Identifying risk factors of post-COVID-19 conditions with machine learning and deep learning algorithms
Guohai Zhou1, Scott P Kelly2, Ling Li2
1Pfizer Global Medical Epidemiology, 500 Arcola Rd, Collegeville, PA 19426, USA.
Global Epidemiology
|October 23, 2025
Summary
Post-COVID-19 conditions (PCC) impact millions. Machine learning models identified age, comorbidity score, and healthcare use as key predictors for developing PCC after SARS-CoV-2 infection.
Area of Science:
- Epidemiology
- Machine Learning
- Public Health
Background:
- Post-COVID-19 conditions (PCC) affect millions globally, necessitating research into their epidemiology and risk factors.
- Understanding real-world drivers of PCC is crucial for early diagnosis and effective management.
Purpose of the Study:
- To quantitatively evaluate the progression to newly developed PCC.
- To identify individual-level risk factors for developing new PCC at various time points following SARS-CoV-2 infection.
Main Methods:
- Applied multiple machine learning and deep learning models to a large US electronic health database (2020-2022).
- Analyzed patients with recent COVID-19 infection to assess PCC development.
- Evaluated risk factors at 60, 74, 90, and 120 days post-infection.
Main Results:
- Patients with new PCC were older, had higher comorbidity scores, and were more likely to smoke or have obesity, hyperlipidemia, or hypertension.
- Machine learning models consistently identified age, Charlson comorbidity score, and early healthcare utilization as leading risk factors.
- Disseminated intravascular coagulation was a significant predictor for cardiovascular or secondary PCC.
Conclusions:
- Machine learning models effectively predict new PCC occurrence.
- Key predictors include Charlson comorbidity score, age, and healthcare utilization frequency.
- These models offer utility for individualized risk prediction of Post-COVID-19 conditions.
