Return to site

Inter-Day Stress Reactivity:

Wearables Provide a New Method to Capture Stress Response Across Multiple Days

The current state of stress response research is becoming outdated, but new technologies might allow us to overcome some of the historical limitations that have plagued the field. For decades researchers have brought participants into laboratory settings, which are novel environments that can be stressful for some participants, and then attempted to stress participants out. This all occurs while participants are connected to cumbersome equipment or are interrupted repeatedly for blood or saliva samples in order to measure participant stress responses.

While this research had led to greater understandings of the stress response, it has a number of inherent limitations. This blog post is an attempt to highlight three (there are many more) limitations of stress research and then show how new technologies can be used to understand the stress response in a new way (Nelson & Allen, 2018; Nelson & Allen, In Press). This is what I will call, "Inter-Day Stress Reactivity," as opposed to "Acute Stress Reactivity."

The Three Problems With Stress Research

1. Ecological Validity (i.e., does research context represent actual lived experience?)

The vast majority of stress reactivity research has taken place within discrete laboratory settings that often lack ecological validity of daily life. In other words, stress research has been limited to collecting data from participants in novel environments that might not generalize to their everyday lives.

What might this look like and how might it differ from participant stress responses in their daily life context?

For simplicity, below is an hypothetical figure that shows a person's stress response measured at three time points across a lab. This is a standard protocol that is assumed to measure baseline stress, peak stress, and stress recovery.

Lab Stress Response

Does the data collected in this hypothetical experiment actually represent what that a participant's stress response would be like in their daily life outside of the lab? Surprisingly, we often implicitly assume that the answer to this question is, "yes", even though we lack data to come to this conclusion. This assumption has potentially large implications for the science of stress.

What if it turns out that this participant's stress response actually looks very different if they experienced the same stress in their home without laboratory equipment and scientists staring at them through cameras?

Below is a hypothetical instance of what this contrast might look like if the participant was stressed at their home, instead of the lab.

Home Stress Response

Here we can see the original stress response in black, which is different from the authentic stress response that happened at home in yellow.

Now, what if we were able to collect a lot more data points at home. Could the "stress response" look any different?

2. Sampling Resolution (i.e., amount of data captured to accurately reflect phenomenon of interest)

Some phenomenon can be captured with low resolution data, but other phenomenon need higher data resolution in order to be accurately captured.

What might this look like?

In the figure below we keep the same resolution in the lab, but increase the sampling frequency at home.

Home Stress Response with Greater Resolution

How might this impact our theories of stress responses? What about our stress response on different days?

3. Intra-Individual Variation (i.e., how our own responses change by day)

Lastly, even if we were to get a better resolution of the stress response that day we would still be limited in our assumptions about that individual's stress response across time, because an individual on day 1 might have a different response than the same individual exposed to the same stressor on day 2.

What might this look like and how might it impact our understanding of stress?

Difference in Home Stress Response with Greater Resolution Across Two Days

Summary of Limitations to Stress Research

  1. The vast majority of stress reactivity research takes place in a laboratory, which doesn't always match real world conditions.
  2. The frequency of data collected is often of low resolution, which may prevent researchers from accurately capturing a stress response.
  3. A small slice of data is collected from a participant, usually within a single day, and this is assumed to represent their stress response any day of the year.

As researchers we often assume that the limited data we collect in the laboratory can then paint a picture of chronic stress responses over longer periods of time. Interestingly, until recently researchers have never been able to actually test the implicit assumption of chronic stress in high temporal detail. This could have major implications for better understanding the transition from healthy to hyper to hypo-reactivity of different stress response systems.

A Solution

Wearable, smartphone, and smart home technology allow for the continuous and passive collection of data that can be used to track stress responses over time and may allow for researchers to capture the dynamic stress response not only within an hour of one day, but across multiple days.

Let's take a look at what this might look like.

Inter-Day Stress Reactivity

Below is a plot of my resting heart rate, collected by an Apple Watch, over the past 4 months that is color coded by month. The red horizontal line depicts my average resting heart rate for the past year, while the dots and lines indicate the variability or change in my resting heart rate across that time. Note that resting heart rate was not collected on every day.

Another way to visualize this is to plot the daily differences in resting heart rate from my average resting heart rate across the past year.

It is interesting to note that each spike in resting heart rate, somewhat resembles what we would think of as a stereotypical stress reactivity response over a short period of maybe an hour, but this image is showing a stress response across multiple days!

From this bird's eye view, we can see that there is some normal variation around my average resting heart rate and that there are 4 instances where there have been massive increases in my resting. This is composed of a few days of anticipatory anxiety where resting heart rate increases, a day of peak resting heart rate, followed by a sudden and steep drop in resting heart rate below the yearly average in the days following the peak. In addition, there are also two smaller increases in resting heart rate in middle and late December.

Well, what significant events were happening during this time?

As you can see from the image above, I had large increases in resting heart rate in the days around flights and interviews. Self-disclosure time: I have always hated flying and it causes massive stress during the week leading up to the flight. I usually experience much less stress on my flight back home as I would have recently been exposed to the stressful stimuli during the first flight. Unfortunately, I don't fly frequently enough to extinguish this phobia and this is what my stress response looks like after years of exposure to flights.... maybe VR can help?

Okay, back to the graph. We can also see a small stress response around my wife's 30th birthday when I surprised her with a snowboarding trip out of town, which coincided with hearing back from interview applications. There is also a small stress response of longer duration that occurred around the holidays when my wife, two dogs, and I drove out of state and stayed with family for holiday.

What was so fascinating about looking at this figure is that when we break the graph down by month, it really starts to resemble what we would think of as the typical stress response within a day. The interesting thing is this is a stress response across days!

Monthly Trajectories


What might inter-day stress reactivity mean for science? Might this new way of conceptualizing stress over long periods of time allow us to better understand how stress "gets under the skin" to influence processes related to disease? What about how stress responses proceed changes in mental health? Might we be able to come up with machine learning algorithms that can identify when a client or patient deviates from their average baseline of stress in order to notify patient and healthcare provider that treatment may be needed? How might incorporating measures of affective sentiment from text messages, facial expressions from selfies, frequency of phone calls, physical activity, and patterns of GPS location improve prediction in these algorithms?

This new data has massive potential to transform clinical health psychology and medicine and potentially improve prevention, intervention, and assessment methods in order to better lives.

Future Directions

I'm just finishing up my Ph.D. program in clinical psychology where I study the comorbidity between psychopathology and biological markers of disease (e.g., cardiac psychophysiology, cortisol, inflammation, and telomere length).

Next year I will be completing my clinical residency at University of Washington School of Medicine in Behavioral Health. After my residency year, my goal is to secure a faculty position in an academic hospital or in industry to integrate wearables into my research investigating mental health-disease comorbidity.

For more information, check out my website.

All Posts

Almost done…

We just sent you an email. Please click the link in the email to confirm your subscription!

OKSubscriptions powered by Strikingly