Digital Monitoring and Chronic Pain Trajectories in Patients Receiving Tertiary Pain Care

Chronic pain management relies on baseline prognostic models, although digital monitoring may better capture long-term patient trajectories.

To describe 6-month chronic pain trajectories using weekly assessments and to evaluate the value of routinely collected baseline sociodemographic and clinical variables in estimating trajectories.

This multicenter mobile health cohort study was conducted in French tertiary pain clinics between September 2019 and 2025. Adults with chronic noncancer pain initiating follow-up in a pain clinic and completing at least 6 months of follow-up were eligible.

Participation in the eDOL digital health program with weekly assessments alongside routine care.

The primary outcomes were weekly bodily comfort, sleep, and mood measured on 11-point numeric rating scale and aggregated into monthly medians. Trajectory stability, time spent in predefined clinical states, latent trajectory classes, and performance of baseline variables in estimating outcomes were assessed using linear mixed-effects models, time-in-state analyses, gaussian mixture models , and machine learning. Informative attrition was addressed using inverse probability of censoring weighting and joint longitudinal-survival modeling.

Among 1477 participants (76.8% [95% CI, 74.6%-78.9%] aged 35-64 years; 82.1% [95% CI, 77.9%-86.3%] women; median pain duration >5 years), most were receiving multimodal treatment for multimorbid chronic pain. Average trajectories remained stable, with no clinically meaningful change in patient-reported outcomes. Three trajectory classes were identified: stable (1266 participants; 85.7% [95% CI, 83.9%-87.5%]), improving (61 participants; 4.1% [95% CI, 3.2%-5.2%]), and worsening (150 participants; 10.2% [95% CI, 8.6%-11.7%]). Baseline characteristics had limited value in estimating pain trajectories.

In tertiary pain clinics patients, trajectories of bodily comfort, sleep, and mood were predominantly stable over 6 months. Baseline characteristics poorly estimated long-term trajectories, supporting continuous digital monitoring and individualized change detection rather than baseline-only prognostic stratification.