are greater among extrafamilial h omicides. To examine these relation ships further multiple l ogistic regression models are analyzed (See Appendix A) . 5.2.2 Family Level Factors Research Question Ib. What specific family level risk factors will be different between intrafamilial and extrafamilial child homicide incidents? The n e xt group of items to be compared between intrafamilial and extrafamilial homi c ide incidents , describe the vict im' s family characteristics . The result s are presented in tables 14 and 15 . As noted earlier in 11 5 this chapter , the item measuring socioeconomic status of the victim ' s family is dropped from the analysis because of missing data . Table 14 presents the resu l ts for marital status of the victim ' s parents , .and the size of the victim ' s family , based on the number of chi l dren. Marital status is similar for both intra- and extrafamilial homicide cases. Most victims are living with single parents regardless of the type of homicide. Seventy- seven percent of the intrafamilial and 92 % of the extrafamilial victims l ived with single parents . Intrafamilial victims have a higher rate of married parents at 23 %, while only 6% of the extrafamilial victims' parents are married . Also , one extrafamilial victim has a widowed parent . In comparing the size of the victim ' s family , the distributions are similar for both homicide categories . Most child homicide victims are the only children in the fami l y , with 52 % of the i ntrafami l ia l, and 53 % of the extrafamilial victims. The second largest fami l y size category for both types of homicide , is victims with more than three s iblings , with 22 % intrafamilial and 21 % of the extrafamilial homicide victims ' families . Also , 15 % of the intrafamilial and 14 % of the extrafamilial victims have one sibling . Fina l ly , 11 % of the intrafamilial 116 Table 14 Child Homicide Categories and Victim Family Characteristics Item Marital Status Married Single Widowed Missing data Family Size Victim Only 1 sibling 2 siblings 3+ siblings Missing data Type of Homicide Intrafamilial Extrafamilial n 6 20 13 4 3 6 (n=26) (n=56) % 23 77 52 15 11 22 n 3 46 1 6 23 6 5 9 13 6 92 2 53 14 12 21 victims and 12 % of the extrafamilial victims have two siblings. Overall , the victim's family size is the same for both types of homicide. In table 15, the results for victims' prior history of child abuse and neglect show differences between each type of homicide. Intrafamilial homicide victims were more likely to experience abuse or neglect before their death. Fifty-six percent of the intrafamilial, while only 28 % of the extrafamilial victims experienced abuse and/or neglect at least once before death. However, for all homicide victims, regardless of type, who experienced 117 Table 15 Child Homi c ide Categor i es and Victims ' and Siblings Prio r Abuse and Ne glect Type of Homicide Intrafamilial Extrafamilial Item Prior History of n Abuse or Neglect* Yes 9 No 7 Missing data 10 Number of Times Prior Abuse\Neglect Reported* 1 report 4 2 reports 3 or more 5 S iblings Prior Abuse No siblings Yes No Missing data Number of Times Sibl ings Prior Abuse Reported 1 report 2 reports 3 or more 1 3 7 4 2 1 6 (n=26 ) (n=56) % 56 44 44 56 63 27 14 86 n 13 33 10 2 2 9 21 11 17 7 2 9 28 72 15 1 5 70 39 61 1 8 82 prior abuse or n eglect, most experienced three or more incidents. At least 56 % of all intrafamilial , and 70 % of all extrafamilial homicide victims experienced abuse and/or n egl ect three or more times . Another 44 % of t h e intrafamilial victims experienced o n e incident of abuse or neglect prior to their d e ath. While only 1 5 % o f the 11 8 extrafamilial victims experienced only one incident , and another 15 % experienced two incidents of prior abuse or neglect. Overall , i f a victim was abused or neglected there were typically more than three incidents . However , more of the intrafamilial victims , at least 25 % more , were victimized prior to death. Also in table 15 are the results for prior child abuse and neglect of the victims ' siblings. It should be noted that almost half of all the homicide victims are the only children in the family. The results are very different for each category of homicide , twice as siblings experienced abuse or neglect prior to the death of an intrafamilial homicide victim. For all of the intra - familial victims ' with siblings , 63 % of those siblings were abused or neglected. However , for all of the extrafamilial victims with siblings , only 39 % of their siblings were abused or neglected. The number of child abuse and neglect reports for siblings were similar for both categories of homicide . The majority of all siblings abused or neglected experienced three or more incidents, with 86 % of the intrafamilial , and 82 % of the extrafamilial homicide siblings. In summary , the results comparing family level risk factors show that victim ' s family characteri s tics are 119 similar with regard to the mar i tal status of the victim ' s parents , and the size of the victim ' s family. Both intrafamilial and extrafamilia l homi cide victim ' s families are typically single parent fami l ies , and the victim is the only child . However , when comparing prior abuse or neglect of both victim ' s and their siblings , there are large differences. Intrafamilia l homicide victims and their siblings experienced more abuse and neglect prior to the victim ' s death , than the extrafamilial victims. However , for all victims or siblings abused or neglected, they typically were abused or neglected three or more times , regardless of the homicide category . To examine further , the re l ationship between intrafamilial and extrafamilial homicides and the family characteristics that include, parent ' s marital status , size of family , and prior history of child abuse and neglect , chi - square values were analyzed for statistical significance. As shown in table 16 , the critical values of chi - square, and the necessary degrees of freedom were reached at the .05 level of significance for two items , marital status of victims' parents and prior abuse and neglect of the victim . Suggesting there is a relationship between these two items and child homicide . The results suggest that the proportion of single parents and prior 120 Table 16 Chi - Square Statistics - Child Homi c ide Type by Victim ' s Family Leve l Factors Item Marital Status" Family S i ze Prior Vic CAN" Prior Sib CAN" 3.28* . 13 2 . 93* . 11 df 1 3 1 1 *p< . 05 **p< . 01 l eve ls of statistical significance dYates Corrected value abuse or n eg l ect o f victims are greater among intra - familial homicide, and the proportion of single parents i s greater among extrafamilial homi c ide s . None of the o ther items wer e s ignifi cant suggesting that the relative frequencies of family size and prior histo ry of abuse a nd neglect of s iblings do not differ for intrafamilial and extra fami lia l ch ild homicide s . To examine these relationships furt h e r multipl e l og i st ic regression models are analyzed (See Appendix A) . 5 . 2 . 3 Cormnuni ty level factors Research Questi on Ic. What specifi c conununi ty level r i sk factors will be different between intrafamilial and extrafami lial child homicide i ncidents? Community l eve l data were onl y avai labl e for the city of Baltimo r e , Maryland . The r e for e , it wa s not possible to 121 compare community leve l factors across the state of Community l evel factors are only analyzed in t h e next phase of ana l ysis ; comparing c i ty child homicide Maryl and . and chi l d abuse and neg l ect data. Summary. In summary , when comparing intrafamilial and extrafamilial child homicide data by victim and suspect profiles, and victims ' family characteristics , the patterns of risk are different. Intrafamilial homicide victims are younger , black , males , and about half are the only child in the familY · The intrafamilial homicide Perpetrator is a caretaker , usually a parent , the victims ' parents are generally single , and both victims and siblings have most l i ke l y been abused or neglected prior to the v i ctim's death . Extrafamilial homicide victims are older , between 10 and 17 years of age , black , males, and more than half are the onlY child in the family. The extrafamilial homicide perpetrator often resembles their victim ' s profile , they are typically black , males , between 14 and 25 years of age. Most homicide victims in general, are from a single parent familY· However , extrafamilial v i ctims and their siblings have not typical ly experienced as much prior abuse or neglect , as intrafamilial victims 122 and their siblings . Based on the differences found between each homicide category, the next phase of analysis will examine each type of child homicide compared with child abuse and neglect data. 5.3 Child Homicide versus Child Abuse and Neglect The next phase of this study compares the characteristics of Baltimore city level child homicide and child abuse and neglect data. The purpose of this analysis is to determine whether there are different risk factor patterns between child homi cide and abuse and neglect incidents, based on three levels of risk, individual, family , and community . Based on earlier findings that show risk differences between intrafamilial and extrafamilial homicide incidents, each homicide category will be compared separately with child abuse and neglect incidents. The basic assumption is that both types of child homicide and child abuse and neglect incidents will have different patterns, across all three levels of risk . through 23. The findings are presented in tables 17 Only Baltimore city child homicide data are compared with Baltimore city child abuse and neglect data. Both 123 homicide types are compared wi th child abuse and neglect inc i dent s . There are a tota l of 55 ch i l d homicide and 210 child abuse and neglect victims d i spersed throughout Baltimore c ity . Of the total number of homicides , 12 are defined as intrafamilial homicides and 43 are defined as extrafamilial h omicides. 5 . 3 . 1 Individual Level Factors Research Question rra . Victim Profi le Are there different individual level risk factors between child homicide, and child abuse and neglect v i ctims? The first group of items compared across each category of child homicide and child abuse and neglect , de scribe the victim ' s profile. With regard to age , the r es ults show that extrafamilial and c hild abuse and neglect vi ct ims have similar patterns . The results in table 17 show that all of the intrafamilial vi c tims are les s than 5 years of age , while 98 % of the extrafamilial and 60 % of the child abuse and neglect victims are between 10 and 17 years . While mo s t homicide victims are mal e , with 88 % of the ext rafamilial and 67 % of the intra - familial , only about half of all the abuse and neglect victims are mal e . Th e majority o f all homicide and abuse 124 Table 17 Child - Homicide and Child Abuse and Neglect Victim Profiles Type of Homicide Item Intrafami lial Extrafamilial Abuse/Neglect --- (n=12) (n=43) (n=210) ---Viet · , ----- --- - --------- ------------- ------------- -----..c im s Age n % n % n % < 5 years 12 100 27 13 5 - 9 yrs 1 2 57 27 l0-17 yrs 42 98 126 60 Gender --=---=-=-=-Male 8 67 38 88 107 51 Female 4 33 5 12 103 49 Race ---=-Black 10 83 40 93 182 86 White 2 17 3 7 26 13 Asian 2 1 B' irthorder ~ 25 10 24 70 irst Born 3 33 Middle Child 7 58 7 13 39 19 Last Born 10 24 59 28 Only Child 2 17 16 39 42 20 and neglect victims are black, with distributions ranging between 83 to 93 %. Also shown in table 17 are comparisons between homicide and abuse and neglect regarding the birth order Position of victims . Patterns f or birth order position Vary across each type of homicide as well as for abuse and neglect victims. Birth order categories for abuse and negl ect victims are distributed across each birth order 125 category . Whereas, most intrafamilial victims are middle children , and most extrafamilia l victims are on l y chi l dren . Eighty-six percent of the intrafamilial homicide victims are middle children , but on ly 13 % of the extrafamilial victims , and 19 % of the abuse and neglect victims. Thirty-nine percent of the extrafamilia l victims are the only children , 20 % of the abuse and neglect, and only 17 % of the intrafamilial victims . The largest birth order category for abuse and neglect victims is for first born children , with 33 %, but only about 25 % of both types of homicide victims are first born children . None of the intrafamilial victims are the last born children , while 24 % of the extrafamilial and 28 % of the abuse and neglect victims are last born chi ldren . In summary , the findings for victim profile items indicate there are differences between both types of child homicide and abuse and neglect victims . Other than race , none of the individual items are similar for either type of homicide and abuse.and neglect victims. Abuse and neglect victims tend to be older than 10 years , black , equally male or female , and of no one particular birth order position. Intrafamilial victims are all less than 5 years of age , black , male, and middle born children. Extrafamilial victims are primarily older than 10 years of 126 - age, black, male, and the only child in their family. Further analyses are performed to establish the relationship between homicide and child abuse and neglect with regard to the items measuring the victim profile . In examining these relationships further , chi-square values are comput ed . The results in table 18 show that for age , gender , and birth order po s ition of the vi ct ims , the computed c hi - squares are l arger than the required cr itical chi - square values . Each item is statistically Table 18 Chi-Square Statistics - Child Homicide and Child Abuse and Ne glect by Victim Profile Items Item Agea Gendera Race Birthorder Position X2 18.59* ** 17 . 76*** 1. 02 17.63*** df 1 1 2 3 *p<.05 **p< .01 ***p<.001 levels of s tati st ical s ignificance dYa t es Corrected value significant at the .001 level o f signifi cance . Based on the finding s we can assume that there is a r e lationship between these victim profile items and child homicide and c hild abuse and neglect. The results suggest that t h e relative frequencies of age , gender , and birthorder 127 intrafami l ial and extrafamilial Position di'ffer for · · The intrafamilial homicide , and child abuse and neglect . Viet · im profile is a young , under 5 years of age , male , who pically the middle child . The typica l extrafamilial i s ty . horn· icide victim profile is male , between 10 and 17 years , is the only child in the family . The child abuse and Who · neglect victim i s typically between 10 and 17 years , the lrst born child , and equally a male or female. To f ' examine these relationships further multiple logistic ssion model s are analyzed (See Appendix A). regre . Susp t ec Profile The next group of items compared, describe the suspects ' profiles. rn comparing age , the results in table 19 show a similar pattern for both intrafamilial homicide and child abuse and neglect suspects. Sixty- seven percent of the intrafamilial and 76 % of the abuse and neglec t suspects are older than 26 years o f age . Whereas , extrafamilial homicide suspects are more typically less than 26 years of age , at 88 %. With regard to gender and race , overall , mo st suspects are black , males - Ninety- four percent of the extrafamilial and 70 % of the abuse and neglect victims are 128 Table 19 Child Homicide and Chi ld Abuse and Neglect Suspect Profiles Type of Homicide Intrafamilia l Extrafamilial Item (n=l2) (n=43) Suspects ' Age 14-25 yrs 26+ yrs Missing data Gender Male Female Missing data Race n 4 8 7 5 Black 9 White 3 Asian Missing data Relation to Victim Natural Parent 6 Foster Parent Step Parent Parent Paramour 3 Other Relative 1 Babysitter 2 Acquaintance Strangers Police Officer Missing data % 33 67 58 42 75 24 50 26 8 1 6 129 n 29 4 10 31 2 10 32 1 10 23 11 2 7 % 88 12 94 6 97 3 63 31 6 Abuse/Neglect (n=2 10) - - -- - n 49 152 9 146 64 182 26 2 165 2 6 25 12 -- ------ % 24 76 70 30 87 12 1 78 1 3 12 6 male , while fewer intrafamilial suspects , with 58 % are male. Also , 75 % of the intrafamilial , 97 % of the abuse and neglect , and 87 % of the extrafamilia l suspects are black. The remaining 24 % of the int r afamilial and 12 % of the abuse and neglect suspects are white . None of the extrafamilial suspects are white , but one is Asian . Also presented in table 19 are resu l ts for victim and s uspect re l ationships . The findings show similarities for both intrafamilial and abuse and negl ect s uspects , in that most s uspects are the victims ' natural parents . Fifty percent of the intrafami lial and 78 % of the abuse and neglect suspects are the natura l parents of the victim . Other typi ca l relati onships for both intrafamilial and child abuse and negl ec t are the parent ' s paramour (most often the mother ' s b oyfriend) , other relatives , and babys itters. For intrafamili al s uspects , 26 % are the parent's paramour , 16 % are the babysitter , and another 8 % are other relatives of the victim. Three percent of the abuse and negl ect suspects are the parent ' s paramour , none are the babysitter , a nd 12 % are other relatives of the victim, also 6 % are friends or acquaintances of the victim. Whil e the most common victim and offender relationships in extrafamilial h omicides are friends and acquaintances to the victim, with 63 %. Th e other 31 % are strangers to the 130 victim victim crime. I I and two incidents involved the police shooting a because the victim was fleeing the scene of a In summary, the findings show there are similar or ot intra ami ia omicide and child SUspect profi' les f b h · f · 1 · 1 h · · e and neglect incidents. f or both ou tcomes the most abus ty . pical suspect profile is a black biological father , an 26 years of age . However , the suspect profile older th extrafamilial homicide incidents is different when for compared with child abuse and neglect. The most typical extrafamilial homicide suspect is a black , male , friend or aintance to the victim, under 26 years of age. acqu · Further examination of the suspect profile items were analyzed using chi - square values. As shown in table 20 , th e required critical chi - square values , and the necessary degrees of freedom are reached at the .001 level of significance for four of the five items . suggesting there is a ' d d relationship between the suspects age , gen er , an relationship with the victim and both categories of homicide and child abuse and neglect incidents . The results suggest that the relative frequencies of age , gender, and relationshiP with victim, differ between extrafamilial homicide and the typical intrafamilial homicide and child abuse and neglect suspect. Both the 131 Table 20 Chi - Square Statistics - Chil d Homi cide and Chi ld Abuse and Neglect by Suspect Profil e I tems Item x2 df - ------- - - - - - --- - - - - - ------- - - Age 38 . 93*** 1 Gender 44.97*** 1 Race 1. 66 2 Relationship 135.3 1*** 7 Caretaker 11 1. 45*** 1 *p< . 05 **p< . 01 ***p< . 00 1 l eve l s of statist i ca l signif i cance intrafami lial homicide and child abuse and neglect suspect profiles include males, who are the bio l ogica l father , and older than 26 years of age . The typica l extrafamilial homic i de s u spect profi l es are of ma l es , between 1 4 and 25 years , who are a friend or acquaintance to the victim . To examine t h ese relationshi ps furt her mul tiple l ogistic regression models are analyzed (See Appendi x A) . 5 . 3.2 Family Level Factors Research Question IIb. Are there different family level risk factors between child homicide, and child abuse and neglect incidents? The next group of items compared between each homicide type and abuse and negl ect describe the victim ' s family characteristics . The results are reported in 132 GI tables 21 and 22. As noted earlier , one item, cioeconomic status of the victim's family , was dropped so . from the analysis because of missing data. The results in table 21 show that the item measuring the marital status e victims' parents is similar for al l three outcomes. of th . . MoS t victims are living in single parent households regardless of whether they are a victim of intrafamilial or extrafamilial homicide , or child abuse and neglect. y - two percent of both intrafamilial and extrafamilial Ninet Viet· ims were living in single parent households , and 86 % of the child abuse and neglect victims were living with a single parent. With the exception of one extrafamilial Victim and one child abuse and neglect victim who were iving with a widowed parent , the remaining victims 1 · . Parents were all married. Distributions comparing the size of the victim's fam · 1 1 l Y, based on the number of children , are a so Presented in table 21 . The results show that family size is somewhat similar for all types of victims , most families are large , with more than three children. More than 30 % of the child abuse and neglect victims have three or more siblings, while 26 % of the extrafamilial and 41 % of th . . have three or more siblings. e intrafamilial victims 133 Table 21 Child H . . ~ omicide and Child Abuse and N 1 t - eg ec Victim Family Type of Homi cide Item Intrafamilial Extrafamilial Abuse/Neglect --- --- (n=l2) (n=43) (n=210) Mar· - - ---- - - - -- - ----- ---- ----- - - - - - -- ----- - - -----~ Married n % n % n % Single 1 8 2 5 27 13 Widowed 11 92 35 92 182 86 1 3 1 1 Missing data 5 F'amily s· ~ Vic~im Only 2 17 17 48 41 19 1 sibling 2 17 5 14 60 29 2 'b si lings 3 25 4 12 41 19 3+ . b .si lings 5 41 9 26 68 33 --!:!_1.ssing data 8 On1 y 19 % of the abuse and neglect and 17 % of the int raf ami· 1i· al · l h · ld homicide victims are on y c i ren. While twice as many (48 %) of the extrafamilial victims are the only children. Slightly less than one- third (29 %) of the abuse and neglect victims have only one sibling, while 17 % Of the · lntrafamilial and 14 % of the extrafamilial victims have only f th b d one sibling. Another 19 % o ea use an neglect victims have two siblings, while 25 % of the int rafamilial and 12 % of the extrafamilial victims have two . Slblings. 134 Table 22 Child H . ~ omicide and Child Abuse and N l t of Victims d s·b1· eg ect Prior Abuse and an i ings Item Type of Homicide Intrafamilial Extrafamili'al Ab ; use Neglect (n=12) (n=43) (n=210) ~-- % n % Yes No n 9 3 % 75 25 n 13 30 30 70 118 92 56 44 Prior c · l AN Reports report ~ reports or more 4 5 Siblings Yes Prior CAN No 7 No · 3 siblings 2 l?~ior CAN Sib1· Reports 1 report ~ reports or more l 6 44 56 48 25 17 14 86 2 2 9 13 13 17 2 11 15 15 70 30 30 40 15 85 20 21 77 124 45 41 18 29 77 17 18 65 59 21 20 15 23 62 Distributions for prior child abuse and neglect of v· ictims and their siblings are presented in table 22. The findings show that, at 75 %, intrafamilial homicide Viet · ims experienced the highest percentage of prior child abuse or neglect. of the 75 %, slightly more than half expe . rienced three or more incidents of abuse or neglect Prior to death. The percentage for child abuse and neglect . . victims who experienced prior abuse and neglect 135 ig tly lower at 56 %. Of the 56 %, approximately was s1· h sixty- five percent experienced prior abuse or neglect three or more times. Only 30 % of the extrafamilial homicide victims experienced prior abuse or neglect. However f , or those who did experience abuse or neglect, 0 xperienced the abuse or neglect three or more times. 70 ~ e . Intrafamilial and child abuse victims have somewhat a use an neg ec , owever , the similar patterns of pri·or b d l t h percentage of prior abuse and neglect for extrafamilial homicide victims is much lower than that of other victims. With regard to the prior abuse or neglect of sib l ings, table 22 shows a slightly higher percentage for siblings of current child abuse and neglect victims. Fifty · 'bl' f t h'ld b - nine percent of the s1 1ngs o curren c 1 a use and neglect victims experienced abuse or neglect, of those , 65 % experienced the abuse or neglect three or more times. The siblings of intrafamilial homicide victims experienced abuse or neglect at a rate of 48 %, of those, 8 6% experienced abuse or neglect three or more times. The lowest percentage of siblings who experienced abuse or neglect . 1 . 1 homi· ci'de victims. Thirty are extrafami 1a Percent of the siblings of extrafamilial homicide victims experienced abuse or neglect , of those , 62 % experienced ~buse or neglect three or more times. In summary , the results show that the majority of all homicide and abuse and neglect victims are from single Parent families . Most extrafamilial victims are the only Children, while most intrafamilial and abuse and neglect v· lctims have at least two siblings. The results suggest intrafamilial and abuse and neglect victims have similar family profiles regarding prior abuse and neglect of Victims and their sibl ings. Although , extrafamilial homicide victims have slightly lower levels of prior abuse and neglect compared with other types of victims , their Siblings experience a similar rate of prior abuse and neglect as both intrafamilial and child abuse and neglect. To further examine the relationships of the victims' family characteristics , chi - square values were analyzed for statistical significance . As shown in table 23 , critical values of chi - square , along with the necessary degrees of freedom were reached at .05 level of significance for the item measuring the victim ' s family size , and at the .01 level of significance for the item measuring the prior abuse or neglect of siblings . These findings suggest that there is a relationship between the size of the family and prior abuse or neglect of siblings and child homicide and child abuse and neglect incidents . The results suggest that the proportion of children in the 137 Table 23 . Ch' ~-Square Statistics Child Homicide and Child Abuse and ~ect by Victim ' s Family Profile Items Item --..::.:.: --- Marital Statusa Family Size Prior CAN v· t· p . lC lm x2 1 . 25 7.85* 2.81 5 . 76** df 1 3 1 1 ;-rior CAN Siblings di<.os **p<.01 levels of statistical significance ates Corrected value Victims' family is greater for intrafamilial homicide and Child abuse and neglect families , than for extrafamilial homicide. Also, the results suggest that the proportion of Prior abuse or neglect of siblings is greater for intrafamilial homicide and child abuse and neglect incidents . Most extrafamilial homicide victims are the only child, indicating fewer incidents of prior abuse or neglect of siblings. The results show no evidence that the relative frequencies for the other two items , marital s tatus of the victims ' parents and prior abuse or neglect of the victim, differ for child homicide and child abuse and neglect. To examine these relationships further multiple logistic regression models are analyzed (S ee Appendix A). 138 5. 3. 3 rmnuni ty Level Factors Co · Re sear h c Questions I I c . Are there different conununity level risk factors between child homicide, and child abuse and neglect incidents? The la s t group of items compared between homicide and and neglect , are the community l evel factors that abuse desc · ribe the victims' community (or census tract) acteristics . The 1990 U.S . Bureau of Census data are char fort he city of Baltimore , Maryland . All of the homicide a use and neglec t victims reside in the urban center anct b of B altimore city . The descriptive results are reported able 24. There are four items included in the int community l eve l factors, each of these items i s mea s ured aggregate level variable , rather than an individual as an The firs t item is the percentage of households beaded by . s i ngle parents in communities where homicide or abuse and neglect victims reside. None of the victims' level variabl e . co:mmuniti·es rate higher than SO %. The had a percentage di s tributions for this item show that 100 % of the intrafamilial victims ' communities , and 98 % of both the extrafami'li'al . d b and neglect communities had and ch1l a use less than 25 % of their househol ds headed bY a single Parent . the percentage of The second item measures Table 24 Ch'l ~ d Homicide and Comm . Child Abuse and Neglect Victim ' s - unity Level Risk Factors Type of Homicide Item Intrafamilial Extrafamilial Abuse/Neglect (n=l2) (n=43) (n=210) ---g_ s. - - - --- - - - ---- ------- --- --- - - - -- -- - -- ~ % n n % n % 25 % 26-50 % 12 100 42 98 207 98 1 2 3 2 %Pave .~ s1-1~o r6 12 100 42 98 185 88 1 2 25 12 %Dncter 18 ~ 26- 50 % 8 67 24 56 109 52 4 33 19 44 101 48 '6 N ~ 0- 50 9. 51 - 1~0 % 8 67 36 58 98 47 4 33 26 42 112 53 Poverty in a community. The results show that the Percent ages across all three victim categories are One hundred percent of the int r afamilial , 98 % of extrafamilial , and 88 % of the child abuse and neglect irict · J.ms' communities had less than 50 % of their residents similar. the liv · J.ng b elow the poverty level in 1990. The third item measuring the percentage of the population under 18 years Of a ge is slightly higher for intrafamilial victims , but Simil ar for extrafamilial and child abuse and neglect 140 Victims. None of the victim's communities had a Percentage rate higher than 50 % of the population. Sixty- seven percent of the intrafamilial communities , 56 % of the extrafamilial, and 52 % of the child abuse and neglect communities had less than 25 % of their residents under 18 Years of age. The last item measures the percentage o f the nonwhite population in a community. The results show similar findings for all three outcomes. Sixty-seven Percent of the intrafamilial, 58 % of the extrafamilial and 4 7 % of the c hild abuse and neglect victims reside in communitie s wh e r e less than 50 % of the population are nonwhite. These findings indicate that most victims reside in culturally mixed communities. In summary, the community level factors show similar Patterns across both categories of homicide and child abuse and negl e ct. Most children who are vi c tims o f Violence reside in fairly mainstream c ulturally mix e d communities. victims' communities have a high rate o f residents living below the poverty level, most communities have 25 to 50 % of their residents living below the pove rty level. Als o , most homicide and abuse and n e gl e ct vi c tims reside in c ommunities with fewer than 25 % of the r es ident s residing in single family households. income , h o us e h o ld types, and cultural diversity. Finally, mos t vi c tims o f 141 homicide and abuse and neglect reside in communities where less than 25 % of the population is under 18 years of age. To further examine these relationships chi-square Values were analyzed for statistical significance. As shown in table 25 , critica l values of chi - square , along With the necessary degrees of freedom were reached at the .OS level of significance for one of the four items measuring the victim's community. The results for the Percentage of the population living below the poverty level suggest that the relative frequencies slightly differ for child homicide and child abuse and neglect. The results suggest that the proportion of the percentage of the population living below the poverty level is greater for child abuse and neglect victims. There is no evidence that the relative frequencies of the percentage Table 25 Chi - Square Statistics - Child Homicide and Abuse and Neglect Incidents by Community Level Risk Factors Item - %SingleHeadHouse %Povertya %Under 18 %Nonwhite 1.63 3.94* .69 .02 df 1 1 1 1 *p< .05 **p< .01 leve ls of statistical significance aYates Corrected value 14 2 of the population who are nonwhite , percentage of the Population under 18 years, or percentage of the population li . . ving in a single headed household differ from child homicide and child abuse and neglect. To examine these relationships further multiple logistic regression models are analyzed (See Appendix A) . ~ection 5.4 Causes and Circumstances of Death or Injury The final phase of the analysis compares the causes and c ircumstances of both child homicide and child abuse and neglect incidents. The purpose of this analysis is to determine whether the characteristics of the causes and c ircumstances of death or injury are different for child homicide and child abuse and neglect. The data are examined using crosstabulations to measure the causes and circumstances of death and injury, based on victim ' s age , race, and gender. section 5.4 is divided into two subsections, one to explain the causes of death and injury, and the other to explain the circumstances of death and injury . state level child homicide and city child abuse and neglect data are used for this phase of the analysis. 143 5 . 4 . 1 Causes of Death and Injury Research Question !!Ia . Child H · . om1c1de Incidents What are the unique causes of child homicide and child abuse and neglect incidents? I n examining the causes of death in statewide child homi · Clde cases , the data presented in table 26 , show that 66 Q. f b O all child homicide v i ctims died as a result o f gunshot wounds . When compari ng the causes of death across each type of homicide there are large variations . The results show that gunshot wounds are the largest category, at 8 6 % for extrafami l ial homicide victims. However , only 19 % of the intrafamil i al victims died as a result of Table 26 Child Homicide categories by Causes of Death ----------------=---~~~~--Type of Homicide A l l Int r afamilial Extrafamilial Item (n=82) (n =26) (n=56) --- ----- - -- ---- - --- -- - ------- - -C ---------- ------~ % n % n % n 49 8 6 BUnshot wounds 54 66 5 19 F'~aten , abused 8 10 7 27 l 2 sire, scalding 7 9 5 19 2 4 · trangula tion/ 16 2 4 asphyxia 6 7 4 Stab Wounds 6 3 11 2 4 Malnutrition/ 5 Neglect 2 2 2 8 144 gunshot wounds. The largest cause of death category for intraf · i· ami ial victims , at 27 % was for fatal physical assaults. Other large categories for intrafamilial Viet· ims included, 19 % who died as a result of arson, 16% Were strangulated, 11 % died as a result of stab wounds, a nd 8 % died from malnourishment or dehydration . None of th e extrafamilial victims died from malnourishment or dehydrat· ion , however , 2 % died from physical abuse . Extraf · 1 · · d lt f ami ial victims also die as a resu o arson , stab Wound h h t t 4 ° s , and strangulation, wit eac ca egory a 6 . Distributions for causes of death also varied when controlling for age , as shown in table 27. Older children Were more likely to die from gunshot wounds , at 87 % while 25 9- 0 Were less than ten years . The findings support most Table 27 Child Homicide Victim's Age by Causes of Death All Victims