Can simple bedside data reveal sepsis types in Uganda?

Sepsis is not a single illness: patients with similar symptoms can have very different underlying biology, or molecular subtypes, that may respond differently to treatment. Identifying those molecular signatures usually requires advanced laboratory testing, which is hard to access in low- and middle-income countries (LMICs). To ask whether basic bedside information could substitute for expensive molecular diagnostics, M. G. Cummings and colleagues performed a secondary analysis of two prospective observational sepsis cohorts in Uganda. The study focused on adults (≥18 years) hospitalized with sepsis at two hospitals — Entebbe Regional Referral Hospital (urban) and Tororo General Hospital (rural). Researchers used available transcriptomic profiling for 355 patients and proteomic profiling for 495 patients to define molecular sepsis subtypes, and then tested whether simpler, bedside-adaptable classifier models could sort patients into those same subtypes. They also compared model performance against molecular classifications developed in high-income countries to see how well bedside data track internationally derived biological patterns.

The team evaluated bedside-adaptable clinical models and clinico-microbiological models that added rapid diagnostic results for HIV, malaria, and tuberculosis. Two cohort datasets were analyzed: RESERVE-U-2-TOR and RESERVE-U-1-EBB. In RESERVE-U-2-TOR, clinical models using demographics and bedside physiological variables achieved moderate discrimination for assigning transcriptomic and proteomic subtypes, with AUROCs of 0.75 (95% CI, 0.69–0.81) and 0.73 (0.66–0.80), respectively. These models showed strong calibration (Integrated Calibration Index E avg ≤0.015). Including rapid HIV, malaria, and tuberculosis test results produced similar AUROCs (0.76 and 0.74) with E avg ≤0.016. In RESERVE-U-1-EBB, discrimination was more variable (AUROC range 0.63–0.75) while calibration remained acceptable (E avg ≤0.053). When the same simple models were evaluated against molecular sepsis frameworks derived in high-income countries, performance patterns were similar: acceptable calibration with only moderate discrimination.

Taken together, the findings show that bedside clinical information — and even bedside data combined with quick pathogen tests for HIV, malaria, and tuberculosis — can provide reasonably well calibrated estimates of molecular sepsis groupings in a resource-limited setting, but they do not reliably distinguish fine-grained molecular signatures. In other words, bedside variables seem to reflect overall illness severity and yield useful risk estimates, yet they incompletely capture the distinct transcriptomic and proteomic patterns that define molecular subtypes. The authors conclude that these bedside-adaptable models are not ready to replace molecular diagnostics as stand-alone tools. To move toward precision sepsis care in LMICs, the study recommends expanding acute-care laboratory capacity and improving access to scalable, low-cost molecular biomarker assays so clinicians can identify biological subtypes and tailor care appropriately.