Application and challenges of the sickle cell outcome grading system (SCOGS) for classifying acute chest syndrome severity Free

Introduction: As a leading cause of hospitalization and mortality in sickle cell disease, acute chest syndrome (ACS) presents with variable severity, ranging from mild hypoxia to multi-organ failure. Management involves supportive care (e.g., oxygen, transfusions) and critical interventions (e.g., mechanical ventilation, vasopressors), but treatment decisions are complicated by heterogeneous presentations and lack of standardized severity assessment (Howard et al., 2015). The Sickle Cell Outcome Grading System (SCOGS) was devised by a panel of adult and pediatric providers using a three-round Delphi process. SCOGS provides definitions, diagnostic criteria, severity grades (1-5) based on the Common Terminology Criteria for Adverse Events (CTCAE), and frequency classifications (acute, chronic, and chronic with exacerbation) for 53 sickle cell disease-related health outcomes. SCOGS classified ACS severity into five tiers based on CTCAE-guided clinical interventions, from minimal support to fatal outcomes. This study evaluates the application in a real-world cohort by identifying challenges in scoring and limitations for the classification of severity in ACS using SCOGS.

Methods: We applied the SCOGS to two cohorts of pediatric and adult patients with SCD, the multicenter Sickle Cell Clinical Research and Intervention Program (SCCRIP) and the single institution Boston Medical Center cohort (BMC cohort). We used existing clinical and demographic variables to quantify ACS event severity. SCCRIP is a large, multi-center longitudinal SCD cohort including pediatric and adult patients with all SCD genotypes (Hankins et al., 2018), and the BMC cohort is a clinical cohort in which ACS events are classified using the International Classification of Diseases version 10. SCOGS defines ACS severity as: Grade 1 (no supplemental oxygen, transfusions, or critical interventions except antibiotics), Grade 2 (FiO2 >21% to <50% or simple transfusion, without non-invasive ventilation, or exchange transfusion), Grade 3 (FiO2 >50%, non-invasive ventilation or exchange transfusion), Grade 4 (exchange transfusion with critical interventions like mechanical ventilation or vasopressors), and Grade 5 (death from ACS). Analyses were conducted with R, version 4.5.1, and SAS 9.4 in SCCRIP and the BMC cohort, respectively. ACS episode and patient-level variables assessed grade distribution and scoring barriers. We investigated the availability of variables necessary for scoring, ease of chart review to extract variables, percentage of the cohort diagnosed with ACS that could be scored, ease of grading, and any observations that provided support for the feasibility of using SCOGS.

Results: In the SCCRIP database 454 patients across 3 sites (with separate electronic health records) experienced 1,088 ACS episodes. Of these, 56.2% were male, the median age at episode was 4 (IQR 2-8), and genotypes were 75.8% HgbSS/Sβ⁰thalassemia, 20.0% HgbSC, and 4.2% others. SCOGS ACS episodes were Grade 1 (796, 73.2%), Grade 2 (222, 20.4%), Grade 3 (14, 1.3%), and no Grade 4 or Grade 5. In SCCRIP, 56 (5.1%) ACS episodes were non-scorable due to missing data in oxygen therapy (93% missing) or the mode of transfusions. In the BMC cohort, 109 ACS episodes were recorded, 51.8% among males. The median age at the episode was 33.5 (IQR 27-39) years old, and 83.9% had the HbSS genotype. SCOGS grades by episode were distributed as Grade 1 (24, 22.0%), Grade 2 (68, 62.4%), Grade 3 (10, 9.2%), Grade 4 (3, 2.7%), and Grade 5 (4, 3.7%). All episodes in the BMC cohort were scorable.

Conclusions: SCOGS provides a consensus-driven framework for ACS severity assessment, but its structure faces significant challenges in real-world application. The skewness of greater ACS severity in the BMC cohort likely reflects the older age of that population and the non-scorable events in SCCRIP, which could have led to potential misclassification, particularly between Grades 2,3, and 4. Comprehensive chart review from a single institution, as seen in the BMC cohort, enabled 100% SCOGS scoring, but difficulties persist in multi-institutional datasets. Missing data and absent variables for FiO2 and non-invasive ventilation led to unclassified cases. Refinements, including improved data capture for oxygen therapy and potential use of Machine Learning algorithms to impute missing data, and automation of SCOGS scoring, are essential to enhance SCOGS's utility in clinical practice and research.