TRUJILLO, Peru — A team of Peruvian and U.S.-based researchers developed a predictive model that could help doctors identify hospitalized COVID-19 patients at higher risk of death, offering a potential tool for improving care and saving lives in future outbreaks.
The study, led by Dr. Ruben Kenny Briceno of Michigan State University's Global Health Institute, conducted in collaboration with physicians from Universidad César Vallejo in Trujillo, Peru, analyzed the medical records of 2,000 patients hospitalized during the first year of the pandemic. The research focused on identifying biological, clinical, and laboratory factors most strongly associated with COVID-19 mortality.
The findings, published in the journal Risk Management and Healthcare Policy, revealed that older age, certain symptoms, and specific lab results were among the strongest predictors of mortality. The model developed by the team predicted mortality with 76% accuracy.
“This model provides a practical and evidence-based way to assess risk early in a patient’s hospitalization,” Briceno said. “It can help clinicians prioritize care and allocate resources more effectively, especially in settings with limited capacity.”
Key Risk Factors Identified
The study found that patient mortality was more likely to be in patients older than age 60, male, and having underlying conditions such as hypertension, type 2 diabetes, obesity, and chronic kidney disease. Common symptoms among those patients included fever, fatigue, shortness of breath, and sore throat.
Laboratory results also played a critical role. Deceased patients had higher levels of leukocytes, neutrophils, urea, and ferritin, which are markers often associated with inflammation and organ stress. Lower levels of hemoglobin, lymphocytes, and platelets were also linked to higher mortality rates. Imaging results further supported the model. Patients with unilateral lung consolidation and severe ground-glass opacities on CT scans had higher mortality rates.
A Model for Future Preparedness
Using logistic regression analysis, the researchers developed a model that incorporates 15 variables, including age, symptoms such as productive cough and fatigue, and laboratory values such as ferritin and urea levels. The model demonstrated strong predictive power, with an area under the curve (AUC) of 0.84, indicating high accuracy in distinguishing between survivors and non-survivors.
The study’s authors emphasized that while the model was developed using data from Peru’s La Libertad region, its structure could be adapted and validated in other areas during future pandemics.
“This research not only helps us understand the factors that contributed to COVID-19 mortality in Peru, but also provides a foundation for improving clinical decision-making in future health emergencies,” said Dr. Irma Luz Yupari-Azabache, one of the study’s co-authors.
The researchers hope their findings will inform public health strategies and hospital protocols, particularly in resource-limited settings. The patient group will continue to be studied, and follow-up studies will be published when more long-term effects are identified.
“This model is a step toward more equitable and data-driven care,” Briceno said. “It’s about using what we’ve learned to be better prepared next time.”