Digital health tech/Canada, United States
Ninety-nine nurses, seventy wearable features, and no control group
A single-arm cohort study of 99 nurses paired one VR session with 12 weeks of wearable monitoring and no control group.
A Toronto and Boston team enrolled 99 frontline nurses in a single-arm prospective cohort trial. After a two-week physiological baseline, participants completed one virtual reality session with a brief psychoeducation intervention between two runs, then 12 weeks of follow-up using weekly and biweekly assessments alongside 70 wearable-derived features.
What happened
A team at St. Michael’s Hospital, Toronto Metropolitan University, the University of Toronto, Boston University and the Massachusetts Veterans Epidemiology Research and Information Center published in JMIR Formative Research a study of stress among frontline nurses.
Ninety-nine nurses were enrolled in 2023: mean age 33.7 with a standard deviation of 8.9, 87% female.
The design is a single-arm prospective cohort trial (NCT05923398). Participants completed at least two weeks of baseline monitoring, then a single virtual reality session consisting of two runs separated by a brief psychoeducation intervention, then 12 weeks of follow-up.
In-session stress was measured with the Subjective Units of Distress Scale and the four-item Moral Injury Outcome Scale, with synchronous heart rate variability.
Longitudinal outcomes combined weekly and biweekly ecological momentary assessments with 70 wearable-derived features.
The test
The measurement design is the interesting part and the study design is the weakness.
Pairing a self-reported distress score with heart rate variability at the same moment inside a VR session is a good idea. Self-reported stress and physiological arousal often diverge, and measuring both simultaneously lets the relationship be examined rather than assumed. Ecological momentary assessment over 12 weeks, rather than a single follow-up questionnaire, captures variation that a one-off measure misses.
The problem is that there is no control group. Every participant received the intervention. That means any change over 12 weeks cannot be attributed to the VR session, because nurses who volunteer for a stress study would be expected to improve over three months through regression to the mean, seasonal variation, the attention of being studied, and ordinary recovery.
The authors chose a single-arm design and the paper’s framing as formative research is consistent with that.
The numbers
| Item | Detail |
|---|---|
| Participants | 99 nurses |
| Mean age | 33.7 (SD 8.9) years |
| Female | 87% |
| Enrolment year | 2023 |
| Design | single-arm prospective cohort trial (NCT05923398) |
| Baseline monitoring | at least 2 weeks |
| Intervention | one VR session, 2 runs with brief psychoeducation between |
| Follow-up | 12 weeks |
| In-VR measures | SUDS, MIOS-4, synchronous heart rate variability |
| Longitudinal measures | weekly and biweekly EMAs, 70 wearable-derived features |
The result
What the study establishes is that the apparatus works: 99 nurses can be recruited, baselined for two weeks, taken through a VR session with simultaneous physiological recording, and followed for 12 weeks with wearables and repeated assessments.
What it cannot establish is that the intervention did anything. A single-arm design with a 12-week follow-up has no way to separate intervention effects from the passage of time.
The analytic surface is also large. Seventy wearable-derived features examined against self-reported outcomes in 99 participants is a wide search. Associations will appear, and with that ratio of variables to participants, some will be chance. Any such finding needs independent replication before it means anything.
The intervention itself is thin: one VR session, with two runs and a brief psychoeducation element between them, carrying 12 weeks of follow-up.
How this compares to what came before
Stress and moral injury among frontline healthcare workers became a measured problem during the pandemic, and most interventions have been surveys, wellness programmes and signposting. Attaching continuous physiological measurement is a genuine methodological advance over asking people how they feel once a month.
Within this batch’s extended reality cluster, the contrast in evidential strength is sharp. A network meta-analysis of 56 randomised trials covering 6,154 patients found extended reality reduced perioperative anxiety by 8.05 points against a minimal important difference of 6.71, at moderate certainty, with RoB 2 and GRADE applied throughout.
That is what evidence for a VR intervention looks like when it exists. This study sits at the other end: a single arm, 99 participants, one session.
The same gradient runs through this batch’s digital health cluster. The interventions with measurable benefit were those tested against sharp comparators: motion sensors against the identical app without them. The ones without comparators, or with weak ones, produced nulls or uninterpretable results.
What it means
For occupational health in healthcare, the useful contribution is the measurement protocol rather than the intervention. An organisation wanting to understand staff stress now has a template: baseline physiological monitoring, repeated in-the-moment assessment, and wearable features over months.
For the VR intervention itself, nothing here supports deploying it. A single session is a small dose for a chronic occupational problem, and the design cannot show it helped.
For research design, this study is a reminder of how much a control arm is worth. The same 99 nurses, randomised to VR or to an equivalent-attention control, would have produced an interpretable result at similar cost.
Business ideas from this paper
Occupational stress measurement platforms for healthcare employers, built on continuous physiological monitoring plus ecological momentary assessment, sold as workforce analytics rather than as treatment.
Moral injury screening and pathway referral, since the construct is now measurable with a four-item instrument and most employers have no process for it.
Controlled evaluation services for workplace wellbeing interventions, where the current standard is uncontrolled pre-post measurement and the bar is low.
What could be wrong with it
There is no control group, so no observed change can be attributed to the intervention. Improvement over 12 weeks would be expected regardless.
Ninety-nine nurses, 87% female, at affiliated hospitals is a convenience sample, and volunteers for a stress study differ from those who decline.
A single VR session is a thin intervention on which to hang 12 weeks of follow-up.
Seventy wearable-derived features against 99 participants is a large analytic surface, and associations found across it need independent replication.
No data availability statement appears in the indexed record.
The study is published as formative research, which is the appropriate framing and means it is not designed to establish effectiveness.
How sure can you be?
Confident that 99 nurses were enrolled, baselined for at least two weeks, taken through a single VR session with synchronous heart rate variability, and followed for 12 weeks.
Confident that the measurement protocol combines physiological and self-reported data in a way that most workplace wellbeing research does not.
Confident that the design is single-arm, because the authors state it.
Less confident about any association between physiological and self-reported outcomes, given 70 features and 99 participants.
Not confident about any effect of the intervention, which a single-arm design cannot establish.
Not confident about generalisation from a convenience sample of volunteers.
Our rubric places this at the floor of the range for this batch: peer-reviewed, many institutions across two countries, a prospective cohort design without a control arm, and no public data deposit.
Do this today
When a study has no control group, read it for its measurements rather than its conclusions. The protocol here is the reusable part.
If you are measuring staff stress, pair the questionnaire with something physiological and repeated. A single annual survey is the current standard and it is very weak.
If you are evaluating a wellbeing intervention, add a control arm. It is the cheapest way to make your result mean something, and this study shows what its absence costs.
Source: Digital Health Monitoring and Intervention Suite for Stress in Frontline Nurses: Prospective Cohort Trial, JMIR Formative Research, 2026-06-11. PMID 42275639 · doi.org ().
Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.
- Single-arm prospective cohort trial (NCT05923398): 99 nurses, mean age 33.7, 87% female, enrolled in 2023.
- Protocol: at least two weeks of baseline monitoring, a single VR session with two runs separated by brief psychoeducation, then 12 weeks of follow-up.
- In-VR stress was measured with the Subjective Units of Distress Scale and a four-item moral injury measure, with synchronous heart rate variability.
- Longitudinal outcomes combined weekly and biweekly ecological momentary assessments with 70 wearable-derived features.
Single-arm prospective cohort trial with physiological and self-reported measurement; no control group, convenience sample, single VR session.
Questions people ask
What does single-arm mean for the findings?
That no effect can be attributed to the intervention. Nurses who enrolled in a stress study would be expected to improve over 12 weeks regardless of what was done to them.
What is moral injury?
Distress arising from acting against one's own moral judgement, or being unable to prevent harm. It became a measured concept in frontline healthcare during the pandemic.
Why measure heart rate variability during the VR session?
Because it provides a physiological read on arousal at the same moment the participant reports their distress, which lets the two be compared directly.
Why is 70 features a concern?
Because with 99 participants, testing associations across 70 variables will produce apparent findings by chance. Any such association needs independent replication.