Naïve Bayes is an interpretable and predictive machine learning algorithm in predicting osteoporotic hip fracture in-hospital mortality compared to other machine learning algorithms
2025

Using Naïve Bayes to Predict In-Hospital Mortality in Elderly Hip Fracture Patients

Sample size: 3625 publication Evidence: moderate

Author Information

Author(s): Jo-Wai Douglas Wang

Primary Institution: The Canberra Hospital, ACT Health, Canberra, Australia

Hypothesis

Can Naïve Bayes effectively predict in-hospital mortality for elderly patients with hip fractures using only sociodemographic and comorbidity data?

Conclusion

Naïve Bayes is a simple and interpretable model that performs comparably to more complex machine learning algorithms in predicting in-hospital mortality for elderly hip fracture patients.

Supporting Evidence

  • Naïve Bayes performed comparably to random forests and logistic regression in predicting mortality.
  • Chronic kidney disease and cardiovascular comorbidities were identified as significant predictors of mortality.
  • The model requires only easily obtainable data, making it practical for early risk assessment.

Takeaway

This study shows that a simple computer program can help doctors figure out which older patients with hip fractures are more likely to get very sick in the hospital.

Methodology

Seven machine learning models were developed and compared using sociodemographic and comorbidity data to predict in-hospital mortality, evaluated through cross-validation and SHAP analysis.

Potential Biases

Potential under-reporting of smoking and drinking habits may introduce bias.

Limitations

The study is based on a single-center retrospective cohort, which may limit the generalizability of the findings.

Participant Demographics

Median age of participants was 84 years, with 75.3% female and 5.2% experiencing in-hospital mortality.

Statistical Information

Statistical Significance

p>0.05

Digital Object Identifier (DOI)

10.1371/journal.pdig.0000529

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication