Discussion

By examining the application of CDC's operational fatal AHT case definition to data on multiple cause-of-death from the National Vital Statistics System, this report demonstrates the use of the definition for public health surveillance activities. The standard fatal AHT case definition can help to identify population subgroups at higher risk for AHT defined by year and month of death, age, sex and race/ethnicity. The results also demonstrate the importance of a standard fatal AHT case definition for young children by presenting comparisons of numbers and rates of fatal AHT cases with those of fatal paediatric TBI (without AHT) cases by selected characteristics. By definition, AHT is a subset of TBI, differentiated by age, nature of injury and cause of injury. Thus, it expected that the epidemiology of AHT would in many ways mirror that of TBI, particularly assault-related TBI.

This is one of a few studies that have examined the demographic and seasonal characteristics of AHT fatalities with direct comparisons to non-AHT TBI. As expected, there were significantly more cases of TBI (without AHT) than AHT for each year of data analysed. Also, the incidence of AHT was relatively constant over the study period from 2003 to 2007. This finding is consistent with several other studies. In their examination of Pennsylvania child abuse registry data from 1996 to 2004, Kesler et al found that the rates of both AHT deaths and injuries were steady over the study period.8 Analyses of California hospital inpatient data also revealed steady rates in AHT in children less than 2 years of age during 1998–2004.27 These consistent findings may reflect a lack of progress toward better detection and classification of fatal AHT cases but may also indicate a truly steady incidence over time.

As was expected, we found that AHT and assault-related TBI (without AHT) deaths occurred most frequently among children less than 1 year of age. Our results also coincide with findings of other studies that demonstrate a peak incidence of AHT early in the first year of life.28 29 However, our study revealed a peak in deaths at 1–2 months of age, while the aforementioned studies found peaks in AHT incidence between 10 and 16 weeks of age. The difference in our findings may represent differences in physiologic vulnerability in early infancy. The youngest infants exposed to AHT (eg, 1–2 months of age) are most likely to die from their injuries, while those who have assault-related head injuries at around 3–4 months of age may be more resilient and more likely to survive their injuries and be represented in hospital-based studies of injuries associated with child abuse.

The age-specific incidence curve for fatal AHT found in this study is similar to that for levels of crying among infants. Studies of crying, a primary stimulus underlying shaking and other behaviours that cause AHT, have found that it increases in the first month after birth, peaks in the second month and decreases thereafter.29–33 This suggests that preventive efforts that help parents deal with the frustration caused by crying infants may be helpful in decreasing fatal AHT.

A consistent finding in the literature on AHT and TBI is the significantly higher incidence among boys.5 8 34 35 This was upheld in the current analysis as well. These findings suggest that boys are not only more frequently AHT victims but also are more likely to be victims of severe AHT that leads to death.

Our findings regarding fatal AHT incidence by race were similar to those of some previous studies. In our study of AHT deaths for the USA, a higher incidence was observed among both non-Hispanic African-American and non-Hispanic AI/AN children; rate estimates for AI/ANs should be interpreted with caution due to small case counts. In analyses of data from Pennsylvania, the highest incidence was observed among African–Americans.8 In another study conducted using California shaken baby syndrome surveillance data, the incidence was highest among Hispanics of any race.27 A higher incidence of AHT in non-Caucasian children has been reported in several studies conducted in North Carolina.5 35

During the study period, seasonal differences were not observed for the incidence of fatal AHT or assault-related fatal TBI without AHT. This finding contributes to mixed results in the existing literature regarding seasonal fluctuation in AHT incidence. One previous study demonstrated significant seasonal fluctuation in incidence of AHT8 with peaks in the winter months (October through December). At least one other study has shown decreases in the incidence of overall physical child abuse during Child Abuse Prevention Month in April.36

There are several limitations in applying the panel's operational case definition for fatal AHT to the National Vital Statistics System mortality data. First, these data are based on death certificates only, which likely resulted in an undercount of cases. Evaluation studies indicate that death certificate data alone underestimate national and state incidence of child maltreatment fatalities.37–39 However, assault-related cases are more likely to be ascertained from death certificate data than cases involving less physical forms of abuse. Crume et al37 found that maltreatment was ascertained in 93 of 109 (85%) cases when the mechanism of death was bodily force or a blunt object, compared to an overall ascertainment rate for maltreatment deaths of 50%. Furthermore, given the similarity in the epidemiology of AHT and assault-related TBI (without AHT), it is possible that some cases in the latter category are misclassified due to incorrect, incomplete and/or non-specific coding of information on the death certificate. However, even if reliance on death certificate data results in underestimation of the incidence of AHT (and assault-related TBI), our study suggests that the operational case definition can be useful for public health surveillance activities, such as monitoring temporal trends and demographic patterns. The AHT cases identified in our study show demographic characteristics similar in distribution to those of other population-based studies of child maltreatment fatalities. Despite the similarities between AHT and assault-related TBI, because AHT is associated with a specific type of behaviour, that is, shaking, and has distinct, long-term consequences and prevention activities that may differ from other assault-related TBIs, we recommend considering the two conditions separately.

A second potential limitation is that some AHT cases may not be identifiable due to medical examiners and coroners lacking sufficient information to confirm and code child maltreatment. A majority of the definite/presumptive AHT cases (64.8%) identified had no ICD-10 maltreatment syndrome codes (T74 or Y07) recorded in their multiple cause listing. Among AHT case records showing Y07 as either the underlying cause code or among the record axis codes, most (93%) did not include information on the specific perpetrator type. This may have resulted from a lack of information about the specific circumstances of the injury event at the time the death certificate was completed. The completeness and specificity of cause of death data, especially pertaining to assignment of abuse/maltreatment codes, could be improved if information from law enforcement records and state and local child death review systems was made available to the medical examiner or coroner prior to when they certify the causes of death on the death certificate.39 40 This would provide additional information (specification of abuse, battering, beating or maltreatment) on the death certificate for mortality coding specialists to assign ICD-10 abuse/maltreatment codes (T74 and/or Y07) to electronic death records for children and teenagers under 18 years of age.41 Also, ICD-10 abuse/maltreatment codes are very limited in specificity. Expanded maltreatment code sets are needed to capture the circumstances surrounding child assault, abuse and maltreatment and to further specify the types of perpetrators. The ICD-11 coding system is currently under development, which could provide an opportunity to develop and implement such code sets for future use.42

Because our study involved national data and explored seasonal trends using death data combined across 5 years, the incidence counts would not likely reflect targeted prevention efforts. Our study data suggest that efforts to prevent assault-related fatal head injuries should be ongoing. On the other hand, our findings suggest that fatal TBI incidence from unintentional causes occur more often in the late spring and summer months when young children are more active outdoors. Similar patterns of peak TBI incidence between April and September have been previously reported.43 This suggests efforts to prevent unintentional fatal TBI, and its sequelae should be most intensive immediately before and during spring and summer months.

Given the similarity in the epidemiology of fatal and nonfatal AHT, potential avenues for prevention are likely to follow similar directions. Current thinking about prevention of AHT is based on the assumption that violent shaking of infants results, at least in part, from lack of information and skills about how to manage prolonged, inconsolable and unpredictable episodes of infant crying.44 45 These periods of crying are a normal feature of child development; however, this is not widely known nor is it appropriately addressed in parental advice literature.7 44 These understandings have led to the development of educational interventions to educate parents about the dangers of shaking, strategies for managing infant crying and the sharing of preventive information with peers.46 Prominent examples of these types of educational programs include a hospital-based shaken baby syndrome education program for parents and the Period of PURPLE Crying.46 47 The hospital-based shaken baby syndrome education program for parents was tested in hospitals in western New York State and found to reduce the incidence of AHT by 47% from the pre- to post-evaluation period, with no comparable decrease found in the comparison site (Pennsylvania).7 Randomised controlled trials of the Period of PURPLE Crying program materials have found them to lead to positive differences in the crying-related knowledge and behaviours of mothers of newborns.44 48 A case control study evaluating the use of an educational video highlighting the dangers of AHT in a level-one trauma center in Utah found that exposure to the video was associated with non-statistically significant reductions in AHT.49 Despite the lack of definitive evidence for the effectiveness of educational programs in preventing AHT, they are being widely used across the USA.50 Currently the Period of PURPLE Crying and the hospital-based shaken baby syndrome education program for parents are being evaluated in North Carolina and Pennsylvania, respectively, with funding support from CDC. Guidance on preventing AHT has been fully described elsewhere.51

In summary, this study demonstrates the use of the standard AHT definition for public health research and surveillance and suggests its utility for evaluation of current and ongoing prevention efforts. Also, several methodological issues that need to be addressed in order to improve those efforts are discussed. Despite the indicated limitations, including lack of specificity of ICD abuse/maltreatment codes and paucity of effective prevention programs, the establishment of a standardised AHT definition has implications as a useful surveillance tool for examining trends and patterns of AHT in the US population.