Meaningful digital biomarkers derived from wearable sensors to predict daily fatigue in MS patients and healthy controls, 2024, Max Moebus et al

Highlights

• Digital biomarkers predict fatigue ratings continuously throughout the day
• Physical and ANS activity while awake and asleep relate to fatigue ratings
• Dysfunction of the ANS alters the effects of biomarkers on fatigue ratings
  

Summary
Fatigue is the most common symptom among multiple sclerosis (MS) patients and severely affects the quality of life. We investigate how perceived fatigue can be predicted using biomarkers collected from an arm-worn wearable sensor for MS patients (n = 51) and a healthy control group (n = 23) at an unprecedented time resolution of more than five times per day. On average, during our two-week study, participants reported their level of fatigue 51 times totaling more than 3,700 data points. Using interpretable generalized additive models, we find that increased physical activity, heart rate, sympathetic activity, and parasympathetic activity while awake and asleep relate to perceived fatigue throughout the day—partly affected by dysfunction of the ANS. We believe our analysis opens up new research opportunities for fine-grained modeling of perceived fatigue based on passively collected physiological signals using wearables—for MS patients and healthy controls alike.

VAS fatigue ratings among different subgroups
Between the control group and MS patients, we found differences in the distribution of VAS fatigue ratings. Table 1 displays average baseline characteristics of the healthy controls (Co), MS patients with a functional ANS (MS I) and MS patients with a dysfunctional ANS (MS II), and whether the differences are statistically significant. Table 1 shows that VAS fatigue ratings significantly differ between MS patients with a dysfunctional ANS and MS patients with a functional ANS (p = 0.026). The two groups also significantly differ in terms of EDSS and FSMC scores, as well as heart rate variability (HRV) metrics (Table S3 in the Appendix). FSMC scores also significantly differ between MS patients with a dysfunctional ANS and the control group but not MS patients with a functional ANS and the control group.

Concluding remarks
In this paper, we have highlighted that state fatigue can be modeled at a time resolution of multiple times a day for healthy individuals and MS patients alike. Based on passively collected data alone, our models clearly outperformed baseline regressors predicting each participant’s average response over our two-week study duration. Dysfunction of the ANS affects the relationship between biomarkers and state fatigue. For healthy individuals, MS patients with a functional ANS, and MS patients with a dysfunctional ANS, state fatigue thus has to be analyzed separately.

VAS fatigue ratings follow a daily upward trend and the time of day and the time participants spent awake were the strongest predictors for state fatigue. Deviations from this daily upward trend might be explained by changes in biomarkers related to cardiac, ANS, electrodermal, and physical activity. The calculated effects linked to the activity of the sympathetic nervous system indicate that emotional states, such as stressful or particularly calming experiences, might affect state fatigue. Further, we find changes in biosignals while asleep to predict state fatigue throughout the following day. This highlights that sleep behavior and its relation to state fatigue should be studied more closely for healthy individuals and MS patients alike.

LINK

 

Predict isn't the right term here. It doesn't predict, it reflects it. Same as the height of a boat relative to the sea bottom doesn't predict the tide, it only matches it.

Oh no that's not possible, we were assured by self-rated experts on fatigue that levels of activity and fatigue have nothing to do with one another. H/T @dave30th. What it seems to show is that although there is a level of activity that seems to match fatigue levels, increasing activity levels has a negative impact:

Interesting how it matches common reports of weather, mostly from atmospheric pressure, having a significant impact:

They stratify between functional and dysfunctional ANS on the basis of:

But the differences in HRV are miniscule. Not sure if there's much basis for that.

This really all mostly adds up to: sick people are functionally limited, some of which is because of fatigue, and sicker people are more limited than less sick people, generally reflected as higher fatigue, even if they have the same disease.

There is the traditional "let's ask about emotions and stuff" that, as usual, doesn't find much. Frankly this should all be stopped, it adds nothing but noise and is clearly about as distracting as a cat strutting in front of obedience dogs in training.

 

Yeah, the word "predict" here is used in the sense of correlation, not like weather prediction. This paper fails even at that since they got the correlation of 0.78 only when the model is applied to the data that the model was derived from. The seemingly simple concept of testing the model is lost on too many people.