Having a large and diverse group of stakeholders together at the table from the very beginning of the process of developing digital technologies is essential as society addresses ethical and privacy issues related to their use, said Husseini Manji. Magali Haas noted that the workshop itself brought together precisely those stakeholders who could play a part in building precompetitive partnerships with patients at the center providing the data.
One of the goals of the workshop, said Manji, was to brainstorm about how to bring diverse stakeholders together in what he called an “ecosystem of partnerships” to advance the use of digital technologies in health care and treatment development. These partnerships may take many forms, including public–private partnerships and through work with foundations.
Building Partnerships with a Disease and Patient Focus
The Michael J. Fox Foundation for Parkinson’s Research (MJFF) has been an exemplar of a foundation that has moved the field forward through partnerships, said Manji. Their scope extends across the research and development continuum and all the way through health care delivery, according to Catherine Kopil. Thus, they engage a wide array of partners—individuals with PD and their families, health care providers, pharmaceutical and biotechnology companies, medical device and app developers, regulators, and payers—to maximize the chance of success over both the long and short terms. They do so, she said, in service to people who are living with this chronic neurodegenerative disease and who are willing to take risks to expedite discovery of a cure.
MJFF takes a portfolio approach to investing both in therapeutic development directly and in projects that tackle field-wide challenges, such as the development of biomarkers and novel endpoints that could be used for clinical trials and to improve clinical care, said Kopil. These endpoints include data from biosensors that could help detect the cardinal symptoms of PD: tremor, rigidity, and bradykinesia. Indeed, she said that in 2014, MJFF decided to take a proactive role in developing mobile tech endpoints by establishing a partnership with Intel to develop a tool that could provide an objective measure of function as well as patient-reported outcomes. One lesson from this early effort was the need to integrate the patient community’s perspective in an iterative fashion throughout every step of the development process, said Kopil. They also encountered challenges that are still ongoing around data standards and privacy, and the challenge of transforming data into usable information.
An unintended consequence of the Intel partnership, said Kopil, was that by aligning themselves with an exclusive partner, they positioned themselves as competitors to other device developers rather than as a facilitator and sharer of information. This led them to convene the Mobile Tech Advisory Council, which gathered 10 industry partners and the Critical Path for Parkinson’s Consortium to work together in a collaborative manner. They have also partnered with Sage Bionetworks to contribute data from the Intel project to a shared data challenge.
To facilitate the collection of real-world data from people with PD, Kopil said MJFF also invested in a digital platform they call Fox Insight, which uses online questionnaires to collect data on the experiences of persons living with PD. In addition to using Fox Insight for this longitudinal observational study, MJFF is also partnering with two NIH-funded clinical trial teams to conduct long-term follow-up of trial participants with PROs, biosensors, mHealth apps, and telemedicine visits.
Building a Partnership to Advance Digital Biomarkers
Developing digital biomarkers is an iterative process, requiring collaboration from a broad range of stakeholders, said Iain Simpson of IXICO. At every step of the way, from biomarker discovery to clinical trials to clinical practice, IXICO brings together clinicians, academic researchers, technology companies, and pharma partners. They evaluate and validate biosensor measures as markers of disease state; translate algorithms for use in clinical research; deploy them in clinical trials in a way that minimizes site and patient burden; collect the data alongside existing
clinical endpoints; and interpret the data to improve clinical outcomes and inform decision making. IXICO is independent and technology agnostic, said Simpson, which allows them to focus on what is important to measure rather than trying to advance a particular technology. “We are sort of a broker between the technology providers and the pharmaceutical companies,” he said.
Building Partnerships That Look to the Future of Technology
Peter Peumans, senior vice president for life science technologies at imec, calls his company an “ecosystem enabler.” By bringing together companies across the entire product development spectrum—from material suppliers to chip designers and manufacturers to application companies that integrate chips into their products—they are figuring out how to build the next generation of nanoelectronic and digital technology and leverage those technologies to provide useful products in the life sciences and health care.
For example, Peumans described a silicon probe that records neural activity from multiple brain structures in freely moving animals. The probe was developed by imec and academic partners, with funding from the Allen Institute for Brain Science, Gatsby Charitable Foundation, the Howard Hughes Medical Institute, and the Wellcome Trust. The system is capable of reading signals from hundreds to thousands of neurons at the single-neuron level simultaneously, thus allowing assessment of neural circuitry at a much lower cost than could be done with other technologies, said Peumans. One of imec’s requirements for this project was that after the technology was developed and validated, it would be made available to the community at large as a standardized tool, he said.
imec is also looking to advance wearables that produce high-quality medical data and has begun to build an ecosystem around using these devices to stimulate behavioral change, said Peumans. Other projects focus on developing better ways to measure the brain directly to enable earlier diagnosis and disease progression monitoring, and developing preclinical tools that will improve understanding of disease and translational success.
imec and its partners at the Universitaire Ziekenhuizen Leuven, the Katholieke Universiteit of Leuven, and the life sciences research institute
VIB have also launched Mission Lucidity to “decode” dementia by creating technologies that enable high-resolution measurement of brain activity,1 said Peumans.
Clinicians and Patients as Partners
People with various neurologic and neuropsychiatric disorders have demonstrated a willingness to use digital tools both as partners with their clinicians in managing their health and as participants in research studies. In many cases, these studies have shown that smartphone apps and other digital tools may more accurately capture intraindividual variations in symptomatology over time, said John Torous, director of the division of digital psychiatry at Beth Israel Deaconess Medical Center in Boston. For example, in a partnership with JP Onnela and colleagues, Torous used a personal smartphone custom app to assess depressive symptoms, including suicidality, in participants with major depressive disorder. The study compared daily scores on questions from the Patient Health Questionnaire-9 (PHQ-9) with scores obtained using the traditional paper form of the PHQ-9. Torous said the scores were closely correlated, although the app scores were higher and with more reports of suicidality than the paper scores, suggesting that people may report more severe symptoms when providing data over a smartphone (Torous et al., 2015). In another study using the Beiwe digital phenotyping platform discussed in Chapter 3, Torous and colleagues showed that passive data collected on a smartphone can signal an oncoming relapse in individuals with schizophrenia (Barnett et al., 2018).
Torous noted that some psychiatrists may be resistant to these technologies, seeing them as coming between them and their patients. He suggested that new clinical models developed in partnerships among clinicians, patients, and device developers may be needed to move these technologies forward. Indeed, he cited a case where a man taking anti-psychotic medication conducted his own smartphone-based study to track the relationship between auditory hallucinations and medication dose (Torous and Roux, 2017). The system he created helped the man understand how his medication was working in a way that made sense to him, said Torous, illustrating the importance of listening to people’s lived experiences and how they are using technology to manage their health problems. Indeed, he said, there are hundreds of apps available already that
patients are using, and clinicians are being called on to provide guidance about the use of these apps. To address this topic, the American Psychiatric Association has developed a hierarchical framework to guide informed decision making around smartphone apps for use in clinical care (Torous et al., 2018).
Digital platforms can also be engineered to enable people to share their personal data with family members, clinicians, and researchers. For example, Ardy Arianpour, CEO of the consumer-driven health care technology platform Seqster,2 said his company has created two platforms—Health One for individuals and the multigenerational platform Health Trust for families—that enable people to aggregate their health data together from apps and wearables, genomic studies, electronic health records, and questionnaires, thus empowering them to take control of their health. The main challenge, said Arianpour, was interoperability. He said that Seqster has integrated cross-platform, multisourced data. They have also built a research portal as a way for participants to share anonymized data for research purposes. This has allowed them to partner with Boston University, the Boston University Ryan Center for Sports Medicine, and the National Football League on a concussion study, and with the Glenner Alzheimer’s Family Centers on a reminiscence technology project. Seqster is built on the idea that individuals have a legal right to accessing their own health data and sharing those data with whomever they choose, said Arianpour. Developing the platform required extensive consultation with lawyers and data scientists to ensure there were no violations of HIPAA or GDPR regulations.
Partnerships Across Disciplines
Munmun De Choudhury added that partnerships across disciplinary boundaries are critical for developing social media-based approaches to diagnose, treat, and prevent mental illnesses. For example, she collaborated with clinical partners to develop a human-centered machine learning approach applied to a linguistic analysis and of Twitter data to identify social media markers of schizophrenia (Birnbaum et al., 2017). This approach, she said, has the potential to reduce the duration of untreated psychosis. She and her collaborators are also studying Facebook and Google archives from a set of young adult psychosis patients to develop a model predictive of early relapse, based on language and behavior expressed through social media.
Data sharing is one of the biggest challenges that limits the ability for partners to work together effectively, said JP Onnela. Simpson agreed, adding that while there will always be mistakes made in the development of new technologies and drugs, by working together and sharing data, it may be possible to minimize those mistakes and avoid making them again.
As an example of how data sharing facilitates the correction of mistakes, Onnela cited the Hubble telescope. When it was first put into orbit in 1990, a major problem was discovered with the mirror, he said. However, a group of scientists working together and sharing their data were able to compensate for the problems by coming up with new image-processing techniques that turned very noisy data into meaningful information, essentially turning blurred images into clear pictures. Similarly, Onnela argued that the “golf cart problem” mentioned in Chapter 3, when a smartphone registered driving over terrain in a golf cart as steps, was not a failure of data but a failure of inference. If you have the underlying data, they can be reanalyzed, he said.
Onnela wondered if there might be a way to collect raw data across different devices, feed those data into openly available algorithms culled from different sources, and reanalyze the data to produce meaningful and harmonized output. The process begins with collecting the data, which requires a collection platform and a way to standardize the data to put it all on a level playing field, said Arianpour. Standardization enables both data sharing and data aggregation, he said.
One problem with data aggregation from digital devices such as wearables is that sensors and devices, especially in the consumer space, provide access only to reduced data such as step counts, which would not be informative for algorithms, said Vaibhav Narayan. Custom-designed devices provide differentiation in the marketplace, which has commercial value for developers. The challenge for researchers, he said, is to convince and/or incentivize consumer tech companies to collect and store high-resolution raw data so they can be compared with data from other devices and used for other purposes. For example, Onnela said that raw data extracted from the accelerometers and gyroscopes in both Android and iOS devices are typically packaged by those devices to provide information on the number of steps taken, but could also be used to estimate the frequency and amplitude of tremors in people with AD or PD. Peumans said that in addition to sharing the raw data, device developers would also need to share information about the sensor itself to enable cross-calibration of the
data. Sharing the context in which data were collected is also important to enable translation of data into meaningful information, said Simpson.
For medical device and app developers interested in having access to research cohorts, Kopil suggested that groups like MJFF can play a role in forming partnerships contingent on the partners’ willingness to share raw data and algorithms. Although there is a long contracting process involved in these partnership agreements, Kopil said access to patient populations and other resources may incentivize developers to share data and algorithms. Simpson added that the bigger the pool of data is, the more valuable it becomes. However, he suggested that the raw data, not the algorithms, are the currency that is important to share because comparison of raw data can often resolve differences among devices and allow recalibration.
The other value of raw data is being able to reanalyze it as the devices and algorithms improve, said Simpson. However, he said some devices are not designed to provide raw data because they do the processing on the device itself and then discard the raw data. Deborah Estrin added that in some contexts, higher level processed data may be sufficient, and as Kopil noted, storing processed data is much more cost effective. Simpson advocated retaining as much raw data as possible, recognizing that there are tradeoffs to be made between the cost of storage now and the cost of lost information in the future.
Arianpour commented that these raw data exist in various forms that may or may not be understandable to researchers depending on the type of data (e.g., genotyping data, geolocation data, etc.). On the one hand, he said there may be platforms that have automated the process of transforming these raw data into a form that is more easily understandable and thus potentially more useful to researchers and other individuals. On the other hand, Narayan said that for researchers searching for a signal, such as might be the case for an investigator working with potential digital biomarker data, getting access to the raw data is very helpful. He suggested that research-grade devices such as those developed by imec may be able to facilitate the collection, storage, and deployment of raw data. Peumans agreed, noting that imec does exactly that: developing such devices with input from clinicians and then sharing those devices with the community. He added that they are very open to further discussions and collaborations in this area.
The desire to make data sharable, while still maintaining usefulness and fidelity, aligns with the patient-centered work of disease-focused foundations such as MJFF and Accelerated Cures for MS, said Magali Haas. Moreover, she suggested that this model could be used to create
public–private consortia around the use of digital medical devices in neurosciences research and health care. Across many neurologic diseases, there are many common constructs and domains, she said, as well as a range of approaches to measure them systematically using next-generation wearable devices. Consortia could facilitate putting all these data in a common pot for multiple researchers to analyze, confirm, and validate. Another benefit of such consortia that take a person-centered approach, said Kopil, is the potential emergence of collective shared responsibility to understand health holistically rather than individually tackling each specific interest group.
One project already operating in this space is Verily’s Project Baseline. This is an effort to map human health, initially by studying 10,000 adults, many of whom have or are at risk for various medical conditions, said William Marks. Another potential opportunity on the table was mentioned by William Potter, senior advisor at the National Institute of Mental Health. The All of Us research program has already begun enrolling what is anticipated to include 1 million participants in this longitudinal, data-driven observational study, although final decisions about how to collect, store, analyze, and interpret the data are still being formulated, according to Potter (see Box 7-1). He said NIH is bringing together not only its various institutes, but also advocacy groups and other stakeholders to ensure alignment of goals and operational aspects of the project.
William Marks concluded his remarks by saying that the convergence of these technologies and opportunities offers the prospect of developing better measures of brain disease. Initially complementary, digital measures may at some point replace more traditional measures, he said. To improve beyond what he called “the crude methodology” currently available, he asserted that the thoughtful collection, organization, and activation of data are critical. Beyond that, he said, there is an immense opportunity to expand research, accelerate therapy development, and ultimately improve outcomes for patients.