Derivation and validation of clinical subphenotypes of adult sepsis in Africa: A latent class analysis of infection site and organ dysfunction
Year: 2026
Type: preprint
Abstract: ABSTRACT Objective Sepsis is a heterogeneous syndrome characterized by substantial variation in infection source, organ dysfunction, and clinical outcomes, and a disproportionate burden in Africa. We aimed to derive and externally validate clinical subphenotypes among adults hospitalized with sepsis in Africa. Design We used latent class analysis with infection source and organ dysfunction as model inputs. We labeled classes post hoc according to their predominant clinical features. We evaluated associations between subphenotypes and 30-day mortality using multivariable logistic regression. Setting Hospitals in Uganda, Rwanda, Liberia, Tanzania, and Malawi. Patients We pooled data from three prospective cohorts of adults hospitalized with sepsis in Uganda, Rwanda, and Liberia for model derivation and independent cohorts from Tanzania and Malawi for external validation. Interventions None. Measurements and Main Results The derivation dataset included 467 adults with sepsis, of whom 128 (27.4%) died within 30 days of admission. Latent class analysis identified four subphenotypes: Class 1 (heterogeneous; n=227; 30.8% mortality), Class 2 (abdominal; n=141; 31.9%), Class 3 (pulmonary; n=68; 16.2%), and Class 4 (soft tissue; n=31; 6.5%). Mortality differed significantly across the four classes (p=0.003). Compared with Class 1, Classes 3 (aOR 0.22, 95% CI 0.10–0.47) and 4 (aOR 0.14, 95% CI 0.02–0.52) had lower odds of 30-day mortality. External validation in 316 adults demonstrated high classification certainty (entropy=0.78; mean maximum posterior probabilities, 0.89-0.92), with similar mortality patterns across corresponding subphenotypes. Conclusions Latent class analysis identified reproducible clinical sepsis subphenotypes defined by infection source and organ dysfunction with distinct mortality risks. These externally validated subphenotypes may improve risk stratification and inform subphenotype-based clinical trials and treatment strategies in Africa. KEY POINTS Question Can routinely available infection-source and organ-dysfunction data identify reproducible clinical sepsis subphenotypes with distinct mortality risks among adults hospitalized in Africa? Findings In a pooled cohort study using latent class analysis, we identified four clinical sepsis subphenotypes among 467 adults and externally validated them in 316 adults from independent African cohorts. The subphenotypes had significantly different 30-day mortality, ranging from 6.5% to 31.9%, with similar mortality patterns in the validation cohorts. Meaning Clinically identifiable sepsis subphenotypes based on routinely available data may provide a pragmatic framework for risk stratification and patient stratification in future sepsis trials in Africa.
Source: medRxiv
Authors: Phoebe Gruccio, Louisa Edwards, Reagan Kakande, Muhammad Zulfiqar, Michael Ndyomugabe, Victor Adejayan, Philemon Ojuman, William S. Girard, Eva Otoupalova, Paulin R. Banguti, Dennis Hopkinson, Aamer Syed, Jean Paul Mvukiyehe, Venance P. Maro, Matthew P. Rubach, John A. Crump, Cozie Gwaikolo, Richard Ssekitoleko, Elisabeth D. Riviello, Edwin Nuwagira, Christopher C. Moore
Institutions: University of Virginia, Stanford University, Mbarara University of Science and Technology, Tulane University, University of Rwanda, Virginia Commonwealth University, Virginia Commonwealth University Medical Center, Kilimanjaro Christian Medical Centre, Duke University, University of Otago, University of California, San Francisco, World Health Organization - Uganda, Beth Israel Deaconess Medical Center
Language: English
Subfield: Epidemiology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7211972305