Introduction:
Chronic kidney disease (CKD) is a condition that affects more than three-quarters of a billion adults worldwide. Nutrition plays a crucial role in the management of CKD patients. The use of Machine Learning (ML) to understand the behavior of different outcomes in dialysis patients has grown in recently. this study aims to develop and compare ML models predictive capability to reach Kt/V cutoff points in hemodialysis (HD) patients, using laboratory and nutritional markers.
Methods:
This is an exploratory analysis of a secondary database, including incident hemodialysis patients in 23 Brazilian clinics followed-up from 2012 to 2016. Predictive capacity of models (Decision Tree, Random Forest, and XG Boost) was develop using Python, including 27 variables of interest. Outcome Kt/V cut-off points were: ≥ 1.2, 1.4, and 1.6.
Results:
A total of 204,590 observations from 1,824 patients were included. Random Forest (RF) model showed superior performance of accuracy, precision, recall, and F1 score metrics compared to other models and different cut-off points. The respective values were: 1.2 (75%, 85%, 81%, 83%); 1.4 (76%, 81%, 79%, 80%); and 1.6 (78%, 82%, 77%, 79%). Notably, 1.2 cut-off point presented highest metrics in this model. Additionally, variable dependence plot analysis pointed out that lower nPCR values had greater impact on RF model’s ability to achieve Kt/V with a cut-off point of 1.2, while higher body fat mass index, hypervolemia, and body mass index had a greater impact on the model's predictive capability.
Conclusions:
Nutritional variables have a significant impact on the effectiveness of HD. This study will contribute to understand the behavior of the available data and, in the future, facilitate the development of more complex models with different outcomes.
I have no potential conflict of interest to disclose.
I did not use generative AI and AI-assisted technologies in the writing process.