Bidirectional Associations Between Autonomic Function, Sleep Quality, and Daily Fatigue and Energy Symptoms in Long COVID: Longitudinal Digital Health Study
Nana Yaw Aboagye; Mark R Baker; Kenneth Baker; Silvia Del Din
BACKGROUND
Long COVID is characterized by persistent fatigue, with disrupted autonomic function and sleep disturbances frequently reported. Whether daily symptom severity influences subsequent sleep and autonomic recovery (the reverse temporal direction) remains underexplored.
OBJECTIVE
This study aimed to examine bidirectional, within-person associations between objectively measured sleep quality, nocturnal heart rate variability (HRV), and daily fatigue and energy levels in individuals reporting post-COVID fatigue symptoms.
METHODS
We conducted a longitudinal observational study using continuous wearable monitoring (Fitbit Inspire 3) in 14 individuals with long COVID over a median of 28 days (range 12‐73 d), yielding 678,057 minute-level observations. Nocturnal HRV parameters (root mean square of successive differences [RMSSD], high-frequency power, low-frequency/high-frequency ratio) and sleep metrics (duration, efficiency, deep sleep, rapid eye movement [REM] sleep) were derived automatically. Fatigue (4-level scale) and energy (5-level scale, 0‐100) were reported via smartphone app (FatigueSense). Two temporal linking strategies were applied: (1) prospective, linking morning/afternoon symptom reports to the previous night’s sleep/HRV; and (2) reverse-direction, linking evening reports to the subsequent night’s sleep/HRV. Repeated-measures correlation (rmcorr) accounted for within-person clustering; person-mean–centered (Mundlak) cumulative link mixed models disaggregated within-person from between-person effects.
RESULTS
Prospective within-person rmcorr analyses (n=260 observations, 14 participants) identified 3 associations surviving Bonferroni correction (α=.0031): sleep duration (=−0.276, 95% CI −0.392 to −0.152; <.001) and REM sleep (=−0.215, 95% CI −0.342 to −0.079; =.002) with next-day fatigue, and sleep efficiency with next-day energy in the counterintuitive negative direction (=−0.196, 95% CI −0.317 to −0.068; =.003). HRV metrics showed no significant within-person day-to-day associations with symptoms (HRV-RMSSD → fatigue: =−0.064, =.36). Mundlak models revealed that HRV-RMSSD associations were predominantly between-person: individuals with chronically higher HRV reported consistently lower fatigue (β=−1.260, 95% CI −2.231 to −0.290; =.01) and higher energy (β=0.539, 95% CI 0.056 to 1.022; =.03). A significant Sleep × Steps interaction (β=0.091 per SD of steps per 1000; =.02) indicated that the protective association of longer sleep on fatigue was attenuated on more active days, consistent with a postexertional malaise pattern. In an exploratory reverse-direction analysis, evening energy (but not fatigue) showed an association with that night’s HRV-RMSSD that was nominally significant but did not persist after Bonferroni correction (rmcorr: =0.327, 95% CI 0.080 to 0.537; =.01; α=.0050) and was supported by a log-linear mixed model (β=0.0069; =.02).
CONCLUSIONS
In this exploratory longitudinal study, sleep duration and REM sleep showed day-to-day (within-person) associations with next-day fatigue, while HRV-RMSSD distinguished individuals with better versus worse average symptom burden. The counterintuitive sleep efficiency-energy association most plausibly reflects chance, given the null sleep efficiency-fatigue association and the multiple-testing context. The preliminary, uncorrected reverse-direction association between evening energy and nocturnal HRV-RMSSD warrants replication. Experimental work is needed to test causal mechanisms and clinical relevance.
Web | DOI | PDF | JMIR Cardio | Open Access
Nana Yaw Aboagye; Mark R Baker; Kenneth Baker; Silvia Del Din
BACKGROUND
Long COVID is characterized by persistent fatigue, with disrupted autonomic function and sleep disturbances frequently reported. Whether daily symptom severity influences subsequent sleep and autonomic recovery (the reverse temporal direction) remains underexplored.
OBJECTIVE
This study aimed to examine bidirectional, within-person associations between objectively measured sleep quality, nocturnal heart rate variability (HRV), and daily fatigue and energy levels in individuals reporting post-COVID fatigue symptoms.
METHODS
We conducted a longitudinal observational study using continuous wearable monitoring (Fitbit Inspire 3) in 14 individuals with long COVID over a median of 28 days (range 12‐73 d), yielding 678,057 minute-level observations. Nocturnal HRV parameters (root mean square of successive differences [RMSSD], high-frequency power, low-frequency/high-frequency ratio) and sleep metrics (duration, efficiency, deep sleep, rapid eye movement [REM] sleep) were derived automatically. Fatigue (4-level scale) and energy (5-level scale, 0‐100) were reported via smartphone app (FatigueSense). Two temporal linking strategies were applied: (1) prospective, linking morning/afternoon symptom reports to the previous night’s sleep/HRV; and (2) reverse-direction, linking evening reports to the subsequent night’s sleep/HRV. Repeated-measures correlation (rmcorr) accounted for within-person clustering; person-mean–centered (Mundlak) cumulative link mixed models disaggregated within-person from between-person effects.
RESULTS
Prospective within-person rmcorr analyses (n=260 observations, 14 participants) identified 3 associations surviving Bonferroni correction (α=.0031): sleep duration (=−0.276, 95% CI −0.392 to −0.152; <.001) and REM sleep (=−0.215, 95% CI −0.342 to −0.079; =.002) with next-day fatigue, and sleep efficiency with next-day energy in the counterintuitive negative direction (=−0.196, 95% CI −0.317 to −0.068; =.003). HRV metrics showed no significant within-person day-to-day associations with symptoms (HRV-RMSSD → fatigue: =−0.064, =.36). Mundlak models revealed that HRV-RMSSD associations were predominantly between-person: individuals with chronically higher HRV reported consistently lower fatigue (β=−1.260, 95% CI −2.231 to −0.290; =.01) and higher energy (β=0.539, 95% CI 0.056 to 1.022; =.03). A significant Sleep × Steps interaction (β=0.091 per SD of steps per 1000; =.02) indicated that the protective association of longer sleep on fatigue was attenuated on more active days, consistent with a postexertional malaise pattern. In an exploratory reverse-direction analysis, evening energy (but not fatigue) showed an association with that night’s HRV-RMSSD that was nominally significant but did not persist after Bonferroni correction (rmcorr: =0.327, 95% CI 0.080 to 0.537; =.01; α=.0050) and was supported by a log-linear mixed model (β=0.0069; =.02).
CONCLUSIONS
In this exploratory longitudinal study, sleep duration and REM sleep showed day-to-day (within-person) associations with next-day fatigue, while HRV-RMSSD distinguished individuals with better versus worse average symptom burden. The counterintuitive sleep efficiency-energy association most plausibly reflects chance, given the null sleep efficiency-fatigue association and the multiple-testing context. The preliminary, uncorrected reverse-direction association between evening energy and nocturnal HRV-RMSSD warrants replication. Experimental work is needed to test causal mechanisms and clinical relevance.
Web | DOI | PDF | JMIR Cardio | Open Access
