THU318 Endocrine Phenotype Drives Mortality Risk At Presentation During Index Hospitalization With SARS-CoV-2 Infection: Differential Trait Signatures May Shed Some Light

Abstract

Abstract Disclosure: K. Baez: None. N. Torres Rivera: None. J. Rebernigg: None. B. Contreras: None. S. Yenari: None. V. Farhangi: None. K. Hamad: None. W. Wiese-Rometsch: None. R. Smith: None. Evidence validates that COVID-19 severity is linked to diabetes mellitus (DM), obesity (Obes) or both (ObesDM). Though to our knowledge, variance explained (VE%) by putative clinical traits evoked by SARS-CoV-2 infection across these metabolic phenotypes has not been reported. We hypothesized that sequenced machine learning could aggregate trait ensembles while computing their attendant VE% associated with hospital mortality in COVID-19 patients who presented for index hospitalization. Demographics, ICD-10-based Elixhauser comorbidity, laboratory results within 24h, COVID-19 directed treatment across 6-surges and administrative data were extracted under IRB exemption from EMR. Generalized regression with LASSO identified traits associated (p<.05) with mortality controlling for age, sex and COVID-19 treatment. Univariate regression modeling created a receiver operating characteristic (ROC) curve computing Youden Index estimating optimal cut-point associated with mortality. Boosted Tree (BT) computed within-phenotype respectively area under ROC (AUROC) representing multi-trait prediction accuracy and proportion of modeled VE% (R2) for each trait. Continuous data summarized with median [IQR] was compared using Kruskal-Wallis test. Discrete data summarized as proportions were compared with chi-square test. Significant two-tailed p-value ≤.0125 was Bonferroni corrected. Among 6275 adults consecutively discharged between March 14, 2020 through July 31, 2022 with < 20% missing data for salient traits, there were 3499 (56%, Control); 986 (16%, DM); 1050 (17%, Obes); and 740 (12%, ObesDm). Results are likewise sequenced. Age (70[54-82], 74[63-82], 59[44-72], 64[54-74] distributed across males (52%, 63%, 49%, 53%), Whites (83%, 78%, 74%, 72%), Blacks (6%, 10%, 12%, 15%) and other races (11%, 11%, 14%, 13%). Similar intergroup comorbidities included hypertension (56%), iron deficiency anemia (23%), chronic pulmonary disease (21%), neurological disease (21%), renal failure (19%), coagulation disease (16%), and heart failure (15%). Cut points were CRP >6.3 mg/dL, ferritin >563 ng/mL, LDH >342 U/L, D-dimer >0.75 μg/mL, glucose >133 mg/dL, albumin <2.96 g/dL, creatinine >1.17 mg/dL, Hb <14.5 gm/dL, ANC/ALC Ratio >6.2, and APC/ALC Ratio >256. Trait contributions sequenced included Control R2 (.58) 9%, 11%, 18%, 17%, 10%, 5%, 10%, 11%, 5%, 4%; DM R2(.63) 7%, 5%, 7%, 16%, 5%, 20%, 5%, 24%, 7%, 4%; Obes R2 (.60) 7%, 5%, 5%, 4%, 10%, 9%, 11%, 39%, 6%, 3.%; and ObesDM R2 (.54) 14%, 9%, 7%, 5%, 11%, 14%, 9%, 21%, 5%, 4%. In conclusion, immunometabolic response contributing to mortality risk varies among endocrinopathies at presentation with SARS-CoV-2 infection. Deeper characterizations at intersection of metabolic diseases and immunometabolic traits may suggest treatment strategies to mitigate dysregulated host response evoked by the COVID -19. Presentation: Thursday, June 15, 2023