Understanding the Characteristics of High-Cost Healthcare Users in a Persistent Concussion Clinic for Youth

Abstract

<h3>Research Objectives</h3> To investigate the predictors of those with high healthcare costs in a Pediatric Persistent Concussion Clinic. <h3>Design</h3> Cohort study. <h3>Setting</h3> A specialty outpatient clinic in a children's rehabilitation hospital in Toronto, Canada. These services included occupational therapy, physical therapy, social work, and/or neuropsychology. <h3>Participants</h3> A referred sample of 461 youth that had been experiencing post-concussion symptoms for at least 4 weeks. Youth were between 3 to 18 years. 47.4% were male and 53.6% were female. <h3>Interventions</h3> Not applicable. <h3>Main Outcome Measures</h3> Total direct healthcare costs from the Persistent Concussion Clinic. Clients whose total direct healthcare costs were above the 80th percentile ($1655.48) were classified as high-cost users. <h3>Results</h3> Four logistic regressions were conducted and compared to each other using likelihood ratio tests. Each model contained variables from the previous model along with an additional group of variables. Variable selection was determined based on previous conceptual frameworks for predicting high-cost users of healthcare systems. Model 1: Demographic Characteristics (age, sex). Model 2: Medical History (psychiatric history, number of previous concussions, history of therapy). Model 3: Symptom Severity (number of symptoms, symptom duration). Model 4: Service Utilization (number of services used, being prescribed new medication). Model 4 revealed clinically significant effect sizes (i.e., OR > 2). Being prescribed new medication, b = 1.13, p < .001, OR = 3.096 (95% CI: 1.634, 6.027) and the number of services utilized, b = 1.55, p < .001, OR = 4.731 (95% CI: 3.368, 7.011) were found to be significant predictors. A likelihood ratio test shows that this model performs significantly better than the logistic regression without service utilization variables, χ2(2) = 154.2, p < .001. <h3>Conclusions</h3> Service utilization is a significant predictor of high-cost users. Future research should incorporate this information into a prediction algorithm, whereby early service utilization patterns may inform us as to who will have higher healthcare costs. <h3>Author(s) Disclosures</h3> The authors have no conflicts of interest to declare.