ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: TH-PO258

Using Machine Learning to Construct a Predictive Model for Hemoglobin in Maintenance Hemodialysis Patients

Session Information

Category: Dialysis

  • 801 Dialysis: Hemodialysis and Frequent Dialysis

Authors

  • Li, Mingzhu, Sichuan Academy of Medical Sciences and Sichuan People's Hospital, Chengdu, Sichuan, China
  • Hong, Daqing, Sichuan Academy of Medical Sciences and Sichuan People's Hospital, Chengdu, Sichuan, China
  • Li, Guisen, Sichuan Academy of Medical Sciences and Sichuan People's Hospital, Chengdu, Sichuan, China
  • Wang, Li, Sichuan Academy of Medical Sciences and Sichuan People's Hospital, Chengdu, Sichuan, China
Background

Constructing prediction models of hemoglobin concentration by machine learning method, to assist clinicians to make clinical decisions in order to achieve individualized and precise treatment for maintenance hemodialysis (MHD) patients with renal anemia.

Methods

The medical records of maintenance hemodialysis patients from January 1, 2021 to January 1, 2023 in Sichuan Provincial People's Hospital were included. Demographic characteristics, test results, medication orders were included to constructed the prediction models. Selection of characteristic variables were mainly based on previous published articles, and all medical data were derived after desensitization. Data were randomly divided into training set (80%) and test set (20%) after desensitization, filling, deletion and other preprocessing. Ten machine learning methods (Linear Regression, K Neighbors Regressor, Support Vactor Regression, Ridge Regression, Lasso Regression, eXtreme Gradient Boosting Regressor , Random Forest Regressor, AdaBoost Regressor, Gradient Boosting Regressor, Bagging Regressor) were used to construct prediction models. Model performance was assessed by assessing the difference between predicted and true values and by fitting with internal validation.

Results

The medical records of 495 patients were finally included. The study included 56% male with a median age of 60 years old. The patients' dry weight was 58.7±11.3 kg and their weight before dialysis was 61.3±11.6 kg. In terms of laboratory results, the patients had a hemoglobin level of 109(96~118) g/L, serum albumin level of 40.8 (38.1~43.9)g/L, serum ferritin of 210.3(130.0~359.4)ng/mL and an average transferrin saturation level of 26.2±17.3 %. The patients' median dose of intravenous iron supplements was 500 mg/month, and their median dose of erythropoiesis-stimulating agents was 5 (4 to 6.25)×104 U/month.Ten machine learning models were used for modeling, Random Forest showed better prediction performance compared with other models, with an RMSE of 9.41 and a coefficient of determination R2 of 0.65 for the training set.

Conclusion

The machine learning-based hemoglobin prediction model of MHD patients can be used to predict hemoglobin concentration to a certain extent, which can contribute to the individualized and precise management of anemia in maintenance hemodialysis patients.

Funding

  • Clinical Revenue Support