Machine Learning Predicts Long COVID Symptoms
Author Information
Author(s): Cordelli Ermanno, Soda Paolo, Citter Sara, Schiavon Elia, Salvatore Christian, Fazzini Deborah, Clementi Greta, Cellina Michaela, Cozzi Andrea, Bortolotto Chandra, Preda Lorenzo, Francini Luisa, Tortora Matteo, Castiglioni Isabella, Papa Sergio, Sona Diego, Alì Marco
Primary Institution: University Campus Bio-Medico of Rome
Hypothesis
Can machine learning accurately predict pulmonary sequelae in COVID-19 patients using clinical data?
Conclusion
The study found that machine learning can predict pulmonary Long COVID sequelae with an accuracy of up to 94%.
Supporting Evidence
- Machine learning approaches can predict Long COVID symptoms effectively.
- The study analyzed clinical data from 152 hospitalized patients.
- Three different machine learning methods were tested for accuracy.
- Results showed a predictive accuracy of up to 94%.
- Data from multiple hospitals were used to enhance the study's robustness.
- Early identification of Long COVID can improve patient management.
- The study contributes to the understanding of Long COVID prognosis.
- Publicly available data repository was created to support further research.
Takeaway
Researchers used computers to look at hospital data and found a way to guess if someone might have breathing problems after COVID-19.
Methodology
Three machine learning approaches were used: traditional classifiers, an ensemble of classifiers, and a multimodal approach based on clinical data.
Potential Biases
The retrospective nature of the study may introduce bias due to incomplete data collection.
Limitations
The study only included hospitalized patients and did not account for home-treated patients or other clinical parameters.
Participant Demographics
The study included adult patients hospitalized for COVID-19, with a mean age of 63 years.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website