Stratification Of Covid-19 Diagnostic Profiles Via Finite Mixture Modeling: Immunology And Rehabilitation Implications

The SARS-CoV-2 pandemic has created a paradigm shift in the understanding of host-pathogen interactions and their sequelae. While acute symptomatic manifestations have been extensively studied, the long-term immunological and comorbid ramifications remain an emergent area of inquiry. PURPOSE: To elucidate heterogeneity in diagnostic profiles among COVID-19 patients, with a focus on secondary pathologies and comorbidities, utilizing finite mixture modeling. METHODS: Protocol was approved by Trinity-Health Ann Arbor Hospital IRB #1. Data were collected from 42,974 PCR-confirmed COVID-19 cases via Electronic Health Records across 31 healthcare facilities. The cohort was predominantly White (73%). The sample was female-dominant (55%). A compendium of 850 diagnostic indicators and comorbidities was scrutinized, but analysis was restricted to pathologies in >5% of the cohort. Finite mixture models were employed to classify the cohort into distinct diagnostic profiles. These were validated based on statistical goodness-of-fit, parsimony, and clinical interpretability. RESULTS: Three diagnostic profiles were returned: (1) Minimal Comorbidity Profile (67%, n = 28,888), (2) Elevated Prevalence of Common U.S. Pathologies like hypertension and hyperlipidemia (24%, n = 10,426), and (3) High Comorbidity Burden, including lung-, heart-, and kidney-related conditions (9%, n = 3,660). Probabilistic assignment to these profiles yielded high classification confidence, with mean posterior probabilities of .98, .89, and .94, respectively. A moderate correlation was observed between these profiles and the multiplicity of recorded SARS-CoV-2 infections (Cramér's V = .3). Specifically, 91% of the Minimal Comorbidity Profile had a singular recorded infection, while 30% in the High Comorbidity Burden profile had multiple. CONCLUSIONS: The findings underscore the potential for targeted immunological and rehabilitative interventions, particularly for profiles with elevated comorbidity burden. This stratification also suggests avenues for focused research into specific comorbidity-COVID-19 interactions. ACKNOWLEDGEMENTS: Data collection was conducted during the active employment of Yager within the Trinity Health organization at St. Joseph’s Hospital in Syracuse, NY.