By Deborah Borfitz
April 26, 2021 | If Stanford University vascular surgeon Oliver Aalami, M.D., gets his wish, passive activity tracking will one day be used for a new primary vital sign along with measures of body temperature, pulse, respiratory rate, and blood pressure. As he imagines it, the “activity index” (Ai) of patients would be recorded on their smartphone and reported as an exponential 14-day moving average taken twice, with the slope showing the longer-term directional trend.
The additional data point could be valuable to physicians when evaluating patients in the emergency room with vague complaints and deciding on degree of diagnostic testing, such as whether to order a CT scan, says Aalami. When faced with surgical decision-making, information on the functional capacity or “frailty” of patients would be critical to determining the procedures for which they would be best suited. This includes their ability to tolerate an invasive procedure, and if they need pre-habilitation.
When managing chronic conditions like peripheral artery disease (PAD), some payers have started requiring that patients get exercise therapy before an intervention will be reimbursed, he adds. Activity trackers have been playing a growing role in such programs.
As a member of the Society for Vascular Surgery (SVS) Health IT Committee, Aalami himself helped the SVS build a Supervised Exercise Therapy (SET) app last year that is now in pilot testing at multiple sites across the country. As clinical associate professor of surgery and director of Biodesign for Digital Health at Stanford, he was already actively working on digital technology to better monitor the movement of patients “in the wild.”
SET is a first-line guideline recommendation in the cardiovascular arena but “few are getting it” due to the scarcity of programs, poor reimbursement, and challenges of participating in person, says Aalami. Add to this that fee-for-service providers are incentivized more to be productive and less to keep patients healthy and outside the expensive four walls of the hospital, he adds. But SET is a perfect fit for digital health where such services can be offered at a low cost, are scalable, and can achieve similar if not better results than site-based programs.
In a study that recently published in PLOS ONE (DOI: 10.1371/journal.pone.0247834), Stanford researchers demonstrated that the frailty and functional capacity of patients with cardiovascular disease could be monitored and evaluated remotely using an Apple Watch and an iPhone preloaded with a homegrown VascTrac research app. A six-minute walk test done at home performed on par with nearest observed in-clinic six-minute walk tests.
The participants were mostly over 65 with shorter stride lengths and a more shuffling gait than the young and healthy triathlete types for which the monitoring devices were designed, Aalami notes. Some patients were using canes and walkers, and phones were “all over the place,” including inside pockets and tucked in brassieres. It was important to let patients carry the phones the way they would in the “real world,” he says.
Although the Apple Watch was included in the study, Aalami says, he is more interested in using just the phone. Wearables come with the “risk of drop-off of use over time.”
The new reliability and repeatability study was preceded by an accuracy study published in npj Digital Medicine (DOI: 10.1038/s41746-018-0073-x) that found the iPhone’s built-in algorithms vastly overestimated distance but did a fairly good job of counting the steps of PAD patients when compared to supervised measurements produced on a walking course, with two cones and someone watching, in the clinic. Although the six-minute walk test traditionally measures distance, steps are more reliable when patients are being monitored remotely because of this error issue, Aalami says.
Next Steps
Unless and until it is built into the workflow of physicians it will be hard to broadly translate the measurement of patients’ activity level into clinical practice, Aalami says. The SVS SET app for home-based exercise therapy, which incorporates health coaches and patient-reported outcome surveys to monitor patients’ quality of life, is an important step forward. The app came in response to abusive overuse of procedures for patients with occlusive PAD, to give providers a low-cost and scalable medical management alternative to try before resorting to surgery.
The case for making the Activity Index a vital sign will be the subject of a forthcoming paper, he adds, and an Android implementation would help move that dream closer to reality. For Ai to even be approved as a new metric will require a LOINC code providing a standard way to electronically exchange activity tracking information. iPhone and Android software would then need to start reporting Ai as a data element that could be easily shared.
But to their credit, Apple and Google have both been following open interoperability standards, says Aalami. The health record elements currently reported (Google Health API for Google and Apple Health Records App for Apple) are faithfully adhering to recommendations coming out of the Argonaut Project. An additional push is coming from the 21st Century Cures Act final rule issued last year, which mandates hospitals to expose all data elements in electronic health records, allowing patients to become stewards of the data and “infusing fuel into the consumer health movement.”
Of the 156 cardiac and vascular surgery patients invited to participate in the most recently published study, 46 did not meet study criteria or opted not to enroll. A common reason was concerns about being monitored, he says. “Otherwise, people were excited about getting the feedback and being connected to the coordinators,” which many of the participants also used as an access point for prescription refills or to report a change in their symptoms.
Passive data collection using mobile applications and sensors should provide a more accurate picture of patients’ functional capacity than the short interview done in the clinic, says Aalami. The pandemic has been a major catalyst for the adoption of telemedicine and remote monitoring, especially in clinical trials. California startup Medable late last year secured an astounding $91 million in investor funding to accelerate the life science industry’s shift to decentralized clinical trials, he notes.
“A big passion and obsession [of mine] are identifying the settings and scenarios where remote monitoring programs can be implemented,” Aalami says. Most such activity is currently happening in the context of integrated delivery networks in regions where payer and provider incentives are aligned. Ochsner Health System in New Orleans is among the most advanced users of remote technologies “simply because 70%-80% of its patients are on Medicare Advantage,” a risk-based payment model that puts providers at-risk for the care of their members.
Clinical Motivation
About a decade ago, Aalami says he started getting frustrated when patients would present with an occluded stent and tell him they noticed months ago that they were experiencing new walking difficulties but thought the problem would get better on its own. That is when he realized the many potential opportunities for monitoring patients at home between scheduled office visits.
His clinical experience has been that the degree of in-stent restenosis directly correlated with level of physical activity, says Aalami, so his “grand idea” was to use the VascTrac app and commercially available activity tracking devices to help predict pending stent failure. “I quickly realized the algorithms on all these devices… were not really developed to measure patients with cardiovascular disease, at least originally.”
As a first step, he had to clinically validate data generated by sensors in the digital devices for this patient population. Although Aalami faced challenges mustering up enthusiasm for the concept in the surgical community, he was able to secure funding for the project and was surrounded by students with the smarts to help him carry it out completely in-house.
In his role as an educator, Aalami is part of a 20-year-old Stanford Biodesign program that teaches a needs-based approach to innovation in healthcare. The focus in recent years has been smart, sensor-embedded devices, he says, including how to securely get data off sensors and return results to patients and providers in a meaningful way. Among recent digital health solutions built by students in the Biodesign for Digital Health program are a remote blood pressure monitoring platform for cardiologists and a medication compliance application for patients who have had a kidney transplant.
One of the lessons taught is that consumer and healthcare products and applications are governed by different sets of rules, Aalami says. In the healthcare realm, the rules are “largely around privacy, security, and data interoperability.” Engagement can be challenging, he adds.
Operating Systems
When it comes to research and trying to normalize all the variables, Aalami says, the Apple iOS ecosystem is far better positioned currently than Google’s Android operating system. His projects at Stanford all use Apple devices and the company’s ResearchKit (open-source framework for creating apps) and HealthKit (central repository for health and fitness data). The Google Cloud platform is used on the backend.
“We open-sourced a lot of the work we’ve done to make it available to the wider world,” Aalami quickly adds. The framework, called CardinalKit, allows anyone with basic iOS skills to create a basic but fully functional research application with all the necessary security components within three hours.
“We are very keen, especially with the CardinalKit framework, to add Android support,” says Aalami. Android still commands the majority share of the market, especially in lower socioeconomic settings. His group is working on cross-platform support to ensure accessibility and help address health disparities.
For the study published in PLOS ONE, all 110 participants were provided with an Apple Watch Series 3 and iPhone 7—each with a current retail price of around $200—that they could keep at the study’s conclusion. “It otherwise would have been very hard to deploy,” says Aalami.
Android has a lot of catching up to do in terms of functionality, says Aalami, noting that the CommonHealth Framework from the Commons Project was only widely deployed last year. It was designed as a cross-platform alternative to Apple Health, on the market for more than five years now, allowing iPhone users to download and share their health records.
While Google Fit launched shortly after Apple’s HealthKit, Google Health Studies only came to market last year in answer to Apple’s ResearchKit. Overall, Apple is the easier place for researchers to start because it is more turnkey, says Aalami. “Android is very bespoke… [with] all these different accelerometers from different manufacturers and the data store for the health data isn’t as secure out of the box.”
No matter what everyday device is being used, a lot of the raw activity data are “really hard to interpret,” says Aalami. “What bothers me about a lot about medicine is that we keep putting people into buckets” based on whether individuals fall above or below a finite number based on population studies. This is not a very… personalized approach.”
Digital health provides an opportunity to monitor patients individually and look for trends, Aalami says. “The absolute value is less important than the change over time… and that’s the dimension you have to add to help with the analysis interpretation.”
Interpretation science is, “still trying to figure out how to digest the data, analyze it, and report it back in a meaningful way,” he adds, “and that’s the whole evolution of this activity index.” Measuring activity as a moving average considers multiple real-world variables—including trips, illness, and whether it is a weekday or the weekend.