Identification of Phenotypes Among COVID-19 Patients in the United States Using Latent Class Analysis

Abstract

Background Coronavirus disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2 or COVID-19) is a heterogeneous disorder with a complex pathogenesis. Recent studies from Spain and France have indicated that underlying phenotypes may exist among patients admitted to the hospital with COVID-19. Whether those same phenotypes exist in the United States (US) remains unclear. Using latent class analysis (LCA), we sought to determine whether clinical phenotypes exist among patients admitted for COVID-19. Methods We reviewed the charts of adult patients who were hospitalized primarily for COVID-19 at Greenwich Hospital and performed LCA using variables based on patient demographics and comorbidities. To further examine the reliability and replicability of the clustering results, we repeated LCA on the cohort of patients who died during hospitalization for COVID-19. Results Two phenotypes were identified in patients admitted for COVID-19 (N = 483). According to phenotype, patients were designated as cluster 1 (C1) or cluster 2 (C2). C1 (n = 193) consisted of older individuals with more comorbidities and a higher mortality rate (25.4% vs 8.97%, p < 0.001) than patients in C2. C2 (n = 290) consisted of younger individuals who were more likely to be obese, male, and nonwhite, with higher levels of the inflammatory markers C-reactive protein and alanine aminotransferase. When we performed LCA on the cohort of patients who died during hospitalization for COVID-19 (n = 75), we found that the distribution of patient baseline characteristics and comorbidities was similar to that of the entire cohort of patients admitted for COVID-19. Conclusion Using LCA, we identified two clinical phenotypes of patients who were admitted to our hospital for COVID-19. These findings may reflect different pathophysiologic processes that lead to moderate to severe COVID-19 and may be useful for identifying treatment targets and selecting patients with severe COVID-19 disease for future clinical trials.

Development and validation of a federated learning framework for detection of subphenotypes of multisystem inflammatory syndrome in children

“…Properly handling the between-site heterogeneity is helpful to improve the estimation and prediction accuracy in multi-site analysis. Most current work that uses centralized data, i.e., pooled data analysis, ignores the between-site heterogeneity and simply applies the standard 6, 1 on the pooled data, which could lead to biased results. 33,34 In contrast, with dMLCA, the between-site heterogeneity is carefully handled, which can provide us with improved estimation accuracy compared to centralized LCA and further facilitates higher prediction accuracy for patient class membership.…”

“…The latent class analysis (LCA) model is a widely used and effective method for disease subphenotyping. [2][3][4][5][6][7][8][9] However, MIS-C is rare while LCA models often involve many parameters that require sufficient samples to guarantee an accurate estimate. In the MIS-C study supported by the PEDSnet, 12,13 the smallest hospital only identified 32 MIS-C patients, which is far from being sufficient for obtaining reliable LCA results.…”

Aplicación de la calculadora de probabilidad fenotípica FEN-COVID en pacientes hospitalizados por COVID-19 en una población latinoamericana

“…11 Un enfoque alternativo lo ha constituido la búsqueda de fenotipos de presentación clínica que por lo general son basados en modelos de agrupación o clustering (por ejemplo, el método de los K vecinos, hierarchical clustering u otras técnicas de aprendizaje no supervisado) o modelos de clases latentes. Mediante este tipo de enfoque se han delimitado varios esquemas que orientan a los pacientes con COVID-19 en el entorno ambulatorio 12 , de hospitalización en sala general 8,[13][14][15][16] o el de cuidados intensivos 6,17 así como en condiciones específicas por ejemplo, SDRA 18,19 o la disfunción multiorgánica. 20 Así mismo se han identificado diferentes procesos fisiopatológicos que conducen a COVID-19 de moderado a grave que puedan ser útiles para precisar objetivos de tratamiento y seleccionar pacientes con enfermedad COVID-19 grave para ensayos clínicos futuros.…”

“…20 Así mismo se han identificado diferentes procesos fisiopatológicos que conducen a COVID-19 de moderado a grave que puedan ser útiles para precisar objetivos de tratamiento y seleccionar pacientes con enfermedad COVID-19 grave para ensayos clínicos futuros. 15 El trabajo de Gutiérrez y col. se posiciona como uno de los más importantes en el contexto de fenotipificación de pacientes hospitalizados, que además cuenta con una validación externa sobre la cohorte COVID-19@HULP incluyendo 2.226 pacientes y aportando una herramienta virtual para el cálculo probabilístico del fenotipo denominada FEN-COVID. 8 En el presente trabajo presentamos la aplicación de esta herramienta en 126 pacientes hospitalizados procedentes de dos instituciones latinoamericanas, se logró documentar la representación de dos de los tres fenotipos descritos con una preponderancia del fenotipo B (83% de los casos), algo similar a la cohorte de validación del trabajo original con 80%.…”

Detection of the Coronavirus

“…Several antibodies are detected using IgM, IgG, and IgA by the ELISA (Enzyme -linked immunosorbent assay) or luminescence Immunoassay. 19 The SARS-CoV-2 not merely disturbs the respiratory function but also causes the pneumonia, gastrointestinal, cardio and neurological problems. It also causes the Kawasaki-like diseases in which the immune system become unable to fight against many inflammatory pathogens and leads to ophthalmic problems, fever, erythema, adenopathy and loss of smell and taste.…”