search text   Advanced Search

This Article Abstract Full Text (PDF) An erratum has been published P3Rs: Submit a response P3Rs: View responses Alert me when this article is cited Alert me when P3Rs are posted Alert me if a correction is posted Citation Map Services E-mail this article to a friend Similar articles in this journal Alert me to new issues of the journal Add to My File Cabinet Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via CrossRef Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Trokel, M. Articles by Sege, R. Search for Related Content PubMed Articles by Trokel, M. Articles by Sege, R. Related Collections Office Practice

Variation in the Diagnosis of Child Abuse in Severely Injured Infants

^a Department of Pediatrics, Boston Medical Center, Boston, Massachusetts
^b Department of Public Health, Boston University, Boston, Massachusetts
^c Biostatistics Research Center
^d Pediatric and Adolescent Health Research Center, Floating Hospital for Children, Tufts–New England Medical Center, Boston, Massachusetts

	ABSTRACT

TOP ABSTRACT METHODS RESULTS DISCUSSION REFERENCES

 
OBJECTIVE. Diagnosis of child abuse is difficult and may reflect patient, practitioner, and system factors. Previous studies have demonstrated potential lethal consequences if cases of abuse are missed and suggested a role for continuing medical education in improving the accuracy of diagnosis of suspected abuse. Although the majority of injured American children are treated at general hospitals, most published studies of severe injury resulting from child abuse have been conducted at children’s hospitals. The objective of this study was to evaluate the role of hospital type in observed variations in the frequency of diagnosis of child physical abuse among children with high-risk injuries.

METHODS. Hospital discharge data were evaluated, and adjusted rates of abuse diagnosis were reported according to hospital type. A regression model estimated the number of cases of abuse that would have been diagnosed if all hospitals identified abuse as frequently as observed at pediatric specialty hospitals. This study consisted of children who were <1 year old and admitted to US hospitals in 1997 for treatment of traumatic brain injury or femur fracture, excluding penetrating trauma or motor-vehicle–related injury. A total of 2253 weighted cases were analyzed.

RESULTS. The proportion of patients with a medical diagnosis of child abuse varied widely between hospital types: 29% of the cases were diagnosed as abuse at children’s hospitals compared with 13% at general hospitals. An estimated 178 infants (39% of total) with these specific injuries would have been identified as abused had they been treated at children’s rather than general hospitals.

CONCLUSIONS. Hospital type was associated with large variations in the frequency of diagnosis of child abuse. This variation was not related to observed differences in the patients or their injuries and may result from systematic underdiagnosis in general hospitals. This result has implications for quality-improvement programs at general hospitals, where the majority of injured children in the United States receive emergent medical care.

Key Words: medical error • hospital discharge data • child abuse • femur fracture • traumatic brain injury • medical error

Abbreviations: TBI—traumatic brain injury • HCUP—Healthcare Cost and Utilization Project • ICD-9-CM—International Classification of Diseases, Ninth Revision, Clinical Modification • NACHRI—National Association of Children’s Hospitals and Related Institutions • ISS—injury severity score • CI—confidence interval • OR—odds ratio

WITH NEARLY 3 million referrals to child protective services and >900000 children found to be victims of child maltreatment in the United States in 2001, abuse continues to have a large impact on the health of children. Approximately 1300 children died as a result of abuse in 2001.¹ Unfortunately, the evaluation and diagnosis of child abuse remains uneven.

Although there is no highly specific test or physical finding that is pathognomonic for child abuse, in premobile children, 2 injuries, femur fracture and traumatic brain injury (TBI), are considered highly suspicious indicators of abuse etiology.²–⁴ Previous studies have found that the diagnosis of abuse has been missed in a substantial portion of abused infants with high-risk injuries.⁵,⁶

The process of suspecting and reporting child abuse is complex; it depends on both physician and patient factors. A survey-based study identified increased suspicion of abuse to be associated with younger children, more severely injured children, single-parent families, and poorly educated mothers.⁷ The physician’s age, postgraduate education, and tolerance for discipline have been shown to affect suspecting and reporting abuse.⁸,⁹ Two recent retrospective chart reviews from single institutions found large variations in the suspicion and diagnosis of child abuse in hospitalized children with head trauma and fractures based on patient factors, including race, age, and injury severity.¹⁰,¹¹ These studies were conducted at children’s specialty hospitals. We are not aware of any published study that evaluated the variation of diagnosis of child abuse between different hospital types across the United States.

The study presented in this report used a large national database of hospital discharge data to evaluate the contribution of hospital type to the variation in diagnosis of child abuse. The investigation focused on patients with injuries frequently caused by child abuse, infants with either femur fracture or TBI that did not result from either penetrating trauma or motor-vehicle accidents. Discharge data concerning children hospitalized in the United States were obtained from the Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database 1997, published by the Agency for Healthcare Research and Quality.¹²

	METHODS

TOP ABSTRACT METHODS RESULTS DISCUSSION REFERENCES

 
The Kids’ Inpatient Database 1997 was developed as part of the Healthcare Cost and Utilization Project (HCUP) to facilitate the study of a wide range of pediatric disorders.¹¹ The data set includes information concerning aspects of hospital care and patient outcomes of children 18 years old from a sample of hospitals across the United States in 1997. Pediatric discharges were sampled from the following 22 states: Arizona, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Iowa, Illinois, Kansas, Maryland, Massachusetts, Missouri, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Tennessee, Utah, Washington, and Wisconsin.

The sample of discharge information was collected from 2521 hospitals and included 1.9 million discharge records. These hospitals included general community hospitals, academic medical centers, and pediatric hospitals. A systematic random sample was drawn from these hospitals consisting of 10% of births and 80% of other pediatric discharges. Once weighted to account for sampling based on HCUP methodology,¹² the national estimates represented >6.5 million discharges. Patient-level data (age, gender, insurance status, diagnosis, and external cause of injury codes based on International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes) and hospital-level factors (teaching status, National Association of Children’s Hospitals and Related Institutions [NACHRI] hospital type, and hospital bed size) were analyzed in this study.

Subjects
From the original data set, we extracted 2253 weighted cases of patients <1 year old with TBI and femur fracture (ICD-9-CM codes: 820.00–821.39) (Fig 1). TBI included all intracranial injury (ICD-9-CM codes: 800.10–800.49, 800.60–800.99, 801.10–801.49, 801.60–801.99, 803.10–803.49, 803.60–803.99, 804.10–804.49, 804.60–804.99, 851.00–854.19, and 950.00–953.9). Children were excluded if they were not admitted through the emergency department (to avoid counting hospital transfers or referrals for specialty care); if they were injured by a motor vehicle, a gunshot, or a knife stabbing; if they did not have any E code assigned; or if they did not have sufficient data available to calculate an injury severity score (ISS).

View larger version (20K): [in this window] [in a new window]   FIGURE 1 Patient inclusion/exclusion flowchart.

The analyzed patient-level factors included age, gender, insurance status, injury type, ISS, mechanism of injury, and disposition. Insurance status was categorized as government assistance (Medicare, Medicaid, self-pay, CHAMPUS, and other government insurance) and all other insurers (BlueCross, commercial, preferred provider organizations, health maintenance organizations, and prepaid health plans). Injury type was initially divided into femur fracture and TBI. TBI was further divided into 3 mutually exclusive categories based on the presence or absence of skull fracture: subdural or epidural hemorrhage with skull fracture (ICD-9-CM codes: 800.20–800.30, 800.70–800.80, and 852.20–852.40), subdural hemorrhage without skull fracture (ICD-9-CM codes: 852.20–852.40), and all other intracranial injury. The ISS was calculated from the ICD-9-CM codes by using ICDMAP software (Tri-analytics, Inc, Bel Air, MD). Mechanism of injury was based on ICD-9 E codes and divided into 4 mutually exclusive groups: child abuse (E codes: 904.0–904.9 and 967.0–967.9); fall (E codes: 880.0–888.9 and 987.0–987.9); struck–not child abuse (E codes: 916.0–921.9, 960.0, and 968.1–968.2); and other. Disposition was divided into routine discharge, discharge to another health care facility, or other. Race was not designated in nearly one third of the cohort and, therefore, was not used in the analysis.

Hospital-level factors that were analyzed included teaching status, urban/rural status, and hospital bed size. Designations for teaching hospitals and urban status were derived from the American Hospital Association Annual Survey of Hospitals by the HCUP. Any hospital with an American Medical Association–approved residency program was categorized as a teaching institution; hospitals in a metropolitan statistical area were considered urban. Hospital bed size was grouped by the Agency for Healthcare Research and Policy into 3 groups nested within teaching status and hospital location (small, medium, and large) as described in the technical documentation available from the agency.

Data Management and Analysis
The cohort was divided according to hospital type. The rates of child abuse diagnosis for all injury types were reported within each group. Child abuse diagnosis rates were compared on the basis of specific injury types. Logistic-regression modeling was used to further evaluate the relationship of the diagnosis of child abuse to hospital type. Child abuse was the response variable, and hospital type was the explanatory variable. An adjusted model was then created, adjusting for both patient- and hospital-level factors including age, gender, insurance type, ISS, injury type, hospital bed size, and hospital teaching status. Injury type was classified into 4 mutually exclusive variables: subdural or epidural hemorrhage without skull fracture; subdural or epidural hemorrhage with skull fracture; femur fracture without any subdural/epidural hemorrhage; and the reference group, which included patients with all other intracranial injuries without femur fractures. The residuals for the 2 continuous variables, age and ISS, were assessed for linearity. Age was found to have an improved fit with a quadratic term. The disposition variable was added to the adjusted model to assess for bias in transfer at time of discharge.

To calculate the number of potential child abuse cases missed in general hospitals, a logistic-regression model was constructed in which child abuse was the response variable and age, gender, ISS, insurance status, and injury types were explanatory. No hospital-specific factors were included in the model. This model was applied only to the reference group: patients treated at children’s hospitals. A regression equation was created by using the coefficients produced from this reference model. When applied to the remaining cohort, the equation generated the probability of being diagnosed with child abuse while adjusting for the previously mentioned patient factors based on the performance of physicians at children’s hospitals. The actual child abuse status was then subtracted from the predicted status to calculate the discrepancy for each patient in both general hospitals and children’s units in general hospitals. The mean discrepancy rates with 95% confidence intervals (CIs) were reported for children seen at general hospitals with and without children’s units. The total number of patients seen at each type of hospital was multiplied by these mean discrepancy proportions to yield the estimated number of children who would have been categorized as abused had they been cared for in a children’s hospital.

	RESULTS

TOP ABSTRACT METHODS RESULTS DISCUSSION REFERENCES

 
The largest group of patients (49%) was admitted to general hospitals, one fourth was admitted to general hospitals with children’s units, and the remaining one fourth was admitted to a children’s hospital (Table 1). Children who were treated at children’s hospitals tended to be younger, more severely injured, and more likely to have private health insurance than those cared for at general hospitals. General hospitals cared for a smaller proportion of children with TBI but a greater proportion of femur fractures than the other 2 hospital types.

Nearly all the hospitals in this cohort were in urban settings. Half of the general hospitals were teaching facilities, as opposed to nearly three fourths of the children’s hospitals. Most hospitals were distributed fairly evenly across the United States. The majority of general hospitals were reported as large, whereas more than half of the children’s hospitals were categorized as small.

The proportion of children who were diagnosed with child abuse varied between hospital types after stratifying by injury (Table 2). Patients with all types of intracranial lesions were identified as victims of child abuse least frequently in general hospitals and most frequently in children’s hospitals. This pattern persisted with lesions more suspicious of abuse. In children with a subdural or epidural hemorrhage with skull fractures, 10% were identified in general hospitals as abused, and nearly one third were identified in children’s hospitals as abused. Infants with subdural hemorrhage without skull fractures were most likely to be diagnosed as having been abused. These included nearly one half of the infants with this condition admitted to general hospitals and three fourths of infants admitted to children’s hospitals. Children with femur fractures were identified as abuse victims in a similar pattern as intracranial injury, least frequently in general hospitals and most frequently in children’s hospitals. In all injury groups, general hospitals with a children’s unit identified abuse more frequently than general hospitals without a children’s unit but less frequently than children’s hospitals.

An estimate was obtained of the number of children who would have been diagnosed as abused if the rate of diagnosis of abuse observed seen at children’s hospitals was applied to all hospitals (Table 3). The c statistic of the reference model of children diagnosed with abuse at children’s hospitals was 0.851. The mean discrepancy for children not diagnosed as abused at general hospitals without children’s units was –10.5% (95% CI: –13.2 to –7.8), showing that an additional 10.5% (114) of these children had characteristics similar to the abused children seen at children’s hospitals. The total number of children diagnosed with abuse at general hospitals without a children’s unit would be 257 (23.6%). General hospitals with a children’s unit had a mean discrepancy rate of –10.8% (95% CI: –15.1 to –6.5), yielding an additional 64 abuse cases. The total number of missed cases of abuse with these 2 specific injury types was 178, which corresponds to nearly 40% of all total cases of abuse among infants under 1 year of age with these 2 specific injury types.

	DISCUSSION

TOP ABSTRACT METHODS RESULTS DISCUSSION REFERENCES

Our study found a large variation in the frequency of diagnosis of abuse based on NACHRI hospital type. Hospitals that focused exclusively on pediatric patients were more likely to detect and report child abuse; this discrepancy was not explained by the patient’s age, injury severity, or insurance status. This pattern was noted in all injury patterns evaluated. There are several potential explanations for this difference. It is possible that injured children were preferentially evaluated and admitted to children’s hospitals simply because they were suspected to have been abused. Additionally, more severely injured children may be preferentially evaluated for abuse in children’s hospitals because children’s hospitals care for more severely injured children than general hospitals. ICD-9 E codes may not be an accurate method of determining population-based estimates for any diagnosis or condition. To help reduce these biases, we excluded all patients who were transferred between facilities, thereby reducing the possibility that more complex cases would be cared for at specialty centers. We also controlled for patients who were discharged from the hospital to another facility and did not find considerable changes in the ORs of the final model. Although children’s hospitals cared for more severely injured children (as measured by the abbreviated injury score), the discrepancy in frequency of diagnosis of abuse remained after injury severity was accounted for in the statistical analysis. Nonchildren’s hospitals had a lower rate of E coding and therefore had a higher proportion of cases excluded from the study.

The definitive diagnosis of child abuse is difficult to ascertain; this study used the clinical diagnosis of child abuse as defined by each participating institution. This clinical diagnosis may well be incorrect, and this study may be subject to misclassification bias in that nonabused patients may have been categorized as abused and truly abused patients may never have been considered to be abuse victims. Previous studies have identified a tendency to underdiagnose child abuse⁵,⁶; therefore, it may be reasonable to assume that more abused patients were misclassified as not abused than nonabused patients were labeled as abused. It is encouraging to note that previous studies suggest that the true rate of child abuse more closely approximates the rate detected at children’s hospitals. Hettler and Greenes¹³ found that 30% of children 0 to 3 years old with acute traumatic intracranial injury were victims of abuse. Reece and Sege² described 50% of subdural hemorrhages resulting from abuse, and Billmire and Myers¹⁴ reported that 92% of intracranial bleeds without skull fractures in children under 1 year of age were the result of abuse. Femur fractures have also been described to be the result of abuse in rates similar to that of children’s hospitals.¹⁵,¹⁶ The most likely explanation for the observed difference in the likelihood of reaching a diagnosis of child abuse seems to be a systematic underdiagnosis of child abuse at general hospitals.

Several studies have highlighted the danger of missing an abusive etiology among injured children. Child abuse is often a chronic condition that recurs until the victim and perpetrator are no longer in contact. One of the late cardinal signs of abuse is multiple wounds or fractures in various stages of healing. Several case series of child abuse show that 18% to 80% of abused children show some signs of prior abuse.³,¹⁷–¹⁹ These abused children most likely saw a physician at some point before receiving the diagnosis of abuse but were not diagnosed appropriately. Jenny et al⁵ reported that 31% of abused patients admitted with acute head injuries had been recently evaluated by physicians for complaints consistent with acute head injury and were incorrectly diagnosed. The most common erroneous diagnoses were viral gastroenteritis followed by accidental head injury and ruled-out sepsis.

Dalton et al⁶ reported that in a cohort of children <3 years old with femur fractures, 20% of the fractures resulting from abuse were diagnosed at a later date, well after the initial hospitalization. Other case reports of misdiagnosed child abuse have been published.²⁰,²¹

Although the process of identifying and reporting suspected child abuse is complex, recent reports have highlighted the importance of specific continuing medical education. In a cohort of primary care pediatricians, Flaherty et al⁷ identified both patient and physician factors that are associated with increased frequency of suspecting abuse as a mechanism of injury. The only practitioner-level factor associated with increased frequency of suspicion of abuse was recent physician education about child abuse. In another report, Flaherty et al²² described a 10-fold increase in physicians reporting child abuse when they had been exposed to abuse-specific continuing medical education. A study of military physicians showed that, although pediatricians had a similar amount of child maltreatment training in residency to emergency physicians, pediatricians reported a significantly greater amount of continuing medical education in abuse than other specialties after residency.⁸ The only factor associated with increased reporting of child abuse in that cohort was the presence of postgraduate abuse training. If this pattern were true among civilian doctors, one might reasonably suppose that one factor in the underdiagnosis of child abuse in general hospitals might be the result of lower levels of abuse-specific continuing medical education.

The quantitative burden of the underdiagnosis of child abuse is likely to be substantial, because the majority of children are cared for at general hospitals with or without pediatric units. Even among this cohort of severely injured infants, general hospitals cared for the largest proportion of patients. Hospital emergency departments are complex environments and are likely to require quality-improvement programs to ensure that not only are clinical staff trained in the recognition of abuse but also that systems are in place to support the identification, medical and social evaluation, and management of abused infants.

	ACKNOWLEDGMENTS

 
This study was supported by the Harvard Injury Control Research Center grant R49/CCR1186-03 from the Centers for Disease Control and Prevention (to R.S.), the Agency for Healthcare Research and Quality grant T32 HS00060 (to M.T. and J.G.), and grant D43TW001083-04 from the Fogarty International Center at the National Institutes of Health (to A.W.)

	FOOTNOTES

 
Accepted Jun 30, 2005.

Address correspondence to Robert Sege, MD, PhD, NEMC, Box 351, 750 Washington St, Boston, MA 02111. E-mail: rsege{at}tufts-nemc.org

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

	REFERENCES

TOP ABSTRACT METHODS RESULTS DISCUSSION REFERENCES

 

1. US Department of Health and Human Services. Child Maltreatment 2001. Washington, DC: US Government Printing Office; 2003
2. Reece RM, Sege R. Childhood head injuries: accidental or inflicted? Arch Pediatr Adolesc Med. 2000;154:11–15[Abstract/Free Full Text]
3. Ewing-Cobbs L, Kramer L, Prasad M, et al. Neuroimaging, physical, and developmental findings after inflicted and noninflicted traumatic brain injury in young children. Pediatrics. 1998;102:300–307[Abstract/Free Full Text]
4. Caniano DA, Beaver BL, Boles ET Jr. Child abuse. An update on surgical management in 256 cases. Ann Surg. 1986;203:219–224[ISI][Medline]
5. Jenny C, Hymel KP, Ritzen A, Reinert SE, Hay TC. Analysis of missed cases of abusive head trauma. JAMA. 1999;281:621–626[Abstract/Free Full Text]
6. Dalton HJ, Slovis T, Helfer RE, Comstock J, Scheurer S, Riolo S. Undiagnosed abuse in children younger than 3 years with femoral fracture. Am J Dis Child. 1990;144:875–878[Abstract]
7. Flaherty EG, Sege R, Mattson CL, Binns HJ. Assessment of suspicion of abuse in the primary care setting. Ambul Pediatr. 2002;2:120–126[CrossRef][ISI][Medline]
8. Lawrence LL, Brannen SJ. The impact of physician training on child maltreatment reporting: a multi-specialty study. Mil Med. 2000;165:607–611[ISI][Medline]
9. Badger LW. Reporting of child abuse: influence of characteristics of physician, practice, and community. South Med J. 1989;82:281–286[ISI][Medline]
10. Lane WG, Rubin DM, Monteith R, Christian CW. Racial differences in the evaluation of pediatric fractures for physical abuse. JAMA. 2002;288:1603–1609[Abstract/Free Full Text]
11. McCormick MC, Kass B, Elixhauser A, Thompson J, Simpson L. Annual report on access to and utilization of health care for children and youth in the United States: 1999. Pediatrics. 2000;105:219–230[Free Full Text]
12. Healthcare Cost and Utilization Project. Kids’ Inpatient Database (KID) [CD-ROMs]. Rockville, MD: Agency for Healthcare Research and Quality; 1997
13. Hettler J, Greenes DS. Can the initial history predict whether a child with a head injury has been abused? Pediatrics. 2003;111:602–607[Abstract/Free Full Text]
14. Billmire ME, Myers PA. Serious head injury in infants: accident or abuse? Pediatrics. 1985;75:340–342[Abstract/Free Full Text]
15. Gross RH, Stranger M. Causative factors responsible for femoral fractures in infants and young children. J Pediatr Orthop. 1983;3:341–343[ISI][Medline]
16. Beals RK, Tufts E. Fractured femur in infancy: the role of child abuse. J Pediatr Orthop. 1983;3:583–586[ISI][Medline]
17. O’Neill JA Jr, Meacham WF, Griffin JP, Sawyers JL. Patterns of injury in the battered child syndrome. J Trauma. 1973;13:332–399[ISI][Medline]
18. Alexander R, Crabbe L, Sato Y, Smith W, Bennett T. Serial abuse in children who are shaken. Am J Dis Child. 1990;144:58–60[Abstract]
19. Jackson G. Child abuse syndrome: the cases we miss. BMJ. 1972;2:756–757[ISI][Medline]
20. Benzel EC, Hadden TA. Neurologic manifestations of child abuse. South Med J. 1989;82:1347–1351[ISI][Medline]
21. Diamond P, Hansen CM, Christofersen MR. Child abuse presenting as a thoracolumbar spinal fracture dislocation: a case report. Pediatr Emerg Care. 1994;10:83–86[ISI][Medline]
22. Flaherty EG, Sege R, Binns HJ, Mattson CL, Christoffel KK. Health care providers’ experience reporting child abuse in the primary care setting. Pediatric Practice Research Group. Arch Pediatr Adolesc Med. 2000;154:489–493[Abstract/Free Full Text]

This article has been cited by other articles:

				B. Spivack Detection of Infant Abuse in Pediatric vs General Hospitals AAP Grand Rounds, June 1, 2006; 15(6): 74 - 75. [Full Text] [PDF]

				Variation in Diagnosing Child Abuse Journal Watch (General), April 21, 2006; 2006(421): 6 - 6. [Full Text]

P³Rs:

This Article Abstract Full Text (PDF) An erratum has been published P3Rs: Submit a response P3Rs: View responses Alert me when this article is cited Alert me when P3Rs are posted Alert me if a correction is posted Citation Map Services E-mail this article to a friend Similar articles in this journal Alert me to new issues of the journal Add to My File Cabinet Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via CrossRef Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Trokel, M. Articles by Sege, R. Search for Related Content PubMed Articles by Trokel, M. Articles by Sege, R. Related Collections Office Practice

	Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved.