BACKGROUND AND OBJECTIVES:

Child maltreatment (CM) is recognized as a major public health concern, and an important number of children suffer injuries related to abuse and neglect that result in death. We sought to identify risk factors for CM fatalities among hospitalized children that can provide clinicians with information to recognize at-risk children and reduce further death.

METHODS:

In this study, we included cases from the 2012 Kids’ Inpatient Database with diagnosis codes related to CM who were <5 years of age and were not transferred to another facility. Potential demographic and clinical risk factors were identified and compared to child fatality in the hospital by using bivariate and multivariate analyses. To assess how cases coded specifically for maltreatment differed from similar cases that only suggested maltreatment, a reduced-model multivariable logistic regression for fatality was created.

RESULTS:

We found 10 825 children <5 years who had inpatient diagnoses coded in their medical record for CM. Most demographic variables (age, race, and sex) were not significantly associated with fatality, whereas clinical variables (transferring in, drowning, ingestions, and burns) were significantly associated with fatality. There were regional differences on the basis of hospital location as well as significantly more chronic conditions, procedure charges, and longer lengths of stay among children who died. Controlling for significant risk factors, those with diagnoses specific for physical abuse had ∼3 times the odds of dying (odds ratio = 2.797; 95% confidence interval: 1.941–4.031).

CONCLUSIONS:

In this study, although infancy and decreased income were associated with increased risk for fatality, more important factors were the types of injuries the child endured and whether the inpatient clinician had identified specific injuries indicating physical abuse.

In 2016, >1700 children died of abuse and neglect in the United States, a number that has increased since 2012.1  However, it is widely believed that child maltreatment (CM) fatalities are underreported nationwide.2,3  Fatalities are the most tragic consequences of CM and have been recognized as a major public health concern. In 2013, the US government acknowledged that this issue should be addressed and initiated the Commission to Eliminate Child Abuse and Neglect Fatalities.4  The commission published a final report in 2016 with a call to action for many improvements to be made in policy, research, and health care, stating that “identifying children and families most at risk of a maltreatment fatality is key to knowing when and how to intervene.”4

Authors of previous studies have researched the nature of CM fatalities in a variety of ways.58  Authors have used data from Child Protective Services (CPS) reported to the National Child Abuse and Neglect Data System,5  from a national emergency department (ED) database,9  a national trauma data bank and pediatric trauma registry,10  a national database of child death reviews,11  and data from national inpatient databases.8  Each data set provides different insight into the risk factors for CM fatalities at the different stages of treatment. Fatalities in EDs have been assessed by using the National Emergency Department Sample,12  but pediatric cases in inpatient hospital settings have had limited investigation. Studies that have used the Kids’ Inpatient Database (KID), a national pediatric database, have limited their research to physical abuse.8,13

It is important to identify risk factors for CM fatalities in children who are admitted to the hospital because, although they have not yet died, they have suffered serious injuries and may be saved. Patterns of serious injuries such as head trauma, abdominal trauma, poisoning, and malnutrition have been identified among fatal CM cases.11  In studies of CPS reports, children who died because of maltreatment tended to be male, identified as African American, and had families that experienced more financial and housing instability than those who did not die.5  Physically abused patients in the ED who were younger, female, and had suffered from intracranial injuries or internal crushing injuries were found to have greater odds of mortality.9

By identifying the demographic and clinical variables that are associated with maltreatment fatalities, we hope to provide hospital-based clinicians with the information to recognize the at-risk population to better target interventions to reduce CM fatalities among pediatric inpatients. In this study, we used the KID to answer 2 research questions related to fatality once a child is hospitalized: (1) What are the risk factors for fatal outcomes in the inpatient setting when compared to nonfatal outcomes? (2) How do risk factors for someone admitted to the hospital with a condition specifically coded as maltreatment differ from someone admitted to the hospital for a similar condition that is not specifically coded as maltreatment? Given previous studies looking at general or ED populations, we hypothesized that certain clinical factors in addition to demographic factors (age, sex, race, and income) would be associated with CM fatalities.

We chose to use the KID to retrospectively study CM requiring inpatient medical attention. The KID is prepared by the Agency for Healthcare Research and Quality, as part of the Healthcare Cost and Utilization Project.14  The KID from 2012 includes >3 million pediatric discharges (<20 years old) from 4179 community, nonrehabilitation hospitals in 44 states. It contains numerous demographic and clinical variables in addition to hospital variables such as hospital size and region. It is the last version of the data set that used only International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes. As a measure of socioeconomic status, the KID reports a quartile classification on the basis of the median income for the patients’ home zip code.

The methodology for using a national database in this study has been used in previous CM fatality literature and can be reproduced in future studies in the field.7,12  The KID is applicable for our study because it contains a large sample of discharges for pediatric patients (0–19 years old) and is useful for assessing patients sustaining severe maltreatment who require hospitalization for their injury or illness.

To identify child abuse and neglect cases, we used ICD-9-CM codes and external injury codes related to maltreatment as defined by Schnitzer et al.15  Certain ICD-9-CM codes were specific for child abuse and neglect (eg, ICD-9-CM 995.50–995.59), and other diagnostic codes were determined to be suggestive of CM. A diagnosis was determined to be suggestive of maltreatment if information in the medical records strongly suggested a possibility of maltreatment in ≥66% of the total sampled hospital discharges and ED visits containing that diagnostic code.15  These ICD-9-CM codes for specific maltreatment diagnoses and suggestive maltreatment diagnoses were split into 3 categories for physical abuse, sexual abuse, and neglect (Supplemental Table 5).12  To further assess how the type of abuse or neglect impacts the risk of fatality, these codes were broken down into trauma and injury categories by practicing child abuse and neglect expert physicians. Categories were created for head trauma, thorax trauma, abdominal trauma, burns, drowning, poisoning, and malnutrition and/or dehydration.

We chose to focus on children <5 years of age because the literature reveals that younger children are more likely to experience fatal maltreatment, especially those <1 year of age.2,8  These young children represent a disproportionate percentage of children hospitalized for abuse and neglect-related conditions.8,11,13  Additionally, the review process for the selection of the ICD-9-CM codes for specified and suggestive maltreatment by Schnitzer et al15  often excluded older children. However, in an effort to keep our population consistent, we did not modify the ages for each specific ICD-9-CM codes as was done by Schnitzer et al15  because there were relatively few diagnosis codes with limited ages among children <5 years. All patients who were transferred to other facilities or institutions were excluded because we did not have sufficient outcome data for these patients and wanted to prevent duplications of patients within our data set.

Statistical analyses were conducted by using SPSS, version 23 (IBM SPSS Statistics, IBM Corporation). To assess how categorical variables differed between those who died in the hospital and those who did not die, we used Pearson’s χ2 test. Continuous variables were compared to fatality by using an independent-samples t test. Statistical significance was defined by P < .05. Many of the variables we examined were provided in the KID by Healthcare Cost and Utilization Project (sex, age, race, income, region, payer, ED services, transfer in, number of diagnoses, number of chronic conditions, number of procedures, length of stay, and total charges), and we used diagnosis codes for system injury and CM (head trauma, thoracic trauma, abdominal trauma, burn, drowning, ingestion, malnutrition and/or dehydration, suggestive sexual abuse, suggestive physical abuse, and suggestive neglect) as noted above.

To assess how risk factors differ for someone admitted to the hospital with a condition specifically coded as maltreatment from someone admitted to the hospital for a similar condition that was not coded specifically as maltreatment (suggestive maltreatment group) we used a logistic regression with fatality as the outcome variable. All significant clinical variables were included in the regression model and then any variable that was no longer significant in the model was removed. We did this first using just specified CM cases and then added in the variable for specified maltreatment in models for all cases. All variables were first added and then nonsignificant variables were removed in a stepwise fashion resulting in a reduced model of significant clinical and demographic variables.

This study was considered exempt from review by Creighton University School of Medicine’s Institutional Review Board because all data in the KID had been deidentified.

Using the KID, we were able to construct a large historical cohort of children admitted to hospitals across the United States with specified and suggestive maltreatment diagnoses. As shown in Fig 1, the 2012 KID contains >3 million cases from 44 179 hospitals in 44 states. Of children <5 years of age who were admitted to the reporting hospitals in the 2012 KID, 10 825 had diagnoses suggesting maltreatment. Overall, there was a 3.34% mortality rate in the sample, with almost double the rate among the specified maltreatment group (4.5% vs 2.8%).

FIGURE 1

Sample of injury and maltreatment cases from the 2012 KID.

FIGURE 1

Sample of injury and maltreatment cases from the 2012 KID.

Close modal

The sample of 10 825 was split into 2 groups: those with specific maltreatment diagnoses (n = 3472) and those with suggestive maltreatment diagnoses (n = 7353). In both groups, bivariate comparisons revealed that sex and race were not significantly different for those who died when compared to those who did not die. As displayed in Table 1, the fatality rate was lower for infants and in the northeast region of the United States. We found there was a higher fatality rate in lower income percentiles, but the difference lacked significance when independently compared to fatality. Additionally, those with no insurance or “other” forms of insurance had significantly higher rates of fatality. Those with public insurance had the lowest rates.

TABLE 1

Pearson’s χ2 Test Results for Demographic Variables

Demographic VariableSpecified MaltreatmentSuggestive Maltreatment
N = 3472, nDied (Mean = 4.5), %PN = 7353, nDied (Mean = 2.8), %P
Sex
Male 19 740 4.4 .848 4390 3.1 .771
Female 1497 4.5 — 2956 2.4 —
Age
0–12 mo 21 044 4.3 .468 3677 1.5 <.001
1–4 y 1368 4.8 — 3674 4.2 —
Racec
White 1461 5.1 — 3444 2.8 —
African American 793 3.8 .146 1139 2.5 .889
Hispanic 560 3.8 .191 1435 2.9 .849
Asian or Pacific Islander 44 6.8 .629 186 1.1 .166
Native American 53 1.9 .288 71 1.4 .490
Other race 206 4.4 .679 370 1.9 .326
Regiona
Northeast 508 2.4 — 1084 2.3 —
Midwest 985 5.3 .008 721 2.3 .975
South 1287 4.8 .019 2714 3.5 .051
West 692 4.2 .085 1832 2.6 .663
Payer
Public 2833 4.0 .006 4529 2.6 .046
Private 354 4.8 — 2182 2.8 —
No insurance 113 7.1 — 249 5.6 —
Other 151 9.9 — 355 3.7 —
Income
0–50th percentile 2298 4.8 .144 4393 2.9 .699
51–100th percentile 1074 3.8 — 2958 2.8 —
Demographic VariableSpecified MaltreatmentSuggestive Maltreatment
N = 3472, nDied (Mean = 4.5), %PN = 7353, nDied (Mean = 2.8), %P
Sex
Male 19 740 4.4 .848 4390 3.1 .771
Female 1497 4.5 — 2956 2.4 —
Age
0–12 mo 21 044 4.3 .468 3677 1.5 <.001
1–4 y 1368 4.8 — 3674 4.2 —
Racec
White 1461 5.1 — 3444 2.8 —
African American 793 3.8 .146 1139 2.5 .889
Hispanic 560 3.8 .191 1435 2.9 .849
Asian or Pacific Islander 44 6.8 .629 186 1.1 .166
Native American 53 1.9 .288 71 1.4 .490
Other race 206 4.4 .679 370 1.9 .326
Regiona
Northeast 508 2.4 — 1084 2.3 —
Midwest 985 5.3 .008 721 2.3 .975
South 1287 4.8 .019 2714 3.5 .051
West 692 4.2 .085 1832 2.6 .663
Payer
Public 2833 4.0 .006 4529 2.6 .046
Private 354 4.8 — 2182 2.8 —
No insurance 113 7.1 — 249 5.6 —
Other 151 9.9 — 355 3.7 —
Income
0–50th percentile 2298 4.8 .144 4393 2.9 .699
51–100th percentile 1074 3.8 — 2958 2.8 —

—, not applicable.

a

Race compared to white; region compared to the northeast.

As shown in Table 2, some trauma and injury categories were significantly related to fatality. Head trauma was significant among specified diagnoses. Drowning and traumatic injuries to the head, thorax, and abdomen had the highest rates of fatality, whereas poisoning, malnutrition and/or dehydration, and burn diagnoses had lower rates of fatality. Cases with a physical abuse diagnosis had higher rates of fatality, whereas sexual abuse and neglect had lower rates of fatality.

TABLE 2

Pearson’s χ2 Test Results for Clinical Variables

Clinical VariableSpecified MaltreatmentSuggestive Maltreatment
N = 3472, nDied (Mean = 4.5), %PN = 7353, nDied (Mean = 2.8), %P
ED services 2295 3.80 .004 5288 2.2 <.001
Transfer in 826 7.5 <.001 1641 4.1 <.001
Head trauma 1210 10.7 <.001 4018 1.8 <.001
Thoracic trauma 495 7.7 .002 581 5.7 <.001
Abdominal trauma 216 5.6 .423 274 6.9 <.001
Burn 140 2.1 .175 97 1.00 <.282
Drowning 16 12.5 .112 908 11.9 <.001
Ingestion 34 .205 1383 0.29 <.001
Malnutrition and/or dehydration 26 .269 NA NA
Physical abuse 3150 4.8 .826 4935 2.0 <.001
Sexual abuse 53 1.9 <.001 1483 0.0 .036
Neglect 606 1.2 <.001 2292 4.9 <.001
Clinical VariableSpecified MaltreatmentSuggestive Maltreatment
N = 3472, nDied (Mean = 4.5), %PN = 7353, nDied (Mean = 2.8), %P
ED services 2295 3.80 .004 5288 2.2 <.001
Transfer in 826 7.5 <.001 1641 4.1 <.001
Head trauma 1210 10.7 <.001 4018 1.8 <.001
Thoracic trauma 495 7.7 .002 581 5.7 <.001
Abdominal trauma 216 5.6 .423 274 6.9 <.001
Burn 140 2.1 .175 97 1.00 <.282
Drowning 16 12.5 .112 908 11.9 <.001
Ingestion 34 .205 1383 0.29 <.001
Malnutrition and/or dehydration 26 .269 NA NA
Physical abuse 3150 4.8 .826 4935 2.0 <.001
Sexual abuse 53 1.9 <.001 1483 0.0 .036
Neglect 606 1.2 <.001 2292 4.9 <.001

NA, not applicable.

As shown in Table 3, children who died in the hospital had significantly more diagnoses, procedures, and chronic conditions in their discharge medical record than those who did not die. Analysis of lengths of stay actually revealed different results in the 2 groups. Those with specific maltreatment diagnoses who died had significantly shorter lengths of stay in the hospital than those who did not die. Conversely, those with suggestive maltreatment diagnoses had longer lengths of stay before death.

TABLE 3

t Test Results for Continuous Clinical Variables

Clinical VariableSpecified MaltreatmentSuggestive Maltreatment
Did Not Die (Mean)Died (Mean)PDid Not Die (Mean)Died (Mean)P
No. diagnoses 7.26 12.29 <.001 3.98 9.68 <.001
No. chronic conditions 1.34 3.51 <.001 0.683 2.96 <.001
No. procedures 1.71 5.13 <.001 0.929 4.75 <.001
Length of stay, d 7.56 4.28 <.001 4.18 5.71 .232
Total charges, $60 891 99 224 <.001 40 049 116 785 <.001 Clinical VariableSpecified MaltreatmentSuggestive Maltreatment Did Not Die (Mean)Died (Mean)PDid Not Die (Mean)Died (Mean)P No. diagnoses 7.26 12.29 <.001 3.98 9.68 <.001 No. chronic conditions 1.34 3.51 <.001 0.683 2.96 <.001 No. procedures 1.71 5.13 <.001 0.929 4.75 <.001 Length of stay, d 7.56 4.28 <.001 4.18 5.71 .232 Total charges,$ 60 891 99 224 <.001 40 049 116 785 <.001

As shown in Table 4, reduced-model multivariable logistic regressions obtained independent odds ratios (ORs) for fatality while accounting for all other significant variables. When modeling fatality among specified CM cases (those with a specified diagnosis of physical abuse, sexual abuse, or neglect), a number of variables were associated with increased odds of fatality, including transferring in, head trauma, the number of chronic conditions, the number of procedures, and total charges. Increased income and length of stay were associated with decreased odds. The model using all CM cases produced an OR of 2.797 (95% confidence interval [CI]: 1.941–4.031) for specified physical abuse and 0.221 (95% CI: 0.009–0.707) for specified neglect.

TABLE 4

Reduced-Model Multivariable Logistic Regressions for Fatality, by Specified or All Maltreatment

VariableSpecified MaltreatmentAll Maltreatment
OR95% CIOR95% CI
Infant (age <12 mo) — — 1.66** 1.19–2.31
Race — — 0.847** 0.750–0.956
Income percentile (0%–50% vs 50%–100%) 0.528*** 0.311–0896 — —
Transfer in 2.11* 1.27–3.49 — —
Head trauma 10.86** 6.15–19.18 — —
Burn — — 0.840** 0.016–0.432
Drowning — — 15.16*** 10.05–22.88
No. chronic conditions 1.84*** 1.62–2.00 1.84*** 1.70–1.99
No. procedures 1.51*** 1.39–1.64 1.61*** 1.51–1.70
Length of stay, d 0.531*** 0.48–0.587 0.648*** 0.611–0.682
Total charges, $1.00*** 1.00–1.00 1.00*** 1.00–1.00 Specified physical abuse — — 2.80*** 1.94–4.03 Specified neglect — — 0.221* 0.009–0.707 VariableSpecified MaltreatmentAll Maltreatment OR95% CIOR95% CI Infant (age <12 mo) — — 1.66** 1.19–2.31 Race — — 0.847** 0.750–0.956 Income percentile (0%–50% vs 50%–100%) 0.528*** 0.311–0896 — — Transfer in 2.11* 1.27–3.49 — — Head trauma 10.86** 6.15–19.18 — — Burn — — 0.840** 0.016–0.432 Drowning — — 15.16*** 10.05–22.88 No. chronic conditions 1.84*** 1.62–2.00 1.84*** 1.70–1.99 No. procedures 1.51*** 1.39–1.64 1.61*** 1.51–1.70 Length of stay, d 0.531*** 0.48–0.587 0.648*** 0.611–0.682 Total charges,$ 1.00*** 1.00–1.00 1.00*** 1.00–1.00
Specified physical abuse — — 2.80*** 1.94–4.03
Specified neglect — — 0.221* 0.009–0.707

—, not applicable.

*** P < .001; ** P < .01; * P < .05.

Using a pediatric inpatient database, we were able to examine the numerous factors that may contribute to the risk of fatality in hospitalized children after CM. On the basis of our analysis, child demographic variables by themselves do not appear to be discriminating risk factors for CM fatality once children are admitted to the hospital. However, clinical variables are important risk factors for CM fatality. The multivariable regression analysis allowed us to recognize that a young child admitted to the hospital with a condition specifically coded as abuse or neglect had significantly increased odds of dying compared with a child admitted to the hospital with a similar condition not specific for abuse or neglect. This information can be vital for training inpatient clinicians to identify CM cases at risk for fatality, especially when they have little insight into CPS information or the details of the patient’s home life.16  An inpatient hospital stay is but a “snap shot” of the patient’s life. Clinicians do not often know how many people are living in the home, the parent’s marital status, or if siblings have been removed from the home previously, but they do know what is in the medical record. They are able to see if a diagnosis of maltreatment has been made. A medical diagnosis specific for maltreatment can allow inpatient clinicians to recognize that the patient has increased risk for fatality and escalate their care and awareness to prevent death.

We found no significant association for many demographic factors with fatality among maltreated children. It is reported in national data that boys have a higher fatality rate than that of girls2  and that African American children die of abuse or neglect ∼2.5 times more than white children.4  Our study is focused on the inpatient population, which may change the significance of risk factors contributing to death. Our results also contrast with results previously found using the National Child Abuse and Neglect Data System, which found age, sex, and being African American to be significantly related to CM fatality.5  However, that study included children of all ages, whereas we only included those <5 years old, and it has been well shown in the literature that when looking at all pediatric cases, the odds of mortality significantly decrease with increasing age.9  In a study using only patients <5 years of age but limited to 2 Texas trauma centers, the authors actually found a significantly higher rate of fatality among the older (ages 1–4 years) patients but found no significant association between race and sex with fatality.6  Our focus on young children in the inpatient setting highlights the importance of biological factors rather than the previously identified socioeconomic factors as risk factors for CM fatality.

The method of payment for medical treatment was significantly associated with fatality for both those with specified maltreatment and suggestive maltreatment. Those with public insurance had the lowest rates of fatality, which is surprising given a previous study assessing child abuse in the United States using the KID that found that the primary payer for hospitalization for child abuse was most commonly Medicaid (65.88%).13  However, in a study using CPS data from California, the authors found that children who were enrolled in public health insurance had lower odds of serious maltreatment (resulting in hospitalization or death) and it was hypothesized that “public health insurance my increase access to healthcare providers, a potentially protective factor.”17  We similarly hypothesize that public health insurance can increase access to health care and help improve outcomes and limit fatalities. We also found regional differences in fatality rates that may be confounded by poverty and access to health care.

The clinical risk factors for CM fatality are similar to the risk factors that have been reported in previous studies. For those with specified abuse, head trauma had the highest rate of fatality among the significant clinical variables. These results are consistent with the literature revealing the majority of physical abuse injuries are trauma to the head, and these head injuries result in the greatest risk for fatality.6,11,18  Unlike others, we did not specifically control for injury severity score in our study, which was found by others to be correlated with mortality.10  It is possible that it is the severity of the injury and not the type of injury that contributes more to the risk for CM fatality. However, the “risk of mortality” and “severity of illness” variables included in the KID are proprietary and computed from procedures and diagnosis codes by using complex formulas. Because the details for the formulas were not fully explained, we chose to not include them in our analysis.

The patients who were transferred from another facility or had some ED service in their discharge medical record were more likely to experience fatality than those who did not. We can only speculate that these cases were more severe and required immediate medical attention in the ED before admission and/or required transfer to a hospital better equipped to provide more advanced care.

It is not surprising that the seriously maltreated children who died appeared to have a more complex clinical picture, with significantly more diagnoses, chronic conditions, procedures, and total charges in their discharge record. These variables, however, have not been well studied in the CM fatality literature. In general, among trauma patients, children with nonaccidental trauma have been found to have longer hospital stays and higher rates of fatality than do children with accidental trauma, even after controlling for injury severity.19  Authors of another recent study of nonaccidental trauma did find that patients with fatal outcomes had higher hospital charges and shorter lengths of stay.6  We similarly found that children with specific maltreatment diagnoses who died had significantly shorter lengths of stay in the hospital than those who did not and speculate that this occurred because children who survived required more time in the hospital to recover and to stabilize before discharge.

Accounting for income and the many significant clinical variables such as drowning and all the types of trauma, we found that diagnosing a case with specified physical abuse (child physical abuse 995.54, shaken infant syndrome 995.55, injury by unarmed fight or brawl E960.0, and injury by assault E961.0–E969.0) significantly contributed additional risk of death after CM. We speculate that the additional risk of fatality related to these specific physical abuse diagnoses may stem from several sources, including illness severity not captured with injury coding, inaccurate history being provided by caregivers, or delay in seeking care, all of which could increase mortality rates. This additional risk related solely to the diagnosis of physical abuse in the medical record has not previously been reported. It highlights the importance of the recognition of physical abuse during a hospital stay and the potential for earlier intervention once a child is hospitalized. Clinicians should consider that physical abuse increases the risk for fatality independent of biological and other social factors.

This study has several important limitations. We performed a secondary analysis of a database using diagnostic codes that were created for billing purposes; problems have been noted with the accuracy of medical codes for CM identification and they may bias our results because coders assigned these codes on the basis of the clinical notes. Additionally, the KID is missing many important child, family, and community factors that have been shown to be associated with CM. For example, we had no information on previous CPS investigations, history of domestic violence or abuse in the home, the parents’ history of drug use, or if the child attended school and/or day care, and we could not link records in the KID with current or past CPS reports. Authors of previous studies have found that CPS received the highest number of referrals for families that had a previous child die of child abuse or neglect,20,21  and that information is important in determining CM risk during a hospital stay. We also have no data about those who died before hospital admission or those who died at home or at another institution not included in the KID. It is also possible that cases could be counted multiple times if the same patient was admitted multiple times in the study year; the KID does not provide individual identifiers, which would allow us to remove these duplicates. This study was also challenged to abide by the strict validated definition of specific and suggestive maltreatment by using ICD-9-CM codes.15  These criteria are also limited in that the suggestive CM definition includes codes related to CM for which only 1 case was used for their determination.

In future studies, we hope to assess diagnostic codes individually to look at how specific diagnoses (complex skull fracture versus simple skull fracture) compare to fatal outcomes. The KID also contains proprietary variables on risk of mortality and severity of illness, which are computed from procedures and diagnosis codes by using complex formulas. These could be further explored to control for severity of injury and to better assess cases that did not have a fatal outcome but were a “near fatality.” With the data from this study and others, we hope to better target at-risk individuals and families with interventions. Given that a recent study in which authors found that children could be identified at birth as having increased risk for postnatal inflicted injury death, it would be useful to assess whether earlier identification of fatality risk could help clinicians implement interventions to prevent CM deaths before hospitalization for CM injury.22

We conclude that for pediatric inpatients <5 years of age, demographic and social variables are less important for inpatient clinicians to focus on compared to the types of injuries and whether a diagnosis of physical abuse has been made. Children with diagnosis codes specific for physical abuse had almost 3 times the odds of dying compared with those with diagnoses not coded as maltreatment. Additional research is needed to assess potential risk factors that are not included in the KID, assess the effects of illness severity, and target at-risk individuals with interventions to improve services and decrease CM fatalities.

We thank the Health Sciences Multicultural and Community Affairs department and Dr Sade Kosoko-Laski at Creighton University as well as the Pediatric Summer Research Program at New York University School of Medicine for facilitating this research opportunity.

Dr Kennedy conceptualized and designed the study, coordinated and conducted the data analyses in SPSS (IBM SPSS Statistics, IBM Corporation), drafted the initial manuscript, and revised the manuscript; Dr Lazoritz assisted in designing the study and critically reviewed the manuscript; Dr Palusci conceptualized and designed the study, coordinated and supervised data analysis, conducted the secondary data analyses, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.

FUNDING: Creighton University Community-Oriented Primary Care Research Endowment National Institutes of Health grant (5S21MD001102). Funded by the National Institutes of Health (NIH).

1
US Department of Health and Human Services
.
Child maltreatment 2016. Available at: https://www.acf.hhs.gov/cb/resource/child-maltreatment-2016. Accessed March 10, 2019
2
US Department of Health and Human Services
.
Child maltreatment 2012. Available at: https://www.acf.hhs.gov/cb/resource/child-maltreatment-2012. Accessed March 10, 2019
3
Berger
RP
,
Sanders
D
,
Rubin
D
;
Commission to Eliminate Abuse and Neglect Fatalities
.
Pediatricians’ role in preventing child maltreatment fatalities: a call to action
.
Pediatrics
.
2015
;
136
(
5
):
825
827
4
US Department of Health and Human Services
.
Within our reach: a national strategy to eliminate child abuse and neglect fatalities: a final report from the Commission to Eliminate Child Abuse and Neglect Fatalities. Available at: https://www.acf.hhs.gov/cb/resource/cecanf-final-report. Accessed March 10, 2019
5
Douglas
EM
,
Mohn
BL
.
Fatal and non-fatal child maltreatment in the US: an analysis of child, caregiver, and service utilization with the National Child Abuse and Neglect Data Set
.
Child Abuse Negl
.
2014
;
38
(
1
):
42
51
6
Yu
YR
,
DeMello
AS
,
Greeley
CS
,
Cox
CS
,
Naik-Mathuria
BJ
,
Wesson
DE
.
Injury patterns of child abuse: experience of two level 1 pediatric trauma centers
.
J Pediatr Surg
.
2018
;
53
(
5
):
1028
1032
7
Farst
K
,
PB
,
King
AJ
,
Bird
TM
,
Robbins
JM
.
Trends in hospitalization rates and severity of injuries from abuse in young children, 1997-2009
.
Pediatrics
.
2013
;
131
(
6
).
8
Leventhal
JM
,
Martin
KD
,
Gaither
JR
.
Using US data to estimate the incidence of serious physical abuse in children
.
Pediatrics
.
2012
;
129
(
3
):
458
464
9
Allareddy
V
,
R
,
Lee
MK
, et al
.
Hospital based emergency department visits attributed to child physical abuse in United States: predictors of in-hospital mortality
.
PLoS One
.
2014
;
9
(
2
):
e100110
10
Burd
RS
,
D
.
The impact of injury coding schemes on predicting hospital mortality after pediatric injury
.
.
2009
;
16
(
7
):
639
645
11
Palusci
VJ
,
Covington
TM
.
Child maltreatment deaths in the U.S. National Child Death Review Case Reporting System
.
Child Abuse Negl
.
2014
;
38
(
1
):
25
36
12
King
AJ
,
Farst
KJ
,
Jaeger
MW
,
Onukwube
JI
,
Robbins
JM
.
Maltreatment-related emergency department visits among children 0 to 3 years old in the United States
.
Child Maltreat
.
2015
;
20
(
3
):
151
161
13
Bullock
DP
,
Koval
KJ
,
Moen
KY
,
Carney
BT
,
Spratt
KF
.
Hospitalized cases of child abuse in America: who, what, when, and where
.
J Pediatr Orthop
.
2009
;
29
(
3
):
231
237
14
Agency for Healthcare Research and Quality
Healthcare Cost and Utilization Project
.
Overview of the Kids’ Inpatient Database (KID). Available at: www.hcup-us.ahrq.gov/kidoverview.jsp. Accessed June 6, 2016
15
Schnitzer
PG
,
Slusher
PL
,
Kruse
RL
,
Tarleton
MM
.
Identification of ICD codes suggestive of child maltreatment
.
Child Abuse Negl
.
2011
;
35
(
1
):
3
17
16
Lavigne
JL
,
Portwood
SG
,
Warren-Findlow
J
,
Brunner Huber
LR
.
Pediatric inpatient nurses’ perceptions of child maltreatment
.
J Pediatr Nurs
.
2017
;
34
:
17
22
17
Miyamoto
S
,
Romano
PS
,
Putnam-Hornstein
E
,
Thurston
H
,
Dharmar
M
,
Joseph
JG
.
Risk factors for fatal and non-fatal child maltreatment in families previously investigated by CPS: a case-control study
.
Child Abuse Negl
.
2017
;
63
:
222
232
18
Nuño
M
,
Pelissier
L
,
Varshneya
K
,
MA
,
Drazin
D
.
Outcomes and factors associated with infant abusive head trauma in the US
.
J Neurosurg Pediatr
.
2015
;
16
(
5
):
515
522
19
Livingston
JK
,
Grigorian
A
,
Kuza
CM
, et al
.
Non-accidental trauma increases length of stay and mortality in pediatric trauma
.
Pediatr Surg Int
.
2019
;
35
(
7
):
779
784
20
Berson
IR
,
Yampolskaya
S
.
Factors predicting child maltreatment fatalities: a competing risk model
.
.
2013
;
6
(
3
):
173
186
21
Jonson-Reid
M
,
Chance
T
,
Drake
B
.
Risk of death among children reported for nonfatal maltreatment
.
Child Maltreat
.
2007
;
12
(
1
):
86
95
22
Vaithianathan
R
,
Rouland
B
,
Putnam-Hornstein
E
.
Injury and mortality among children identified as at high risk of maltreatment
.
Pediatrics
.
2018
;
141
(
2
):
e20172882

## Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.