JMIR Publications

ABSTRACT

Background:

Heart rate variability (HRV) is a widely used non-invasive measure of autonomic nervous system (ANS) regulation and has been associated with physiological stress and recovery processes. The increasing availability of smartphone-based photoplethysmography (PPG) enables large-scale longitudinal HRV monitoring in real-world settings; however, the temporal association between individualized HRV reductions and subsequent self-reported symptoms remains insufficiently characterized.

Objective:

To examine whether episodes of low HRV, measured using smartphone PPG and normalized relative to individual baseline values, are associated with contemporaneous and subsequent self-reported symptoms collected through a digital health platform.

Methods:

We performed a retrospective observational analysis of 499,723 PPG-derived HRV measurements from 9,394 users of a mobile health platform. HRV was quantified using the root mean square of successive differences (RMSSD) and standardized within individuals using z-score normalization. User-generated free-text symptom reports were processed using a rule-based natural language processing (NLP) pipeline and classified into four symptom domains: acute autonomic, prodromal/systemic, neuropsychological, and sleep-related. Associations between low-HRV episodes and symptom occurrence were evaluated across predefined temporal windows (0–1, 3–7, and 7–14 days) using odds ratios (ORs) and Fisher’s exact tests. Symptom-free survival following low-HRV episodes was assessed using Kaplan–Meier analysis.

Results:

Low-HRV episodes (RMSSD z-score < −1) were associated with increased symptom occurrence across multiple temporal windows. Acute autonomic symptoms showed the strongest association, with higher odds during low-HRV periods (OR = 2.38, p < 0.001). Prodromal/systemic symptoms occurring 1–3 days after low-HRV episodes were also associated with increased occurrence (OR = 1.58 - 1.46, p < 0.001). Neuropsychological symptoms demonstrated statistically significant association within the 7 days window (OR = 1.69, p = 0.002). Sleep-related symptoms showed a positive association but did not reach statistical significance (OR = 1.28, p = 0.45). Kaplan–Meier analysis demonstrated reduced symptom-free survival following low-HRV episodes (log-rank χ² = 7.99, p = 0.0047), with a 19% higher hazard of symptom onset during the subsequent 14-day period.

Conclusions:

Individualized reductions in HRV measured by smartphone PPG were associated with immediate and delayed changes in self-reported symptom occurrence across multiple symptom domains. These findings suggest that personalized HRV monitoring may provide a scalable approach for detecting temporal patterns of physiological variation in free-living populations and support further prospective validation of HRV-based digital biomarkers.