JMIR Publications

ABSTRACT

Background:

Sickle cell disease (SCD) is an inherited red blood cell disorder affecting millions worldwide that results in many potential medical complications throughout the lifecourse. The hallmark of SCD is pain. Many patients experience daily chronic pain, as well as intermittent, unpredictable acute vaso-occlusive painful episodes called pain crises. These pain crises often require acute medical care and can lead to a day hospital or emergency department (ED) visit, or even a hospitalization [1]. Over one in four patients seen for acute pain in the ED or day hospital are admitted, with efforts focused on palliative pain control and hydration for management [2]. However, mitigating pain crises is challenging for both the patients and their providers, given the perceived unpredictability and subjective nature of pain.

Objective:

We aim to use physiologic changes during an acute pain crisis to objectively measure and predict pain scores.

Methods:

For this pilot study, we enrolled 20 adult patients presenting to the Adult Day Hospital with acute pain. Before beginning pain treatment, each participant was given a wearable device (Microsoft Band 2) that collected physiologic measurements, and pain scores were recorded from our mobile app Technology Recordings to Understand Pain (TRU-Pain). We used this data to describe changes in pain scores and physiologic measures. Then, we constructed regression and classification machine-learning models to predict pain scores collected from mobile applications with features extracted from wearable signals.

Results:

Patients were monitored for an average of 3.79 hours (SD: +/- 2.23 hours) with an average of 5,826 (SD: +/- 2,667) objective data values per patient. As expected, we found that pain scores and heart rate decreased for most patients during the course of hospitalization. Using the wearable data, we were able to create a regression model to predict subjective pain scores with a root mean error of 1.430 and correlation between observations and predictions of 0.706. Furthermore, we verified the hypothesis that the regression model outperformed the classification model by comparing the performances of the Support Vector Machines and the Support Vector Machines for Regression.

Conclusions:

The combination of the TRU-Pain app and wearable device provided an efficient and cost-effective method for collecting objective, physiological markers of an acute pain crisis and change in pain intensity for adult patients with SCD. Applying machine learning in this small and highly imbalanced dataset, demonstrated the potential for objectively measuring pain in SCD.