Beyond HRV: What Your Heart's Electrical Shape Reveals About Stress
HRVECGMachine learningStress

Beyond HRV: What Your Heart's Electrical Shape Reveals About Stress

A machine learning study on combat-exposed soldiers found that the shape and timing of ECG waves predicted anxiety more reliably than HRV alone — because anxiety lives in the body — yet current consumer wearables ignore most of these electrical signals.

Published: March 29, 2026

More Than the Gap Between Beats

If you wear a fitness tracker, you probably know your heart rate variability. HRV — the millisecond-level fluctuations between heartbeats — has become the default metric for stress, recovery, and readiness on devices from Garmin to Apple Watch to Whoop. Meta-analyses consistently show that people under chronic stress, anxiety, or depression tend to have reduced HRV, reflecting a shift toward sympathetic ("fight or flight") dominance and impaired vagal tone [1, 2]. Your wearable distills this into a single number, and you check it each morning.

But that number captures only one dimension of what your heart is doing. Every heartbeat produces a complex electrical waveform — the familiar PQRST trace on an ECG — with distinct peaks and valleys that reflect atrial contraction, ventricular depolarization, and recovery. The specific heights, durations, and ratios of these waves carry information about how stress physically reshapes cardiac activity. A new study led by Illya Chaikovsky at Ukraine's Glushkov Institute of Cybernetics suggests this waveform detail matters just as much as HRV for detecting psychological distress — and potentially more [3].

The researchers examined 90 Ukrainian servicemen (average age 38) entering rehabilitation after at least three months in a combat zone. None had detectable heart disease on routine ECG screening. Using a compact, portable 6-lead ECG device called "Cardio + P," the team captured high-resolution cardiac data and processed it through a proprietary scoring system that translates both standard ECG parameters and HRV measures onto a standardized 100-point scale. They compared these physiological readings against four established psychological assessments: the Beck Anxiety Scale, the PTSD Checklist for DSM-5 (PCL-5), the Patient Health Questionnaire-9 for depression (PHQ-9), and structured psychologist interviews. A sequential feature selector — a machine learning algorithm that iteratively tests combinations of features — identified the top 40 ECG and HRV parameters that best predicted each psychological score, using linear regression for interpretability and five-fold cross-validation to guard against overfitting.

The Shape of Stress

The clearest result was for the Beck Anxiety Scale, which achieved a cross-validated R² of 0.520 on training data and 0.359 on the held-out test set — meaning the model explained roughly 52% and 36% of the variance in anxiety scores, respectively. The other psychological measures showed weaker but still positive predictive relationships (PCL-5 test R² of 0.294; PHQ-9 test R² of 0.248).

The Beck scale's advantage has a logical explanation: 15 of its 21 items measure physical symptoms — numbness, dizziness, heart pounding, difficulty breathing. It is, in effect, a somatic anxiety inventory. When anxiety manifests physically, the heart's electrical signature naturally reflects it. The PTSD and depression scales lean more heavily on cognitive and emotional items that the cardiovascular system mirrors less directly. The model worked best precisely because anxiety lives in the body.

The most interesting finding was which cardiac features the model selected. While HRV parameters like PNN50 (the percentage of successive heartbeat intervals differing by more than 50 milliseconds) appeared in the top features for several scales, so did amplitude-time parameters of the ECG itself — the actual heights of Q-waves and S-waves in specific leads, the area under the P-wave, and ratios between wave amplitudes. Composite indices reflecting overall myocardial condition and regulatory reserves also ranked highly. In correlation analysis, the Beck Anxiety Scale produced 26 statistically significant Pearson correlations with ECG/HRV parameters at p < 0.01 — far more than any other psychological measure [3]. In short, the shape of the heartbeat carried stress information that beat-to-beat timing alone did not capture. The researchers attribute this to known but understudied effects of chronic stress on the myocardium: elevated cortisol, inflammation, oxidative damage, and microvascular flow disturbances can alter the amplitude and timing of ECG waves even in people without clinical heart disease [3, 4, 5].

What This Means for Wearables — and What It Doesn't

Here's the gap that matters for anyone relying on consumer devices. The Apple Watch records a single-lead ECG equivalent to Lead I of a standard 12-lead setup. It is FDA-cleared for detecting atrial fibrillation and classifying sinus rhythm — and it does this well. But its heavily filtered single-lead signal attenuates P-waves, introduces artifacts around the QRS complex, and cannot assess the multi-lead amplitude relationships this study relied on [6]. Most other wearables don't record ECG at all; they estimate HRV from optical sensors at the wrist, losing waveform morphology entirely. The Chaikovsky study used a portable but research-grade 6-lead device capturing Q-wave amplitudes across leads II and aVR, S-wave amplitudes in leads aVR and III, P-wave area in lead I — measurements no current consumer wearable provides.

Several caveats also temper the findings themselves. The sample was small (n = 82–84 after outlier removal) and specific: male combat veterans in rehabilitation. Whether these ECG-anxiety correlations generalize to civilian populations, female subjects, or people without trauma exposure is unknown. The study was cross-sectional, capturing one moment rather than tracking changes over time, so it cannot show whether waveform shifts predict future anxiety or merely reflect current state. A test-set R² of 0.359 is scientifically meaningful but leaves nearly two-thirds of anxiety variance unexplained — far from clinical-grade prediction. And the proprietary scoring system, while internally validated against 1,112 healthy volunteers, has not been independently replicated [3].

That said, the technology trajectory points toward closing the hardware gap. Next-generation wearable ECG systems are exploring flexible multi-lead electrodes, AI-powered waveform analysis, and soft sensor materials that could bring clinical-grade morphological analysis to everyday devices [7]. If and when that arrives, the kind of stress signatures this study identified could become part of a morning readiness check — not just HRV, but the full electrical fingerprint of each heartbeat. Your wrist sensor already listens to the rhythm. The question is when it starts listening to the shape.

References

  1. Chalmers, J.A. et al. "Anxiety Disorders Are Associated with Reduced Heart Rate Variability: A Meta-Analysis," Frontiers in Psychiatry 5 (2014) 80. doi:10.3389/fpsyt.2014.00080
  2. Kim, H.-G. et al. "Stress and Heart Rate Variability: A Meta-Analysis and Review of the Literature," Psychiatry Investigation 15 (2018) 235–245. doi:10.30773/pi.2017.08.17
  3. Chaikovsky, I. et al. "Analysis of Heart Rate Variability and Subtle ECG Changes Based on Machine Learning for Objective Assessment of the Psychological State of Military Personnel," Frontiers in Psychology 17 (2026) 1688230. doi:10.3389/fpsyg.2026.1688230
  4. Lampert, R. "ECG Signatures of Psychological Stress," Journal of Electrocardiology 48 (2015) 1000–1005. doi:10.1016/j.jelectrocard.2015.08.005
  5. Osborne, M.T. et al. "Disentangling the Links Between Psychosocial Stress and Cardiovascular Disease," Circulation: Cardiovascular Imaging 13 (2020) e010931. doi:10.1161/CIRCIMAGING.120.010931
  6. Saghir, N. et al. "A Comparison of Manual Electrocardiographic Interval and Waveform Analysis in Lead 1 of 12-Lead ECG and Apple Watch ECG: A Validation Study," Cardiovascular Digital Health Journal 1 (2020) 30–36. doi:10.1016/j.cvdhj.2020.07.002
  7. Next-Generation Wearable ECG Systems: Soft Materials, AI Integration, and Personalized Healthcare Applications, Chemical Engineering Journal (2025). sciencedirect.com