Sign in Subscribe
Machine learning

Machine Learning Models for Accurate Graft Loss Identification in Kidney Transplant Records

Strategic Insights

— 1 min read

Researchers from Pontifical University Javeriana in Bogota, Colombia, have developed and validated three machine learning models combined with Natural Language Processing (NLP) techniques to accurately identify graft loss in Electronic Medical Records (EMRs) of kidney transplant recipients. The study analyzed 2712 patients transplanted between July 2008 and January 2023, examining 117,566 unstructured medical records written in Spanish. The team used NLP and machine learning models to classify the status of kidney allografts, achieving significant results in terms of area under the curve (AUC) and F1 score.

Key Takeaways:

  • The researchers developed and validated three machine learning models combined with NLP techniques for unstructured texts written in Spanish, achieving high performance on the validation set.
  • The Random Forest model achieved the highest AUC (0.98) and F1 score (0.65), but had a modest sensitivity (0.76) and a relatively low Positive Predictive Value (PPV) (0.56).
  • The Neural Network model performed well with a high AUC (0.98) and reasonable F1 score (0.61), but had a low PPV (0.49), indicating more false positives.
  • The Logistic Regression model had the lowest AUC (0.91) and F1 score (0.49), but showed the highest sensitivity (0.83) with the lowest PPV (0.35).
  • The study concluded that the models could be adapted for clinical practice, though they may require additional manual work due to high false positive rates.
  • The research was conducted by Juliana Cuervo-Rojas, Andrea Garcia-Lopez, Juan Garcia-Lopez, and Fernando Giron-Luque from the Department of Clinical Epidemiology and Biostatistics at the Pontifical University Javeriana.

Statistics:

  • 2712 patients were included in the study, transplanted between July 2008 and January 2023.
  • 117,566 unstructured medical records were analyzed.
  • The Random Forest model achieved an AUC of 0.98 and F1 score of 0.65.
  • The Neural Network model achieved an AUC of 0.98 and F1 score of 0.61.
  • The Logistic Regression model achieved an AUC of 0.91 and F1 score of 0.49.
  • The study showed a sensitivity of 0.83 for the Logistic Regression model, but a PPV of 0.35.
  • The three machine learning models showed modest performance on the test set due to data imbalance.

Sources:

  • PLOS One, 2025;20(5)
  • Pontifical University Javeriana, Bogota, Colombia
  • Public Library Science, 1160 Battery Street, Ste 100, San Francisco, CA 94111, USA
  • Juliana Cuervo-Rojas, Andrea Garcia-Lopez, Juan Garcia-Lopez, and Fernando Giron-Luque, Department of Clinical Epidemiology and Biostatistics, Pontifical University Javeriana.

Read more