Dynamic indicators of resilience in COVID-19 patients: a comprehensive analysis of heart rate dynamics.

Objective Physiological resilience-the ability to maintain or regain homeostasis after a stressor-is rarely quantified in clinical practice. Complex-systems theory proposes dynamic indicators of resilience based on variabilityvariation, autocorrelation, and complexity ofin physiological signals. We examined whether such indicators, derived from routine ward-monitor heart-rate (HR) data, relate to were associated with in-hospital COVID-19 outcomesseverity and long-term survival.
Methods Continuous 1-minute HR tracesApproach Routine ward-monitor HR data from 181 hospitalized COVID-19 patients were analyzed. Fourteen at 1-minute resolution. Dynamic indicators were calculated for a short window (the first valid 270 min-minute interval after admission) and for an extended window (first 1620 min, -minute-summary based on the first six consecutive segments).sequential valid 270-minute intervals. Logistic regression assessedwas used to assess associations with an in-hospital composite outcome of COVID-19-related complications, ICU admission, or death;. Cox models evaluated long-term mortalitywere used to examine survival; systematic post-discharge follow-up was conducted only in patients ≥ aged ≥60 years. Models were adjusted for age and sex.
Results Main results Greater HR variability (minute-level HR variation during the 270-minute interval, reflected by standard deviation, and range) over 270 min, was associated with lower odds of in-hospital disease severity (standard deviation: OR=0.6762, p=0.0402; range: OR=0.64, p=0.01) and better survival (HazR=0.51, p=0.045). Over 1620 min these associations weakened and reversed direction, consistent with noise accumulation. for standard deviation). Short-lag autocorrelations at 270 min correlatedduring the 270-minute interval were positively associated with acute severity, whereas higher entropy over 1620 min showed a protective trend both for qSOFA, but were not clearly associated with the in-hospital severity (p = 0.06) and survival. Fractal dimension exhibited a non-significant protective trend for composite. Sample entropy showed time-scale-dependent associations with in-hospital severity but was significantly: higher sample entropy during the 270-minute interval was associated with higher odds of severe disease (OR=1.46, p=0.04), whereas higher sample entropy in the extended 1620-minute summary was associated with lower odds of severe disease (OR=0.68, p=0.02). Higher fractal dimension in the extended summary was associated with increased long-term mortality at 1620 min (HR in age-interaction models (HazR=1.93, p=0.04). Age modified several associations, including the protective-dependent patterns were observed, most clearly for the association of long-lag autocorrelation and the Lyapunov exponentbetween fractal dimension and long-term mortality.
Conclusions Significance Minute-level HR dynamics may contain complementary resilience signals: short-window variability clinically relevant information beyond co.