ID:15982 Chronic Pain Patients During COVID-19: Machine Learning Reveals Attributes Leading to Pandemic Susceptibility and Resilience

Introduction: The novel coronavirus has disrupted chronic pain patients’ care, on-going clinical studies, interrupted daily routines and pain management plans, as well as halted social/extracurricular activities. These disturbances may contribute to increased pain intensity, worsening disability, and deteriorating mood in a population with mental and physical health comorbidities. COVID-19 presented a unique opportunity to observe patients’ pain experience, including quality-of-life (QoL) and daily activities, as well as identify and characterize individuals who are potentially susceptible to changes during a substantial stressor. Methods: As part of on-going multi-site Boston Scientific studies prospectively observing up to 1700 chronic leg and back pain patients’ responses to spinal cord stimulation (SCS), we used smartphones to collect daily self-reported pain intensity, mood, sleep, medications, and activities. We also obtained in-clinic questionnaires and objective measures from smartwatches, sleep sensors, weekly voice recordings, and SCS usage. To evaluate changes during COVID-19, we defined two 6-week periods: “COVID” (03/6/2020—04/17/2020), “Pre-COVID” (12/20/2019- 01/30/2020). Since patients may be differentially impacted, we performed multivariate analyses integrating changes in self-reported variables between periods, which were normalized and K-means clustered to identify sub-cohorts. We also administered questions to assess patients’ emotional state during the pandemic, analyzed with natural language processing (NLP). Results: In our results (Figures 1-4), we found no differences in self-reports between pre-COVID and COVID for the entire cohort (n=70/159). However, clustering identified 3 sub-cohorts: individuals whose pain worsened (pain-susceptible), whose activities decreased (ADLsusceptible), and whose mood, sleep, medication, and activities remained the same or improved (QoL-resilient) during COVID. Partial correlations between changes in self-reports also showed differences as a function of period and sub-cohort. Sensor data indicated that NLP-identified fear related speech content during COVID was lower for the QoL-resilient group, who also had greater watch step counts during pre-COVID, supporting the idea that they had the best overall wellbeing or initial behaviors of the 3 groups. There were no differences in clinical assessments or SCS usage between sub-cohorts or periods. Conclusion: Our results indicate the existence of 3 patient sub-cohorts that diverge in their behaviors during COVID-19. We find each sub-cohort has a characteristic signature that allows us to predict the response an individual patient had to the pandemic. These findings demonstrate the importance of multi-dimensional digital monitoring with important implications for telemedicine, clinical trials and neuromodulation system management. Disclosure: Richard Rauck, MD: Boston Scientific: Contracted Research: Self, Medtronic: Contracted Research: Self, Mainstay: Contracted Research: Self, Saluda: Contracted Research: Self, Stimwave: Contracted Research: Self, SPR Therapeutics: Contracted Research: Self, Nevro: Contracted Research: Self, Neuros: Contracted Research: Self, Sara Berger, PhD: IBM: Employee:, Guillermo Cecchi, PhD: IBM: Employee:, Carla Agurto, PhD: IBM: Employee:, Elif Eyigoz, PhD: IBM: Employee:, Kristen Lechleiter, MS: Boston Scientific: Employee:, Dat Huynh, PhD: Boston Scientific: Employee:, Brad Hershey, BS: Boston Scientific: Employee:, Eric Loudermilk, MD: None, Julio Paez, MD: None, Louis Bojrab, MD: None, John Noles, MD: None, Todd Turley, MD: None, Mohab Ibrahim, MD: None, Amol Patwardhan, MD: None, James Scowcroft, MD: Nevro: Contracted Research:, Boston Scientific: Contracted Research:, Saluda: Contracted Research:, Rene Przkora, MD: Boston Scientific, Abbott, Nevro, Medtronic: Contracted Research:, Nathan Miller, MD: None, Gassan Chaiban, MD: Boston Scientific: Consulting Fee:, Matt McDonald, MS: Boston Scientific: Salary/Employee: Self, Jeffrey Rogers, PhD: IBM: Employee: [Formul presented] [Formula presented] [Formula presented]