Forecasting COVID-19 inpatient mortality using fundamental parameters in resource-constrained settings: a countrywide...

Accurate mortality prediction was essential for guiding hospitalization and management during the COVID-19 pandemic, particularly in low-resource settings with already fragile health systems. In such regions, political instability and weakened healthcare infrastructure amplified viral spread while constraining response capacity. Hospitalization was, therefore, mostly restricted to the most critical cases due to shortages of beds, staff, and equipment, leaving frontline physicians to make rapid triage decisions under pressure. Yet, most existing prognostic tools were unsuitable for use in such resource-constrained contexts. To address this gap, we developed and validated a simplified, resource-directed score for predicting COVID-19 mortality.

This nationwide multicenter cohort study involved prospective data collection and retrospective analysis of hospitalized and non-hospitalized COVID-19 patients in Syria. The study was conducted across nine healthcare centers in four major Syrian cities, reflecting the diverse levels of care available in the war-torn country. It included 3,199 hospitalized and 293 non-hospitalized patients. Comprehensive datasets were collected, including demographic characteristics, clinical presentation, vital signs, laboratory tests, and imaging findings. The primary outcome was in-hospital mortality prediction, with secondary outcomes including mortality prediction for intensive care unit (ICU) patients and those managed at home. To develop the scores, we employed a regression coefficient-based scoring system methodology. Finally, the performance of the developed score was compared with established mortality prediction scores.

LR-COMPAK, utilizing six easily obtainable variables (age, comorbidities such as kidney disease and malignancy, pulse rate, oxygen saturation, and consciousness), showed superior predictive performance with an area under the receiver operating characteristic curve of 0.88 [0.87-0.90] and explained 52% of mortality variance, demonstrating applicability to non-hospitalized patients as well. Regional and temporal disparities in severity scores and mortality highlighted variations in healthcare capacity. The LR-ALBO-ICU score, additionally incorporating lactate dehydrogenase and bicarbonate levels, effectively predicted ICU mortality, showing great value in critical care decision-making.

LR-COMPAK and LR-ALBO-ICU offer practical, effective tools for mortality prediction in COVID-19 patients, particularly in resource-limited settings. These tools can guide hospitalization decisions, optimize resource allocation, and improve patient outcomes, facilitating translation between resource-rich and resource-limited healthcare environments.

Not applicable.