OHDSI Home | Forums | Wiki | Github

'Variability' or 'chaos' in physiological signals

Hi everyone, I’m interested in ‘variability’ as a term describing the degree or amount of irregular temporal structure contained in physiological signals. It’s most applicable in “Heart Rate Variability”, or HRV, which has thousands of peer-reviewed papers focused on this one measure.

An example is fetal HRV, the loss of which indicates strong probability of stillbirth. In adults low HRV is also associated with low “vagal tone”, frailty, sleep apnea, obesity, diabetes, heart failure etc.

I note three problems in literature. First, nobody really knows what ‘variability’ is: it can be a linear measure, like standard deviation, or can be some non-linear metric like ‘entropy’ (amount of information needed to describe a pattern). Everybody agrees that more variability is good, indicating robust physiological response, but nobody can agree what is actually being measured. The second problem is that there are a host of terms. “Chaos” was from the first heyday in the late 20th century… I found that in term in Atlas. But it was discredited 20 years ago as mathematically unsupportable, and is rarely used. Now we read of “irregularity”, “variability”, “complexity”, “heterogeneity” or “dispersion”. I prefer “aperiodicity”, which perhaps describes it most clearly in mathematical terms. The third problem is that in different modalities, ‘variability’ means different things. In heart rate, it means the irregular pattern of the beat-to-beat heart period. With blood pressure, variability often means visit-to-visit or month-to-month, and is a linear measure equalling standard deviation, or variance about the mean.

I’ve noted two terms in Atlas: fetal heart rate ‘variability’, and resting heart period ‘chaos’. Also in Athena I find R-R interval standard deviation by EKG, and R-R interval.coefficient of variation by EKG. Those last two terms are perhaps opposed to each other, one being linear and the other non-linear. This goes to my point - variability should be conceptually uniform.

1 Like

Another term that I just found: “temporal structure intricacy”, which actually is a fairly well-placed descriptive term, in and of itself. Wang S-C, Ting C-K, et al. Discovery of temporal structure intricacy in arterial blood pressure waveforms representing acuity of liver transplant and forecasting short-term surgical outcome via unsupervised manifold learning. 2021, arxiv.org.

Then there’s ‘fractal dimension’, ‘correlation dimension’, etc etc. All these disparate and occasionally-related measures to assess person’s functional status.

Hello, Ebenezer

In OMOP we mostly work with existing terminology systems, called vocabularies, and trying not to add terms if there is another way. If you see some concepts as essential and beneficial for the community, you can try to suggest a term to LOINC or SNOMED. I think your issue should be addressed there.

For now, if you urgently need specific concepts, add them as 2 bil+ concepts. There are a lot of instructions here on the forum and in OHDSI Book.