Sleep2 logo Sleep2 logo
Case Study

Sleep Monitoring in Special Operations Forces: 23,000 Nights with sleep²


Category

longitudinal field study

Scope

23,000+ nights, 15 months

Participants

Jagdkommando, including control group

Why Sleep Determines Performance in Special Forces

Special Operations Forces (SOF) work under extreme physical and mental demands – often with sleep deprivation and irregular sleep schedules. Sleep quality and duration are the foundation for cognitive performance and decision-making, emotional stability and stress regulation, as well as physical recovery. Lack of recovery worsens reaction time and judgment – with potentially severe consequences in critical situations. Thus, sleep is not a wellness topic, but a variable of operational readiness.

The Study: the Human Peak Performance Project (HP³)

The HP³ research team led by sleep and brain researcher Manuel Schabus and project manager Caroline Rakowitz presented the results at the SOMA 2026 conference in Raleigh (North Carolina) – the world's largest meeting for “Special Operations Medicine” and “Human Performance Optimization”. For the first time, an Austrian team presented its own field data on sleep, recovery, and performance of special forces on this international stage.

Over approximately 15 months, the sleep of operators from the Jagdkommando – the special forces unit of the Austrian Armed Forces – was recorded almost daily in the field: a total of more than 23,000 objectively evaluated nights. This is not a short laboratory study, but a long-term observation under real conditions and the largest study of its kind to date.

Measurement was not done through movement but through a precise heart rate and heart rate variability sensor on the upper arm (Polar Verity Sense) in combination with the sleep² app (University of Salzburg). AI models derive sleep phases and recovery parameters from this, which would otherwise only be accessible in a sleep lab (polysomnography) – from sleep duration and efficiency to sleep onset duration and nighttime awakenings to deep and REM sleep, nighttime heart rate, and HRV.

The study proceeded through several phases: initially a baseline without feedback, then a first intervention phase with daily objective feedback, and finally a second phase with additional recommendations oriented towards cognitive behavioral therapy for insomnia (CBT-I) – such as sleep rhythm, sleep pressure, relaxation, and sleep-promoting routines. A control group was a fixed part of the design.

The underlying measurement methodology was previously validated against 90 polysomnography nights at the Paris-Lodron University Salzburg and compared with 8 other wearables. The results showed the highest accuracy for the sleep² algorithms combined with the Polar Verity Sense sensor (Topalidis et al., 2025).

The Results: Measurably Better Sleep Despite Constant Stress

Over the study phases, several sleep parameters improved – remarkable because stresses in SOF are known to build up over the year. The daily objective feedback and subsequent sleep interventions helped stabilize and partly improve sleep and recovery, while the control group showed increasing sleep fragmentation over time.

 

MetricResult
Subjective Sleep Qualitysignificantly improved (p < 0.001)
Objective Sleep Score (SSC)significantly improved (p < 0.007)
Sleep Stability (NOA2)Intervention stabilized, control participants increasingly restless (Time × Group: F(2, 18663) = 25.10; p < 0.001)
Sleep Efficiencyincreased under intervention (Time × Group: F(2, 18663) = 21.33; p < 0.001)
Sleep Onset Latency (SOL10)shortened in intervention groups (F = 2.77; p < 0.033)

 

Longitudinal effects over three study phases. Source: Schabus & Rakowitz (2026), HP³.

Sleep behaved in these data like a true operational parameter: It changed over time and responded to targeted intervention. Untreated sleep deprivation, on the other hand, accumulated into a growing risk.

What the Data Reveals About This High-Performance Group

The operators are explicitly not "poor sleepers". Low nighttime heart rate and high heart rate variability indicate an extraordinary physical condition and recovery ability. On free days, they significantly caught up on sleep – measured against their stress profile, they tend to sleep too little (comparable high-performance groups like top athletes often aim for eight to nine hours). Notably, there was also a high proportion of REM sleep, which can be associated with increased emotional processing, motor learning, and pronounced stress regulation in SOF.

For us, the special aspect is that we could make sleep visible not only in the lab but directly in the everyday life of this high-performance group.  — Manuel Schabus

The data shows that sleep is measurable, changeable, and highly relevant for health and resilience.  — Caroline Rakowitz

Relevance Beyond the Military Context

The fundamental question concerns many professional groups: What happens to sleep when stress increases over weeks and months – and what helps to stabilize it again? Insights from this particularly stressed group can be long-term applied to shift work, police, emergency services, medical personnel, and top sports. The central message: Sleep can be objectively made visible, understood, and specifically improved today over a long period.

Recommendation

  • Treat sleep like any other mission-critical variable: measure, understand, and then actively manage.
  • Use objective, HRV-based monitoring – not just actigraphy – to detect performance drops early, instead of noticing them only in critical situations.
  • Scalably improve sleep – through daily feedback and digital interventions oriented towards CBT-I as implemented in sleep².

 

Sources:
Schabus, M., & Rakowitz, C. (2026). Daily Sleep Monitoring in SOF: Longitudinal Insights from Over 23,000+ Nights Using Wearable HRV and Digital Sleep Intervention. Presented at the SOMA 2026 Scientific Assembly, Raleigh (NC), Apr 27 - May 01, 2026 . Human Peak Performance Project (HP³).

Validation of the measurement methodology: Topalidis et al. (2025), PsyArXiv. https://doi.org/10.31234/osf.io/27wun_v1