Concussion in youth has received heightened attention owing to emerging evidence that this common injury can affect academics and cause behavioral changes and neurocognitive deficits in working memory, concentration, processing speed, and eye and motor function.1-3 Reported estimates of the number of youth with concussion have increased in recent years, likely owing in part to increased involvement in youth sports, introduction of return-to-play legislation, and increased awareness of concussion by clinicians, parents, and youth themselves.1,4-6
Concussion diagnosis remains symptom-based and does not require advanced diagnostic tools such as imaging.7 Thus, unlike many other types of traumatic injuries, children with concussion potentially enter the health care system through a variety of portals, including primary care or specialty care such as sports medicine or neurology, in addition to the emergency department (ED) and urgent care. A previous study highlighted the diversity of health care use for youth with concussion and minor head injury based on private insurance data, emphasizing the importance of primary care clinicians.8 However, pediatricians noted that they lack up-to-date concussion training and/or resources needed for timely and accurate diagnosis and ongoing case management.9 A more complete understanding of the distribution of point of health care entry for children with concussion is needed to guide both health care networks and clinicians where targeted training and resources need to be deployed.
In addition, an Institute of Medicine (IOM) report highlighted the lack of data on concussions treated outside the ED setting or sustained outside organized high school or collegiate athletics.1,10-12 Prior efforts have either focused on a single sport, used EDs as the sole site of identification, or studied a broader spectrum of head injury.13-22 In response to the IOM report, several initiatives are under way that are attempting to better estimate youth concussion incidence. To give guidance as to the needed breadth of those efforts and to identify locations in the health care system that need to be skilled in concussion diagnosis and initial management, we leveraged the linked electronic health record (EHR) system at The Children’s Hospital of Philadelphia (CHOP) to comprehensively describe point of entry into a large, regional, pediatric health care network for CHOP primary care patients who sustained concussions from July 2010 through June 2014; distributions were assessed over time across broad age, demographic, and socioeconomic characteristics.
Box Section Ref IDKey Points
Question Across a broad health care network, where do children initially seek health care for concussion?
Findings In this descriptive epidemiology study, 82% of patients sought initial care for concussion with primary care and only 12% within the emergency department. Younger children and those insured by Medicaid were more likely to use the emergency department as their initial point of health care entry.
Meaning Efforts to measure the incidence of concussion cannot solely be based on emergency department visits, and primary care clinicians must be trained in concussion diagnosis and management.
Description of CHOP Network
The CHOP network encompasses more than 50 locations throughout southeastern Pennsylvania and southern New Jersey, including 31 primary care centers, 14 specialty care centers, a 535-bed inpatient hospital, 2 EDs, and 2 urgent care centers supporting more than 1 million visits annually. The CHOP network serves a socioeconomically and racially/ethnically diverse population and accepts most insurance plans, including Medicaid. In addition to primary care clinicians, sports medicine, neurology, and trauma care clinicians at CHOP evaluate and manage concussions using a linked EHR system (EpicCare; Epic Systems Inc). The EHR is used for all aspects of clinical care, from inpatient to outpatient settings, as well as initial and follow-up office visits.
This study was reviewed and approved by the CHOP institutional review board. Consent was waived because the research was limited to existing data and involved no more than minimal risk to the patient, and the research could not be practicably carried out without the waiver.
We queried CHOP’s EHR database to identify all CHOP primary care patients who were aged 0 to 17 years and who had an initial in-person medical visit at CHOP related to concussion from July 1, 2010, through June 30, 2014. Concussion visits were defined as those assigned an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code indicative of concussion (Table 1). Primary care patients were identified as those with at least 1 primary care visit to a CHOP primary care network location during the study period; the study was limited to patients whose primary care is normally delivered by the CHOP network because we expect the distribution of entry locations for this group to be a more accurate representation of patient behavior in using the CHOP health care system for concussion than all patients of the CHOP health care network. Initial visits were defined as the first clinical encounter for a concussion as determined by date and time of the encounter. Because we could not separate multiple concussions that may have been sustained by an individual patient owing to limitations of CHOP’s EHR system, each patient was counted only once in analyses.
Given that the focus of this analysis was on patients whose concussion was their only traumatic brain injury, as well as to minimize concussion misclassification, we excluded patients who were also assigned an ICD-9-CM code for a more severe traumatic brain injury (codes included the following key words/phrases: contusion, laceration, subarachnoid, subdural, epidural, other and unspecified intracranial hemorrhage, and moderate or prolonged loss of consciousness) within 2 weeks of the initial concussion visit (n = 20). In addition, patients with clinically important injuries to body regions other than the brain were excluded (n = 298); this was assessed by identifying ICD-9-CM codes corresponding to other injuries on the same day as the initial concussion visit. Last, patients who were receiving ongoing concussion treatment during the study period but whose initial concussion visit occurred prior to July 1, 2010, were excluded (n = 121).
Potential points of entry within the CHOP health care network included (1) 1 of the primary care network offices; (2) outpatient specialty care, including orthopedics/sports medicine, trauma, and neurology; (3) 1 of 2 EDs or 2 urgent care centers; and (4) direct admission to the hospital. Race/ethnicity, sex, age, and payor (Medicaid, private, vs self-pay) at time of initial concussion visit were ascertained from the EHR.
The distribution of initial concussion visit location was described by age, sex, payor, and race/ethnicity; Wald 95% CIs based on a normal approximation to the binomial distribution were calculated using the surveyfreq procedure in SAS. In addition, during May and June 2012, extensive training was provided to CHOP primary care clinicians on the current standard of care and approach for diagnosing and managing concussions. Therefore, analyses were conducted looking both at the overall 4-year period, as well as stratified by date of initial concussion visit: July 2010 to June 2012 and July 2012 to June 2014. All analyses were conducted in SAS version 9.3 (SAS Institute Inc).
A total of 8083 children aged 17 years and younger had an initial in-person clinical visit for concussion (without a more severe head injury or other clinically significant injury) within the CHOP network during the 4-year study period. The frequency and proportion of relevant characteristics at the time of initial concussion visit are shown in Table 2. The median age of patients was 13 years (interquartile range, 10-15 years); 4.6% (n = 368) were aged 0 to 4 years, 30.8% (n = 2492) were aged 5 to 11 years, 34.9% (n = 2820) were aged 12 to 14 years, and 29.7% (n = 2403) were aged 15 to 17 years. Most were non-Hispanic white (n = 5729; 70.9%) and had private insurance (n = 6652; 82.3%). Almost two-thirds of all visits occurred from July 2012 to June 2014 (n = 5026).
Overall, 81.9% (95% CI, 81.1%-82.8%; n = 6624) of patients had their first concussion visit at CHOP within primary care, 5.2% (95% CI, 4.7%-5.7%; n = 418) within specialty care, 11.7% (95% CI, 11.0%-12.4%; n = 947) within the ED, and 1.2% (95% CI, 0.9%-1.4%; n = 94) direct admit to the hospital (Table 3). The proportion of CHOP primary care patients who sought care at their primary care office increased 13% between July to September 2010 and April to June 2014 (72.4%, 184/254 to 81.7%, 503/616), while the proportion seeking care at the ED decreased 16% over the same period (15.4%, 39/254 to 13.0%, 80/616) (Figure).
Table 3 illustrates the distribution of health care entry location by age, sex, race/ethnicity, payor, and period. Location of entry varied significantly by age. More than half of children aged 0 to 4 years entered the CHOP network via the ED (51.9%; 95% CI, 46.8%-57.0%; 191/368), a significantly higher proportion than for older children (14.9% [371/2492] of those aged 5-11 years [95% CI, 13.5%-16.3%]; 7.9% [222/2820] of those aged 12-14 years [95% CI, 6.9%-8.9%]; and 6.8% [163/2403] of those aged 15-17 years [95% CI, 5.8%-7.8%]). Conversely, more than three-quarters of those aged 5 to 17 years (5-11 years: 1995/2492; 12-14 years: 2415/2820; and 15-17 years: 2056/2403) initially sought care at CHOP via primary care. There were also important differences by race/ethnicity and payor: 42.4% (95% CI, 39.8%-45.0%; 586/1383) of non-Hispanic black patients entered via the ED compared with 4.9% (95% CI, 4.3%-5.4%; 280/5729) of non-Hispanic white patients. Similarly, 37.1% (95% CI, 34.4%-39.7%; 478/1290) of children insured by Medicaid and 24.1% (95% CI, 17.1%-31.2%; 34/141) who were self-pay entered through the ED compared with 6.5% (95% CI, 5.9%-7.1%; 435/6652) with private insurance.
While our understanding of concussion as an important public health issue among children has grown tremendously over the last decade, as highlighted in a recent IOM report,1 the epidemiology of youth concussion has been limited primarily to those who are high school age or older, participate in organized sports, and/or are treated in EDs. In this study, we leveraged the strength of a linked EHR system throughout a large health care network to provide a comprehensive description of the point of health care entry specific for pediatric concussion across the developmental age spectrum—the first such US assessment to include a diverse demographic and socioeconomic sample. By doing so, we discovered important insights into the variations in point of entry by age, payor, and race/ethnicity that will provide guidance to the future development of youth concussion surveillance systems being designed to provide national incidence estimates. In addition, our results highlighted the potential of EHR systems to facilitate research in this field.
First, the findings confirmed that concussions occur among children of all ages: approximately one-third of patients (2860/8083) were younger than 12 years. In contrast, most pediatric concussion research has focused on youth of high school age, resulting in limited knowledge about the natural history of concussion in a middle or elementary school population.23 Owing to the evolving neuropsychology and neurophysiology across the pediatric age range, it is likely that children manifest concussion symptoms differently across the age continuum.24 Our findings underscore the critical need for future research ensuring validated, age-appropriate diagnostic and treatment strategies are available across the entire pediatric age range.
Second, our results illustrate that more than three-fourths of the patients (6624/8083) had their first concussion visit within the CHOP network with primary care and just over 10% (947/8083) within the ED, suggesting that concussion estimates based solely on ED visits likely substantially underestimate the true incidence of this injury in children. A study linking multiple population-based health administration databases in Canada25 and a study using data on private insurance claims in Massachusetts8 documented a similar increase over time in concussions and minor head injuries treated during office visits (eg, primary care and specialty care) compared with treatment within EDs. Our study uniquely extends these findings in 2 ways. First, we used a stricter definition of concussion by excluding minor head injuries that were not diagnosed as concussions. More importantly, by capturing cases across broad demographics and socioeconomics throughout an entire health care network, we identified important variations in health care use by insurance payor and other demographic subgroups.
Our results highlight the critical importance of primary care clinicians in concussion care. This may be driven in part by insurance reasons or the fact that patients can often be scheduled for office appointments in the primary care setting sooner and at lesser expense than in specialty care or the ED. Regardless of the reason, these data provide critical guidance to health care networks, alerting them that the primary care setting is increasingly being used for concussion care and those clinicians may be in need of augmented training or increased resources. For primary care clinicians, these data provide evidence they can use to advocate for concussion clinical decision support tools in their health care system. Because specialized equipment is currently not needed for diagnosis, primary care clinicians can be well positioned to provide the initial evaluation for most patients with concussion. In some geographic areas, they may be the only available clinical resource. Preliminary findings suggest that primary care clinicians can in fact provide quality care to children with concussion.26 The care partnership between specialists and primary care that has occurred in the diagnosis and treatment of pediatric mental health disorders, such as attention-deficit/hyperactivity disorder, suggests there are models for successful clinical collaboration.
We also observed important variations in health care use among different subgroups. Children younger than 5 years old were more likely than older children to have the ED as their initial point of entry into the CHOP system for concussion. This aligns with previous research describing high rates of ED use for injuries in younger children.27 Concussion is challenging to diagnose in infants and toddlers as they are often unable to reliably relay their symptoms to the clinician, which may make their parents more likely to seek ED care for more urgent evaluation and concern for more serious brain injuries. Others have noted that ED use can be influenced by parental understanding of the primary care clinician as a comprehensive source of care,28 parental anxiety during the relatively high-stress period of infancy,29 or misperceptions of illness/injury severity.30 Our findings suggest the need for specialized training for concussion diagnosis and/or development of special screening tools in EDs for this difficult-to-diagnose population. These assessments may need to be age modified as evidence suggests different symptom constellations among pediatric age groups.24 Enhancement of the triage process that occurs in most primary care settings via telephone nurse triage may be advantageous in reassuring parents that many instances of minor head trauma, even in the very young, can be appropriately managed by primary care. Future work might focus on delineation of concussion characteristics in this youngest age group aimed at providing the foundation for these interventions.
Patterns of health care entry point also varied by payor and race/ethnicity, a finding congruent with previous research related to pediatric injuries in general.27,31 Patients with Medicaid were disproportionately more likely to use the ED setting as an initial point of entry into the CHOP network for concussion, as were non-Hispanic black patients. This may be owing to the fact that specific state-level Medicaid provisions may direct the patient where they can seek care, or it may be owing to existing beliefs as to the appropriateness of the ED for injury care for particular demographic subgroups. This suggests that concussion epidemiology based on ED records may overestimate the incidence of concussion among these groups and underestimate the incidence among other groups.
There were several limitations of this study. First, inherent to the use of a single health care network’s EHR system as a data source is the fact that visits outside that network are not included. Some patients may have had a prior visit at a community ED or non-CHOP provider before seeking care at CHOP and, therefore, this analysis may not fully characterize the health care use of a well-defined underlying population. To explore the extent of this limitation, we conducted sensitivity analyses that limited the sample only to patients of primary care practices that were within 20 miles driving distance from the CHOP ED as these might be more likely to use the CHOP ED rather than another community ED. Despite being more likely to be minority and receiving Medicaid than the entire sample, 69% of these patients first sought care at their primary care office, reinforcing our primary findings. Previous analyses indicated that the CHOP primary care population is similar in key demographic variables to the Philadelphia metropolitan area32 and likely is representative of the general patient population in the Mid-Atlantic region. Second, this assessment did not take into account those who did not seek care for their concussion, either because they did not recognize or want to disclose their symptoms or because their family did not perceive their injury required medical care. As highlighted by the IOM, this “culture of resistance” is prevalent and, although not the focus of this analysis, is a significant societal issue in need of further study. Third, only the initial CHOP visit for the first concussion occurring during the study period was identified for each patient. Certainly, some proportion of patients experienced multiple concussions during the study period. Unfortunately, the EHR was not set up to systematically separate multiple concussions within a given patient’s record. To estimate the extent to which we underestimated the number of unique concussions, we selected 3% of concussion patients (n = 218) for manual medical record review; we randomly selected patients within strata based on the number of concussion visits and the length of time between the first and last visits; patients with more visits and a longer time between the first and last visits were more likely to be selected. Based on this review, we estimated that 16% of our patients (1289/8083) sought treatment for more than 1 concussion during the study period, indicating that we likely underestimated the true number of unique concussions by that percentage over the 4-year period. A more complete assessment of the patterns of health care use for the entire course of care, including subsequent concussions, is under way to complement the analyses contained herein. Last, it is important to note that these analyses were conducted on patients with a primary care clinician; the distribution of health care location initially used for concussion may be different in those without a primary care clinician.
In summary, using a novel method to leverage rich data captured in a unified EHR system on a diverse demographic and socioeconomic population, this study suggests that incidence estimates of pediatric concussion that rely solely on ED records substantially underestimate the true incidence of this injury. Most pediatric patients with concussion sought their initial concussion care within the CHOP network with a primary care clinician, illustrating the need to provide up-to-date training and clinical decision support tools to these clinicians. Important variations in the point of health care entry by age, race/ethnicity, and payor also suggest the accuracy of incidence estimates based on a single point of health care entry may vary for these different subgroups. These differences may lead to targeted interventions to improve recognition and management of concussion in these populations. In addition, the increased use of EHRs may provide an opportunity to leverage health records for research into the natural history of concussion that may lead to improvements in the prevention, diagnosis, and management of this common childhood injury.