Using machine learning models for predicting monthly iPTH levels in hemodialysis patients.
Chih-Chieh Hsieh1, Chin-Wen Hsieh2, Mohy Uddin3
1Anhsin Health Care, Pingtung, Taiwan; Graduate Institute of Biomedical Informatics, College of Medical Science and Technology, Taipei Medical University, Taipei, Taiwan; Division of Nephrology, Department of Internal Medicine, Pingtung Christian Hospital, Pingtung, Taiwan.
Computer Methods and Programs in Biomedicine
|December 5, 2024
Summary
Machine learning accurately predicts intact parathyroid hormone (iPTH) levels in hemodialysis patients. This approach aids in early identification of high-risk individuals, reducing reliance on traditional blood tests.
Area of Science:
- Nephrology
- Medical Informatics
- Artificial Intelligence
Background:
- Secondary hyperparathyroidism (SHPT) monitoring in hemodialysis patients relies on intact parathyroid hormone (iPTH) levels.
- Accurate and timely iPTH assessment is crucial for managing SHPT in this population.
Purpose of the Study:
- To develop and evaluate machine learning (ML) models for predicting monthly iPTH levels in hemodialysis patients.
- To identify high-risk patients for SHPT through predictive modeling.
Main Methods:
- A retrospective study utilizing data from the TSN-KiDiT registration system and PTCH medication records.
- Five ML models were employed to classify patients into three iPTH level categories (<150, 150-600, >600 pg/ml).
- Data processing involved varying durations (1-month vs. 3-month continuous) and feature sets (52 features vs. 20 SHAP-identified features).
Main Results:
- The XGBoost model, using three months of continuous data and all 52 features, achieved the highest Weighted AUROC of 0.922.
- The model demonstrated particular accuracy in predicting high iPTH levels (≥600 pg/ml).
Conclusions:
- Machine learning models are highly effective for predicting iPTH levels in hemodialysis patients.
- This predictive capability facilitates early identification of high-risk patients, potentially improving SHPT management.
- Future work should focus on explainable AI and integrating ML frameworks into clinical workflows.
Related Concept Videos
01:27Dialysis
264
Renal failure occurs when the kidneys lose their ability to filter waste products from the blood effectively. It can be classified into two types: acute renal failure (ARF) and chronic renal failure (CRF).
Acute kidney injury develops suddenly and can be caused by pre-renal causes (e.g., hypovolemia, shock), intrinsic renal causes (e.g., acute tubular necrosis), or post-renal causes (e.g., urinary obstruction). In contrast, chronic renal failure progresses gradually over time and is often...
Acute kidney injury develops suddenly and can be caused by pre-renal causes (e.g., hypovolemia, shock), intrinsic renal causes (e.g., acute tubular necrosis), or post-renal causes (e.g., urinary obstruction). In contrast, chronic renal failure progresses gradually over time and is often...
264
01:21EPS and iPS Cells in Disease Research
2.8K
Embryonic and induced pluripotent stem cells are excellent models for disease research because of their ability to self-renew and differentiate into most cell types. Somatic cells from a patient are isolated and reprogrammed into induced pluripotent stem cells or iPSCs. These iPSCs are later differentiated into the desired cell type, which mirrors the diseased cell of the patient. In this way, disease models have been created for investigating diseases such as Down syndrome, type I diabetes,...
2.8K