Incidence and Age Distribution of Hospitalized Presumptive and Possible Abusive Head Trauma of Children Under 12 Months Old in Japan

Background Abusive head trauma (AHT) is the leading cause of fatal maltreatment among young children. The incidence of AHT in Japan, however, remains unknown. This study examined the incidence and distribution of age in months among young children under 12 months of age hospitalized with intracranial injury in Japan. Methods We conducted a multicenter cross-sectional study of children under 36 months old admitted with intracranial injury to hospitals that employed the Diagnostic Procedure Combination (DPC) payment system between 2010 and 2013. Presumptive and possible AHT were defined using the combination of ICD-10 codes modified from the coding system recommended by the United States Centers for Disease Control and Prevention. Results The average incidence was 7.2 (95% confidence interval [CI], 7.18–7.26) for presumptive and 41.7 (95% CI, 41.7–41.8) for possible AHT per 100,000 children less than 12 months old from 2010 to 2013. The distributions of age in months for both presumptive AHT and possible AHT had peaks at around 2 and 8 months. Conclusions This is the first study to report the incidence of hospitalized children with presumptive and possible AHT using population-based data. Further datasets are needed to evaluate the incidence and specific preventive strategies to prevent AHT in infants during the months of highest risk.


INTRODUCTION
Abusive head trauma (AHT), or shaken baby syndrome (SBS), is the leading cause of death due to child maltreatment 1,2 and may lead to neurological sequelae. 3 To evaluate prevention strategy, the incidence of AHT needs to be reported repeatedly. Using active surveillance, the incidence of admitted AHT children (per 100,000 children) was reported from 12.5 to 38.8 in several Western countries, such as United States (children under 1 year old), 4,5 United Kingdom (children under 2 years old), 6 Australia (children under 2 years old), 7 or New Zealand (children under 2 years old). 8 These active surveillance systems are expensive, however, and recent studies have demonstrated the validity of passive surveillance systems using the algorithm of International Classification of Diseases, Tenth Revision (ICD-10) codes for hospitalization records. 9,10 Although the epidemiology of AHT shares similarity among countries, it has not been sufficiently examined in Japan. The only "self-reported" prevalence for shaking was reported as 3.9% in a population-based study, 11 which was similar to that in Western countries. However, there are few studies that examine the incidence of AHT using hospitalization records in Japan.
Japan has a unique health care system and medical facilities. Universal health coverage was established in 1958, 12 and everyone is insured and has unlimited access to all health care facilities at a relatively low cost. In addition, Japan has one of the highest penetration rate of computed tomography (CT) and magnetic resonance imaging (MRI) in developed countries. 13 The number of CT and MRI scans was approximately three-to fourfold higher than the average for Organization for Economic Cooperation and Development (OECD) countries. Thus, Japanese hospitalization records may have the potential capacity to detect intracranial injury among young children. A nationwide study investigating the incidence of admitted AHT is necessary to confirm the similarity of the incidence of AHT regardless of health care system or medical equipment. Furthermore, Fujiwara et al 14 reported that the prevalence of AHT diagnosed with CT showed two peaks: around 2 to 4 months and around 7 to 9 months. The second peak around 7 to 9 months is not clearly reported in other countries (ie, United States, [15][16][17]18 or China 19 ), and the mechanisms and preventive strategies for later infancy remain unknown. Since study subjects in the previous study are limited to a single national central hospital, 14 the distribution of ages in month for AHT using nationwide samples with hospital-based data need to be confirmed. Therefore, this study examined the incidence of hospitalized AHT for infants under 12 months old and described the distribution of ages in month for AHT children under 36 months old using hospital-based medical claim data.

Data source
This study utilized the Diagnosis Procedure Combination (DPC) database, which includes patient discharge and administrative claims. 20 The DPC database does not include cases due to motor vehicle accidents because administrative claims do not cover such. All advanced treatment hospitals have been directed to adopt the DPC in 2003, and other general hospitals have adopted it voluntarily. Advanced treatment hospitals have capabilities to provide advanced medical care and technologies with 400+ beds and 16+ clinical departments. The number of participating hospitals increased from 82 hospitals with 68,982 inpatient beds in 2003 to 1,585 hospitals with 492,206 inpatient beds in 2014. This number includes more than half of all general hospital beds nationwide. 21 The DPC database includes the demographics of patients, diagnosis, and comorbidities using the ICD-10 codes, treatments and surgeries, length of stay, and discharge status. 22,23 Physicians in charge must code the diagnosis according to medical records, and medical clerks or licensed medical information managers record all drugs, devices, and procedures during the admission. 24

Identifying suspected AHT
This study targeted children under 36 months old who were admitted to a hospital in the DPC database due to intracranial injury from 2010 through 2013. We extracted children with suspected AHT who had at least one diagnosis from multiple categories of disease diagnosis ("main diagnosis," "admissionprecipitating diagnosis," "most resource-consuming diagnosis," "second most resource-consuming diagnosis," "comorbidities present at time of admission," and "conditions arising after admission"). 22,23 We defined AHT based on certainty of AHT, that is, "presumptive AHT" definitions following the recommendation by the United States Centers for Disease Control and Prevention (CDC) 9,25 and "possible AHT" definition instead of "probable AHT" used in previous studies 5,9,10,27 as shown in Table 1. We decided to use the term "possible" AHT in this study because DPC data did not clearly have codes for unintentional injuries. The considerations for these differences among definitions are mentioned in the discussion section. Presumptive AHT represented children who had intracranial injuries (ICD-10 intracranial injury code: S06.0-S06.9, T90.5) and either an intentional injury codes (physical abuse: T74.1, other maltreatment syndromes, T74.8, maltreatment syndrome, unspecified, T74.9, assault: Y00, Y01, Y04, Y08, Y09; other maltreatment; Y07.0, Y07.1, Y07.2, Y07.3, Y07.8, Y07.9) or retinal hemorrhage (H356), without any kind of exclusion criteria for plausible causes including coagulation defects, purpura and other hemorrhagic conditions (D65-D69), deficiency of vitamin K (E56.1), birth trauma (P10-P15), intracranial nontraumatic hemorrhage of the fetus and newborn (P52), and hemorrhagic disease of the fetus and newborn (P53). Possible AHT included, in addition to presumptive cases, children who had intracranial injury (S06.0-S06.9, T90.5) without unintentional injury code (transport accident [V01-V99], other external causes of accidental injury including falls [W00-X59], sequelae of transport accident [Y85], and sequelae of other accidents [Y86]), applying the above excluding criteria (eg, coagulation defects). If the children were admitted multiple times in the same year, we selected one admission due to presumptive AHT; poorer outcome at discharge, if multiple diagnoses were the same (presumptive or possible); or a younger age, if the other conditions were the same. In addition to 83 advanced treatment hospitals, there were 36 children's hospitals, which belonged to the Japanese Association of Children's Hospitals and Related Institutions (JACHRI). These hospitals can provide comprehensive pediatric care with 100+ pediatric beds, 20+ pediatricians per 100 beds, and CT=MRI facilities. Twenty-five hospitals (69.4% of children's hospitals and related institutions) and all 83 advanced treatment hospitals employed the DPC payment system. We assumed that all the AHT admission cases would be captured in these advanced treatment hospital and JACHRI. In total, 108 (ie, 83 plus 25) out of 119 (ie, 83 plus 36) advanced treatment hospitals and children hospitals and related institutions were covered in the DPC payment system (90.8%), which is the same during the observed period (ie 2010-2013). Therefore, we used the coverage rate (Z) as 90.8% for all years.
Next, we examined the differences between presumptive and possible AHT cases among children under 36 months old. We compared demographics (ie, sex, age), clinical conditions (ie, Japan Coma Scale [JCS], duration of admission, transfer from other hospital, use of ambulance), outcome at discharge with six categories (ie, cured, improved, remission, remain unchanged, exacerbation, and death), and admission cost. We dichotomized outcomes into poor (remain unchanged, exacerbation, or death) and better (cured, improved, or remission) outcomes. Under the DPC payment system, the total admission cost, which is based on diagnosis group, includes basic hospital stays and any treatments during the admission, such as laboratory tests, medications, injections, and surgical procedures. In addition, for sensitivity analysis, the month-age distribution of epidural hemorrhage (EDH; S06.4) was compared between presumptive AHT, possible AHT and possible AHT excluding EDH, because EDH is more likely to occur after falls. 26 Lastly, we described the distribution of age in months for presumptive AHT, possible AHT, and possible AHT excluding EDH. We used STATA MP, version 14.0 (Stata Corp LP, College Station, TX, USA). The ethical review board at Tokyo Medical and Dental University approved this study. Table 2  however, the proportion of children aged 24-35 months old increased from 23.2% to 25.5%. Among possible AHT, 11.1% to 13.5% were presumptive AHT. Table 3 and Figure 1 shows the incidence of presumptive and possible AHT per 100,000 infants, adjusted by the changes in the number of admissions in DPC hospitals from 2010 through 2013 and the coverage of DPC system among hospitals where can provide advanced pediatric treatment. The overall incidence was 7.22 (95% confidence interval [CI], 7.18-7.26) per 100,000 infants for presumptive AHT, 41.7 (95% CI, 41.7-41.8) per 100,000 infants for possible AHT. The incidence for presumptive AHT increased (P for trend = 0.042), but the incidence of possible AHT was stable between from 2010 through 2013 (P for trend = 0.950). Table 4 shows the differences between children less than 36 months age admitted with presumptive and possible AHT. Children with presumptive AHT were younger, had a poorer JCS score, longer hospitalization, were more often transferred from another hospital, more often used an ambulance, and showed poorer outcome at discharge than children with possible AHT. The proportion of EDH was higher among possible AHT cases than in presumptive AHT (18.8% vs 6.8%). The medians of the admission costs were JPY 754,000 (approximately USD 7,540) and JPY 240,000 (approximately USD 2,400) for presumptive and possible AHT per case, respectively. Figure 2 shows the distribution of age in months for presumptive AHT, possible AHT, and possible AHT excluding EDH. For both presumptive and possible AHT, we observed two peaks, at 1-3 and 7-9 months old. Further, possible AHT excluding EDH cases also had two peaks at 1-3 and 7-9 months old. After 12 months old, the number of both presumptive and possible AHT cases became relatively stable.

DISCUSSION
This is the first study to describe the incidence of presumptive and possible AHT using ICD-10 codes for admissions in Japan. The incidences were 7.2 for presumptive AHT and 41.7 for possible AHT per 100,000 infants less than 12 months old from 2010 Yamaoka Y, et al. through 2013, Both presumptive and possible AHT had two peaks, at 1-3 and 7-9 months old.

Incidence of presumptive AHT
The incidence of presumptive AHT was comparatively lower than that of previous studies, which was 13.0-33.4 per 100,000 infants based on ICD-9 or 10 codes in other counties. 4,5,[8][9][10]18,27 The definitions of AHT to estimate incidences varied among studies. The passive surveillance to capture AHT cases from claim data was obtained from the hospital discharge records used in several studies. 5,9,10,27 Berger et al reported a high accuracy of the CDC operational definition of AHT using ICD-9CM codes, compared with AHT defined by child protection teams. 28 However, Japanese coders may not have sufficient training in coding for cause of intentional injury, such as physical abuse (T74.1, T74.8, T74.9), assault (Y00, Y01, Y04, Y08, Y09), or other maltreatment syndromes (Y07.0, Y07.1, Y07.2, Y07.3, Y07.8, Y07.9) that were used in other studies. 5,9,10,27 Therefore, the insufficient coding for intentional injury might lead to underestimate the incidence of presumptive AHT in Japan. On the other hand, the incidence of possible AHT in this study indicated the possibility of AHT cases at most, because we only could obtain AHT    Intracranial Injury of Children Under 12 Months Old without unintentional injury code. It might include some intracranial injuries that physicians or coders did not code for the cause of injury. Therefore, to obtain accurate incidence for AHT and to establish effective passive surveillance, sufficient education for physicians and coders is needed to ensure the effective evaluation of AHT.
Another method to obtain better estimates for AHT is to utilize active surveillance with multiple sources of information. 4,8,18 For example, Keenan et al 4 abstracted information from medical charts and medical examiner records, reviewed radiological examinations, and utilized determination of AHT by social service agency. Because AHT is a rare event and a large population or multiple years of followup is needed to assess its incidence, conducting active surveillance becomes expensive and time-consuming. Previous studies using passive surveillance showed similar incidence of AHT with that of active surveillance. 9,10 Therefore, improving better passive surveillance is epidemiologically essential to assess the incidence of AHT. From a practitioner's point of view, detailed information from medical records and child protection services would be helpful for better understanding and prevention of AHT. There are both needs to establish passive and active surveillances that capture the occurrence of AHT in hospitals across the country and share information from multidisciplinary organizations.

Characteristics of presumptive AHT
This study revealed that the children with presumptive AHT were younger, had severe medical conditions at hospitalization and at discharge, EDH was less frequent, and higher admission cost than possible AHT cases. Longer hospitalization, 3 poor outcomes, 3,29 and higher admission cost 30 have also been reported as characteristics of AHT. In terms of clinical presentations, seizures at presentation or within 24 hours, and apnea at presentation were significant characteristics of AHT that could lead to a severe JCS score, 31 as seen in presumptive AHT in our study. Therefore, we can confirm that our study subjects had similar clinical severities to that of presumptive AHT cases, and not possible AHT, in the other studies.

The distribution of age in months
We found two peaks for both presumptive and possible AHT, at  1-3 and 7-9 months old. The second-peak distribution in 7-9 months was similar to that of possible AHT without EDH, suggesting that both presumptive and possible AHT during 7-9 months may not have been caused by direct impact, such as falls. A plausible explanation for the first peak at 2 months for presumptive and possible AHT is "the crying period", which is same as in previous studies. 15,32 The possible interpretations for the second peak at 8 months for presumptive AHT need further consideration. This second peak has not been reported in other countries, and is consistent with a previous study conducted by identifying AHT using CT in Japan. 14 Other mechanisms need to be considered for presumptive AHT in late infancy. One possible explanation is that different types of parenting stress or challenges occur in the later months of infancy, for example, the transition from milk to solid food 33 ; the increased supervision needed as a result of developmental milestones, such as crawling or pulling up 34 ; or anxiety in regard to the delay of those development milestones. Even though Japan has easy access to radiographic examination, as it has the highest rates of CT and MRI in developed countries as well as universal health insurance coverage, the second peak at 8 months for AHT is unique and its underlying mechanisms remain unclear. Further research is needed to target presumptive AHT, especially in the later months of infancy, to determine the possible reasons and preventive strategies in Japan.

Limitations and future direction
There are several limitations and challenges for further research. First, we utilized the DPC database, which covered more than half of all general hospital beds nationwide. All university hospitals and all 34 children's hospitals and related institutions that provide specialized comprehensive pediatric care were included. We might, however, have missed AHT cases admitted to general hospitals, cases that died before transfer to pediatric tertiary hospitals, or cases that died at home. Moreover, we could not distinguish the children who were admitted or who had transferred hospitals over the years, although we can track the children who were admitted multiple times in the same year. Therefore, future research should utilize the national database for claim records and track patients over years and between hospitals. Second, the incidence of presumptive AHT in this study may have been underestimated due to poor training for physicians and coders to code intentional injury. A multidisciplinary approach for information integration may help to accurately capture the cause of injury in DPC data. Lastly, this study examined the incidence of AHT in population-based samples obtained from the DPC database. However, the DPC database did not include patients who died in the emergency department without being admitted to the hospital. There is a possibility that fatal presumptive AHT cases are underestimated. Information from forensic autopsy records must be added to include fatal AHT cases to determine a more accurate incidence in the future.

Conclusion
This is the first study to report the incidence among hospitalized children with presumptive and possible AHT under 12 months old in Japan, which was 7.2 and 41.7 per 100,000 infants, respectively. Two peaks for distributions of age in months among children under 36 months old were found at 1-3 and 7-9 months old, a finding which is unique to the Japanese population. Further active surveillance is needed to validate the accuracy of passive surveillance using DPC data. In the meantime, DPC data may be useful to evaluate prevention strategies against AHT in Japan.