JMIR Publications

ABSTRACT

Background:

Symptom burden during active cancer treatment is prevalent and clinically significant, yet frequently goes unmonitored between scheduled care visits. AI-powered virtual care platforms offer a promising approach to continuous real-world symptom management, though longitudinal evidence in these settings remains limited.

Objective:

This study aimed to: 1) characterize the demographic and clinical profile of enrolled patients and their baseline symptom burden, 2) estimate population-level symptom trajectories over 180 days, and 3) examine whether platform engagement and cancer-related characteristics moderated symptom change over time.

Methods:

This retrospective study analyzed secondary data from 2,214 oncology patients enrolled in Reimagine Care between December 2022 and May 2026. Symptom burden was assessed using an adapted version of the Edmonton Symptom Assessment System (10 items, 11-point scale), administered via electronic patient-reported outcome (ePRO) surveys. Longitudinal trajectories were estimated using linear mixed-effects models with repeated ePRO observations nested within patients over 180 days. Moderation by virtual care center (VCC) engagement intensity and metastatic disease status was examined using time-by-moderator interaction terms.

Results:

The analytic cohort had a mean age of 63.7 years (SD = 12.9), was predominantly female (58.9%; n = 1,304/2,214), with a mean of 4.22 non-cancer comorbidities. Breast cancer was the most prevalent diagnosis (28.5%; n = 630/2,214), and 25.3% (n = 561/2,214) had documented metastatic disease. Tiredness was the most severe symptom at baseline (mean = 4.21). Total symptom burden declined significantly over 180 days (β = -0.0236, p < .001), with reductions of 2.13 and 4.25 points at 90 and 180 days, respectively. The largest symptom-level improvements occurred in tiredness (β = -0.0056), appetite loss (β = -0.0040), drowsiness (β = -0.0035), and nausea (β = -0.0023). Greater VCC engagement was associated with faster overall symptom improvement (β = -0.0050, p < .001) and improvements in appetite loss (β = -0.00078, p = .009) and nausea (β = -0.00087, p < .001), but also with worsening depression (β = 0.00052, p = .009) and anxiety trajectories (β = 0.00044, p = .027), consistent with greater utilization among psychologically burdened patients. Patients with metastatic disease demonstrated attenuated improvement in tiredness (β = 0.00167, p = .011), drowsiness (β = 0.00153, p = .012), appetite loss (β = 0.00149, p = .027), and nausea (β = 0.00105, p = .048) vs non-metastatic patients with no significant moderation of total symptom burden (β = 0.0024, p = .332).

Conclusions:

Total symptom burden declined significantly over 180 days among oncology patients enrolled in Reimagine Care, an AI-powered virtual care platform, with physical symptoms showing the greatest improvement. Platform engagement and disease stage emerged as clinically relevant moderators of symptom trajectories. Controlled designs using validated instruments are needed to establish causal claims and evaluate clinical significance. Clinical Trial: N/A