Research square

Using Multiple Types of Data Improves Machine Learning Prediction of Long COVID

Updated

Abstract

Integration of electronic health record data with survey and genomic information improved model performance, achieving an AUROC of 0.748.

  • Long COVID affects a significant number of those infected with SARS-CoV-2, with over 778 million individuals impacted globally.
  • A diverse cohort of SARS-CoV-2-infected individuals (n>17,200) was used to evaluate predictive models for long COVID.
  • Combining different data sources led to improved risk prediction compared to using electronic health record data alone.
  • Active-duty service status, self-reported fatigue, and a specific genetic variant were identified as important predictors for long COVID outcomes.
  • The study suggests that using multiple data types can enhance risk assessment and support tailored interventions for individuals with long COVID.

Simplified

Full Text

Full text is available at the source.

Funding

Competing interests

Declarations The authors declared no competing interests for this work. Additional Declarations: There is NO Competing Interest.
PubMed

What Lands in Your Inbox Each Week:

  • 📚7 fresh studies
  • 📝plain-language summaries
  • ✅direct links to original studies
  • 🏅top journal indicators
  • 📅weekly delivery
  • 🧘‍♂️always free