JMIR Publications

ABSTRACT

Background:

Depressive symptoms in older adults are amplified by social isolation and limited access to clinic-based mental health care. Ecological momentary assessment (EMA) permits remote self-monitoring and unobtrusively captures response times (RTs) that may index psychomotor and cognitive functioning.

Objective:

This study investigated utility of EMA-based RT dynamics for predicting symptom change and profiling potential responders for repeated self-monitoring in late-life depression.

Methods:

Forty-nine community-dwelling adults aged ≥65 years (mean 70.7, SD 5.8; 72% female) with a history of major depressive disorder received case management incorporating daily EMA. Participants provided self-reports of mood, appetite, sleep quality, and general well-being. Pre- and post-assessments included GDS-15, CESD-R, PHQ-9, and BAI. RTs were cleaned with an asymmetric interquartile-range rule, z-standardized within person × response level, and modeled with exponential decay curves over successive EMA trials. Efficacy of EMA-adjunctive care was evaluatated using pre-post comparisons of symptom scales. We then examined associations between RT-derived features and symptom change using correlational analyses. Finally, Bayesian multilevel modeling was applied to assess the clinical relevance of RT dynamics, including group differences in adaptation patterns.

Results:

Older adults at risk for depression showed significant symptom reductions over the 4-week EMA-adjunctive care period across all four psychological scales (CESD-R: mean Δ = 11.5, rank-biserial r = 0.78; GDS-15: mean Δ = 2.14, Cohen’s d = 0.76), alongside high EMA adherence (>90%). In correlational analyses, descriptive EMA-score metrics and raw response times showed modest, symptom-specific associations with symptom change (ΔCESD-R: |r| ≈ 0.29; ΔPHQ-9: |r| ≈ 0.32; ΔBAI: |r| ≈ 0.35), but were not significantly related to change in geriatric depression (ΔGDS-15: |r| ≈ 0.24). In contrast, exponential-decay model parameters derived from standardized RT were significantly associated with geriatric depressive symptom change (Δ GDS-15), with the strongest effects observed for the Feeling item (e.g., decay rate: r = -0.398, asymptote: r = -0.321). Bayesian multilevel modeling further indicated that EMA-adjunctive care responders showed faster RT adaptation than non-responders (median decay-rate ratio ≈ 4.9, 95% CrI [1.44, 14.31]), whereas differences in post-adaptation RT levels were smaller and uncertain (median post-adaptation RT ratio ≈ 1.25, 95% CrI [0.95, 1.58]). Sensitivity analyses showed consistent decay-rate effects across alternative specifications.

Conclusions:

Dynamic characteristics of EMA-based response times emerged as a sensitive proxy for monitoring changes in depressive symptoms among older adults at risk. These findings highlight the potential utility of response time as a digital biomarker derived from brief, low-burden EMA self-monitoring, supporting the development of scalable and personalized mental health interventions for geriatric populations.