Behavioral health and biophysical complexity as predictors of healthcare utilization in adults with sickle cell disease

Abstract

Abstract Background: Identifying predictors of high healthcare utilization in adults with Sickle Cell Disease (SCD) remains a clinical and operational priority. This study examined whether behavioral health (Bh) and Biophysical complexity (Bio) scores predict SCD utilization patterns and costs, independent of prior healthcare use. Methods: We conducted a retrospective analysis of 606 adults with SCD, including 47 high utilizers (≥3 inpatient admissions/year). Descriptive statistics, t-tests, and Spearman/Pearson correlations assessed relationships between Bh and Bio scores and healthcare encounters. Bh scores were the sum of ICD10 documented behavioral health diagnoses (11 items) at any visit, while Bio scores were the sum of 26 items: ICD10 documented complications (10 items) plus ICD9-documented comorbidities (16 items), each item scored as 0 or 1. Zero-inflated negative binomial regression evaluated the predictive effects of Bh and Bio scores on outpatient visits, emergency department (ED) use, inpatient days, discharges, and total costs. Statistical significance was set at p < 0.05. Results: High utilizers had significantly higher Bh and Bio scores than non-high utilizers (Bh: mean difference = 0.43, p = 0.023; Bio: mean difference = 1.52, p = 0.002). Spearman correlations showed that both Bh and Bio scores were positively correlated with ED usage, inpatient days, inpatient discharges, outpatient visits, and total costs (p < 0.0001). Zero-inflated models predicting utilization outcomes from Bh and Bio indicated that, in the dominant component of the sample (component mixing probability ≥ 90%), both Bh and Bio significantly predicted ED use (Bh: β = 0.32, p = 0.002; Bio: β = 0.25, p < 0.001). This corresponds to a 38% increase in ED usage for every one-unit increase in Bh (exp (0.32) ≈ 1.38) and a 28% increase for every one-unit increase in Bio (exp (0.25) ≈ 1.28). Only Bio significantly predicted inpatient discharges (β = 0.46, p < 0.001) for 59% of patients and inpatient days (β = 0.14, p < 0.001) for 25% of patients. Among high utilizers, Bh was not a significant predictor of utilization. In contrast, Bio remained a consistent predictor, accounting for 48% of patients' inpatient days (β = 0.84, p = 0.0002) and 62% of patients' inpatient discharges (β = 0.40, p < 0.0001). Conclusions: Bh and Bio scores are significantly associated with healthcare utilization in adults with SCD. Bio scores, in particular, demonstrate consistent predictive value across utilization types and cost, especially among high utilizers. Integrating Bh and Bio metrics into SCD team care assignment models may optimize resource allocation and improve outcomes for individuals with complex SCD needs.