Parkinson disease (PD) is a progressive, neurodegenerative disorder that is increasing in prevalence [1]. Falls and cognitive impairment are major sources of disability in PD, associated with loss of independence, hospitalization, long-term care facility placement, and mortality [2-7], contributing to the estimated US $52 billion annual cost for PD care in the United States [8]. The early identification of individuals at risk for falls and cognitive decline is critical to enable appropriate intervention, which may take the form of home modifications, medication changes, or other adjustments, and to improve long-term outcomes. While many PD-specific and nonspecific risk factors for falls have been identified, the single best predictor for falls remains a prior history of falls, and new, more accurate models for fall prediction are needed [9-11]. Similarly, while many PD-specific and nonspecific risk factors for dementia have been identified [12-17], few studies have examined predictors of the new onset of PD-related mild cognitive impairment [18-20].

Digital tools, which enable frequent and objective in-home assessment of motor [21-24] and nonmotor function [25-29], stand to improve the management of PD and may improve the prediction of falls and cognitive impairment. However, they are not routinely used in clinical practice to aid in the prediction of falls or cognitive impairment. Many fall prediction models incorporate performance-based tests, which are typically conducted in a traditional clinic setting. However, relevant metrics, such as gait speed, cadence, and stride time variability, can be measured with digital tools in the home [11,30,31]. Studies have demonstrated the potential to use digital tools to predict falls and time to a first fall [32,33]. Digital tools have also been used to demonstrate that slower gait speed, lower levels of physical activity, less time spent outside of the home, and fewer trips outside the home are associated with cognitive impairment in older adults [34,35]. Novel methods of capturing the most bothersome symptoms and their functional impact [36] can generate a participant-reported natural history of PD [37] that may contextualize and inform understanding of data obtained from objective in-home assessment. Studies are needed to determine whether in-home digital metrics and remote participant reporting can improve the prediction of important PD-related clinical milestones.

In a previous study, AT-HOME PD (Assessing Telehealth Outcomes in Multiyear Extensions of Parkinson Disease Trials), we remotely characterized motor and nonmotor progression of participants with early PD using annual video visits, quarterly smartphone-based motor assessments, and quarterly online surveys over 2 years of follow-up [38]. AT-HOME PD was designed to assess a novel framework for the remote follow-up of clinical trial participants, to compare digital and traditional outcomes, and to identify novel digital disease measures. AT-HOME PD enrolled 226 participants (mean age 64.7, SD 8.8 y; 40% female; mean time since diagnosis 3.8, SD 1.2 y) from 42 US states and 1 Canadian province, who participated in either STEADY-PD III (NCT02168842) or Study of Urate Elevation in Parkinson’s Disease, Phase 3 (SURE-PD3; NCT02642393)—both multicenter, randomized, controlled, phase III studies of potential disease-slowing agents [38-40].

In this study, we will leverage this well-characterized cohort and describe the protocol for AT-HOME PD2 (Assessing Telehealth Outcomes in Multiyear Extensions of Parkinson Disease Trials–2), a decentralized observational study with clinical and digital phenotyping that will: (1) evaluate the extent to which digital tools and remote participant reporting can improve the prediction of falls and cognitive impairment compared with traditional models; (2) quantify change in physical activity and gait over time; and (3) explore the relationship between physical activity, development of falls or cognitive impairment, and symptom progression over time. We hypothesize that digital assessment of gait, sleep, and physical activity in the real world will improve the prediction of falls and cognitive impairment compared with a model composed of traditional measures.

AT-HOME PD2 is a decentralized, observational PD study with up to 3 years of follow-up. All visits and assessments are conducted remotely in the participant’s natural environment. Participants complete 2 video visits per year, wear 2 sensors (Fitbit Charge 5 and ActiGraph CentrePoint Insight Watch) for 1 week each month, complete smartphone-based motor tasks (using the mPower 2.0 app) each quarter, and complete online surveys (via Fox Insight) each quarter (Figure 1).

A notice of award from the National Institutes of Health was received on April 3, 2023 (Multimedia Appendix 1). The study was approved by the University of Rochester Institutional Review Board in August 2023. Study recruitment was initiated on September 15, 2023, and is ongoing. The original pool of potential participants included 215 individuals from the AT-HOME PD study and 256 individuals from the STEADY-PD III or SURE-PD3 studies who had provided consent for future contact but did not participate in the AT-HOME PD study. Participation was opened to individuals from the PD GENEration study on November 18, 2024. As of January 15, 2025, 142 participants have enrolled. Of these, the average age is 69.2 (SD 8.7) years, 85 (59.9%) are male, 137 (96.5%) are Caucasian, 2 (1.4%) are Hispanic or Latino, and participants have a mean disease duration of 8.9 (SD 1.3) years. Among enrolled participants, 110 (77%) participated in the AT-HOME PD study, 9 (6%) in STEADY-PD III only, 8 (6%) in SURE-PD3 only, and 15 (11%) were recruited from the PD GENEration study. Among enrolled participants, 135 reside in 34 US states and the District of Columbia, and 7 reside in 1 Canadian province. Enrollment is anticipated to be completed by April 2025.

AT-HOME PD2 has enrolled 71% of the target number of participants from 3 research cohorts and is expected to generate new insights into the prediction of falls and cognitive impairment, inform the understanding of the role of physical activity in the management of midstage PD, and extend the longitudinal characterization of early PD clinical trial participants to up to 10 years through a participant-friendly, decentralized framework. Remote clinical and digital phenotyping in AT-HOME PD2 will complement in-person observational cohort studies, such as the Parkinson’s Progression Markers Initiative [64] and the Oxford Parkinson’s Disease Centre Discovery cohort [65], and long-term follow-up studies of other clinical trial cohorts, such as Levodopa in Early PD [66], and will provide insights into PD progression in the current treatment era. The inclusion of participants from PD GENEration will enable comparison of progression between clinical trial and observational research cohorts.

In AT-HOME PD2, traditional clinical assessments are complemented by digital tools, including a smartphone-based app, a commercial wrist-worn sensor, and a research-grade wrist-worn sensor. Many larger-scale digital phenotyping efforts have focused on populations with early-stage, untreated PD [67-69]; however, digital tools also have the potential to support the development of new treatments for and the care of individuals with later-stage PD. In AT-HOME PD2, we focus on a population with later-stage disease and use a combination of active motor assessments and passive data collection to generate insights into PD motor impairment, gait, physical activity, steps taken, and sleep. Our primary aim is centered on improving the prediction of falls and cognitive impairment—2 patient-centered disease milestones that are highly relevant to individuals with later-stage disease. To this end, we will use rigorous methods to capture the occurrence of falls and cognitive impairment. Falls are captured via a weekly automated telephone call, which will limit the potential for recall bias. Cognitive diagnosis is determined by a consensus cognitive committee comprised of experienced physicians who will use cognitive, functional, and clinical data to determine cognitive status [70].

Nonetheless, we recognize that this approach has limitations; for example, we anticipate that adherence to weekly fall reporting will decline over the study period. In addition, we recognize that some participants will have already experienced a fall or developed cognitive impairment at the time of study entry. While we will primarily focus on the prediction of time to first fall and development of mild cognitive impairment and dementia after study entry, secondary analyses will evaluate the prediction of fall frequency, time to first fall requiring hospitalization, time to first fall resulting in injury, time to first-ever fall, time to a recurrent fall, time to development of mild cognitive impairment, and time to development of dementia. In addition, we acknowledge that we do not capture every potential risk factor—for example, objective assessment of autonomic function is limited to the capture of heart rate, and neither of the wrist-worn sensors captures blood pressure. We also acknowledge the potential for differential attrition, with participants with greater physical or cognitive impairment more likely to be lost to follow-up. To partially address this, we collect contact information for a secondary informant at baseline.

Beyond the prediction of clinically meaningful events, this study will yield 3-year longitudinal data on change in motor and nonmotor digital measures, which will inform understanding of progression in this understudied stage of PD. Increased physical activity in PD is likely associated with slower decline [71,72] and a lower risk of mortality [73]; yet, studies to date on longitudinal change in physical activity have largely been limited to early PD or relied on self-report [68,74]. We will use active motor tasks and passive monitoring to objectively measure physical activity, steps taken, and gait parameters in a midstage PD sample. In turn, we will identify measures that capture progression over up to 3 years and examine the relationship with change in traditional clinical measures (eg, the MDS-UPDRS), and with the occurrence of clinically meaningful events. An improved understanding of the impact of physical activity on clinical outcomes will inform the counseling of individuals with midstage PD.

In addition, we will compare the performance of the consumer-oriented Fitbit with that of the research-grade ActiGraph device. Digital devices and algorithms must be validated in the appropriate context of use, which in this case is the monitoring of midstage PD in the home environment. While commercial devices are widely available, easily accessible, and may be acceptable for at-home assessment of physical activity in PD, they require validation before integration into clinical care or clinical trials. In this study, participants will wear the Fitbit and ActiGraph simultaneously on the same wrist, which will enable comparisons between devices. We will compare the feasibility of remote wrist-worn sensor assessment through an evaluation of wear-time adherence. We will also examine interdevice reliability through a comparison of physical activity, step count, and sleep parameters. The identification of a consumer device that accurately and reliably captures activity and sleep in PD would provide a potentially scalable method for monitoring these variables as part of clinical care.

In conclusion, in AT-HOME PD2, we will remotely clinically and digitally phenotype 200 participants with midstage PD, with the goal of improving the prediction of falls and cognitive impairment, characterizing long-term changes in physical activity, and examining the relationship between physical activity and clinical progression.