Using machine learning involving diagnoses and medications as a risk prediction tool for post-acute sequelae of COVID-19 (PASC) in primary care

• Published in: BMC medicine Vol. 23; no. 1; pp. 251 - 12
• Main Authors: Lee, Seika, Kisiel, Marta A., Lindberg, Pia, Wheelock, Åsa M., Olofsson, Anna, Eriksson, Julia, Bruchfeld, Judith, Runold, Michael, Wahlström, Lars, Malinovschi, Andrei, Janson, Christer, Wachtler, Caroline, Carlsson, Axel C.
• Format: Journal Article
• Language: English
• Published: England BioMed Central Ltd 30.04.2025; BioMed Central; BMC
• Subjects: Adolescent; Adult; Aged; Cardiovascular disease; Case-Control Studies; Chronic illnesses; Codes; Complications; Coronaviruses; COVID- 19; COVID-19 - complications; COVID-19 - diagnosis; COVID-19 - epidemiology; Decision trees; Development and progression; Diabetes; Drugs; Evidence-based medicine; Fatigue; Female; Females; Hospitalization; Hospitalization - statistics & numerical data; Humans; Learning algorithms; Machine Learning; Male; Males; Middle Aged; Missing data; Population; Population studies; Post-Acute COVID-19 Syndrome; Post-acute sequelae of COVID-19 (PASC); Prediction models; Primary care; Primary Health Care; Regression analysis; Risk Assessment - methods; Risk Factors; SARS-CoV-2; Stochastic gradient boosting; Sweden - epidemiology; Type 2 diabetes; Variables; Young Adult; Sweden; COVID- 19; Primary care; Stochastic gradient boosting; Machine learning; Post-acute sequelae of COVID- 19 (PASC)
• ISSN: 1741-7015; 1741-7015
• DOI: 10.1186/s12916-025-04050-w

The aim of our study was to determine whether the application of machine learning could predict PASC by using diagnoses from primary care and prescribed medication 1 year prior to PASC diagnosis. This population-based case-control study included subjects aged 18-65 years from Sweden. Stochastic gradient boosting was used to develop a predictive model using diagnoses received in primary care, hospitalization due to acute COVID- 19, and prescribed medication. The variables with normalized relative influence (NRI) ≥ 1% showed were considered predictive. Odds ratios of marginal effects (OR ) were calculated. The study included 47,568 PASC cases and controls. More females (n = 5113) than males (n = 2815) were diagnosed with PASC. Key predictive factors identified in both sexes included prior hospitalization due to acute COVID- 19 (NRI 16.1%, OR 18.8 for females; NRI 41.7%, OR 31.6 for males), malaise and fatigue (NRI 14.5%, OR 4.6 for females; NRI 11.5%, OR 7.9 for males), and post-viral and related fatigue syndromes (NRI 10.1%, OR 21.1 for females; NRI 6.4%, OR 28.4 for males). Machine learning can predict PASC based on previous diagnoses and medications. Use of this AI method could support diagnostics of PASC in primary care and provide insight into PASC etiology.