BACKGROUND
Sickle cell disease (SCD) is a genetic red blood cell disorder associated with severe complications including chronic anemia, stroke, and recurrent episodes of pain called vaso-occlusive crises (VOCs). VOCs are unpredictable, difficult to treat, and the leading cause of hospitalization. Recent efforts have focused on the use of mobile health technology (mHealth) to develop algorithms to predict pain in SCD. Combining the data collection abilities of a consumer wearable, such as the Apple Watch, and machine learning techniques for patients living with SCD, may help us better understand the pain experience and find trends to predict pain from VOCs.
OBJECTIVE
The aim of this study is to: (1) evaluate machine learning models for accuracy for people with SCD with VOCs admitted to the Day Hospital, and (2) predict the pain scores of VOCs with the Apple Watch using machine learning in people with SCD.
METHODS
Following approval of the institutional review board, patients with SCD, over 18 years of age, and admitted to the Duke University SCD Day Hospital for a VOC between July 2021 to September 2021 were approached to participate in the study. Participants were provided with an Apple Watch Series 3, worn for the duration of their visit. Data collected from the Apple Watch included heart rate, heart rate variability (calculated), and calories. Pain scores, vital signs, and analgesics were collected from the Electronic Medical Record. Data was analyzed using 3 different machine learning models (Multinomial Logistic Regression Model, the Gradient Boosting Model, and the Random Forest Model), to assess the accuracy of pain scores.
RESULTS
We enrolled 20 patients, all of which identified as Black or African American, and consisted of 12 females (60%) and 8 males (40%). All participants had confirmed diagnoses of SCD, with 14 individuals diagnosed with HbSS (70%), 5 with HbSC (25%), and 1 with HbSOArab (5%). The median age of the population was 35.5 years. The median time each individual spent wearing the Apple Watch was 2 hours and 17 minutes and a total of 15,683 data points were collected across the population. The best-performing model was the Random Forest model, which was able to predict the pain scores with an accuracy of 84.5%, and a root-mean-square error (RMSE) of 0.84.
CONCLUSIONS
The high level of accuracy from this model validates the ability to utilize a non-invasive device, the Apple Watch, to predict pain scores during VOCs. It is a novel approach and presents a low-cost method that could benefit clinicians and individuals with SCD in the treatment of VOCs.