Integrating Clinical and Histopathological Data to Predict Delayed Graft Function in Kidney Transplant Recipients Using Machine Learning Techniques

2024

Predicting Delayed Graft Function in Kidney Transplant Recipients Using Machine Learning

Sample size: 354 publication 10 minutes Evidence: high

Author Information

Author(s): Tirasattayapitak Sittipath, Ratanatharathorn Cholatid, Thotsiri Sansanee, Sutharattanapong Napun, Wiwattanathum Punlop, Arpornsujaritkun Nuttapon, Sirisopana Kun, Worawichawong Suchin, Rostaing Lionel, Kantachuvesiri Surasak, Verran Deborah

Primary Institution: Ramathibodi Hospital, Mahidol University, Thailand

Hypothesis

Can integrating clinical and histopathological data improve the prediction of delayed graft function in kidney transplant recipients?

Conclusion

Machine learning models, particularly the XGBoost model, can effectively predict the risk of delayed graft function in kidney transplant recipients.

Supporting Evidence

Among 354 DDKT recipients, 64 (18.1%) experienced delayed graft function.
The random forest model had a specificity of 99.96% and an AUROC of 0.9323.
The XGBoost model achieved a sensitivity of 89.1% and an accuracy of 97.9%.
Predictive models can guide acceptance decisions and avoid risky biopsies.
Integrating clinical and histopathological data improved predictive accuracy.
Donor diabetes and prolonged cold ischemia time were significant risk factors.
Machine learning models outperformed traditional statistical models in prediction.
External validation is needed to confirm the generalizability of the findings.

Takeaway

This study shows that using smart computer programs can help doctors figure out if a kidney transplant will work well or not, which is really important for patient care.

Methodology

A retrospective cohort study using clinical and histopathological data from kidney transplant recipients to develop and validate machine learning models.