Real-World Effectiveness and Resource Utilization of Red Blood Cell Exchange in Adult Sickle Cell Disease: a Nationwide Robust Analysis from 2017-2022

Abstract Background: Red blood cell exchange (RBCx) is a critical, guideline-recommended therapy for severe, life-threatening complications of sickle cell disease (SCD), including acute chest syndrome and stroke. Despite its established biological rationale, there is a paucity of large-scale, contemporary evidence on its real-world utilization patterns, clinical effectiveness outside of trial settings, and associated healthcare burden. This study sought to evaluate these metrics using a nationally representative cohort of hospitalized adults with SCD.Methods: We performed a retrospective cohort study using the Healthcare Cost and Utilization Project (HCUP) National Inpatient Sample (NIS) from 2017 to 2022. The NIS is the largest all-payer inpatient database in the United States. Adult (≥18 years) hospitalizations were included if any listed ICD-10-CM diagnosis code was for SCD (D57.xx). The intervention, RBCx, was identified using the specific ICD-10-PCS procedure code 6A550Z0. To enhance reproducibility, we employed a doubly robust estimation strategy to control confounding by indication. This involved two stages: first, a multivariable logistic regression model was built to calculate a propensity score for receiving RBCx. This model included patient demographics (age, sex), payer status, hospital characteristics (teaching status, bedsize), and clinical factors (admission type, primary vs. secondary SCD diagnosis, and flags for eight critical comorbidities: acute chest syndrome, stroke, pulmonary hypertension, avascular necrosis, venous thromboembolism, sepsis, chronic kidney disease, and acute kidney injury). Second, these scores were used to generate stabilized inverse probability of treatment weights (IPTW), which were truncated at the 1st and 99th percentiles to prevent undue influence from extreme weights. Primary outcomes were analyzed using multivariable models adjusted for the same confounders, with the final estimates weighted by the product of the IPTW and the NIS discharge weights (DISCWT). Due to rare mortality events, a penalized logistic regression (Firth's logit) was used for the mortality outcome. Skewed, positive continuous outcomes (LOS, charges) were analyzed using survey-weighted Gamma generalized linear models with a log link. All statistical analyses were conducted using StataNow 19.5 MP.Results: The unweighted sample comprised 185,332 adult SCD hospitalizations. Applying discharge weights, this represents a national total of 926,660 hospitalizations over the 6-year study period, or an annual average of 154,443 hospitalizations. RBCx was performed in a small fraction of cases, corresponding to a national total of 2,775 procedures, or an annual average of 463 procedures. Patients receiving RBCx were significantly more likely to have a primary SCD diagnosis, present with acute stroke, and be treated at large, urban teaching hospitals. After doubly robust adjustment, the association between RBCx and in-hospital mortality was not statistically significant (adjusted Odds Ratio [aOR] 1.36, 95% CI 0.68-2.73, p=0.381). In contrast, RBCx was associated with a significant increase in healthcare resource utilization. Patients undergoing the procedure had significantly longer hospital stays (adjusted Rate Ratio [aRR] 1.45, 95% CI 1.30–1.62, p<0.001) and incurred substantially higher total hospital charges (aRR 1.84, 95% CI 1.65–2.06, p<0.001), representing an adjusted average of 45% longer stays and 84% higher costs, respectively.Conclusions: In a large, nationwide analysis, RBCx is utilized as a highly targeted but infrequent therapy for the most severely ill adult patients with SCD. The absence of a statistically significant mortality signal should be interpreted cautiously; while the therapy likely mitigates a high baseline risk of death, this finding must be viewed in the context of deep confounding by indication. Although we used rigorous statistical methods, the potential for residual confounding from unmeasured variables (e.g., laboratory data, vital signs) remains a significant limitation of this administrative database study. Notwithstanding these limitations, our findings confirm that this clinically vital intervention is associated with a substantial and statistically significant increase in resource utilization. This real-world evidence is crucial for informing clinical guidelines, shaping healthcare policy, and guiding future cost-effectiveness analyses for this life-saving procedure.