Applying a Novel Diagnostic Code System to Identify Postconcussive Symptom Subgroups Among Service Members With Mild Traumatic Brain Injury

OBJECTIVE: First, to describe the development of a novel postconcussive symptom (PCS) code set for identifying symptoms via medical records. Second, to apply it to a population-based cohort of service members with a history of mild traumatic brain injury (mTBI) per Military Health System (MHS) records, to identify distinct subgroups based on patterns of risk for specific PCS. SETTING: MHS. PARTICIPANTS: Population-based sample of service members with mTBI who served in the Army, Air Force, Navy, and Marine Corps and received a diagnosis within the MHS (n = 148 293). DESIGN: Retrospective cohort study using medical record data from the MHS spanning 2002 to 2021. MAIN MEASURES: In collaboration with clinical experts, we iteratively refined a novel PCS code set comprised of ICD-9/10 codes based on Neurobehavioral Symptom Inventory categories, when possible. We used latent class analysis (LCA) with a split-sample cross-validation procedure to identify subgroups of service members with mTBI based on probability of receiving each PCS diagnosis. RESULTS: The final PCS code set included 20 symptoms, spanning vestibular, sensory, cognitive, and mood/behavioral-related symptoms. The LCA supported 5 distinct subgroups, the most prevalent being the Minimal subgroup (59%), characterized by low probability of all PCS. The next most common class was the Headache class (16%), followed by the Mood-Behavioral (15%), Headache-Sleep (6%), and Headache-Mood-Sleep (5%) classes. CONCLUSIONS: Using a novel PCS code set leveraging routinely collected data, we identified 5 clinically meaningful and statistically distinct subgroups based on symptom patterns in a population-based cohort of service members with a history of mTBI. These subgroups provide a nuanced, person-centered understanding of symptoms among those with a history of TBI and can inform targeted interventions and policies aimed at meeting the needs of these service members. Further, findings establish a foundation for investigating risk factors and outcomes across subgroups, informing prognostication.