The Antwerp Activity Index threshold of 67.5 accurately predicted active days based on WHO criteria with a ROC-AUC of 0.98, 96.0% sensitivity, and 91.8% specificity.
Observational (n=146)
Does the Antwerp Activity Index (AAI) derived from chest strap heart rate monitors accurately classify active days in cardiac rehabilitation and healthy individuals?
The Antwerp Activity Index provides a highly accurate, heart rate-based method for objectively classifying daily physical activity levels according to WHO criteria.
Effect estimate: ROC-AUC 0.98
Abstract Background The introduction of wearable heart rate monitors (HRMs) could enable continuous, transmural, and objective tracking of physical activity (PA). Despite the widespread availability of HR monitors, translating HR data into meaningful, personalized metrics that effectively quantify PA remains an ongoing challenge. Aims To develop and evaluate a novel HR-based score, the Antwerp Activity Index (AAI), for quantifying PA. To determine the optimal daily AAI threshold for classifying active days. Methods This prospective study involved participants from two groups (cardiac rehabilitation and healthy individuals) who wore a Polar H10 chest strap during PA for 12 to 14 weeks. The HR data collected from this HRM was transmitted via an application programming interface (API) to our secure database, where it could be used to calculate the daily Antwerp Activity Index (AAI). The percentage of heart rate reserve (%HRR) was used to determine moderate (≥40-69 %HRR) and vigorous (≥70-100 %HRR) intensity activity zones. Active days were classified based on a predefined threshold of ≥21.43 minutes (i.e. 150 min/week divided by 7) of moderate to vigorous activity (MVA), as recommended by the World Health Organisation (WHO). Minutes spent in the moderate zone counted as 1 MVA-minute, while minutes in the vigorous zone counted double (1 minute = 2 MVA-minutes). The optimal daily AAI threshold to predict active days was determined using Receiver Operating Characteristic Curve – Area Under the Curve analysis (ROC-AUC). Performance of the threshold is evaluated using sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV). Results The average time of device usage was 11.70±6.7 weeks. A total of 146 participants (73.5% male; 48.4±13.4 years old) collected at least 1 AAI value. The median daily AAI value was 73.4 (IQR: 24.7). The AAI score exhibited a strong positive correlation with MVA duration (Spearman's rank correlation coefficient, SCC: 0.961; p.001). The median AAI score on active days significantly differed from the median AAI score on non-active days (80.6 vs. 50.9; p.001). ROC curve analysis calculated an ideal AAI threshold of 67.5 with a ROC-AUC of 0.98. This threshold showed high sensitivity (96.0%), high specificity (91.8%), and high accuracy (94.4%) in predicting active days based on WHO criteria. Conclusions The AAI score provides a robust, HR-based method for objectively classifying daily PA levels. Its strong correlation with time spent in MVA zones and its high predictive performance suggest that it can serve as a valuable tool for real-time activity monitoring in both clinical and general populations.
Buyck et al. (2025) conducted an observational in Cardiac rehabilitation and healthy individuals (n=146). Antwerp Activity Index (AAI) via Polar H10 chest strap vs. WHO criteria for moderate to vigorous activity was evaluated on Optimal daily AAI threshold to predict active days (ROC-AUC 0.98). The Antwerp Activity Index threshold of 67.5 accurately predicted active days based on WHO criteria with a ROC-AUC of 0.98, 96.0% sensitivity, and 91.8% specificity.