Continuous Physiologic Markers of Heart Rate Variability Derived From Bedside Electrocardiogram Precede Onset of Acute Respiratory Distress Syndrome: A Physiologic Modeling Study.

Summary

Using continuous ECG-derived cardiorespiratory features, particularly heart rate variability, the model predicted ARDS up to 48 hours before onset. A combined waveform+EMR model achieved AUROC 0.92 and PPV 0.58 at a 12-hour horizon, outperforming waveform-only and the Lung Injury Prediction Score.

Key Findings

• Waveform-derived heart rate variability features predicted ARDS up to 48 hours before onset.
• Combined waveform+EMR model achieved AUROC 0.92 (95% CI 0.91–0.93) and PPV 0.58 (95% CI 0.55–0.62) at 12 hours.
• Performance exceeded waveform-only (AUROC 0.86; PPV 0.49) and LIPS (maximum AUROC 0.88; PPV 0.18).
• ECG-based markers provided sufficient dynamical information even without EMR data.

Clinical Implications

If prospectively validated, bedside ECG analytics could trigger earlier ARDS-focused management (e.g., lung-protective strategies, conservative fluids, infection control) before clinical onset.

Limitations

• Retrospective single-system cohort with limited ARDS cases (n=62)
• No external validation; potential overfitting and center-specific biases

Future Directions

Prospective, multi-center validation with real-time deployment, fairness assessment across demographics, and evaluation of clinical impact via interventional trials.