Behavioral Health, Biological Complications, and Hospital Utilization in Adults With Sickle Cell Disease: Limits of Statewide Administrative Data
Daniel M Sop, Wally R Smith, Yue M Zhang, Shirley JohnsonAbstract
Background
In our prior statewide analysis, we found that ICD-defined care-related complications were the strongest drivers of prolonged hospitalization and cost among adults with sickle cell disease (SCD). We next examined the role of behavioral health (BH) conditions known contributors to healthcare utilization to understand whether they independently drive hospital outcomes or primarily act through biological complications. Large statewide discharge datasets like Virginia Health Information (VHI) are often used to guide care-delivery planning, yet they lack clinical granularity for SCD-specific complications such as acute chest syndrome, stroke, or organ damage phenotypes. These limitations may obscure how behavioral and biological factors interact to drive outcomes in SCD.
Methods
We conducted a retrospective analysis of 15,737 adult inpatient admissions (6,170 patients) with sickle cell disease from 2016–2018 to evaluate the impact of behavioral health conditions, including psychiatric and substance use disorders, on hospital utilization and outcomes. Length of stay was modeled using negative binomial generalized estimating equations. Total hospital charges were modeled using generalized linear models with a gamma distribution and log link. In-hospital complications were modeled as a binary outcome using logistic generalized estimating equations. Models adjusted for age, admission type, ZIP-level income score, payer type, and comorbidity burden, with repeated admissions accounted for using an exchangeable correlation structure.
Results
Patients averaged 2.5 admissions over 3 years, with a highly skewed distribution that included one patient with 444 admissions. Mean age was 37 years, mean LOS 5 days, mean hospital charges $34,322, and complication rate 3.9%. Behavioral health conditions were present in 13% of admissions (psychiatric 4%, substance use 9.8%). Patients with BH conditions had slightly lower measured comorbidity burden than those without (mean Elixhauser score 2.49 vs 2.96, p ≈ 0.001). Psychiatric disorders were modestly associated with complications after adjustment (p = 0.0016), while substance use disorders were not. In repeated-measures models, substance use disorders were associated with slightly shorter LOS (estimate ≈ –0.092, p = 0.007), and neither BH condition significantly predicted hospital charges. In contrast, overall comorbidity burden strongly predicted LOS and charges (LOS estimate ≈ 0.104, p < 0.0001).These findings suggest that while behavioral health conditions influence patterns of utilization, biological comorbidity burden remains the dominant determinant of hospital outcomes. However, inspection of discharge coding revealed a critical limitation: the dataset under-captured SCD-specific complications associated with high mortality; such as acute chest syndrome, neurologic injury, and severe organ damage; while grouping them into broader administrative categories. This misclassification likely attenuates the apparent impact of both behavioral and disease-specific complications.
Conclusions
Behavioral health conditions contribute to healthcare utilization among adults with SCD, but biological complications appear to drive the severity, cost, and downstream mortality risk. Statewide administrative datasets like VHI provide useful population-level signals but fail to capture key SCD-specific complications, potentially leading to underestimation of disease burden and misallocation of healthcare resources. Taken with our earlier results, this work highlights why SCD care planning should be informed by clinically curated, longitudinal registries designed for SCD (and capable of supporting quality improvement and guideline adherence monitoring), rather than relying solely on general discharge datasets; an approach exemplified by GRNDaD and related multi-site efforts.