Data were analyzed using the 2018 and 2019 MarketScan Commercial Claims and Encounters (CCAE) database for the commercially insured population for both inpatient and outpatient claims. MarketScan captures person-specific utilization and enrollment for inpatient (patients admitted to the hospital) and outpatient claims (eg, emergency department visits, primary care visits, urgent care visits). The CCAE database includes data from active employees, early retirees, Consolidated Omnibus Budget Reconciliation Act (COBRA) continuees, and dependents insured by employer sponsored plans. While not nationally representative, the CCEA generally contains data on more than 25 million commercially insured beneficiaries each year across the United States.

Patients with TBI were identified using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes: S02.0, S02.1, S02.80, S02.81, S02.82, S02.91, S04.02, S04.03, S04.04, S06, S07.1, and T74.4.²² Only initial encounters (which indicates a new injury) were included where the seventh character was “A,” “B,” or missing.²³ Patients who were diagnosed with a TBI in 2018 were considered to be patients with TBI and were followed for 1 year following the TBI.

Prevalence of selected demographic characteristics and outcomes for the TBI and non-TBI groups were calculated. Demographic characteristics included age group (0–17, 18–24, 25–34, 35–44, 45–54, and 55+ years), sex, region (northeast, north central, south, and west), urbanicity, and insurance status (comprehensive, preferred provider organization (PPO), and other). Other plan type is defined as Health Maintenance Organization (HMO), Exclusive Provider Organization (EPO), Point of Service (POS) with capitation, Consumer-Driven Health Plans (CDHP), and High Deductible Health Plans (HDHP). Urbanicity was assigned using the metropolitan statistical area of the primary beneficiary. Beneficiaries were defined as residing in an urban area if their county of residence was classified as a metropolitan statistical area; beneficiaries residing outside of a defined metropolitan statistical area were considered to reside in rural areas.

A matched retrospective cohort study design was used to compare patients with TBI to patients who did not have a TBI in 2018. Individuals were identified in 2018 if they had a TBI and matched to individuals in 2018 who did not have a TBI during the 2018 data year. Patients with a TBI were exact matched 1:1 to patients without a TBI on age, sex, region, mental health and substance abuse (MHSA) coverage indicator, insurance status, and month of index date to account for baseline differences in health care resource use. Patients without a TBI were randomly assigned an index date within 2018 that matched patients with a TBI, the index date was the date that the matched TBI patient was seen for the TBI.

New diagnoses among children (eg, learning disorders, attention-deficit/hyperactivity disorder, speech/language problems, developmental delay, bone/joint/muscle problems, anxiety problem) and adults (eg, diabetes, stroke, hypertension, alcoholism, drug addition, anxiety), as well measures of health care utilization (ie, outpatient visits, ED visits, and hospital admissions) were examined by TBI status and according to demographic characteristics in the year following an index health care encounter.

We calculated counts and percentages of key demographic characteristics by TBI status. Additionally, t-test statistics were run for demographic characteristics and outcomes by type of health care utilization and TBI group. Effect sizes were also computed for each t-test using Cohen’s d and interpreted in accordance with Cohen²⁸ to determine the associations’ practical significance. An r of 0.2 represents a small effect size, an r of 0.5 represents a medium effect size, and an r of 0.8 represents a large effect size.²⁸ Of note, only associations with a medium effect size or greater will be described in the text (though all are available in the table). In addition, the 10 most common ICD-10-CM codes for new diagnoses assigned to patients were identified to show other diagnoses related to the visit. This was stratified by TBI/non-TBI group, then by outpatient/inpatient, and finally by children/adults, and prevalence was reported. Analyses were performed in Stata Version 17 (Stata Corp LP, College Station, TX). CDC human subjects’ review of the protocol determined that it was not research involving human subjects. Thus, Institutional Review Board approval was not required.

The descriptive statistics of the sample are reported in Table 1. Each sample contained 72,338 patients and prevalence for age, sex, region, and insurance type were identical by design. The majority of the sample were children (47.8%) and young adults 18 to 24 years of age (18.9%), and the sample was split evenly between males (50.6%) and females (49.4%). Around one-third of the sample (37.8%) lived in the South, the majority resided in an urban area (>89%), and approximately half had PPO insurance (51.5%).

Within 1 year of the initial index date, the majority of patients with TBI (96.7%) and matched non-TBI patients (86.1%) had at least one outpatient visit (Table 2). A greater proportion of patients with TBI, as compared to non-TBI patients had an ED visit in the year following the index date (28.5% and 13.1%, respectively). Among visits for children, 70.7% had a new diagnosis within 1 year of the initial encounter among individuals who were diagnosed with a TBI compared to 48.9% of non-TBI individuals. For visits among adults, 83.9% had a new diagnosis among the TBI group and 60.2% among the matched non-TBI group.

Table 3 summarizes the findings of health care utilization, including outpatient visits, ED visits, and hospital admissions. The percent of outpatient visits, ED visits, and hospital admissions was higher for those with TBI compared to the matched sample of those without a TBI in total and also within each demographic variable examined.

For outpatient visits, those with a TBI had a significantly higher percentage of at least one outpatient visit (P < .0001), and this was also true for patients of all ages (P < .0001), both males and females (P < .0001), all regions (P < .0001), both rural and urban residents (P < .0001), all insurance types (P < .0001), and for new diagnoses in the year following the index encounter, for both children (P < .0001) and adults (P = .0004) compared to those in the non-TBI group. For ED visits and hospital admissions, those with a TBI had a significantly higher percentage, compared to those without a TBI, of overall ED visits in the year following the index encounter (P < .0001). This held for all ages (P < .0001), with larger effects (medium effect and above) for patients ≥25 years of age (Cohen’s d’s = 0.5); both males and females (P < .0001), with a larger effect for females (Cohen’s d = 0.9); all regions (P < .0001); both rural and urban residents (P < .0001); and all insurance types (P < .0001). And among those who had an ED visit, the proportion having a new diagnosis in the following year was also higher for both children and adults (P < .0001) with a TBI, compared to those without a TBI. The same patterns of significantly higher percentages were evident for hospital admissions for those with a TBI compared to those without a TBI.

Table 4 presents the top 10 new diagnoses within 1 year of the initial encounter for both the TBI and non-TBI group, stratified by type of visit (outpatient or inpatient) and by children/adults. For outpatient visits, among children in the TBI group, the most common secondary diagnosis was headache (19.4%) followed by cough (16.1%). For adults in the TBI group, the most common secondary diagnosis was hypertension (22.4%), followed by headache (18.2%). For both children and adults with TBI, there were smaller percentages for post-concussive syndrome and other conditions such as anxiety disorder. The pattern for the top 10 new diagnoses within 1 year of the initial encounter for the non-TBI group was similar with the exclusion of post-concussive syndrome and the inclusion of nasal congestion among children and obesity among adults for outpatient visits. For inpatient visits (eg, hospital admissions) among both TBI and non-TBI groups, the top 10 new diagnoses were similar (but varied in order and percentage) and also included vomiting and suicidal ideations among children and nicotine dependence and shortness of breath among adults.

There are some limitations to consider. First, for inclusion in the study, individuals had to have medical insurance. Having insurance and seeking medical attention may be influenced by several factors, such as age, gender, race, spoken language, health insurance coverage, income level, ease of access, and geography.³⁸ For example, individuals who do not have insurance or speak English experience barriers and are less likely to seek care^39–41 and thus are more likely to have less representation in health care datasets. Additionally, MarketScan CCAE does not capture all medical insurance (eg, those who are uninsured or who have government insurance). Second, for our study, being in the TBI group is dependent on seeking medical care and receiving a TBI diagnosis, as the present study used ICD-10 TBI diagnosis to define the injury. Though ICD codes are sensitive for diagnosis of severe TBI,⁴² the use of ICD codes to define brain injury in general may result in under diagnosis.⁴³ Moreover, not all individuals with TBI seek health care evaluation for the injury and may remain undiagnosed for a TBI,^44–46 especially for those that are mild. Thus, it may be the case that the TBI group in this study may be not capturing as many mild TBIs. This study captures those with an initial diagnosis of TBI by relying on the seventh character of the ICD-10-CM code. While many insurers require completion of billable codes up to 7 digits where appropriate, it is possible that a TBI that was not initial was captured within this studies definition due to incomplete coding. Additionally, some individuals in the comparison group may have had undiagnosed mild TBI in 2018, and by nature of the study design, those in the comparison group may have had a history of TBI that occurred prior to 2018. This may result in potential for misclassification bias and have attenuated the results. Third, diagnoses were categorized as new if no visit occurred for that ICD-10 code in the 3 months prior to the initial encounter. It is possible that some diagnoses classified in this study as new were actually made prior to the 3 month cut-point. Fourth, it is also vital to consider that ICD codes are primarily collected for billing rather than research purposes, and their use may be inconsistent, inaccurate, and incomplete.^47–49 Fifth, MarketScan data are not nationally representative, and results cannot be generalized to the entire population of the United States, though they do comprise a very large convenience sample. Sixth, interpretation of higher health care utilization was not examined in detail. For example, patients who have been diagnosed with a new TBI and receive follow-up care following the injury may be a sign of good clinical care (eg, obtaining appropriate rehabilitation and care) or it may serve as a proxy for worse outcomes (eg, higher risk for comorbidities or increased cost). Thus, increased health care utilization should be interpreted with caution and not go beyond the scope of the study. Last, causality between TBI and health outcomes was not studied, rather association was.

The present findings of increased visits and new diagnoses within 1 year following TBI indicate future research needs. Examining these patterns over further years post-injury would be informative, as would the relationship of visits and new diagnoses with potential risk factors for adverse long-term health outcomes. Similar to Eliacin et al,¹⁰ this data source can be used to examine the proportion of super-utilizers and risks of super-utilization. Additionally, future research could explore the potential of proactive education and follow-up care to reduce new diagnoses and resource use, as well as examining health care outcomes among those who are uninsured or who have government insurance, a critical area for future work given higher rates of TBI among those who are uninsured.⁵⁰