Development of the pain (pain AI intervention) platform for patients at home

The National Institute of Neurological Disorders and Stroke (NINDS) awarded Mayo Clinic Jacksonville $2,277,111 on September 15, 2026, under the Drug Use and Addiction Research Programs (CFDA 93.279) to develop the PAIN (Pain AI Intervention) platform, a wearable sensor-based artificial intelligence system for at-home continuous pain monitoring and personalized opioid management.

The project addresses limitations in patient-reported pain scales by deploying validated deep learning algorithms that recognize physiologic patterns associated with pain responses. A multimodal wearable device will capture electrocardiogram, photoplethysmography, electrodermal activity, body movement, temperature, barometric pressure, and oxygen saturation data, supplemented with blood pressure cuff and pulse oximeter measurements, to build objective pain biomarkers and detect signs of over-medication. The platform aims to create real-time, at-home continuous physiologic monitoring data from which machine learning models will estimate pain intensity and opioid response, enabling personalized medication regimens in a closed-loop system.

The research plan involves three objectives: characterizing multimodal physiologic and psychosocial signatures of postoperative pain across 500 patients recruited at preoperative consultation; validating machine learning and artificial intelligence models to estimate pain intensity and opioid response; and evaluating acceptability, appropriateness, feasibility, and integration potential of the AI ecosystem in outpatient care. Performance occurs at Mayo Clinic Jacksonville's facility in Jacksonville, Florida. The project period runs from September 15, 2026, through September 14, 2029. This is a project grant, an assistance type supporting defined research activities for a specified period.

Abstract

Pain measurement relies on patient-reported scales, prone to four crucial obstacles: high subjectivity, the need for consciousness, the evaluation at only one point, and the inability to describe pain's complexity. The core innovation within this project centers on deploying our previously developed and validated deep learning algorithms leveraging continuous physiologic data captured using a wearable sensor. We hypothesize that the recognition of physiologic patterns driven by pain responses (i.e., physiological data such as heart rate, respiratory rate, and heart rate variability (HRV)) will allow for accurate and safe therapeutic intervention in a closed-loop individualized sensor-based artificial intelligence (AI)-enabled system. This project will allow the creation of a real-time, at-home continuous physiologic monitoring database from which a pain biomarker will be created using artificial intelligence. The overarching goal is to personalize pain medication regimens leveraging sensor-driven, digital ecosystems that can recognize subtle markers of pain and over-medication. Multiple sensors in a single device (ECG, PPG, EDA, body movement, temperature, barometric pressure, and oxygen saturation) will be used, supplemented with a blood pressure cuff and finger pulse oximeter, thereby building a composite and objective profile to determine pain needs and safety indicators. To test the hypothesis, we plan on 1) characterizing multimodal physiologic and psychosocial signatures of postoperative pain, 2) validating machine learning/artificial intelligence (ML/AI) models to estimate pain intensity and opioid response, and 3) evaluating the acceptability, appropriateness, feasibility, and integration potential of the omni-AI ecosystem in outpatient care. A total of 500 patients will be recruited at their initial preoperative consultation and general anxiety levels and pain catastrophizing attitudes will be measured using the state-trait anxiety inventory (stai) and pain catastrophizing scale (PCS). The patients will evaluate their pain intensity using the numerical rating score (NRS) concomitantly. After the surgery, the wearable omni devices will continuously record physiologic signals, which will be stored on-device during the 5-day home monitoring period and securely transferred to mayo clinic servers after device return. During daytime hours (8 am, 12 PM, 4 PM, and 8 PM), patients will be prompted by a mayo clinic-developed mobile APP to submit an NRS pain evaluation and record medication consumption, with the option to enter additional reports at any time when they experience pain. We will create a data bank with subject-specific time series, including physiological variables and self-reported pain scores. To measure the acceptability, appropriateness, and feasibility of our proposed intervention, we will use the validated acceptability of intervention measure (aim), intervention appropriateness measure (IAM), and the feasibility of intervention measure (FIM) surveys. This will be achieved by performing a phone interview five days after the study discharge. The consolidated framework for implementation research (cfir) will also be deployed to assess integration potential.