Abstract
Objectives. We examined variations in children’s risk of an unintentional or intentional fatal injury following an allegation of physical abuse, neglect, or other maltreatment.
Methods. We linked records of 514 232 children born in California from 1999 to 2006 and referred to child protective services for maltreatment to vital birth and death data. We used multivariable Cox regression models to estimate variations in risk of fatal injury before age 5 years and modeled maltreatment allegations as time-varying covariates.
Results. Children with a previous allegation of physical abuse sustained fatal injuries at 1.7 times the rate of children referred for neglect. Stratification by manner of injury showed that children with an allegation of physical abuse died from intentional injuries at a rate 5 times as high as that for children with an allegation of neglect, yet faced a significantly lower risk of unintentional fatal injury.
Conclusions. These data suggest conceptual differences between physical abuse and neglect. Findings indicate that interventions consistent with the form of alleged maltreatment may be appropriate, and heightened monitoring of young children referred for physical abuse may advance child protection.
In the United States in 2011, 6.2 million children were referred to child protective services (CPS) for abuse or neglect.1 When the demographic profile of referred children is compared with that of the general population, it becomes clear these children face multiple individual,2–4 family,5–7 and community8,9 risk factors. Among children referred to CPS, however, the accurate identification of those for whom the threat is most immediate and consequential has proven difficult.10 High rates of maltreatment rereferrals among children with initially unfounded allegations,11–13 and child maltreatment deaths despite CPS involvement,1 point to the challenge of accurately assessing children’s current and future risk of abuse and neglect.
One possible indicator of the nature and severity of the physical threat faced by a child is the type of alleged maltreatment. Nationally, more than three quarters of children are referred to CPS for neglect (78.5%); far fewer are referred as possible victims of physical abuse (17.6%).1 Despite high rates of concurrence between maltreatment types,14–17 it is notable that physical abuse is alleged for only a minority of children. Because of the ambiguity surrounding what constitutes child neglect (broadly defined as acts of parental omission that endanger children),17 an allegation of physical abuse for a child younger than 5 years may be a more reliable marker of safety concerns that necessitate CPS intervention. In other words, although many children referred for neglect may also experience varying degrees of physical abuse, if the physical abuse is so severe or chronic that it is explicitly alleged, this may be an important signal of risk.
Public health researchers use variations in rates of death as population-based indicators reflective of broader group disparities in health, safety, and well-being.18,19 Similarly, variable rates of fatalities among children previously referred for maltreatment may provide a means of differentiating among high-risk subsets of children. Although death is a relatively rare event, group differences in fatality rates suggest variable exposures to antecedent risk factors. Previous research indicates that children with a history of CPS referrals have an increased risk of death by injury and other causes.20–24 We examined whether children previously referred for physical abuse had an increased risk of both unintentional and intentional fatal injury compared with children referred for neglect and children referred for other forms of maltreatment.
METHODS
We linked birth records, administrative CPS records, and death records from California to create the analysis data set. We obtained confidential vital birth and death records from the California Department of Public Health and administrative CPS records from the California Department of Social Services. Death records provided information concerning child fatalities, CPS records allowed us to examine the timing and form of alleged maltreatment, and birth records permitted adjustments for a range of risk factors present at birth and associated with both CPS involvement and death.
Record Linkages
We matched records with the Link Plus probabilistic linkage software program version 2.0 (Registry Plus, Atlanta, GA), which employs a statistical model for weighting the likelihood that 2 records represent the same individual.25,26 Following an extensive review of probabilistically generated record pairs, we set lower-bound and upper-bound thresholds for declaring a given record pair a match. We then clerically reviewed record pairs that fell between the lower- and upper-bound thresholds to assign final match status.27–29
In the first linkage phase, we assigned death records from 1999 to 2007 to a file of deaths involving children born between 1999 and 2006 who died before their fifth birthday (n = 25 987). We then linked these death records to birth records for all children born in California between 1999 and 2006 (n = 4 317 216). Overall, we successfully matched 98.1% of death records to a birth record. In a separate linkage, we extracted CPS records involving children born between 1999 and 2006 and referred for maltreatment before age 5 years (n = 596 962) from the state’s administrative data system and linked them to birth records. In total, we successfully matched 86.6% of children referred for maltreatment to a birth record. Unmatched CPS records reflected children who were born in another state and then referred for maltreatment in California, as well as children for whom incomplete information prevented a successful match. To reduce the possibility of identifying a biased subset of children with more serious CPS referrals resulting in death, we established no direct CPS–to–death record linkages. Further details regarding the linkages underlying this data set are available elsewhere.3,21,30
Variables
Dependent variable.
The dependent variable in our analysis was time to injury death among children referred for maltreatment. We classified deaths from injury according to International Classification of Diseases, 10th Revision external cause-of-death codes as reported in death records.31 We included all mechanisms of fatal injury (e.g., drowning, fall, assault). We modeled the manner of injury (i.e., unintentional, intentional, undetermined) jointly and then stratified into deaths that were unintentional and intentional. To capture age dynamics, we coded children’s entry into the study by the date of birth; we coded maltreatment referrals as time-varying covariates such that children were considered to be at risk for death only after the first referral was received. We censored observations when no death occurred by December 31, 2007, or the child’s fifth birthday, whichever came first. We censored children who died from other noninjury causes (e.g., disease, cancer) on the date of death.
Independent variable.
We classified alleged maltreatment into 3 categories: physical abuse, neglect, and other maltreatment. We derived maltreatment type from established administrative codes used by CPS workers to enter data into the state’s case management system. Physical abuse is a unique administrative code; the neglect category comprised maltreatment codes for both general and severe physical neglect, as well as caretaker absence or incapacity. Our other maltreatment category consisted of a seemingly heterogeneous group of 3 allegation types: emotional abuse, sexual abuse–exploitation, and abuse of a sibling. We aggregated these to a single category both because stratified analyses demonstrated a consistently low risk of death and because the small number of injury deaths associated with these allegation types restricted our ability to generate stable estimates.
We entered maltreatment types into our models as time-varying covariates. Although we followed all children from birth, a given child only entered into a hierarchically organized maltreatment risk set (i.e., physical abuse, neglect, other maltreatment) beginning on the date of the first allegation to CPS. We included a child in the maltreatment risk set corresponding to this first allegation type until an allegation higher in the hierarchy was received, at which point we reclassified the child to a new maltreatment risk set. For example, a child who was first referred to CPS for an allegation of neglect at 6 months of age and was later referred for physical abuse at age 2 years would fall into the neglect risk set from the date of the neglect allegation (age 6 months) until the date of the abuse allegation (age 2 years); this child would then be moved to the physical abuse risk set from age 2 years until either death or the end of the observation window.
We incorporated all maltreatment allegations, not only those that were investigated or substantiated. We made this decision because of research highlighting the unreliability of the substantiation process as a means of distinguishing maltreated from nonmaltreated children.11–13,32 In addition, the question for our analysis was not whether a child was previously identified as a legally substantiated victim of maltreatment, but whether among children referred to CPS, the form of alleged maltreatment signaled a different level of physical risk, as reflected by subsequent fatal injury. Consistent with our interest in maltreatment type as a marker of future harm, we examined only children whose first referrals to CPS were nonfatal.
Other covariates.
To control for possible confounding attributable to differences in child and family risk factors associated with specific forms of maltreatment, we adjusted all multivariable models for 8 birth record covariates: child gender (male vs female), child health (risk present vs not present), health insurance used for the birth (public vs private), maternal race/ethnicity (Black vs White, Hispanic vs White, Asian/Pacific Islander vs White, Native American vs White), maternal age (modeled as a continuous variable), maternal education (≤ high school vs ≥ some college), paternity (missing vs established), and birth order (later born vs first born). We coded a child as having a health risk if 1 or more birth abnormalities were recorded or birth weight was less than 2500 grams. In California, mothers not covered by any health insurance at the time of birth are retroactively enrolled in the state’s Medicaid (public health insurance) program. Rates of missing variables for all covariates were low, ranging from 0.4% (child gender) to 2.2% (maternal education). We also included a covariate for the number of maltreatment referrals.
Analysis
We computed descriptive statistics for the full population of children with a first (nonfatal) referral of maltreatment to CPS; we also present the characteristics of the full population of live births in California as a point of reference. For each birth record covariate, we used the χ2 test to assess (1) whether children fatally injured following a referral to CPS differed significantly from those who were not fatally injured and (2) whether among fatally injured children, there were significant covariate differences by manner of fatal injury (i.e., intentional vs unintentional).
We then fit 3 multivariable Cox proportional hazard regression models. For our Cox models, we examined maltreatment type as a predictor of (1) all manners of fatal injury, (2) intentional fatal injury, and (3) unintentional fatal injury. We reported Cox estimates as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). HRs measure how often a particular event occurs in 1 group (e.g., physically abused children) compared with a reference group (e.g., neglected children) over time. An HR of 1 indicates that the incidence of death between the 2 groups does not differ over time; an HR of 2, for example, means that the incidence of death in the first group is twice that of the reference group.
For our multivariable models, we evaluated the proportional hazard assumption both graphically and according to Schoenfeld residuals.33 We included injury fatalities of undetermined intent (n = 22) in descriptive and multivariable examinations of all injury deaths but excluded them from manner-of-death comparisons. We conducted all analyses with Stata version 12 (StataCorp LP, College Station, TX).
RESULTS
Table 1 summarizes the sociodemographic characteristics of the overall population of children born in California between 1999 and 2006 (n = 4 317 216) and the characteristics of children referred to CPS for maltreatment before age 5 years who were successfully linked to a birth record (n = 514 232). Children referred for maltreatment had a unique sociodemographic profile. Larger shares of referred children were covered by public health insurance (66.7% vs 43.5%), had a health risk (14.0% vs 10.7%), and had no established paternity (24.0% vs 9.3%). Referred children were also more likely to have mothers who were younger than 25 years (49.1% vs 32.9%), had no more than a high school degree (78.8% vs 57.3%), were Black (13.2% vs 6.0%), and had previously given birth (70.9% vs 61.2%).
TABLE 1—
Descriptive Statistics for Live Births, Children Referred for Maltreatment, and Subsequent Intentional and Unintentional Fatal Injuries Before Age 5 Years: California, 1999–2007
| Nonfatal First Referrals to Child Protective Services |
Fatal Injuries by Mannerb |
||||||
| Variable | Live Births 1999–2006 (n = 4 317 216), % | No Fatal Injury (n = 513 840), No. (%) | Fatal Injury (n = 392), No. (%) | Pa | Intentional (n = 123), No. (%) | Unintentional (n = 247), No. (%) | Pa |
| Allegation type (most severe) | <.001 | <.001 | |||||
| Physical abuse | … | 86 029 (16.7) | 95 (24.2) | 60 (48.8) | 26 (10.5) | ||
| Neglect | … | 330 536 (64.3) | 275 (70.2) | 60 (48.8) | 203 (82.2) | ||
| Other maltreatment | … | 97 677 (19.0) | 22 (5.6) | 3 (2.4) | 18 (7.3) | ||
| Gender | <.001 | .247 | |||||
| Male | 51.1 | 263 390 (51.3) | 233 (59.4) | 69 (56.1) | 154 (62.4) | ||
| Female | 48.9 | 250 428 (48.7) | 159 (40.6) | 54 (43.9) | 93 (37.7) | ||
| Health risk | .014 | .427 | |||||
| Present | 10.7 | 72 148 (14.0) | 72 (18.4) | 24 (19.5) | 40 (16.2) | ||
| Not present | 89.3 | 441 656 (86.0) | 320 (81.6) | 99 (80.5) | 207 (83.8) | ||
| Insurance for birth | .054 | .627 | |||||
| Public | 43.5 | 341 016 (66.7) | 276 (71.3) | 84 (68.9) | 174 (71.3) | ||
| Private | 56.5 | 170 246 (33.3) | 111 (28.7) | 38 (31.1) | 70 (28.7) | ||
| Birth order | .69 | .145 | |||||
| First born | 38.8 | 149 401 (29.1) | 110 (28.2) | 40 (32.5) | 62 (25.3) | ||
| Later born | 61.2 | 363 593 (70.9) | 280 (71.8) | 83 (67.5) | 183 (74.7) | ||
| Maternal age, y | .003 | .041 | |||||
| < 25 | 32.9 | 252 448 (49.1) | 222 (56.6) | 78 (63.4) | 129 (52.2) | ||
| ≥ 25 | 67.1 | 261 392 (50.9) | 170 (43.4) | 45 (36.6) | 118 (47.8) | ||
| Maternal education | .334 | .702 | |||||
| ≤ high school | 57.3 | 394 099 (78.8) | 308 (80.8) | 99 (82.5) | 194 (80.8) | ||
| ≥ some college | 42.7 | 105 918 (21.2) | 73 (19.2) | 21 (17.5) | 46 (19.2) | ||
| Maternal race/ethnicity | <.001 | .067 | |||||
| Black | 6.0 | 67 968 (13.2) | 79 (20.2) | 32 (26.0) | 42 (17.0) | ||
| White | 31.3 | 154 959 (30.2) | 135 (34.4) | 33 (26.8) | 94 (38.1) | ||
| Latino | 50.1 | 260 460 (50.8) | 156 (39.8) | 48 (39.0) | 100 (40.5) | ||
| Asian/Pacific Islander | 11.8 | 23 721 (4.6) | 8 (2.0) | 4 (3.3) | 3 (1.2) | ||
| Native American | 0.8 | 6122 (1.2) | 14 (3.6) | 6 (4.9) | 8 (3.2) | ||
| Paternity | <.001 | .297 | |||||
| Missing | 9.3 | 123 151 (24.0) | 126 (32.1) | 45 (36.6) | 77 (31.2) | ||
| Established | 90.7 | 390 689 (76.0) | 266 (67.9) | 78 (63.4) | 170 (68.8) | ||
Note. Numbers may not sum to reported total because of missing data for some variables. Allegation type is coded to reflect the most serious allegation made to child protective services. Allegations are organized hierarchically with physical abuse at the top (most serious–severe) and other maltreatment at the bottom (least serious–severe). Only birth record variables are reported for the full population of live births; allegation type is excluded as indicated by ellipses.
Determined by the χ2 test.
Excludes fatal injuries of undetermined intent (n = 22).
Fatal Injuries
Subsequent to a first referral for maltreatment, 392 children were fatally injured before age 5 years. A majority of fatal injuries were unintentional (63.0%). Fatally injured children differed significantly (P < .05) from other referred children across all sociodemographic covariates, with the exceptions of health insurance (P = .054), birth order (P = .69), and maternal education (P = .334). When we stratified fatal injuries by manner of injury, only maternal age differences emerged as significant (P < .05).
Relative to other referred children, a significantly larger share (P < .001) of those who were fatally injured had previously been referred for physical abuse (24.2% vs 16.7%) or neglect (70.2% vs 64.3%) rather than other forms of maltreatment (5.6% vs 19.0%). Differences in the distributions of maltreatment type by the manner-of-injury death also emerged (P < .001). Children referred to CPS for neglect were more likely to have died from unintentional rather than intentional injuries (82.2% vs 48.8%). A larger share of children with a previous allegation of physical abuse had a fatal injury classified as intentional rather than unintentional (48.8% vs 10.5%).
Multivariable Models
Table 2 presents adjusted HRs and accompanying 95% CIs for our 3 multivariable Cox models. When we modeled all fatal injuries, children with a previous allegation of physical abuse died at a rate 70% greater than that of children referred for neglect (HR = 1.70; 95% CI = 1.34, 2.17).
TABLE 2—
Multivariable Cox Proportional Hazard Models Estimating Children’s Risk of Intentional or Unintentional Fatal Injury Before Age 5 Years: California, 1999–2007
| Variable | All Injury Deaths (n = 392), HR (95% CI) | Intentional Injury Deathsa (n = 123), HR (95% CI) | Unintentional Injury Deathsa (n = 247), HR (95% CI) |
| Allegation type (most severe) | |||
| Physical abuse | 1.70 (1.34, 2.17) | 5.22 (3.61, 7.57) | 0.59 (0.39, 0.90) |
| Neglect (Ref) | 1.00 | 1.00 | 1.00 |
| Other maltreatment | 0.27 (0.17, 0.42) | 0.18 (0.05, 0.56) | 0.30 (0.18, 0.49) |
| Referral count | 0.90 (0.85, 0.97) | 0.92 (0.83, 1.03) | 0.91 (0.84, 0.99) |
| Gender | |||
| Male | 1.36 (1.11, 1.67) | 1.15 (0.80, 1.65) | 1.56 (1.20, 2.03) |
| Female (Ref) | 1.00 | 1.00 | 1.00 |
| Health risk | |||
| Present | 1.32 (1.01, 1.71) | 1.45 (0.92, 2.30) | 1.09 (0.76, 1.55) |
| Not present (Ref) | 1.00 | 1.00 | 1.00 |
| Insurance for birth | |||
| Public | 1.17 (0.93, 1.48) | 0.95 (0.64, 1.42) | 1.20 (0.89, 1.61) |
| Private (Ref) | 1.00 | 1.00 | 1.00 |
| Birth order | |||
| First born (Ref) | 1.00 | 1.00 | 1.00 |
| Later born | 1.46 (1.13, 1.88) | 1.32 (0.85, 2.03) | 1.59 (1.14, 2.20) |
| Maternal age | 0.96 (0.94, 0.98) | 0.95 (0.92, 0.99) | 0.97 (0.95, 0.99) |
| Maternal education | |||
| ≤ high school | 0.95 (0.72, 1.25) | 1.07 (0.64, 1.77) | 0.96 (0.68, 1.36) |
| ≥ some college (Ref) | 1.00 | 1.00 | 1.00 |
| Race/ethnicity | |||
| Black | 1.15 (0.86, 1.54) | 1.75 (1.05, 2.89) | 0.93 (0.64, 1.35) |
| White (Ref) | 1.00 | 1.00 | 1.00 |
| Latino | 0.65 (0.51, 0.83) | 0.80 (0.50, 1.27) | 0.60 (0.45, 0.81) |
| Asian/Pacific Islander | 0.43 (0.21, 0.88) | 1.85 (0.30, 2.40) | 0.24 (0.08, 0.76) |
| Native American | 2.55 (1.47, 4.43) | 4.52 (1.89, 10.84) | 2.04 (0.99, 4.21) |
| Paternity | |||
| Missing | 1.30 (1.04, 1.62) | 1.54 (1.04, 2.27) | 1.28 (0.96, 1.70) |
| Established (Ref) | 1.00 | 1.00 | 1.00 |
Note. CI = confidence interval; HR = hazard ratio.
Excludes injury fatalities of undetermined intent (n = 22).
A consistent but more extreme picture emerged for intentional fatal injuries. A previous allegation of physical abuse was associated with a more than 4-fold increase in a child’s risk of an intentional fatal injury relative to children referred for reasons of neglect (HR = 5.22; 95% CI = 3.61, 7.57). When we modeled unintentional injury deaths, however, children referred for physical abuse were significantly less likely to die than children referred because of neglect (HR = 0.59; 95% CI = 0.39, 0.90). Across all 3 models, children referred for other types of maltreatment had a significantly reduced risk of death from injury (P < .05) relative to both neglected and physically abused children (data comparing other maltreatment to physical abuse not shown).
DISCUSSION
We identified differences in fatal injury risk among infants and young children referred to CPS for maltreatment. We first examined all manners of fatal injury. We based our decision to look at all injury deaths—rather than only those that were coded as intentional—on literature suggesting a notable underascertainment of intentional injuries stemming from maltreatment in death records,34–38 coupled with evidence highlighting the preventability of unintentional fatal injuries.39,40 We found that children with a previous referral of physical abuse were fatally injured at rates significantly higher than those of children referred for neglect or other forms of maltreatment.
We then restricted our analysis to injury deaths classified as intentional, reflecting a subset of fatalities conceptually consistent with the physical risk suggested by an allegation of abuse and the most unambiguously preventable. When we modeled intentional injuries, the heightened risk associated with a previous allegation of physical abuse was even more pronounced. Finally, we examined unintentional injury fatalities—a manner of death that may logically stem from acts of omission frequently characterized as child neglect (including inadequate supervision and caregiving)—and found that children referred for physical abuse faced a significantly lower risk of unintentional fatal injury than children referred for neglect.
The increased risk of death from injury following an allegation of physical abuse has been suggested in data from other sources,1,22,23 although it has received little discussion. Our findings indicate that an allegation of physical abuse involving a child younger than 5 years signals a significantly greater risk of fatal injury, particularly intentional injury, than does an allegation of neglect or other maltreatment. Although many neglected children may be victims of unreported physical abuse,14,17 an allegation of physical abuse emerged in these data as a prospective marker of a child’s risk of fatal injury, perhaps reflecting a different level of either chronic abuse exposure or acute abuse severity.
Although the exact etiology of risk differences is unknown, the nonrandom distribution of fatal injuries across maltreatment types suggests that the 16.7% of children in our data who were at some point referred to CPS for physical abuse were differentially exposed to parental or environmental hazards associated with a fatal injury, particularly an intentionally inflicted fatal injury. In other words, if one assumes that it is largely random whether an inflicted injury proves fatal on the first, fifth, or 20th strike, then the significantly higher rate of intentional injury fatalities observed for children with a previous referral of physical abuse indicates that, on average, these children also experience more frequent nonfatal assaults. If one instead assumes that less severe assaults escalate to an eventual fatality, these data still suggest that children with an allegation of physical abuse are uniquely exposed to risks associated with an inflicted fatal injury. Either way, if the goal is to advance child protection through the identification of children at greatest physical risk, relative differences in rates of overall injury fatalities, and specifically intentional injury fatalities, point to the potential for targeted work with the proportionately small subset of young children referred for physical abuse.
Strengths and Limitations
Ours was the first study to prospectively follow children known to CPS to ascertain differences in risk of fatal injury by maltreatment type and manner of death and with covariate adjustments. Although a previous population-based analysis used these linked data to identify a referral to CPS as an independent predictor of fatal injury,21 no analyses to date have examined fatality risk differences by maltreatment type. Despite the strengths of these linked, longitudinal data, and the prospective design that was employed, several aspects of our analysis must be considered when interpreting findings.
First, during the creation of the linked data set, 14% of CPS records meeting linkage inclusionary criteria were not matched to a birth record. Although a large share may have been true false negatives (i.e., children born in another state), we missed some matches because of incomplete information in the underlying records.
Second, we chose to code maltreatment allegations hierarchically, reflecting our interest in examining specific maltreatment allegation types as a means of differentiating physical risk among young children. This epidemiological approach was consistent with our research objectives but may have obscured the cumulative or potentially interacting effects of concurrent maltreatment.41,42
Third, undercounts of intentional injury fatalities have been well established in the literature. We cannot rule out the possibility that children with a previous allegation of neglect experienced injury fatalities that manifested in manners subject to systematic classification errors that were specific to that group. However, we also modeled all injury deaths and found that the overall risk of fatal injury was significantly greater for children referred for physical abuse.
Fourth, we restricted our analysis to referrals and injury fatalities among children younger than 5 years. Our findings address only the relationship between covariates, maltreatment type, and injury fatalities during the first 5 years of life. These dynamics may differ among older children and adolescents. Likewise, the risk associated with a given maltreatment type may be quite different in an examination of noninjury fatalities.
Finally, the use of data from California restricts the generalizability of findings. Referral practices and CPS policies differ across states, as do legal guidelines for what constitutes the maltreatment subtypes we examined.1 Although cautious extrapolations are warranted, we used the broadest possible definition of CPS contact—any referral of maltreatment, including referrals screened out over the phone without investigation—to minimize the influence of local policy variations in CPS mandates for accepting or substantiating maltreatment referrals. Furthermore, we posit that an allegation of physical abuse is inherently less ambiguous than an allegation of neglect and is therefore a more consistently credible marker of physical safety risks for young children. As such, we expect that referrals for physical abuse are subject to fewer definitional variations across jurisdictions.
Conclusions
Data from California indicate that children with a previous allegation of physical abuse faced a significantly greater risk of an intentional fatal injury before age 5 years than did children referred for neglect, but a lower risk of unintentional fatal injury. These findings are consistent with conceptual understandings of physical abuse as an act of commission and neglect as an act of omission.
Injury fatality differences suggest that targeted CPS practices and policies for investigating, monitoring, and serving the proportionately small number of young children for whom physical abuse has been alleged may advance efforts to protect children from physical harm. Meanwhile, attempts to reduce the absolute number of child injury deaths must focus on the broader population of neglected children and incorporate strategies to prevent unintentional injuries.
Acknowledgments
This study was supported by the Maternal and Child Health Research Program, Maternal and Child Health Bureau (Title V, Social Security Act), Health Resources and Services Administration, Department of Health and Human Services (grant R40MC25689). The linkage of underlying data sources was funded by the Harry Frank Guggenheim Foundation.
We thank the California Department of Social Services, the Stuart Foundation, and Casey Family Programs for their support of the California Child Welfare Indicators project. We also acknowledge Jill Duerr Berrick, Neil Gilbert, Sophia Rabe-Hesketh, Nick Jewell, and Joseph Magruder for their feedback on previous iterations of this analysis.
Human Participant Protection
This study was approved by the California committee for the protection of human subjects, the institutional review board at the University of California, Berkeley, and was reviewed by the California Vital Statistics Advisory Board.
References
- 1. Administration for Children and Families. Child maltreatment 2011. Available at: http://www.acf.hhs.gov/sites/default/files/cb/cm11.pdf. Accessed March 14, 2013.
- 2.Wu SS, Ma CX, Carter RL et al. Risk factors for infant maltreatment: a population-based study. Child Abuse Negl. 2004;28(12):1253–1264. doi: 10.1016/j.chiabu.2004.07.005. [DOI] [PubMed] [Google Scholar]
- 3.Putnam-Hornstein E, Needell B. Predictors of child protective service contact between birth and age five: an examination of California’s 2002 birth cohort. Child Youth Serv Rev. 2011;33(11):2400–2407. [Google Scholar]
- 4.Sullivan PM, Knutson JF. Maltreatment and disabilities: a population-based epidemiological study. Child Abuse Negl. 2000;24(10):1257–1273. doi: 10.1016/s0145-2134(00)00190-3. [DOI] [PubMed] [Google Scholar]
- 5.Parrish JW, Young MB, Perham-Hester KA, Gessner BD. Identifying risk factors for child maltreatment in Alaska: a population-based approach. Am J Prev Med. 2011;40(6):666–673. doi: 10.1016/j.amepre.2011.02.022. [DOI] [PubMed] [Google Scholar]
- 6.Theodore AD, Chang JJ, Runyan DK, Hunter WM, Bangdiwala SI, Agans R. Epidemiologic features of the physical and sexual maltreatment of children in the Carolinas. Pediatrics. 2005;115(3):e331–e337. doi: 10.1542/peds.2004-1033. [DOI] [PubMed] [Google Scholar]
- 7.Berger LM. Income, family characteristics, and physical violence toward children. Child Abuse Negl. 2005;29(2):107–133. doi: 10.1016/j.chiabu.2004.02.006. [DOI] [PubMed] [Google Scholar]
- 8.Coulton CJ, Korbin JE, Su M, Chow J. Community level factors and child maltreatment rates. Child Dev. 1995;66(5):1262–1276. [PubMed] [Google Scholar]
- 9.Drake B, Pandey S. Understanding the relationship between neighborhood poverty and specific types of child maltreatment. Child Abuse Negl. 1996;20(11):1003–1018. doi: 10.1016/0145-2134(96)00091-9. [DOI] [PubMed] [Google Scholar]
- 10.Gilbert R, Kemp A, Thoburn J et al. Recognising and responding to child maltreatment. Lancet. 2009;373(9658):167–180. doi: 10.1016/S0140-6736(08)61707-9. [DOI] [PubMed] [Google Scholar]
- 11.Drake B, Jonson-Reid M, Way I, Chung S. Substantiation and recidivism. Child Maltreat. 2003;8(4):248–260. doi: 10.1177/1077559503258930. [DOI] [PubMed] [Google Scholar]
- 12.Kohl PL, Jonson-Reid M, Drake B. Time to leave substantiation behind: findings from a national probability study. Child Maltreat. 2009;14(1):17–26. doi: 10.1177/1077559508326030. [DOI] [PubMed] [Google Scholar]
- 13.Drake B. Unraveling “unsubstantiated.”. Child Maltreat. 1996;1(3):261–271. [Google Scholar]
- 14.Ney PG, Fung T, Wickett AR. The worst combinations of child abuse and neglect. Child Abuse Negl. 1994;18(9):705–714. doi: 10.1016/0145-2134(94)00037-9. [DOI] [PubMed] [Google Scholar]
- 15.Turner HA, Finkelhor D, Ormrod R. Poly-victimization in a national sample of children and youth. Am J Prev Med. 2010;38(3):323–330. doi: 10.1016/j.amepre.2009.11.012. [DOI] [PubMed] [Google Scholar]
- 16.Jonson-Reid M, Drake B, Chung S, Way I. Cross-type recidivism among child maltreatment victims and perpetrators. Child Abuse Negl. 2003;27(8):899–917. doi: 10.1016/s0145-2134(03)00138-8. [DOI] [PubMed] [Google Scholar]
- 17.Mennen FE, Kim K, Sang J, Trickett PK. Child neglect: definition and identification of youth’s experiences in official reports of maltreatment. Child Abuse Negl. 2010;34(9):647–658. doi: 10.1016/j.chiabu.2010.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hajat A, Kaufman JS, Rose KM, Siddiqi A, Thomas JC. Long-term effects of wealth on mortality and self-rated health status. Am J Epidemiol. 2011;173(2):192–200. doi: 10.1093/aje/kwq348. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Putnam-Hornstein E. Preventable injury deaths: a population-based proxy of child maltreatment risk in California. Public Health Rep. 2012;127(2):163–172. doi: 10.1177/003335491212700205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Barth RP, Blackwell DL. Death rates among California’s foster care and former foster care populations. Child Youth Serv Rev. 1998;20(7):577–604. [Google Scholar]
- 21.Putnam-Hornstein E. Report of maltreatment as a risk factor for injury death: a prospective birth cohort study. Child Maltreat. 2011;16(3):163–174. doi: 10.1177/1077559511411179. [DOI] [PubMed] [Google Scholar]
- 22.Jonson-Reid M, Chance T, Drake B. Risk of death among children reported for nonfatal maltreatment. Child Maltreat. 2007;12(1):86–95. doi: 10.1177/1077559506296722. [DOI] [PubMed] [Google Scholar]
- 23.Sabotta EE, Davis RL. Fatality after report to a child abuse registry in Washington State, 1973–1986. Child Abuse Negl. 1992;16(5):627–635. doi: 10.1016/0145-2134(92)90101-v. [DOI] [PubMed] [Google Scholar]
- 24.Sorenson SB, Peterson JG. Traumatic child death and documented maltreatment history, Los Angeles. Am J Public Health. 1994;84(4):623–627. doi: 10.2105/ajph.84.4.623. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Division of Cancer Prevention and Control. Link Plus User’s Guide, Version 2.0. Atlanta, GA: Centers for Disease Control and Prevention; 2007. [Google Scholar]
- 26.Campbell KM, Deck D, Krupski A. Record linkage software in the public domain: a comparison of Link Plus, the Link King, and a “basic” deterministic algorithm. Health Informatics J. 2008;14(1):5–15. doi: 10.1177/1460458208088855. [DOI] [PubMed] [Google Scholar]
- 27.Herzog TN, Scheuren FJ, Winkler WE. Data Quality and Record Linkage Techniques. New York, NY: Springer; 2007. [Google Scholar]
- 28.Winkler WE. Overview of Record Linkage and Current Research Directions. Washington, DC: US Census Bureau; 2006. [Google Scholar]
- 29.Clark DE. Practical introduction to record linkage for injury research. Inj Prev. 2004;10(3):186–191. doi: 10.1136/ip.2003.004580. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Putnam-Hornstein E, Webster D, Needell B, Magruder J. A public health approach to child maltreatment surveillance: evidence from a data linkage project in the United States. Child Abuse Rev. 2011;20(4):256–273. [Google Scholar]
- 31.World Health Organization. International Statistical Classification of Diseases and Related Health Problems. 10th revision. Geneva, Switzerland: World Health Organization; 1992. [Google Scholar]
- 32.Hussey JM, Marshall JM, English DJ et al. Defining maltreatment according to substantiation: distinction without a difference? Child Abuse Negl. 2005;29(5):479–492. doi: 10.1016/j.chiabu.2003.12.005. [DOI] [PubMed] [Google Scholar]
- 33.Cleves M, Gould WW, Gutierrez RG, Marchenko YV. An Introduction to Survival Analysis Using STATA. 3rd ed. College Station, TX: Stata Press; 2010. [Google Scholar]
- 34.Crume TL, DiGuiseppi C, Byers T, Sirotnak AP, Garrett CJ. Underascertainment of child maltreatment fatalities by death certificates, 1990–1998. Pediatrics. 2002;110(2):e18. doi: 10.1542/peds.110.2.e18. [DOI] [PubMed] [Google Scholar]
- 35.Schnitzer PG, Covington TM, Wirtz SJ, Verhoek-Oftedahl W, Palusci VJ. Public health surveillance of fatal child maltreatment: analysis of 3 state programs. Am J Public Health. 2008;98(2):296–303. doi: 10.2105/AJPH.2006.087783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Ewigman B, Kivlahan C, Land G. The Missouri Child Fatality Study: underreporting of maltreatment fatalities among children younger than five years of age, 1983 through 1986. Pediatrics. 1993;91(2):330–337. [PubMed] [Google Scholar]
- 37.McClain PW, Sacks JJ, Froehlke RG, Ewigman BG. Estimates of fatal child abuse and neglect, United States, 1979 through 1988. Pediatrics. 1993;91(2):338–343. [PubMed] [Google Scholar]
- 38.Herman-Giddens ME, Brown G, Verbiest S et al. Underascertainment of child abuse mortality in the United States. JAMA. 1999;282(5):463–467. doi: 10.1001/jama.282.5.463. [DOI] [PubMed] [Google Scholar]
- 39.Rimsza ME, Schackner RA, Bowen KA, Marshall W. Can child deaths be prevented? The Arizona Child Fatality Review program experience. Pediatrics. 2002;110(1):e11. doi: 10.1542/peds.110.1.e11. [DOI] [PubMed] [Google Scholar]
- 40.Schnitzer PG, Ewigman BG. Household composition and fatal unintentional injuries related to child maltreatment. J Nurs Scholarsh. 2008;40(1):91–97. doi: 10.1111/j.1547-5069.2007.00211.x. [DOI] [PubMed] [Google Scholar]
- 41.Edwards VJ, Holden GW, Felitti VJ, Anda RF. Relationship between multiple forms of childhood maltreatment and adult mental health in community respondents: results from the Adverse Childhood Experiences study. Am J Psychiatry. 2003;160(8):1453–1460. doi: 10.1176/appi.ajp.160.8.1453. [DOI] [PubMed] [Google Scholar]
- 42.Teicher MH, Samson JA, Polcari A, McGreenery CE. Sticks, stones, and hurtful words: relative effects of various forms of childhood maltreatment. Am J Psychiatry. 2006;16(6) doi: 10.1176/ajp.2006.163.6.993. 3:993–1000. [DOI] [PubMed] [Google Scholar]