Leveraging Digital Biomarkers from Wearable Devices for Early Prediction of Acute Exacerbations in COPD Patients

Prof. Andreas Schmidt; Dr. Anya Sharma; Dr. Mustafa Kemal Ergün

Abstract

Background: Chronic Obstructive Pulmonary Disease (COPD) is characterized by recurrent acute exacerbations (AECOPD) that lead to significant morbidity, mortality, and healthcare costs. Early detection of impending AECOPD is critical for timely intervention and improved patient outcomes. Wearable devices offer a novel opportunity to continuously monitor physiological parameters, generating 'digital biomarkers' that could predict exacerbations in advance. This study aims to evaluate the utility of digital biomarkers derived from commercially available smartwatches for early prediction of AECOPD. Methods: A prospective observational study recruited 200 stable COPD patients who wore a commercial smartwatch (monitoring heart rate, heart rate variability, sleep patterns, activity levels) for 12 months. Anonymized data were collected alongside daily symptom logs. A machine learning model (Random Forest) was trained to identify patterns indicative of impending AECOPD, using data from up to 72 hours prior to a clinically confirmed exacerbation. Results: The model achieved an AUC of 0.85 (95% CI: 0.83-0.87) for predicting AECOPD within 48 hours. Key predictive features included increased resting heart rate variability (decreased SDNN, p<0.01), disrupted sleep patterns (increased fragmentation index, p<0.001), and a significant reduction in overall activity levels (steps/day, p<0.01). The sensitivity was 78% and specificity was 72%. Conclusion: Digital biomarkers from wearable devices demonstrate significant potential for early, non-invasive prediction of AECOPD, offering a proactive approach to managing this chronic condition and potentially reducing hospital admissions.

Keywords: digital biomarkers, wearable devices, COPD, acute exacerbation, predictive analytics

Acute exacerbations of Chronic Obstructive Pulmonary Disease (AECOPD) represent a major clinical and economic burden, frequently leading to hospitalizations, accelerated disease progression, and impaired quality of life. Current methods for detecting impending exacerbations rely on patient-reported symptoms, which often manifest late. Continuous, passive monitoring via wearable technology offers a transformative paradigm shift, enabling the development of objective digital biomarkers for early warning irrespective of patient reporting. This presents a significant opportunity to intervene proactively and improve patient management.

The aim of the study.

To identify and validate digital biomarkers derived from commercially available wearable devices for the early prediction of acute exacerbations in COPD patients, and to quantify their predictive performance in a real-world setting.

Material and methods.

A 12-month prospective observational cohort study was conducted with 200 adult patients diagnosed with moderate to severe COPD (GOLD Stage II-IV). Participants were instructed to wear a commercial smartwatch (Fitbit Sense) continuously and to complete a daily symptom diary using a smartphone application. The smartwatch collected continuous data on heart rate (HR), heart rate variability (HRV - SDNN, RMSSD), sleep duration and quality (sleep stages, fragmentation), and physical activity (steps per day, active minutes). Clinically confirmed AECOPD events were identified by an independent pulmonologist based on predefined criteria and hospital records. A Random Forest classifier was trained on a dataset comprising 72-hour windows of physiological data preceding an AECOPD event or a matched non-exacerbation period. Feature engineering involved calculating daily averages, standard deviations, and trends. Model performance was evaluated using a nested cross-validation strategy, with outcome measures including Area Under the Receiver Operating Characteristic (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

The Random Forest model achieved an AUC of 0.85 (95% CI: 0.83-0.87) for predicting an AECOPD within 48 hours, demonstrating strong predictive capability. At an optimized threshold, the model yielded a sensitivity of 78% and a specificity of 72%. The PPV was 65%, and the NPV was 83%. Feature importance analysis revealed that decreased daily SDNN (a measure of HRV, p < 0.01), increased sleep fragmentation index (p < 0.001), and a significant drop in mean daily steps (p < 0.01) were the most influential digital biomarkers predicting an impending exacerbation. Changes in resting heart rate were also significant (p < 0.05) but less important. The model was able to provide a median of 36 hours (IQR: 24-48 hours) lead time before symptomatic onset.

This study clearly demonstrates that digital biomarkers extracted from widely accessible wearable devices hold significant promise for the early prediction of acute exacerbations in COPD. The identified physiological shifts in HRV, sleep patterns, and physical activity provide a powerful, non-invasive, and continuous warning system. This proactive monitoring approach has the potential to enable earlier clinical interventions, reduce the severity and duration of exacerbations, improve patient quality of life, and ultimately decrease healthcare utilization. Further research is warranted to develop robust clinical decision support systems based on these findings and to evaluate their impact in larger, diverse patient populations within real-time clinical workflows.

1. Sgalla G, et al. Acute exacerbation of chronic obstructive pulmonary disease: current and future perspectives. Eur Respir Rev. 2021;30(161):210086.

2. Inan OT, et al. Wearable Technologies for Chronic Disease Management. Annu Rev Biomed Eng. 2018;20:539-563.

3. Li X, et al. Early prediction of COPD exacerbations using digital health technologies: a systematic review. NPJ Digit Med. 2022;5(1):145.

4. Minet A, et al. Wearable sensors for continuous monitoring of vital signs in COPD patients: a systematic review. Sensors (Basel). 2020;20(17):5015.

5. Spruit MA, et al. An Official American Thoracic Society/European Respiratory Society Statement: Key Concepts and Advances in Pulmonary Rehabilitation. Am J Respir Crit Care Med. 2013;188(8):e13-e64.