Tracking Physical Energy
In this post, I explain how I track my physical energy, and share some interesting insights I have learned so far from my data.
Physical energy is another essential life indicator that I track on a regular basis. Not to be confused with caloric expenditure, this metric is just a quick measure of how tired/energetic I feel at any given moment.
How to Measure Fatigue Scientifically
Just like with all other life metrics, I did a thorough review of the research literature to choose the best measure of energy. Most of the current methods of measuring fatigue/energy levels fall under one of the following categories:
- indirect (modeled)
The subjective methods involve the person responding to a question “how tired are you”. The response scales vary from the simple visual analog scale to a Likert-type scale with words-anchors (e.g, Samn-Perelli fatigue scale or the Karolinska Sleepiness Scale).
The objective methods require subject to perform a task in response to a stimulus (e.g., psychomotor vigilance test, or PVT). Polysomnography, a more intrusive and impractical approach involves capturing brain waves and eye movements, and requires specialized equipment. Despite its complexity, polysomongraphy results are considered a gold standard in fatigue measurement.
Finally, the indirect/modeled approach involves modeling fatigue using on subject’s previous night sleep data. The most common approaches use SAFTE (Sleep, Activity, Fatigue, Task and Effectiveness) model developed by US Army. SAFTE model is widely used by US Department of Defense and Department of Transportation.
Tracking Energy in My Daily Log
After thorough testing and consideration, I chose the visual analog scale as the simplest and fastes way to log my energy levels on a regular basis.
The measure is set up as a sliding scale in rTracker, with values ranging from -5 (lowest) to 5 (highest). It takes less than 10 seconds to assess and record my energy level at any given moment.
This is how my average energy levels looked in 2019 in the mornings, afternoons and evenings (scores have been normalized and centered around 0):
Oura Ring Scores and Energy
One way to validate the energy metric is to look at the correlation between the average daily energy score and Oura’s “readiness score” for that day:
One would expect the correlation to be positive, and indeed, it is, albeit small:
> cor(ouralife$readiscoredt,ouralife$energydt,method=c("spearman")) [,1] [1,] 0.1564503 >
The sleep quality also seems to be correlated with various stages of sleep - you can see it by fitting a simple linear regression model (using detrended data):
> OuraEnergy<-ouralife %>% select(energydt,lightdt,remdt,deepdt) > mod<-lm(energydt~.,data=OuraEnergy) > tidy(mod) # A tibble: 4 x 5 term estimate std.error statistic p.value <chr> <dbl> <dbl> <dbl> <dbl> 1 (Intercept) -0.00250 0.00318 -0.787 0.432 2 lightdt 0.118 0.0265 4.46 0.00000923 3 remdt 0.0599 0.0262 2.29 0.0223 4 deepdt 0.0170 0.0303 0.563 0.573 >
The light sleep stage seems to have the highest impact on my energy during the day:
Overall, I am happy with the Energy metric, and plan to continue using it in my daily logs.