JMIR Publications

ABSTRACT

Background:

Mortality prediction is important yet challenging in palliative care. Limited evidence exists on the use of artificial intelligence (AI) and wearable devices in palliative cancer patients.

Objective:

To explore the potential of using wearable devices and AI to predict death events in palliative cancer patients.

Methods:

This prospective study was conducted National Taiwan University Hospital. Patients diagnosed with cancer and receiving palliative care were invited for enrollment in wards, outpatient clinic and home-based care setting. Each participant was given a smartwatch to collect physiological data, including steps taken, heart rate, sleep time, and blood oxygen saturation. Clinical assessments were conducted weekly. Participants were followed until the end of life or up to 52 weeks. With these input features, we evaluated the prediction performance of several machine learning-based classifiers and deep neural network (DNN) in 7-day death events. We used area under the receiver operating curve (AUROC), F1 score, accuracy, and specificity as evaluation metrics. Shapley value analysis was performed to further explore the good performance models.

Results:

From September 2021 to August 2022, 1,657 data points were collected from 40 patients with a median survival time of 34 days, with the detection of 28 death events. Among the proposed models, Extreme-gradient boost (XGBoost) and deep learning yielded the best results, with an AUROC of 96.9%, F1-score of 77%, accuracy of 93%, and specificity of 97% on the testing set. Shapley value analysis identified average heart rate on the day as the most important feature. Other important features included steps taken, functional level, urination status, and clinical care phase.

Conclusions:

With wearable devices and AI, our model successfully predict the death of patients in the following 7 days. Compared to previous study, we further show the potential to integrate AI and wearable devices into clinical palliative care. Machine learning approach may provide valuable insights into the end-of-life stage, support clinical decision-making, and enable personalized care. Further studies are required to validate the approach and assess its impact on clinical care. Clinical Trial: Study protocol registered on ClinicalTrial.gov (NCT05054907)