Wearable AI Forecasts Prolonged Sitting in Women With Chronic Pelvic Pain

Researchers at the Icahn School of Medicine at Mount Sinai have developed an artificial intelligence framework capable of forecasting prolonged sedentary periods in women with chronic pelvic pain disorders. Published on September 30 in npj Women’s Health, the study demonstrates how everyday wearable technology can function as a predictive early-warning system for physical inactivity, enabling just-in-time adaptive health interventions. Chronic pelvic pain, affecting roughly one in seven women and often linked to endometriosis and uterine fibroids, frequently forces patients into extended sitting due to pain and fatigue. Traditional digital health advice lacks the temporal precision needed for this demographic. To address this, the Mount Sinai team analyzed minute-by-minute activity, heart rate, and sleep data collected over 90 days from 134 participants with chronic pelvic pain and 61 healthy controls using Fitbit devices. By training on ten days of per-participant data, they built personalized models to forecast activity levels one hour ahead, specifically targeting 15-minute waking windows suitable for brief movement breaks. The research challenges the industry assumption that advanced deep-learning architectures are necessary for accurate health predictions. Lightweight, interpretable machine learning models matched the performance of computationally intensive networks while maintaining high accuracy despite missing or asynchronous data. This robustness indicates that future forecasting tools could operate directly on personal devices rather than relying on cloud servers, significantly lowering computational overhead and enhancing user privacy. The practical application lies in delivering timed, personalized prompts to move before prolonged inactivity begins. Rather than flooding users with generic notifications, the system would issue minimal, strategically timed reminders to reduce alert fatigue. Senior author Dr. Ipek Ensari highlighted the shift from reactive advice to proactive, context-aware coaching. Lead author Dr. Jannes Jegminat noted that model interpretability facilitates easier integration into consumer wearables and smartphones. The team is now transitioning to clinical validation, embedding the forecasting framework into a just-in-time adaptive intervention to determine whether AI-driven prompts effectively reduce sedentary time and improve quality of life. Success could establish a new standard for privacy-preserving, edge-computing health AI adaptable to other chronic conditions exacerbated by prolonged sitting.