JMIR Publications

ABSTRACT

Background:

Hemodialysis (HD) therapy is an indispensable tool used in critical care management. Patients undergoing HD are at risk for intradialytic adverse events, ranging from muscle cramps to cardiac arrest. So far, there is no effective HD device-integrated algorithm to assist medical staff in response to these adverse events one step earlier during HD.

Objective:

We aim to develop algorithms of machine learning to predict intradialytic adverse events in an unbiased manner.

Methods:

Three-month dialysis and physiological time-series data were collected from all patients who underwent maintenance HD therapy at a tertiary-care referral center. Dialysis data were collected automatically by HD devices, and physiological data were recorded by medical staff. These time-series datasets were feature-extracted using linear and differential analyses to de-dimensionalize time.

Results:

Time-series dialysis data were collected during the four-hour HD session in 108 patients who underwent maintenance HD therapy. There were a total of 4,221 HD sessions, 406 out of which developed at least one intradialytic adverse event. Models were built by two-class averaged perceptron and evaluated by four-fold cross-validation using machine learning. The developed algorithm can predict overall intradialytic adverse events with an area under the curve (AUC) of 0.83, muscle cramps with an AUC of 0.85, and blood pressure elevation with an AUC of 0.93. In addition, the model built based on ultrafiltration-unrelated features predicts all types of adverse events with an AUC of 0.81, indicating ultrafiltration-unrelated factors also contribute to the onset of adverse events.

Conclusions:

Our results demonstrated that algorithms combining linear and differential analyses with two-class classification machine learning predict the intradialytic adverse events in quasi-real-time with high AUCs. Such methodology implemented with local cloud computation and real-time optimization by personalized HD data could alarm clinicians to take timely actions in advance.