Critical Slowing Down as a Wearable Monitoring Tool
Prediction 4 from the dynamical systems analysis—that critical slowing down (CSD) occurs before transitions between disease states—has immediate practical application. CSD manifests as increased variance and autocorrelation in fluctuating observables as a system approaches a tipping point (Scheffer et al. 2009) (Scheffer et al. 2012). In physiological systems, CSD has been detected in HRV before cardiac arrhythmia transitions (Olde Rikkert et al. 2016) and in mood time series before depressive episodes (Wichers, Groot, and Group 2016) Leemput et al. (2014).
For ME/CFS, the relevant observable is HRV (heart rate variability), which reflects autonomic nervous system regulation and is continuously measurable with consumer wearables. Escorihuela et al. demonstrated that reduced HRV predicts fatigue severity in ME/CFS (Escorihuela et al. 2020), establishing HRV as a validated correlate of disease state in this population. The proposed monitoring system:
Algorithm design.
- Compute rolling 7-day HRV variance (\(\sigma^2_\text{HRV}\)) and lag-1 autocorrelation (\(r_1\)).
- Alert when: \(\sigma^2_\text{HRV}\) increases \(> 2 \sigma\) above personal baseline for \(> 3\) consecutive days and \(r_1 > 0.7\).
- Discriminate direction using concurrent context: HRV destabilization during treatment ramp-up → approaching recovery transition; HRV destabilization during activity increase → approaching crash.
- Required hardware: any HRV-capable wearable (Oura, Garmin, Apple Watch with appropriate sampling rate).
The key scientific question is whether the CSD signal is detectable above noise in real patient data—whether the theoretical prediction survives contact with the messy reality of free-living HRV measurement.
Can critical slowing down signatures (increasing variance and autocorrelation in HRV time series) be reliably detected in free-living ME/CFS patients using consumer wearables, and can the direction of the approaching transition (crash vs. recovery) be discriminated from concurrent activity and treatment context?
Why it matters: If CSD is detectable, it provides a personalized early warning system for crashes (enabling preventive pacing) and a quantitative signal of treatment response (enabling earlier dose optimization). The clinical impact would be substantial: crash prevention is the single most important pacing goal, and current methods rely on subjective symptom monitoring.
Minimum study design: \(n >= 50\) ME/CFS patients, \(>= 12\) months continuous HRV monitoring, capturing \(>= 5\) detected state transitions (crashes or treatment responses) per participant. Requires time-stamped activity logs and treatment records for direction discrimination.
Analogous successes: CSD detected in depression mood time series (Wichers, Groot, and Group 2016); CSD detected before ICU patient deterioration (Olde Rikkert et al. 2016); CSD theory validated in ecological and climate systems (Scheffer et al. 2012).
Key uncertainties: Wearable HRV measurement noise may overwhelm the CSD signal. Free-living conditions (exercise, caffeine, sleep position) introduce confounders. The ME/CFS disease attractor may not exhibit clean CSD if the transition is gradual rather than abrupt.