Unlocking the potential of wearable sensors to measure physical activity and predict clinical worsening in pediatric pulmonary hypertension - pediatric pulmonary hypertension (PH) carries a nearly 30%…

The National Heart, Lung, and Blood Institute awarded The Children's Hospital of Philadelphia $1,068,774 on August 11, 2026, under the Cardiovascular Diseases Research program (CFDA 93.837) to develop a cloud-based platform for measuring physical activity via wearable sensors and predicting clinical worsening in pediatric pulmonary hypertension.

The project addresses a critical gap in pediatric pulmonary hypertension treatment: pediatric mortality remains near 30% over five years, yet only two medications are approved for children compared to twelve for adults. Standard adult clinical trial endpoints such as the six-minute walk distance cannot be administered to most children due to age or developmental status, blocking the design of pediatric medication trials. The recipient will deploy open-source computational tools (GGIR for raw accelerometer analysis and MIMS for movement summary) within a cloud-based, modular platform to standardize physical activity data processing across wearable sensor devices, wear locations, and age groups. The platform will assign age and sex-specific physical activity percentile rankings derived from nationally representative National Health and Nutrition Examination Survey data and will develop machine learning models to detect physical activity patterns indicating disease progression. This work overcomes limitations of prior pediatric studies constrained by proprietary sensor software, black-box algorithms, and narrow reference populations that limited generalizability across devices and age groups.

Performance occurs in Philadelphia, Pennsylvania. The award period runs through May 31, 2031.

Unlocking the potential of wearable sensors to measure physical activity and predict clinical worsening in pediatric pulmonary hypertension - pediatric pulmonary hypertension (PH) carries a nearly 30% 5-year mortality. However, only 2 medications are approved for children while 12 are approved for adults. Treatments remain limited as there are no validated clinical trial endpoints that accommodate the pediatric population to enable medication testing. Since physical activity (PA) limitations are characteristic of PH, adult trials often test improvement in functional measures such as the 6-minute walk distance (6MWD), but this test is impossible in most children due to age or ability. There is an urgent need to identify new endpoints applicable to all children regardless of age or developmental status to design clinical trials, increase treatments, and improve outcomes in pediatric PH. PA estimated by a wearable sensor correlates with 6MWD and other disease markers in adult studies and small, single-center pediatric studies. But prior studies are inadequate due to major obstacles including 1) the absence of standardized data processing methods with use of proprietary sensor software, "black box" data processing algorithms, and narrow reference populations that limit generalizability across devices, wear locations (HIP, arm, etc.) and age groups, and 2) lack of tools to process multidimensional raw sensor data beyond local desktop computers to scale to larger clinical networks. Our team's approach will overcome the inadequacies of prior studies by using an innovative cloud-based, modular platform to 1) apply open-source computational tools and scoring algorithms including ggir (raw accelerometer data analysis) and mims (monitor independent movement summary) to raw PA data to assign age and SEX specific PA percentile rankings from nationally representative national health and nutrition examination survey data and 2) develop machine learning models to detect PA patterns associated with clinical status in pediatric PH. We are poised for translation and scaling of this methodology to a well- established clinical network, the pediatric pulmonary hypertension network registry. Our multi-disciplinary team with expertise in pediatric PH, PA, wearables, bioinformatics, and machine learning will analyze longitudinal raw sensor data from 200 registry participants from 6 sites to 1) establish PA metrics including mims-PA percentile ranking and ggir-derived metrics (volume, number/duration of PA bouts, intensity gradient) as indicators of disease severity (aim 1) and 2) train and validate a novel machine learning model incorporating PA metrics to predict clinical worsening in pediatric PH (aim 2). This proposal is strengthened by my foundational experience processing raw sensor data in pediatric PH patients and healthy children through K23HL150337. Aligned with nhlbi's strategic vision and research priorities, these studies will "identify (DIGITAL) biomarkers of PH that could better identify individuals at high risk…and guide treatment." upon successful completion of these aims, we will standardize methods for estimating PA from a wearable sensor and establish the clinical relevance of PA in pediatric PH. Our approach will facilitate testing of new therapies in PH clinical trials and can be generalized to other cardiopulmonary conditions to augment patient monitoring, expand treatments, and improve outcomes.