Predicting death from COVID-19 using pre-existing conditions: implications for vaccination triage

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Global SARS-CoV-2 vaccine shortages necessitate prioritizing high-risk populations for inoculation.
• Identifying individuals most vulnerable to COVID-19 mortality is crucial for effective public health strategies.

Purpose of the Study:

• To develop and validate a predictive model for COVID-19-related death using clinical data.
• To compare the model's performance against traditional age-based risk stratification.

Main Methods:

• Longitudinal clinical data from 15,502 laboratory-confirmed SARS-CoV-2 patients in metropolitan Detroit were analyzed.
• Least absolute shrinkage and selection operator (LASSO) regression was used to develop a parsimonious prediction model incorporating 36 pre-existing conditions and demographic variables.
• The model was prospectively validated on a separate cohort of patients.

Main Results:

• The final prediction model comprised 14 variables, including 11 comorbidities, and was developed using data from 11,635 patients (6.1% case fatality ratio).
• Prospective validation on 3,867 patients (4.8% case fatality ratio) demonstrated the model's effectiveness.
• The 14-variable model identified 6% more individuals at risk of death compared to a threshold of 65 years, though it offered no advantage below age 45.

Conclusions:

• A validated predictive model can enhance the identification of individuals most likely to benefit from COVID-19 vaccination.
• Targeted vaccination strategies based on predictive modeling may lead to significant reductions in COVID-19 mortality.