Data are from weighted estimates based on the Nationwide Emergency Department Sample (2006-2018).
Data are from weighted estimates based on the Nationwide Emergency Department Sample (2006-2018). Three visits with an ED charge above $100 000 in 2011 were considered to be outliers and thus were excluded from the analysis.
eTable. ICD Diagnosis Codes
Customize your JAMA Network experience by selecting one or more topics from the list below.
Shah YS, Iftikhar M, Justin GA, Canner JK, Woreta FA. A National Analysis of Ophthalmic Features and Mortality in Abusive Head Trauma. JAMA Ophthalmol. 2022;140(3):227–234. doi:10.1001/jamaophthalmol.2021.5907
What demographic and clinical factors are associated with higher mortality in abusive head trauma (AHT)?
In a nationally representative cross-sectional study of an estimated 12 287 emergency department cases of children younger than 5 years from 2006 to 2018, patients diagnosed with AHT in lower-income zip codes, from the midwestern United States, or older than 1 year had higher mortality rates. Retinal hemorrhages, orbital fractures, subarachnoid hemorrhages, subdural hemorrhages, intracranial hemorrhages, hypoxic ischemic brain injury, cerebral edema, and skull fractures were also associated with higher mortality.
These findings suggest that public health efforts aimed at reducing cases of AHT should be targeted toward low-income areas, taking into consideration the indicators for mortality identified in this investigation.
It is important to recognize presenting features and factors associated with mortality in abusive head trauma (AHT) owing to the severity of the diagnosis and the necessity for prompt action.
To describe the prevalence and economic burden of AHT and identify factors associated with mortality.
Design, Setting, and Participants
This retrospective, cross-sectional study used the Nationwide Emergency Department Sample database to identify all emergency department visits in the US for patients younger than 5 years with a primary diagnosis of abusive head trauma between January 1, 2006, and December 31, 2018. This study was conducted in 2021.
Main Outcomes and Measures
Prevalence, demographic characteristics, clinical characteristics, mortality, and economic burden associated with AHT. Weighted national estimates were calculated using sampling weights provided in the Nationwide Emergency Department Sample database.
From 2006 to 2018, there were an estimated 12 287 cases of emergency department visits in the US for patients younger than 5 years with a primary diagnosis of AHT. The estimated number of AHT cases decreased by 672 (95% CI, 403-940; P < .001) from 2006 to 2018. The incidence decreased by 6.7% each year (incidence rate ratio, 0.93; 95% CI, 0.93-0.94; P < .001) between 2006 and 2018. During the course of a hospital visit, 646 patients (5.3%) died. The majority of patients with a diagnosis of AHT were younger than 1 year (n = 7046; 57.3%), were male (n = 7268; 59.2%) and had Medicaid insurance (n = 8585; 70.0%). After controlling for demographic characteristics, factors associated with increased mortality were age greater than 1 year (odds ratio [OR], 2.45; 95% CI, 1.50-3.99; P < .001), first or second income quartile (OR, 1.78; 95% CI, 1.08-2.91; P = .02), midwestern United States (OR, 2.04; 95% CI, 1.04-4.00; P = .04), level 1 trauma center (OR, 2.69; 95% CI, 1.07-6.75; P = .04), orbital fracture (OR, 15.38; 95% CI, 2.41-98.18; P = .004), cerebral edema (OR, 8.49; 95% CI, 5.57-12.93; P < .001), intracranial hemorrhage (OR, 4.27; 95% CI, 1.71-10.67; P = .002), hypoxic ischemic brain injury (OR, 4.16; 95% CI, 2.13-8.10; P < .001), skull fractures (OR, 3.20; 95% CI, 1.76-5.82; P < .001), subarachnoid hemorrhage (OR, 2.43; 95% CI, 1.22-4.83; P = .01), retinal hemorrhage (OR, 2.17; 95% CI, 1.40-3.38; P < .001), and subdural hemorrhage (OR, 2.05; 95% CI, 1.05-3.98; P = .04).
Conclusions and Relevance
This study’s findings suggest that health care disparities may be present in the treatment of AHT. Recognizing factors suggested in this investigation to be associated with higher mortality, public health efforts should be targeted toward low-income areas and in the midwestern United States.
Abusive head trauma (AHT) is a form of trauma that involves the intentional shaking and/or blunt impact of an infant or young child.1 Although the incidence of AHT is not clear, there are an estimated 1000 to 1500 cases a year,1,2 and between 5% and 35% of all children diagnosed with it will die.1
Because of its relative rarity, signs of AHT are often missed,1 leading to increased mortality3 and continued trauma.3 Many of the signs of intentional trauma are also easily missed,1 as they require additional imaging3 or an assessment by an ophthalmologist. One of the most common ophthalmic signs of AHT is the presence of retinal hemorrhages, which should be further evaluated in the emergency department (ED) setting. To our knowledge, there are no studies analyzing trends of AHT, the most common ophthalmic presenting signs, and how they relate to mortality in a large, nationally representative database.
It is important to understand the incidence of AHT, especially at the ED level, given the seriousness of the diagnosis and requirement for prompt action by the physician. It is similarly important to understand the presenting clinical signs and their association with mortality for the patient in order to make the best care decisions possible. Therefore, the objective of this study was to examine the incidence of AHT and characteristics of patients with this diagnosis presenting to the ED over 12 years. Furthermore, we investigated which demographic characteristics and presenting ophthalmic and nonophthalmic clinical signs were associated with increased mortality in children.
The Nationwide Emergency Department Sample (NEDS) is the largest all-payer database for hospital-based ED visits in the United States and is sponsored by the Agency for Healthcare Research and Quality as part of the Healthcare Cost and Utilization Project. It consists of the administrative billing records of approximately 20% of all hospital-based ED visits in the US, totaling about 30 million ED visits annually.4 The sample is stratified based on hospital characteristics such as geographic region, trauma center designation, rural vs urban location, teaching status, and hospital ownership.4 Poststratification survey weighting can allow estimation of ED visits nationwide.4 It uses the International Classification of Diseases (ICD) system to code diagnoses. The first diagnosis in the database is known as the primary diagnosis and is defined as the condition that is chiefly responsible for requiring medical care. All other diagnoses associated with a patient are considered secondary diagnoses. More information on this database is publicly available.4
This retrospective, longitudinal cross-sectional study was conducted in 2021 using the NEDS to identify all ED visits in the US with a primary diagnosis of abusive head trauma from January 1, 2006, to December 31, 2018, with an age of 5 years or younger. The ICD codes used to identify the visits and classify diagnoses are based on the CDC’s recommendations5 and are described in the eTable in the Supplement.
The study was approved by the Johns Hopkins Institutional Review Board and adheres to the tenets of the Declaration of Helsinki6 and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. The NEDS database does not contain any direct patient identifiers, so informed consent was not required or obtained. Data on race and ethnicity are not available in the NEDS database and thus were not included in this study. Estimates for a particular characteristic are not shown if the number of visits is 10 or fewer in order to preserve patient confidentiality, in accordance with the Healthcare Cost and Utilization Project data user agreement.
National estimates were calculated using the sampling weights provided in the database. Descriptive statistics were calculated for patient demographic characteristics (age, sex, insurance, and median household income quartile of the residents in the patient’s zip code), hospital characteristics (region, teaching status, and trauma level), and ED charges. Incidence was calculated using US Census data of population estimates of children younger than 5 years. The income quartile ranges differ based on the year and can be found online.7 In 2018, quartile 1 was defined as a median household annual income of the residents in the patient’s zip code between $1 and $45 999, quartile 2 from $48 000 to $58 999, quartile 3 from $59 000 to $78 999, and quartile 4 at $79 000 and above. Income quartiles 1 and 2 were combined to form a “first/second” income quartile, and quartiles 3 and 4 were combined to form a “third/fourth” income quartile. In 2011, 3 visits had an ED charge above $100 000. These were considered outliers and excluded from the analysis. Descriptive statistics were also calculated for ophthalmologic findings (retinal hemorrhage, retinal detachment, vitreous hemorrhage, papilledema, optic nerve sheath hemorrhage, scleral hemorrhage, optic nerve injury, orbital fracture, and orbital hemorrhage), neurologic findings (subarachnoid hemorrhage, subdural hemorrhage, extradural hemorrhage, intracranial hemorrhage, hypoxic ischemic brain injury, cerebral edema, intracranial hypertension, hydrocephalus, and seizures), musculoskeletal findings (skull fractures, long bone fractures, rib fractures, cervical spine injury, and bruising/ecchymoses), procedures (computed tomography of the head, magnetic resonance imaging of the head, ultrasound of the head, lumbar puncture, and skeletal assessment), mortality, and disposition from the ED. Univariable and multivariable logistic regression models were used to assess demographic characteristics and clinical variables associated with mortality. Statistical significance was defined as P < .05. Stata, version 16 (StataCorp LP) was used for all statistical analyses.
From 2006 to 2018, there were 2660 ED visits for patients younger than 5 years with a primary diagnosis of AHT in the NEDS sample, representing an estimated 12 287 cases. The estimated number of ED visits with AHT as a primary diagnosis decreased by 672 (95% CI, 403-940, P < .001) from 2006 to 2018. The incidence decreased by 6.7% each year (incidence rate ratio, 0.93; 95% CI, 0.93-0.94; P < .001), from 5.4 (95% CI, 4.1-6.6) cases in 2006 to 2.1 (95% CI, 1.7-2.4) cases in 2018 per 100 000 population. The number of patients who died during the same period stayed relatively stable, from 65 (95% CI, 22-108) in 2006 to 31 (95% CI, 6-56) in 2018. The incidence and mortality per year can be seen in Figure 1.
Table 1 shows patient demographic characteristics for all cases with a primary diagnosis of AHT. In terms of age, 7046 patients (57.3%) were younger than 1 year, and 5240 (42.7%) were between 1 and 5 years. A total of 7268 patients (59.2%) were male, 5017 (42.7%) were female, and 8585 had Medicaid insurance (70.0%). More patients were in the first/second income quartiles (n = 8252; 68.6%) compared with the third/fourth quartiles (n = 3784; 31.4%). Regionally, the South region had the most cases (n = 4879; 39.7%) of AHT compared with other regions. Most cases were at a metropolitan teaching hospital (n = 8625; 70.2%) and a level 1 trauma center (n = 4902; 39.9%).
The mean ED charge per patient throughout the study period was $2758 (95% CI, $2540-$2976), and the total ED charge for all patients was $26 500 000 (95% CI, $23 600 000-$29 400 000). The mean ED charge increased from $2081 (95% CI, $1691-$2472) in 2006 to $4706 ($3573-$5838) in 2018, as shown in Figure 2.
Table 2 shows the clinical characteristics of all cases with a primary diagnosis of AHT. The most commonly documented ophthalmologic finding was retinal hemorrhage (n = 2685; 21.8%). Other findings included retinal detachment (n = 18; 0.1%), vitreous hemorrhage (n = 87; 0.7%), papilledema (n = 73; 0.6%), optic nerve sheath hemorrhage (n = 17; 0.1%), and orbital fracture (n = 15; 0.1%). The most common neurologic findings were subdural hemorrhage (n = 3256; 26.5%) and seizures (n = 1557; 12.8%). In terms of musculoskeletal findings, the most common findings were skull fractures (n = 2656; 21.6%) and bruising/ecchymoses (n = 2630; 21.4%). Less common findings were long bone fractures (n = 954; 7.8%) and rib fractures (n = 854; 7.0%). In terms of imaging and procedures, patients most commonly got a computed tomography scan of the head (n = 2516; 20.5%). Most patients were admitted to the hospital from the ED (n = 5604; 45.6%). An estimated 646 patients (5.3%) died during the courses of their visit.
A multivariable regression that assessed demographic factors and clinical signs showed several factors that were associated with mortality (Table 3). Children aged 1 year or older were more likely to die compared with children younger than 1 year (odds ratio [OR], 2.45; 95% CI, 1.50-3.99; P < .001). Patients were more likely to die if they were in the first/second income quartiles compared with the third/fourth income quartiles (OR, 1.78; 95% CI, 1.08-2.91; P = .02). Compared with patients in the northeastern US, patients in the midwestern US were more likely to die (OR, 2.04; 95% CI, 1.04-4.00; P = .04). Compared with patients who visited a hospital that was not a trauma center, patients were more likely to die at a level 1 trauma center (OR, 2.69; 95% CI, 1.07-6.75; P = .04). Clinical findings associated with death included orbital fracture (OR, 15.38; 95% CI, 2.41-98.18; P = .004), cerebral edema (OR, 8.49; 95% CI, 5.57-12.93; P < .001), intracranial hemorrhage (OR, 4.27; 95% CI, 1.71-10.67; P = .002), hypoxic ischemic brain injury (OR, 4.16; 95% CI, 2.13-8.10; P < .001), skull fractures (OR, 3.20; 95% CI, 1.76-5.82; P < .001), subarachnoid hemorrhage (OR, 2.43; 95% CI, 1.22-4.83; P = .01), retinal hemorrhage (OR, 2.17; 95% CI, 1.40-3.38; P < .001), and subdural hemorrhage (OR, 2.05; 95% CI, 1.05-3.98; P = .04).
In this cross-sectional study, we used a nationally representative sample to examine the prevalence, demographic characteristics, clinical characteristics, economic burden, and outcomes associated with AHT from 2006 to 2018. During this period, we found that the incidence of AHT decreased by about two-thirds while the mortality of AHT stayed constant.
To our knowledge, few studies have examined the trend of incidence of AHT on a national scale. Previous studies have shown individual hospitals or regions experiencing an increase in the incidence of AHT,8,9 whereas some national studies have shown no change in the incidence of AHT.10,11 A large number of initiatives have been implemented in the past decade to reduce the prevalence of AHT, with strategies focused on raising awareness of this form of abuse, reducing infant crying, and promoting a caregiver’s emotional regulation.12 However, many of these studies could not systematically monitor and investigate the impact on the incidence of AHT.12 Our study shows that these public health efforts may have had a role in decreasing the prevalence of AHT between 2006 and 2018. Other studies have shown that the health of the economy plays a role in the incidence of AHT,13 but our results did not show an uptick in cases during the economic recession between 2007 and 2009. Importantly, there was a decrease in AHT cases from 2015 to 2016, when there was a shift from ICD-9 to ICD-10 codes. After the switch in ICD coding, the number of cases stayed relatively stable from 2016 to 2018, indicating that the decrease in cases in 2016 may simply be due to a difference in coding rather than a true decline in diagnoses. Studies have demonstrated that the shift to ICD-10 codes may underestimate the true number of cases, particularly in diagnoses relating to trauma.14 Furthermore, 1 study compared the use of ICD-10 codes of physical child abuse to actual cases within a hospital and found that ICD-10 coding severely underestimated the number of true cases.15 Similarly, future studies using ICD-10 codes may underestimate the true burden of AHT.
Abusive head trauma was more prevalent in children younger than 1 year, in boys, in patients with Medicaid insurance, and in patients of low socioeconomic status, findings that have been replicated in other studies.10,16,17 We found that the southern region of the US had the most cases of AHT, followed by the Midwest. One study found that the highest rate of cases between 2000 and 2009 was in the Midwest,10 but this discrepancy may be due to a difference in data collection or recent changes in regional trends.
Despite the decreasing prevalence of AHT, the number of patients with AHT who died stayed relatively stable throughout the entire study period, with no statistical difference between 2006 and 2018. Therefore, it may be important to determine what features influence mortality of patients with AHT so clinicians can target these patients for aggressive treatment. Many studies have analyzed factors associated with the mortality of patients with AHT in smaller cohorts, but ours evaluated patients in the ED, which serves as the first point of contact for most patients. Older age was associated with death, which has been shown in previous studies and is thought to be due to the anatomy of the more developed brain and skull of an older child, along with the potential accumulation of injuries from AHT.18 Patients in a lower income quartile were more likely to die compared with those in higher income quartiles, possibly indicating disparities in care. In the NEDS database, income quartile is assigned based on the median income of the zip code the patient resides in. Our results suggest that zip codes with lower median incomes may be facing health care disparities, possibly due to a lack of resources to treat AHT, increased severity of patients presenting with AHT, or lower awareness of AHT. Patients were also more likely to die in the midwestern regions of the US. Previous studies have found a higher number of cases in the Midwest and hypothesized that this may be due to fewer prevention programs, differences in economic trends compared with other regions, or differences in ICD-9 and ICD-10 coding.10 However, to our knowledge, there may have not been any studies analyzing the reason for higher mortality in the Midwest.
Relevant ophthalmologic findings that were documented in patients with AHT were retinal hemorrhage, retinal detachment, vitreous hemorrhage, papilledema, optic nerve sheath hemorrhage, scleral hemorrhage, optic nerve injury, orbital fracture, and orbital hemorrhage, all of which have been identified in previous studies.19-23 Of these clinical features, retinal hemorrhages and orbital fractures were associated with higher mortality. Although other ophthalmologic findings may be important for diagnosis and overall visual outcomes, these findings suggest clinicians should pay special attention to retinal hemorrhages and orbital fractures in a child already diagnosed with AHT, as they are readily assessable and indicative of survival. Retinal hemorrhages are thought to occur in AHT owing to repetitive acceleration-deceleration forces that cause traction on the retina via the vitreous body, resulting in hemorrhages.24 More severe cases may be associated with the presence of retinal hemorrhages, thus explaining the association with higher mortality. Although studies have noted the prevalence of orbital fracture in AHT,23 there are not many hypotheses on why it is associated with AHT. The presence of orbital fractures may correlate with increased severity of abuse and a higher degree of force required to fracture a child’s orbit, thus also correlating with a higher mortality rate.
Other clinical features, such as subarachnoid hemorrhage, subdural hemorrhage, intracranial hemorrhage, hypoxic ischemic brain injury, and cerebral edema, were also associated with higher mortality, indicating key features to look for on neuroimaging. While these neurologic findings are a well-known phenomenon in AHT,1 to our knowledge, few studies have analyzed their association with mortality. Subarachnoid, subdural, and intracranial hemorrhages likely occur because of tearing of the superficial cortical veins from acceleration-deceleration forces. Hypoxic ischemic brain injury is thought to occur because of local damage to respiratory control centers in the brainstem, resulting in apnea.25 Cerebral edema is thought to occur because of trauma to the brain tissue, resulting in swelling. Similar to retinal hemorrhages, increased force may be necessary to cause hemorrhages and damage to the brainstem and/or other brain tissue, resulting in higher mortality rates. The only musculoskeletal finding associated with a higher rate of mortality was the presence of skull fractures. Previous studies have found that the frequency of skull fractures is similar among accidental and abusive head trauma,26 but our results show that when it is present in AHT, it may be associated with mortality.
It also may be important to take note of the increase in mean charge per visit in our study. This increase may reflect the national trend of rising health care costs, as indicated by studies showing an increased mean ED charge from 2010 to 2016.27 However, this change also may reflect an increase in imaging and testing or an increased severity or complexity of cases.
There are several limitations to our study. Our study only includes patients with a primary diagnosis of AHT, excluding those with AHT as a secondary diagnosis. Additionally, because the NEDS database uses ICD codes, there may be errors in documentation, missing data, and misrepresentation of a patient’s diagnosis in the health record. The database also does not include the severity of the diagnosis or the laterality or specific location of a physical examination finding, limiting our interpretations and conclusions. Finally, the database collects data based on encounters, not patients, making it possible that patients who had recurrent visits are represented more than once, artificially increasing the prevalence of the disease.
In conclusion, this cross-sectional study found that health care disparities may be present in the treatment of AHT. These data suggest that public health efforts to raise awareness and decrease prevalence of AHT should be targeted toward low-income areas and the midwestern US, taking into consideration the indicators for mortality identified in this investigation.
Accepted for Publication: November 23, 2021.
Published Online: January 20, 2022. doi:10.1001/jamaophthalmol.2021.5907
Corresponding Author: Fasika A. Woreta, MD, MPH, Wilmer Eye Institute, Johns Hopkins School of Medicine, 1800 Orleans St, Baltimore, MD 21287 (firstname.lastname@example.org).
Author Contributions: Ms Shah had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Shah, Iftikhar, Woreta.
Acquisition, analysis, or interpretation of data: Shah, Iftikhar, Justin, Canner.
Drafting of the manuscript: Shah, Iftikhar, Justin.
Critical revision of the manuscript for important intellectual content: Shah, Iftikhar, Justin, Canner, Woreta.
Statistical analysis: Shah, Iftikhar.
Administrative, technical, or material support: Canner.
Supervision: Iftikhar, Justin, Woreta.
Conflict of Interest Disclosures: None reported.
Meeting Presentation: Data from this manuscript were accepted for a virtual oral presentation at the American Society of Ophthalmic Trauma on June 5, 2021; a virtual poster presentation at the Women in Ophthalmology Conference from August 26 to 29, 2021; and an in-person poster theater presentation at the American Academy of Ophthalmology in New Orleans, Louisiana, on November 14, 2021.