C67-16 Phenotypic Comorbidity Clusters Predict Glucocorticoid Outcomes in Hospitalized Influenza: A Learning Health System Analysis

Rationale Systemic glucocorticoids treatment in patients hospitalized with influenza remains controversial. While guidelines, including ATS, caution against their use due to harm potential, this recommendation is driven by data that often treats patient population as uniform. We hypothesized that risk-benefit profile of glucocorticoids is not uniform but varies across patient phenotypes defined by functional comorbidity burden. This study, conducted as a learning health system (LHS) landscape analysis, aimed to identify these phenotypes and evaluate association of glucocorticoid use and its outcome. Methods Under IRB exemption, a retrospective landscape analysis of 886 consecutively hospitalized adults with laboratory-confirmed influenza A or B between November 2021 and September 2025. Patients receiving steroids for non-influenza indications were excluded. K-means clustering integrated 38 Elixhauser comorbidity categories with laboratory traits that reflected pathophysiological status to identify distinct patient phenotypes. Three patient clusters were identified. Within each, we compared primary and secondary outcomes between patients who did and did not receive systemic glucocorticoids. Results Demographics (pooled) with clinical course and outcomes are presented in Table. Cluster 1 (N = 683; Low Comorbidity Burden) exhibited a low median Elixhauser Comorbidity Index (ECI) of 3 and high functional status (median HALP score: 22). Within this cluster, glucocorticoid use (n = 307) versus non-use (n = 376) was associated with a minimal difference in hospital mortality (1% vs. 0%), no difference in ICU admission (2% vs. 2%) or hospice discharge (2% vs. 2%). Cluster 2 (N = 182; High Comorbidity Burden) had a high median ECI of 5 and high comorbidity prevalence, including Congestive Heart Failure (46%) and Anemia (49%), with lower functional status (median HALP score: 16). In this phenotype, glucocorticoid use (n = 94) versus non-use (n = 88) was not associated with increased mortality (1% vs. 0%) but was associated with lower rates of ICU admission (4% vs. 8%) and hospice discharge (3% vs. 7%). Cluster 3 (N = 21; Severe Inflammatory Phenotype) was characterized by profound lymphopenia and high inflammation. No differences were observed, though the group was distinct. Conclusions The association of glucocorticoid therapy with clinical outcomes in hospitalized influenza appears heterogeneous and dependent on the patient's phenotype. Within Cluster 1, steroids provided no observable benefit, aligning with current ATS guideline. Conversely, in patients within Cluster 2, glucocorticoid use was not associated with harm and showed a potential signal for improved secondary outcomes. While statistically underpowered, our LHS landscape analysis suggests that a 'one-size-fits-all' guideline for glucocorticoid therapy is insufficient, supporting a phenotypic approach to better risk-stratify patients.