A Machine Learning–Based Prediction Model for Acute Kidney Injury in Patients With Community-Acquired Pneumonia: Multicenter Validation Study
2024

Predicting Acute Kidney Injury in Pneumonia Patients Using Machine Learning

Sample size: 6371 publication 10 minutes Evidence: high

Author Information

Author(s): de Azevedo Cardoso Taiane, Chua Horng-Ruey, He Runnan, Xin Yantao, Liu Bin, Xiang Wentao, Ma Mengqing, Chen Caimei, Chen Dawei, Zhang Hao, Du Xia, Sun Qing, Fan Li, Kong Huiping, Chen Xueting, Cao Changchun, Wan Xin

Primary Institution: Nanjing Medical University

Hypothesis

Can machine learning algorithms accurately predict acute kidney injury in hospitalized patients with community-acquired pneumonia?

Conclusion

The study developed a machine learning model that accurately predicts acute kidney injury in pneumonia patients during hospitalization.

Supporting Evidence

  • 15.8% of pneumonia patients developed acute kidney injury.
  • The deep forest model achieved an AUC of 0.89 in internal validation.
  • The model identified 11 key indicators for predicting AKI.
  • A web-based prediction platform was developed for clinical use.
  • Statistical significance was determined with a p-value of less than 0.05.

Takeaway

Doctors can use a computer program to help figure out which pneumonia patients might get kidney problems, so they can help them sooner.

Methodology

The study used five machine learning algorithms to analyze data from pneumonia patients and identify risk factors for acute kidney injury.

Potential Biases

Selection bias may have influenced the results due to the retrospective design.

Limitations

The study was retrospective, which may affect data accuracy, and urine output data was incomplete.

Participant Demographics

The study included 6371 patients, with a mean age of 67 years, and 40.2% were female.

Statistical Information

Confidence Interval

95% CI 0.84-0.92

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.2196/51255

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication