Healthcare (Basel, Switzerland)

Using AI to Predict Burnout, Long COVID, and Long Sick Leave in Healthcare Workers

Updated

Abstract

Essence

A unified AI framework predicted and extended leave strongly and more modestly in healthcare workers.

Evidence

Predictive modeling study analyzed 1244 Romanian healthcare professionals with 14 structured features and found stacked-ensemble ROC-AUC values of 0.70 for burnout and 0.93 for Long COVID and extended leave on held-out test sets.

Caveat

The models were developed on one SMOTE-balanced Romanian occupational health dataset, so external prospective validation and the binary burnout construct remain important limits.

Simplified

Key numbers

0.93
ROC AUC for
Performance of the stacked ensemble model.
0.93
ROC AUC for Extended Leave
Performance of the stacked ensemble model.
1244
Sample Size
Number of healthcare professionals analyzed.

Full Text

What this is

  • Healthcare workers face significant risks from , , and extended sick leave, which often share predictors.
  • This research develops an integrated AI framework to predict these outcomes using a structured dataset from 1244 Romanian healthcare professionals.
  • The framework employs machine learning models to improve workforce risk monitoring and support proactive interventions.

Essence

  • An integrated AI framework predicts , , and extended sick leave in healthcare workers using a unified dataset. The stacked ensemble model achieved high performance, particularly for and extended leave, enhancing workforce risk monitoring.

Key takeaways

  • The stacked ensemble model achieved ROC AUC scores of 0.93 for and extended leave, indicating strong predictive performance. This model outperformed traditional logistic regression significantly, suggesting the effectiveness of advanced machine learning techniques in occupational health.
  • Feature analysis identified key predictors for , , and extended leave, including age, job role, and chronic conditions. This underscores the importance of understanding multifactorial risks in healthcare settings.
  • The deployment of a real-time web application enables healthcare teams to assess individual risk profiles efficiently, facilitating proactive interventions and enhancing occupational health management.

Caveats

  • The study's dataset is limited to a single hospital in Romania, which may affect the generalizability of the findings to other healthcare systems. Multi-institutional validation is necessary for broader applicability.
  • Self-reported outcomes for and could introduce bias, affecting the reliability of the predictions. Future research should incorporate validated measures for these conditions.
  • The binary classification of complex conditions like and oversimplifies their multidimensional nature. Future studies should explore graded assessments to capture the spectrum of these syndromes.

Definitions

  • Burnout: A syndrome resulting from chronic workplace stress, characterized by emotional exhaustion, depersonalization, and reduced personal accomplishment.
  • Long COVID: Persistent symptoms lasting weeks to months after acute SARS-CoV-2 infection, affecting physical and cognitive health.
  • Extended medical leave: Prolonged absence from work due to health-related issues, often linked to burnout or chronic conditions.

Simplified

Funding

Competing interests

The authors declare no conflicts of 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