JMIR Publications

ABSTRACT

Background:

Sighing has been proposed as a primary respiratory reset mechanism that is often linked to emotional regulation. However, evidence linking sighing to anxiety relies heavily on laboratory studies, which lack ecological validity. It remains unclear whether ambulatory sighing dynamics in free-living settings reflect momentary (state) symptom fluctuations or enduring (trait) pathology. Evidence from daily-life monitoring is needed to evaluate sighing dynamics as a candidate digital biomarker for anxiety disorders.

Objective:

This study aimed to evaluate daily-life sighing dynamics as a candidate digital biomarker of anxiety by disentangling state and trait anxiety–sigh associations and comparing these dynamics between individuals with anxiety disorders and healthy controls. A secondary objective was to assess the feasibility, signal quality, and joint data coverage of a synchronized ecological momentary assessment (EMA) and respiratory inductance plethysmography (RIP) protocol.

Methods:

We conducted an intensive longitudinal study integrating smartphone-based EMA with continuous RIP using a Hexoskin smart shirt. Thirty-eight adults were enrolled (n=15 with anxiety disorders; n=23 healthy controls) and completed four 36-hour intensive monitoring blocks distributed over 1-2 weeks. Participants wore the Hexoskin RIP smart shirt during each 36-hour block and completed 6 randomly timed EMA prompts per day during the daytime hours (09:00–21:00). Sighs were operationally defined as breaths with tidal volume ≥2× each participant’s median tidal volume. Each EMA entry was linked to the preceding 5-minute respiratory window, and the primary outcome was sigh proportion (sigh breaths/total breaths) per window. Multilevel generalized linear mixed models (GLMMs) were used to analyze anxiety–sigh coupling, decomposing anxiety into within-person (state) and between-person (trait) components.

Results:

The analysis included 1279 synchronized psychophysiological windows from 33 participants. Five participants were excluded due to insufficient valid synchronized windows. In the primary beta-binomial model of sigh proportion, higher between-person (trait) anxiety was significantly associated with a lower overall sigh proportion (OR 0.80, 95% CI 0.74–0.89; P<.001), while healthy controls showed a lower baseline sigh probability than the anxiety disorder group (OR 0.78, 95% CI 0.65–0.93; P=.005). The within-person anxiety-by-group interaction was significant (OR 1.14, 95% CI 1.03–1.26; P=.01), indicating that sighing tended to increase with higher momentary anxiety in healthy controls but was attenuated in participants with anxiety disorders. Feasibility was high (EMA completion 79.5%; high-quality respiratory samples 84.1%; EMA–RIP linkage 81.1%).

Conclusions:

Daily-life sighing dynamics showed a state–trait dissociation, with reduced state-dependent coupling in anxiety disorders versus healthy controls, supporting sighing as a candidate digital biomarker of anxiety. The synchronized EMA–RIP protocol was feasible and yielded high-integrity real-world data.