Researchers at Mount Sinai developed personalized AI models that use wearable-device data to forecast prolonged sitting among women with chronic pelvic pain. The study analyzed data from 134 women with chronic pelvic pain disorders and 61 healthy participants, providing a foundation for future digital tools designed to deliver more timely movement prompts.

Key Takeaways

• Mount Sinai researchers developed AI models to forecast upcoming periods of prolonged sitting among women with chronic pelvic pain
• The study analyzed wearable data from 134 women with chronic pelvic pain disorders and 61 healthy participants
• Participants wore Fitbit devices for up to 90 days, generating activity, heart-rate and sleep data
• Simpler forecasting models performed comparably with more computationally intensive deep-learning models
• Prospective clinical trials are still needed to determine whether personalized movement prompts reduce sedentary time or improve symptoms and quality of life

Mount Sinai Researchers Develop AI Wearable Forecasting Models

Researchers at the Icahn School of Medicine at Mount Sinai developed personalized AI models that use wearable-device data to forecast periods of prolonged sitting among women with chronic pelvic pain.

The research examined whether information collected from consumer wearable devices could be used to anticipate sedentary behavior. The models were designed to forecast activity patterns rather than diagnose or treat chronic pelvic pain.

The study included 134 women with chronic pelvic pain disorders and 61 healthy participants. Wearable devices provided information on activity and other physiological measures that researchers used to identify individual patterns.

Researchers then developed forecasting models tailored to each participant to determine whether recent wearable data could help predict when someone was likely to remain seated for an extended period.

The study was published in npj Women’s Health, adding a women-specific application to research involving artificial intelligence and wearable health data.

Chronic pelvic pain can occur alongside conditions such as endometriosis. Related endometriosis recognition efforts have also explored how health technology may help clinicians identify patterns associated with complex women’s health conditions.

The current research does not establish that AI wearables reduce chronic pelvic pain or improve health outcomes. Instead, it demonstrates a method for forecasting sedentary behavior that can be evaluated in future intervention studies.

Study Analyzes Wearable Data From Women With Chronic Pelvic Pain

Participants wore Fitbit devices for up to 90 days. The devices supplied activity, heart-rate and sleep data, allowing researchers to examine patterns over extended periods rather than relying on isolated measurements.

These data provided information about participants’ daily movement and inactivity. Researchers used the resulting datasets to train models designed to predict upcoming activity patterns.

The study included women with chronic pelvic pain disorders as well as healthy participants, giving researchers data from people with and without chronic pelvic pain conditions.

Chronic pelvic pain can affect daily physical activity, making sedentary behavior a relevant area for investigation. The study focused specifically on forecasting prolonged sitting rather than determining the medical causes of pain.

Wearable devices also allowed information to be collected during participants’ everyday activities outside a clinical setting. This approach generated repeated measurements that could be used to identify patterns unique to each individual.

Because heart-rate information formed part of the wearable dataset, questions around fitness device heart-rate accuracy remain relevant to the broader use of consumer wearables for health monitoring and research.

The individual-level dataset became the foundation for the personalized forecasting models evaluated in the study.

Personalized Models Predict Sedentary Periods Up to an Hour Ahead

The researchers developed models that could forecast activity levels approximately one hour ahead. The models were also used to identify periods of sitting lasting at least 15 minutes.

Forecasting activity in advance differs from simply recording whether someone has already been inactive. A prediction could potentially identify an upcoming period of prolonged sitting before it occurs, creating the technical basis for a future movement prompt.

The models used each participant’s wearable data to make individualized predictions. This allowed researchers to examine whether recent activity patterns could help forecast near-term behavior.

Approximately 10 days of individual data could be used to train the forecasting models. Researchers compared different modeling approaches to determine how effectively they predicted subsequent activity.

The forecasting process incorporated several types of wearable information. Activity data provided direct information about movement, while heart-rate and sleep measurements added further information for the models.

The goal was to evaluate personalized forecasting rather than establish a treatment protocol. Future research will need to determine whether predictions can be translated into effective interventions.

Simpler AI Models Perform Comparably With Deep-Learning Approaches

The study compared relatively simple forecasting models with more computationally intensive deep-learning approaches. The simpler models performed comparably with the more complex systems for the forecasting tasks researchers evaluated.

The finding suggests that predicting sedentary behavior may not require the most computationally intensive modeling approach tested in the study.

Researchers also found that individualized prediction could be achieved with models less complex than deep-learning systems. This provides a basis for evaluating different levels of model complexity when developing wearable-based forecasting tools.

The models were assessed using wearable information collected during participants’ everyday lives. Researchers wanted to determine whether these systems could anticipate activity patterns using data generated outside controlled laboratory settings.

The findings remain specific to the forecasting tasks evaluated in the study. Comparable performance between simpler and deep-learning models does not establish that either approach can improve chronic pelvic pain symptoms.

The research also did not determine whether receiving a movement prompt would cause participants to sit less or experience better health outcomes. Those questions require testing an intervention rather than forecasting behavior alone.

Future Trials Will Test Personalized Movement Prompts

The research provides a foundation for future studies examining whether personalized movement prompts are useful for women with chronic pelvic pain. Such prompts could potentially be timed around an individual’s predicted period of prolonged sitting.

Prospective clinical trials are needed to determine whether these prompts actually reduce sedentary time. Researchers would also need to evaluate whether changes in sitting behavior are associated with improvements in symptoms or quality of life.

The distinction between prediction and intervention is central to the findings. The current models forecast behavior, but the study does not demonstrate that acting on those predictions changes health outcomes.

Future research can build on the forecasting method by testing personalized prompts in controlled settings and measuring both behavioral and health-related outcomes.

For women with chronic pelvic pain, the study highlights a specific aspect of daily activity that wearable technology can measure over time. The next stage of research will determine whether those forecasts can support interventions that produce measurable changes in sedentary behavior, symptoms or quality of life.

Frequently Asked Questions

What did the Mount Sinai AI wearable study find?

Researchers developed AI models that could forecast periods of prolonged sitting among women with chronic pelvic pain. The models used wearable-device data and could predict activity approximately one hour ahead.

How were wearable devices used in the chronic pelvic pain study?

Participants wore Fitbit devices for up to 90 days. The devices collected activity, heart-rate and sleep data that researchers used to develop and evaluate personalized forecasting models.

How many women participated in the AI wearable study?

The study analyzed wearable data from 134 women with chronic pelvic pain disorders and 61 healthy participants.

Can AI wearables currently treat chronic pelvic pain?

The study did not establish an AI wearable treatment for chronic pelvic pain. It examined whether wearable data could predict periods of prolonged sitting.

What research is needed before personalized movement prompts can be used clinically?

Prospective clinical trials are needed to determine whether personalized movement prompts reduce sedentary time and whether any resulting behavioral changes affect symptoms or quality of life.

Disclaimer: This article is for informational purposes only and is not intended to provide medical advice, diagnosis, or treatment. Wearable devices and AI-based health technologies should not replace guidance from a qualified healthcare professional. Anyone experiencing chronic pelvic pain or other concerning symptoms should consult an appropriate medical provider.