Sleep duration and Nocturnal vs. Standing HRV recovery from COVID

Many of us have learned the hard way to stop training when we’re sick, and to ease back into it as we recover. In 2012, I contracted the hand/foot/mouth (HFMD) virus from my nephews. It was awful. After symptoms resolved, my training resumed with a deload week. Taking it easy, it took ~9 days after symptoms cleared for my 7-day rolling HR and LnRMSSD averages to return to within 1 standard deviation (SD) of pre-infection values (14-d baseline) (Fig. 1). These HRV values were obtained in the standing position after waking.

Fig 1. RMSSD and HR responses to HFMD virus.

I’ve observed the same general standing HRV pattern in every subsequent illness, with the time-course of HRV recovery proportional to the severity of illness. I recently had COVID. Predictably, I observed a very similar HRV response. It took ~10 days post-symptom resolution for my standing values to normalize. Training was even more conservative than post-HFMD since I’m getting older (and somewhat wiser, hopefully).

However, unlike in previous cases of illness, this time I also had nocturnal HRV and sleep data from my Oura ring. Without careful interpretation, it would seem that HRV responses (standing vs. nocturnal) reveal largely contradictory responses. This is problematic if one were trying to use HRV, or “Readiness” scores based largely on HR or HRV, to guide training decisions post-illness (or any time for that matter). Which should you follow? The data and my interpretations are shared below.

Supine vs. Standing HRV

First, I’ll address why nocturnal and standing HRV are different. Nocturnal HRV is acquired during sleep, representing parasympathetic function under passive conditions, largely undisturbed by our wakeful thoughts and emotions. HRV should peak during sleep, reflecting healthy circadian variation in ANS activity, which is associated with nocturnal blood pressure dipping and a lower risk of cardiovascular diseases. Contrastingly, standing HRV captures the ANS response to a mild challenge. Blood wants to pool in the legs when standing, which would cause hypotension, limit blood supply to the brain, and cause dizziness or fainting if not for properly functioning counteractive mechanisms. Baroreceptors detect reduced blood pressure following postural change, resulting in a reflexive increase in HR and vasoconstriction to maintain blood pressure. In healthy people, there is a sudden HR response (↓parasympathetic activity) followed by a quick recovery. However, when sick or stressed (mental, physical, etc.), the HR response may be exaggerated and slower to recover.

By measuring HR/HRV ~1 min after standing, we can observe how efficiently the ANS is adapting to a minor challenge. In fact, ANS testing in clinical settings typically involves a series of reflex tests. These include HR or HRV responses to deep breathing, orthostatic stress (standing), isometric handgrip, etc. This is because abnormalities are more likely to be revealed when the ANS is challenged. It always seemed intuitive to me that if we wanted to use HRV as an indicator of how we may adapt to physical stress (i.e. training), we should measure HRV in response to a little bit of physical stress.  


Four hours before my flight to San Diego for the ACSM Annual meeting, I was walking the dog when my wife texted me a picture of her positive COVID test. Thus, I cancelled my flights, cared for my ill wife, and waited to get sick. Two days later, I woke up with a sinus headache and both my nocturnal and standing HRV were substantially reduced (good agreement between responses). A sore throat came on day 2 and stuck around for 5-6 days. My average ambulatory HR measured continuously throughout the first few days by Oura was ~10 bpm higher than usual, despite being more sedentary. My appetite increased dramatically to compensate. I attempted some HRV biofeedback sessions with limited success (screenshot comparison of sessions with and without COVID below). Coughing up phlegm was my last symptom to clear.

Fortunately, work stress and other obligations were minimal throughout this period. Light exercise (zone 2 air bike) and desk work resumed on 06/09, corresponding with my first “normal” standing daily RMSSD value. Thereafter is where nocturnal and standing HRV patterns diverge (Fig. 2). Nocturnal RMSSD increases above baseline post-symptoms, whereas standing RMSSD remains mostly suppressed.

Fig 2. Nocturnal and Standing HR and HRV trends pre-, during, and post-COVID symptoms

Thus, Oura is telling me that my “readiness” is high (some of the highest scores I’ve recorded to date) (Fig. 3). Contrastingly, standing RMSSD values are telling me to gear down, that things aren’t quite normal yet.

Fig. 3. Sleep duration and daily Oura Readiness scores.

Why the discrepancy?

Although I was feeling pretty much back to normal post-symptoms, when I resumed work and light exercise, I was exhausted by the end of the day (but not during the day). I was falling asleep on the couch by ~9 pm (normal bed time 10-10:30 pm). I was also sleeping longer than usual (Fig. 3). My average total sleep duration for May was 401 min, whereas for the first 10 days post-symptoms, I averaged 429 min. Clearly, I had increased sleep needs. Interestingly, my total sleep time was the strongest correlate (vs. sleep stages, efficiency, etc.) of my nocturnal RMSSD (Fig 4).

Fig. 4. Association between nocturnal RMSSD and HR
Grey dots represent baseline. Vertical line at 401 min represents mean sleep duration from previous month

However, the association between sleep duration and standing HRV reverses post-symptoms (Fig. 5). You can see in Fig. 2 and Fig 5. that my standing RMSSD trends to baseline as my total sleep duration trends to baseline. Thus, my standing HRV remained suppressed apparently until my sleep duration returned towards normal. Reduced sleep duration in this instance likely reflects that my body no longer needed the extra rest, and by returning to baseline at this time, my standing HRV suggests the same thing. I also stopped falling asleep on the couch and returned to my habitual sleep/wake time.

Fig. 5. Association between standing RMSSD and sleep duration during and post-symptoms.
Vertical line at 401 min represents mean sleep duration from previous month.


My first instinct was that my increase in nocturnal RMSSD was probably a result of reduced saturation effects. This is a poorly understood concept and creates a lot of confused wearable users. Essentially, there is an inverse linear relationship between HR and RMSSD (as HR decreases, RMSSD increases) until ~50-55 bpm, at which point RMSSD starts to decrease, reflecting a quadratic relationship. This phenomenon is well documented in numerous studies, yet is poorly conveyed to users by wearable companies. There’s a strong possibility that if your HR is <50 bpm (i.e., fit individuals), your nocturnal RMSSD is reduced due to saturation. The mechanism seems to be that very high parasympathetic activity saturates cholinergic receptors in the myocardium, resulting in sustained inhibitory effects on the SA node, causing a slow HR with reduced respiratory sinus arrhythmia (thus, reduced RMSSD).

Practically, this means that if you are typically experiencing saturation due to a very low HR (as I often do with nocturnal value), increased stress can increase your HR out of the saturation zone and thus result in increased RMSSD. This is entirely counterintuitive because you’d expect more stress = increased HR and reduced RMSSD. I receive more emails from wearable users over this issue than any other.

Fig. 6 below clearly shows a quadratic association between nocturnal RMSSD and HR, suggesting that reduced saturation effects may be contributing to the increased nocturnal RMSSD post-symptoms. This is unlike the expected linear association observed in my standing values, where HR is mostly > 55 bpm (Fig).

Fig. 6. Association between HR and RMSSD for nocturnal and standing values.

However, when I adjust for HR (by dividing RMSSD by HR), it shows that nocturnal RMSSD was increased, independent of changes in HR (Fig. 7). This suggests that the elevated RMSSD was unlikely due to reduced saturation effects, alone. Thus, in this case, it seems that RMSSD was increased due to greater parasympathetic modulation.

Fig. 7. Nocturnal and standing RMSSD/HR

While it’s tempting to interpret longer sleep duration (often a good thing) and higher nocturnal RMSSD (also often a good thing) as signs of high readiness to train, that’s obviously incorrect in this context. I interpret these responses to reflect higher recovery demands and processes from lingering effects of the illness, plus the additional stress from resuming work and exercise (even at low intensity and volume). This is supported by the fact that my ANS was not yet able to respond as efficiently as usual to the minor stress of standing within the first several days post-symptoms.


Nocturnal values provided valuable insight regarding my sleep patterns and nightly ANS activity during and after COVID. However, taken alone and out of context, one could easily misinterpret these changes to support resumption of intense training. Thus, standing HRV also provided important insight, showing that my ANS was poorly adaptive to a mild physical challenge post-symptoms. Taken together and in context, the appropriate interpretation is likely that I was still adjusting and recovering from the illness (greater sleep/recovery needs and enhanced nocturnal parasympathetic activity to support them), and that intense training would be poorly tolerated (suppressed HRV in response to standing, a minor physical challenge). Consequently, exercise remained light (air bike, body weight circuits, deload-style lifting) until standing HRV finally normalized at ~10 days post-symptoms, which corresponded fairly well with the return of my sleep duration to normal values. Therefore, nocturnal and standing HRV were both valuable, but different. These are not interchangeable values.

I’ll finish with a brief thought on “Readiness” scores and the misguided idea that HRV is somehow analogous to a % recovery meter. I used to think that despite often being wrong and creating false expectations of what HRV is, that readiness scores were relatively harmless. I am now of the opinion that circumstances exist where readiness scores can be harmful. We should use these tools to identify pattern-changes in the data and interpret them in context, but we should not use these tools for their automated algorithms and training advice (e.g. from a wearable below).

Trend Analysis: Importance of Context

It’s pretty well documented that regular aerobic exercise can result in an increase in resting HRV. Generally, you can expect no change or even a slight increase in HRV the day following low to moderate intensity aerobic exercise. However, with higher intensity training, HRV can take up to 48-72 hours to return to baseline, depending on intensity, duration, training status, fitness level, age, gender, etc. (Stanley et al. 2014). I’ve seen this numerous times with my own data where HRV decreases significantly 24 hours following interval sessions (particularly when they are not performed regularly) and increase beyond baseline by 48 hours.

Because low-moderate aerobic work tends to have an acute stimulatory effect of parasympathetic activity, it has been suggested that this would be useful as active recovery following high intensity sessions.

“because (at least) autonomic supercompensation following low-intensity training may occur within 24 h and since cardiac parasympathetic reactivation is delayed by the build-up of metabolites, inclusion of low intensity training subsequent to a high-intensity session may accelerate metabolite breakdown [88]. Athletes who train twice daily may also benefit from the accelerated recovery (metabolic recovery, as reflected by autonomic recovery) afforded by a low-intensity training session” (Stanley et el. 2013)

I recently included moderate intensity aerobic work on off days (approx.10 mins each on treadmill, cycle and rower for a total of 30 mins) and following my training sessions (10-12 mins on cycle) over about a three week period. Previous to this, very little aerobic work was being done, at least not consistently. During this time I lifted on Mon-Tue-Thurs-Fri each week. Below is my HRV data (lnRMSSDx20, standing) that includes a few weeks prior to the inclusion of regular aerobic work as well as the few weeks that followed.

daily trend aerobic weeks Next, I’ve included the weekly mean HRV and %CV (coefficient of variation) values.

mean and CV o2 weeks


I started performing the aerobic work midway through week 3 and continued until week 6. The trends both clearly show an increase in HRV during this time. We also see quite a large change in %CV with the regular aerobic work. In the weeks before and following the aerobic work, there are much bigger day to day changes in HRV which is quite typical for me. The inclusion of regular moderate aerobic work attenuated the daily changes I’d typically see following heavy training sessions. Clearly the post-workout aerobic work and active recovery work on off days was effective at promoting recovery. However,it’s important to clarify that HRV parameters are reflective of cardiovascular-autonomic activity, which does not necessarily include neuromuscular ability, CNS potential, etc.

“changes in cardiac parasympathetic activity are useful for monitoring aspects of recovery that are dependent on cardiovascular function. By contrast, changes in cardiac parasympathetic activity are less useful for monitoring other aspects of recovery such as restoration of muscle and liver glycogen, or repair of damaged muscle tissue” (Stanley et al. 2013)

Therefore, the lack of day to day changes in my HRV following heavy resistance training workouts does not imply that I was fully recovered within 24 hours and could repeat performances (e.g., heavy squats), only that that particular system was recovered. Thus, for strength athletes in particular, HRV is only one marker to consider when assessing daily recovery status. More work needs to be done in this area to determine how useful HRV monitoring is in this population and how it can be used effectively.

This data also shows how interpretation of a trend is context dependent, as mean and %CV values are affected by exercise mode and intensity. Thus, if working with team sport athletes, we may expect larger fluctuation and a lower mean when less aerobic exercise is prescribed and vice versa. Even endurance athletes will experience similar HRV changes when preparing for competition as the amount of high intensity/interval training increases and low-moderate intensity/steady state work decreases. This is often characterized with a bell-shaped HRV trend (example below).

bell shaped trend

The above data is taken from a case study we did of a collegiate endurance athlete over his competitive season (will be in a future edition of JASC). There is clearly a progressive increase in HRV up to a peak, at which point there is a progressive decrease. This is likely a result of more high intensity training and lower volumes of moderate/steady state  work as the athletes prepares to peak, further supporting the need to assess HRV changes in context to training phase, goal, structure, etc. (Buchheit,2014).

Changes in HRV are always context dependent. Decreases in the trend are not always associated with fatigue, nor are increases always associated with higher “readiness”. Nothing is ever as black and white as we’d like it to be. Additional reference to training load, psychometrics and performance will help with interpretation and if necessary intervention.


Buchheit, M. (2014). Monitoring training status with HR measures: do all roads lead to Rome?Frontiers in Physiology5.

Stanley, J., Peake, J. M., & Buchheit, M. (2013). Cardiac parasympathetic reactivation following exercise: implications for training prescriptionSports Medicine43(12), 1259-1277.

HRV in a bit more detail: Part 2

Part 1 of this series provided information on heart function, ECG basics, HRV basics and how the Autonomic nervous system influences heart rate. For Part 2, I’ll discuss and display basic HRV analysis concepts to try and enhance your understanding of HRV.  I’ll relate as much of this discussion as possible to smart-phone based HRV tools as it is unlikely that most readers of this site have access to an ECG.

Athlete vs. Non-Athlete

Fit individuals generally have lower resting heart rates and greater parasympathetic activity at rest. These adaptations to training may be a result of both intrinsic heart adaptations (SA node remodeling, increase in ventricle capacity) and autonomic adaptations (greater vagal activity).

Below is an ECG segment from a collegiate male endurance athlete. This sample is likely capturing the normal fluctuation in heart rate that occurs in response to respiration (breathing). Heart rate tends to speed up on inspiration and slow down on expiration. The technical term used to describe this phenomenon is “Respiratory Sinus Arrhythmia”.

athlete ECG RSA1

Endurance Athlete ECG

For comparison, below is a screen shot of a healthy non-athlete ECG. Here, you can clearly see a higher resting heart rate and less variability.

Non-athlete ECG

Non-athlete ECG

The Excel snap shot below is what R-R interval data looks like once exported from the ECG software (Acqknowledge in this case) to a workbook. Though specialized HRV software is much more functional, it’s certainly possible to perform some time series (statistical) analysis on the R-R interval data with basic excel functions (i.e. Standard Deviation, Mean R-R Interval, RMSSD, etc.). Today, we’ll focus primarily on RMSSD as this parameter appears to be the preferred HRV index for athlete monitoring (See Plews et al. 2013 and this for more on RMSSD).

R-R Intervals

R-R Intervals


With HRV software analysis, ECG recordings are converted to a tachogram, which plot the successive R-R intervals on the y-axis and the number of beats within the ECG segment on the x-axis. This provides a nice visual representation of heart rate variability over a given time and makes for easy software analysis.

Below is a 5 minute ECG segment from an endurance athlete converted to a tachogram with our Nevrokard HRV software. Notice how the R-R intervals vary considerably over a broad range. The time domain values follow.

Endurance Athlete Tachogram

Endurance Athlete Tachogram


In contrast, below is the tachogram and time domain analysis of an age matched non-endurance athlete for comparison. Note how the R-R intervals are relatively stable and within a narrow range, demonstrating less variability.

Non-athlete Tachogram

Non-athlete Tachogram


To put this in perspective for the good folks using ithlete or BioForce, I’ll convert the raw RMSSD values (displayed in the “Summary Statistics” screen shots above) to ithlete/BioForce values. To do this, we simply log transform the raw RMSSD and multiply it by 20 (lnRMSSDx20). Keep in mind that the ithlete uses a 55-sec test and BioForce uses 2.5-min test. The values shown here are from 5-min ECG samples, but you get the idea.

rmssd conversion

Ectopic Beats and Artifacts

Last post, I discussed “normal” beats originating from the SA node. Any beats originating from outside the SA node disturb cardiac rhythm and can therefore impact HRV. These are called ectopic beats. Electrical interference, or excessive noise or movement can create “artifacts” which can also affect the data. It’s important to manually inspect ECG data for ectopic beats or artifacts and correct them (replace with the adjacent “normal” cycle) or discard the ECG sample entirely if there are excessive disturbances. Most smart-phone HRV tools do not provide R-R interval data and therefore manual inspection for ectopic beats is impossible. Conveniently, devices such as ithlete and BioForce are designed to automatically detect and correct irregular beats. For example, the application will detect and replace R-R intervals that are unlikely to occur in healthy, resting individuals (e.g., R-R intervals below 500ms or above 1800ms). I’d assume OmegaWave Pro, SweetBeat and other devices also have this feature.

Below is an example of an ectopic beat from an ECG which appears to be a premature atrial contraction. You’ll see this again in a moment on the tachogram as well.


Measurement Protocol

For athlete monitoring, HRV data is ideally collected as soon after waking in the morning as possible after bladder emptying.  We want as close to resting conditions as possible. It would be wise to avoid checking e-mails, text messages and anything else that can alter mood, excite you, upset you, etc. Even water consumption will have an acute impact on HRV. Consistency of measurement protocol and time of measurement are important for longitudinal monitoring. In addition, being as motionless as possible and undisturbed is equally as important.

In the screen shot below towards the right hand side, we can clearly see when the individual gets restless and moves or adjusts his position. We can also see the ectopic beat that occurs toward the start of the measure (same ectopic beat shown above in the ECG). Slight and subtle movements can clearly impact heart rate so be as still as possible when you perform your measures at home.

ectopic_disturbed tachogram

In the tachogram below, notice how heart rate changes considerably at around the 10 minute mark. This is a result of a researcher entering the exam room where the subject was resting during an ECG recording. Clearly, the resting condition was disturbed as the subject was excited/stressed from the person entering the room. Thus, do your best to remain as undisturbed as possible when performing a measure at home.

disturbed measure 10min

All data shown today were from supine ECG recordings. Next post I’ll discuss and show HRV changes in response to postural change (i.e., from supine to standing).

RMSSD: The HRV Value provided by ithlete and BioForce

Most individuals who take their sport or training very seriously have likely heard of heart rate variability (HRV). Thanks to devices such as the Polar RS800 (Formerly S810) wrist-watch/heart rate monitor and eventually ithlete, the first (to my knowledge) commercially available smart phone HRV application, HRV data can be collected easily and affordably. The recent accessibility of HRV tools has resulted in greater usage, more data and of course greater popularity.

What most folks aren’t aware of however is that HRV is not a solitary figure or value. In fact, numerous HRV parameters exist that are supposedly representative of different autonomic variables. Below is a brief list and description of popular HRV analysis methods and values (many more values exist than described).

Time Domain Analysis: This method includes statistical and geometrical analysis of R-R interval data. Common statistical time domain values include:

  • SDNN – Standard Deviation of Normal to Normal intervals.
  • RMSSD – The square root of the mean squared difference between adjacent N-N intervals.

*Note:  NN or “normal to normal” is used to denote that only “normal” beats originating from the sinus node are measured. Impulses from other areas within the myocardium (non-sinus node impulses) are termed ectopic beats. Ectopic beats disturb normal cardiac rhythm and can therefore affect HRV. Generally 3 or more ectopic beats within a short-term measurement meets criteria for exclusion in many research papers.

Frequency Domain Analysis: This method is considerably more complex than time domain analysis and often requires longer measurement durations. It assesses how variance is distributed as a function of frequency.

  • HF – High Frequency Power: A marker of Parasympathetic Activity
  • LF – Low Frequency Power: A marker of both Parasympathetc and Sympathetic Activity
  • LF/HF – Low Frequncy/High Frequency Ratio: Once thought to represent the balance between sympathetic and parasympathetic activity however this remains a hot topic of debate.

As you can see, saying something along the lines of “My HRV is low today” is really vague. I’m sure I’ve been guilty of this in the past. More often than not, most people are referring to their RMSSD value as this is the same parameter provided by ithlete and BioForce (among other HRV tools).

The RMSSD is commonly used as an index of vagally (Vagus Nerve) mediated cardiac control which captures respiratory sinus arrhythmia (RSA), the frequent changes in heart rate occurring in response to respiration (Berntson et al. 2005). During inhalation, heart rate speeds up. During exhalation, heart rate slows down. RMSSD is an accepted measure of parasympathetic activity and correlates very well with HF of frequency domain analysis (discussed above).

PhD candidate and HRV researcher James Heathers provides a good explanation of why we would want to track changes in RMSSD vs. other HRV values throughout training here. I’d like to add that RMSSD is one of the few meaningful values that we can acquire with ultra-short measurement durations. It’s generally accepted that a 5 minute recording is the gold standard for HRV analysis (Task Force 1996). However, 5 minutes is entirely too long if we expect compliance from athletes or individuals. Thankfully, ample research exists that shows that ultra-short (60 seconds or less) RMSSD values (randomly selected from within a 5 minute recording) highly correlate with RMSSD from the standard 5 minute ECG recording (Katz et al. 1999; Mackay et al. 1980; Nussinovitch et al. 2012; Nussinovitch et al. 2011; Salahuddin et al. 2007; Smith et al. 2013; Thong et al. 2003). Unfortunately no research exists that tested the suitability of ultra-short RMSSD in athletic populations so my colleague Dr. Mike Esco and I went ahead and did this very recently in athletes at rest and post-exercise (paper currently in peer review). I will let you know what we found once it gets published.

Why does my HRV score (from ithlete or BioForce) look different from the values in research?

I hope you are not comparing your ithlete or BioFroce scores to data you see in published research. Simon, the creator of ithlete, decided to modify the RMSSD value collected by ithlete to make for a more intuitive and easily interpretable figure for non-expert users. The value you see from the app is the natural log transformed RMSSD multiplied by 20 (lnRMSSDx20). This modification essentially provides a figure on a 100 point scale (though ithlete scores above 100 are possible in highly fit individuals, though not common).

*Note: lnRMSSDx20 is a patented formula and therefore those interested in using this commercially must acquire a licence.


To be clear, RMSSD is only one HRV parameter. By no means was this article suggesting that other HRV values are meaningless. The purpose of this blog was to simply provide an explanation of the what and why of RMSSD since so many people are using ithlete and BioForce lately. Certainly, ECG derived HRV remains the gold standard and likely multiple HRV parameters provide a more complete picture of training status verses just one. However, for the purposes of convenience in non-expert users, the RMSSD provides an easily acquired and interpretable figure in a short period of time that reflects parasympathetic activity which is quite useful for monitoring the effects of training and in the manipulation of training loads.


Berntson, G. G., Lozano, D. L., & Chen, Y. J. (2005). Filter properties of root mean square successive difference (RMSSD) for heart rate. Psychophysiology,42(2), 246-252.

Camm AJ, Malik M et al. (1996) Heart rate variability: Standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Circ 93(5): 1043-1065

Katz A, Liberty IF, Porath A, Ovsyshcher I, Prystowsky E (1999) A simple bedside test of 1-minute heart rate variability during deep breathing as a prognostic index after myocardial infarction. Am Heart J 138(1): 32-38

Mackay JD, Page MM, Cambridge J, Watkins PJ (1980) Diabetic autonomic neuropathy. Diabetol 18(6): 471-478

Nussinovitch U, Cohen O, Kaminer K, Ilani J, Nussinovitch N (2012) Evaluating reliability of ultra-short ECG indices of heart rate variability in diabetes mellitus patients. J Diabetes Complic 26(5): 450-453

Nussinovitch U, Elishkevitz KP, Katz K, Nussinovitch M, Segev S, Volovitz B, Nussinovitch N (2011) Reliability of ultra‐short ECG indices for heart rate variability. Ann Noninvasive Electrocardiol 16(2): 117-122

Salahuddin L, Cho J, Jeong MG, Kim D (2007) Ultra short term analysis of heart rate variability for monitoring mental stress in mobile settings. Conf Proc IEEE Eng Med Biol Soc 4656-4659

Smith AL, Owen H, Reynolds KJ (2013) Heart rate variability indices for very short-term (30 beat) analysis. Part 2: validation. J Clin Monit Comput E-Pub Ahead of Print

Thong T, Li K, McNames J, Aboy M, Goldstein B (2003) Accuracy of ultra-short heart rate variability measures. Conf Proc IEEE Eng Med Biol Soc 3, 2424-2427