Introduction
The epidemiology of abusive head trauma (AHT) has been difficult to elucidate. Problems quantifying the number of children with this form of physical child abuse have included nonstandard research definitions, inconsistent nomenclature, disagreement about the mechanism of injury, and difficulty with case ascertainment. Some of these challenges have been addressed in the past decade through consensus panels seeking to define cases and standardize nomenclature. This chapter will use the nomenclature “abusive head trauma” (AHT), which recognizes that this subset of all closed head injury includes multiple mechanisms of injury ( Figure 6-1 ).
Abusive head trauma was described first by Dr. John Caffey in 1946. Dr. Caffey reported a case series of six infants with whiplash-shaken infant syndrome, all of whom had subdural hematomas and characteristic bone fractures. In 1962, a seminal paper by C. Henry Kempe brought the battered child syndrome to public attention, including AHT. Shaking was proposed as a mechanism for the injuries seen in these infants by a British pediatric neurosurgeon, Dr. Norman Guthkelch. “Shaken baby syndrome” was more formally described by Caffey in 1972 as a syndrome of intracranial and intraocular bleeding with no external signs of injury caused by vigorous shaking of infants. Caffey questioned whether some of the developmental delays, cerebral palsy, and epilepsy diagnosed in children could be attributed to unrecognized brain damage caused by shaking. This question remains relevant in both developed and developing countries 35 years later.
Quantification of AHT is hampered by difficulties in ascertainment, including misclassification. Studies reporting incidence have used different populations, including patients presenting to pediatric or subspecialty care, children admitted to intensive care units, victims of fatal abuse seen by medical examiners, and large administrative datasets. Each of these data sources has areas of potential bias ( Table 6-1 ). Ascertainment bias occurs when subpopulations of children are not included within the data set. For example, children with “subclinical” injury may not reach medical attention. Zolotar et al performed an anonymous phone survey of mothers with children under 2 years of age from a stratified random sample of birth certificates in North Carolina. Preliminary results show that approximately 1% of parents with a child less than 2 years of age reported shaking their children. Mothers (0.7%) and mothers’ partners (0.6%) shook children at similar rates. Thus shaking and possibly AHT occur more frequently within the population than is suggested by cases diagnosed in a medical setting. The rate of 1% reported by Zolotar would suggest that shaking is 54 times the rate of severe AHT prospectively observed in an earlier study in the same state. ,
| Ascertainment Bias |
| Subclinical injury |
| Prehospital deaths |
| ICU population only (severe injury) |
| Restricted age group |
| Misclassification |
| Misdiagnosis in hospital (not AHT) |
| Misdiagnosis on death certificate |
| Abusive head trauma vs. non-abusive head trauma |
Additional evidence for unrecognized injury is provided by studies documenting old brain injuries in as many as 30% to 45% of children who are diagnosed with AHT. Misclassification may lead to bias that can occur if a child is incorrectly diagnosed either when medical attention is sought or at postmortem. A retrospective review of 51 children with no neurological symptoms who were screened for AHT because of other injuries (rib fracture, healing fractures, or facial injury), revealed that 37% (95% CI: 24%-51%) also had a head injury. The status of the unscreened children having a similar finding is unknown. A case series of children with confirmed AHT showed that 31% of the children had seen a clinician for symptoms of head trauma before a definitive diagnosis and the diagnosis had been initially missed.
Deaths due to maltreatment are frequently not classified as homicides. In a three-state study of death due to maltreatment, underascertainment of child maltreatment fatalities was found in all three states by both child welfare agencies and in death certificate data. The combination of the two data sets correctly identified 90% of fatalities due to maltreatment. Thus both underascertainment and misclassification may falsely reduce the incidence in studies of AHT.
Population-Based Incidence Studies of Abusive Head Trauma
There have been several population-based studies of the incidence of abusive head trauma. Incidence is defined as the number of new cases diagnosed in a predetermined population over a specific amount of time, and is usually expressed in number of cases per unit of time. These incidence studies have all used different populations and slightly different definitions. Remarkably, the incidence estimates have been similar, which may reflect the fact that only the most severe cases are recognized. The prevalence of children in the population suffering from abusive head trauma is unknown.
The first prospective, population-based, incidence study of AHT was performed by Barlow and Minns. This study identified 19 cases of AHT over an 18-month period. The authors collected data from all hospital pediatric departments, pediatric intensive care units, neurosurgical units, and death records in Scotland. The calculated incidence of AHT was 24.6 per 100,000 infants per year (95% CI: 14.9, 38.5). The median age in this population was 2.2 months, with no child over 1 year of age. The main limitation of this study was its relatively small population base.
The first prospective population-based U.S. study of AHT was performed in North Carolina over a 2-year period. The study collected cases of children with head trauma who were less than 2 years of age from all nine pediatric intensive care units in the state. The authors also reviewed the charts of all deaths among children under 2 years of age. Additionally, the three out-of-state hospitals likely to accept referrals for North Carolina residents were surveyed. Whether the case was abusive or non-abusive head trauma was decided by the treating medical personnel at each hospital, but was reviewed by the investigators. A jury mechanism was developed to make decisions for cases that did not have a clear determination. Because of the larger population base, this study was able to provide more precise estimates than the Scottish study: 29.7/100,000 person-years (95% CI: 22.9, 36.7) in children less than 1 year of age; and 3.8/100,000 person-years (95% CI: 1.3, 6.4) for children during the second year of life.
The case fatality rate in this study was 22.5%. The median age at injury was 5.9 months. The median injury age was older in the North Carolina study than in the study of Barlow and Minns; however, this discrepancy is most likely due to the larger population base of the North Carolina study, which allowed case findings in children older than 1 year of age. The key limitation of the North Carolina study was its focus on severely injured children. Excluding children not admitted to an intensive care unit would tend to underestimate the incidence of AHT.
Ellingson et al used the Kids’ Inpatient Database (KID) for the years 1997, 2000, and 2003 in an effort to find a passive surveillance technique that could be used to monitor national trends in incidence rates of AHT. The KID dataset is part of the Healthcare Cost and Utilization Project (HCUP) collected by the U.S. Agency for Healthcare Research and Quality (AHRQ). The database contains an 80% sample of all non–birth-related discharges of children from all hospitals in participating states, which can be weighted to provide national estimates of disease. Using International Classification of Disease 9th Clinical Modification (ICD-9-CM) diagnosis codes to define cases in children less than 1 year of age, the authors calculated the national incidence estimate for KID 2000 at 27.5/100,000 person-years (95% CI: 22.6, 32.3). This incidence rate is remarkably similar to that of infants in the North Carolina study conducted in 2000 and 2001, and the Scottish study ( Figure 6-2 ). Limitations of the KID include exclusion of prehospital deaths, and the inability to verify each case. However, use of an existing surveillance mechanism is vastly less expensive than prospective case ascertainment. A 2008 U.S. Centers for Disease Control and Prevention panel is currently working on guidelines for the use of ICD-9-CM and ICD-10 codes to standardize ascertainment of AHT in hospital discharge datasets for research.
