An estimated seven in ten U.S. adults say they track at least one health indicator, and 21 percent of this group use some form of technology to track their health data, according to data from the Pew Research Center. But these figures are likely to skyrocket thanks to health platforms such as Google Fit, Apple's HealthKit and AT&T ForHealth, which use sensors built into smartphones and wireless fitness devices to record physical activity.
This data deluge is a goldmine for biomedical research and drug development, particularly with the introduction of Apple's ResearchKit. The software, which powers the Stanford-developed MyHeart Counts app, allows users to better understand their health data while providing researchers the opportunity to access it for future studies.
In a recent Huffington Post article, Ida Sim, MD, PhD, professor of medicine at University of California, San Francisco, noted that such technologies hold the potential to encourage the general public to participate in medical studies and make the research community more collaborative and open. "There’s a new movement in academic research called participatory research, where patients are part of the groups that should be asking: ‘What questions are interesting? What should we test?’” Sim said in the piece. "The public could start seeing research as something that isn't imposed on [them], but as an activity that we all do together so that we can learn together."
This May, Sim, who co-directs of Biomedical Informatics at UCSF's Clinical and Translational Sciences Institute, will speak at Stanford's Big Data in Biomedicine Conference on how health information collected on mobile devices holds the potential to inform clinical decisions and transform health care. As a co-founder of non-profit Open mHealth, she and colleagues are leading the charge to build open source software that facilitates sharing and integration of digital health data.
Below she outlines how leveraging mobile health data can improve how physicians diagnose, treat and prevent disease and the challenges in facilitating the sharing and integration of this vast treasure trove of data.
What are the large-scale opportunities to harness the rapidly growing reservoir of information to improve biomedical research and human health?
We can use this data to do a variety of things like combining genomic information and behavior data from wearables to discover new insights into health and disease.
We can also move from what works on average to more tailored programs focused on the idea of what works for me. For example, if we employ A/B-like testing with digital health, genomics, and other data combined, we can understand which interventions work for an individual and under what contexts, allowing for more tailored healthcare.
Finally, we can learn about a person beyond their clinical visit - which is only a small slice of their “health pie." By getting multiple health snapshots, doctors will be able to provide patients with better medical support and preventative strategies that support overall physical and mental well-being.
What are the major challenges in unlocking the potential of digital health data?
When we write a sentence, we construct the sentence with grammar. We use vocabulary to fill in the blanks to give meaning to the sentence. Meaning is lost when either the grammar or the vocabulary is ambiguous or not shared between parties. In a similar way, making sense of data from various digital health devices is challenging when the devices don’t represent data the same way.
Currently, wearable devices and other healthcare tools describe the data they collect using their own languages that are not shared or integrated with other devices. For example, a Wi-Fi enabled weight scale might represent data as “weight: 88” but we have no clue if that means 88 kg, femptograms, lbs, or stones. A calorie counter might represent calories as “calories: 400” but we have no clue if this was calories expended or calories consumed. For clinicians, these kinds of ambiguities are show stoppers that lock up the potential of digital health data.
In addition, data from the devices themselves are stored in silos, meaning that it is not easy for patients or clinicians to combine and view multiple data streams together. Blood pressure from one device isn’t syncing with weight data from another, which can lead to an incomplete picture of a patient’s health over time.
If we strive for greater interoperability with a common language and structure for both understanding and integrating digital health data, we can help to bring clinical and patient needs together for better health-care outcomes.
Can you share with an example of how the Open mHealth platform has been used to help patients with chronic illness?
We have been working with Michael McConnell, MD, at Stanford’s Preventive Cardiology Clinic to use digital health data to aid in the care of patients with cardiovascular conditions. Our product, called Linq, allows a clinician to prescribe a data plan that is specific to that patient. For example, they might prescribe a patient to measure their blood pressure once a day before breakfast for a week, and set the level above which the doctor and the patient will be notified. The patient then syncs their favorite app or device along with that data prescription. The device-agnostic data is then piped through the Open mHealth infrastructure from data storage to data processing to data visualization in a way that allows the patient and clinician to understand and have a joint conversation around that data, which should lead to more data-driven insight and action.
The pilot is ongoing. Patients are being recruited to participate.
How are you and colleagues at Open mHealth working to better facilitate sharing and integration of digital health data?
We are achieving this by working in several distinct areas.
We started by tackling the problem that different devices have different application program interfaces (APIs) that don’t work with each other. Our DSU repository highlights various ways we’re accomplishing this. One such API tool allows an individual to authorize data to be retrieved by other applications, breaking down barriers between different devices. Another tool, API shims, allow us to take in different data points to be understood in a common language. Finally, our API integration server allows anyone to access third party tools (e.g. Fitbit) for free.
At the same time, we’ve been working to create clinically valid schemas that represent important clinical measures with references to existing clinical standards. This helps eliminate confusion as to what any given data point might mean. As our schema library shows, this includes a number of common measures, including body weight, heart rate, body mass index, or calories burned/consumed and we are expanding to more clinical measures such as medication adherence.
We’re also working on modules for various types of analytics and visualization, some of this in partnership with the Mobile Data to Knowledge (MD2K) NIH Center of Excellence. Beyond creating a common language for understanding what data points mean, we must also represent them in ways that allow clinicians to make health-care decisions with the greatest ease and effectiveness.
Beyond creating code and other programmatic language, we broker clinical needs with how toolmakers represent data through Clinical Measure working groups (CLIMEs), focusing on challenging representational areas (like medication adherence) that require input from clinicians and technologists. Finally, we build demonstration products that instantiate the Open mHealth infrastructure.
What do you think about patient privacy when it comes to digital health data? What is your idealized version of patient privacy?
Currently, it is unclear exactly who owns the digital health data that is generated. Often times, toolmakers that produce wearable devices get to hold on to the data after a user stops wearing the device, or this data cannot be transferred over to other systems at all. When individuals leave a particular hospital system, it can be difficult to revoke access to that data or transfer it to any other system.
At Open mHealth, we believe that the focus should be on individual users having ownership of their own data. What does this look like? Patients would, in the future, be able remove data from any digital health system or toolmaker. They would be able to revoke the permissions of these previous systems to access that data, thereby allowing them to easily transfer that data to other systems. Patients should also be able to have much more granular control over their data, deciding which data can be shared with which systems under which circumstances. We are still a ways off from this model of data ownership, but as more and more devices and digital health systems are introduced, it seems likely that our existing, quite limited framework of patient privacy will change.
Previously: Big data used to help identify patients at risk of deadly high-cholesterol disorder, Registration for Big Data in Biomedicine conference now open, Big data approach identifies new stent drug that could help prevent heart attacks and Videos of Big Data in Biomedicine keynotes and panel discussions now available online
Photo by Norbert von der Groeben