Gary Wolf

August 11, 2020

This post is an attempt to capture some of what I’ve learned preparing for a webinar organized by Susannah Fox called “Prioritizing Patient Engagement and Inclusion of Patient-Generated Covid-19 Data.” This is part of a webinar series called Advancing clinical registries to support pandemic treatment and response, which is organized by a nonprofit called Council of Medical Specialty Societies. As is obvious from this pileup of biomedical vocabulary, this webinar is meant for academic and clinical research professionals, which means my participation is necessarily that of an outsider. Other participants in the panel include Gina Assaf and Hannah Davis of Body Politic, and Emily Sirotich of the COVID-19 Global Rheumatology Alliance.

A patient registry is a database that includes diagnosis and treatment details. As Susannah wrote in her introductory post, quoting the AHRQ definition of a registry:

An organized system that uses observational study methods to collect uniform data (clinical and other) to evaluate specified outcomes for a population defined by a particular disease, condition, or exposure, and that serves one or more predetermined scientific, clinical, or policy purposes.

A registry’s typical purpose is to evaluate the effectiveness of medical interventions. Hospital systems, for instance, can look at their patients with heart disease, and make comparisons among various treatment options. Used in this way, registries are crucial for improving the quality of care. However, patient registries serve many other functions besides comparative evaluation of treatment options; for instance, they can help clinical researchers recruit patients for experimental trials, they can be used to asses quality of care in general, to estimate the value of health allied support services, and to guide regional and national policy changes. With these broad uses, and with the growing scale and ubiquity of computing and data science, patient registries have been tempted into expanding their data collection efforts beyond obvious treatment parameters into more general health surveillance.

However, in this expansion into health surveillance, patient registries meet other players equally interested in knowing everything about everybody, including, public health research; big tech product development; and government ministries. Now, faced with the challenge of the COVID-19 pandemic, stewards of patient registries, including the leadership of medical specialty societies that are attending this webinar, are attempting to understand where their data gathering and data access responsibilities begin and end. Since people dealing with COVID-19 may show up in the health care system at various stages of their disease and with symptoms that put them into contact with many different specialties, it’s not even obvious how to prevent them from being double and triple counted in investigations of basic questions like: How many people are getting sick? What remedies have they tried? Did they recover? When did their infection start and end?

When these questions are challenging to answer from health system data, observations from other sources may be able to fill the gap. These observations include data from wearable devices, actively tracked data like entries in a notebook or self-tracking app or spreadsheet, and contextual data like location and proximity, and self-managed records of interventions, such as medicines taken. The potential benefits of bridging these two sources of observations – the clinical and the personal – have lead to countless proposals and projects, some of which I’ve advised, and many more of which I’ve watched from up close or at a distance and managed to learn from. I’m going to outline some of the more important non-obvious challenges in bridging these two different sources of health-relevant observations: clinical observations and self-collected observations. The most important of these challenges is that self-observation and clinical observation are made by people whose interests aren’t aligned. Since their interests aren’t aligned, their priorities are different. Since limited resources mean only the most important priorities are addressed, the biomedical and person-centered approaches inevitably clash. At the end of this post I’ll introduce the concept of “self-registration” as a potential bridge approach firmly anchored on both sides of this divide.

Modern administrative workflows leverage powerful technologies that have become ubiquitous in our everyday life. A speaker in an earlier webinar in the series on Advancing clinical registries, Michael Howell, who works as a Principle Scientist at Google, gave his health care listeners some practical advice: Take tools that work in your regular life and use them to create new tools for understanding your health records. Since Howell is at Google, the everyday stuff he had in mind included cloud storage and computing, search, and new machine learning capacities for working with unstructured data like images and free text. These kinds of everyday tools offer the prospect of dramatic process improvements, allowing more clinical data, of more different kinds, to be integrated into fundamentally familiar observational research approaches.

This advice is reasonable when understood as coming from an enterprise platform provider to enterprise customers in health care. It’s a top down approach that takes data from individuals and aggregates it into a resource for professional knowledge making. But Google is also a company that builds tools for individuals, and Howell invoked the possibility of health researchers being able to understand health and disease by linking all the disparate information about cases to each specific patient, an approach he called “patient centered registries.” This would give a much more complete picture of COVID-19 and other public health threats.

In a patient-centered registry, it isn’t just the everyday tools of pubic administration, marketing, and management that can contribute data, but also the everyday tools used by individuals to understand their own health and fitness. In 2013 Pew Research reported that 33% of U.S. adults track health indicators or symptoms, like blood pressure, blood sugar, headaches, or sleep patterns. Last year 30 million Apple Watches were sold, and 19 million Fitbit devices. Our everyday tools for thinking about our health include wearable tech, symptom tracking apps, online advice and support from peers, and scientific and popular knowledge available on the Internet. So if we want to know about typical and atypical symptoms of COVID-19, why wait until people show up at the doctors’ office or emergency room and then ask them to tell us: When did you first feel sick? It’s reasonable to want to build on top of our everyday tools, and track the development of the disease as it occurs. I want to underline what tends to be forgotten in our product-obsessed culture: these tools are not simply measurement instruments and wearables; they include the social and cognitive tools individuals are using to understand and manage their own health.

We obviously do not have patient centered registries today. Why not?

The reason is that integration of detailed personal data into conventional research workflows is technically complex, legally fraught, culturally suspect in many communities, scientifically difficult due to measurement, analytical, and a provenance issues, expensive, and controlled by a fragmented healthcare system. In other words, it’s a mess out there. At every discussion of this problem, the people with the most confidence in a technological solution are the least experienced in the field; a sure sign of hard problems waiting just beneath the surface. And even when some of the process barriers are overcome, resulting in access to continuous time series data from large numbers of individuals, operating this kind of system in the wild requires continuous engagement, rather than one time consent. Devices fail and go out of date, surveys get ignored, apps get deleted, and research staff with crucial knowledge drops away when grant funding runs out. You have to have watched projects come and go over several cycles to fully appreciate the challenge.

A clash of approaches stands between biomedicine and the potential for discovery in today’s everyday tools. This clash is not just a clash of design approaches and it can’t be solved by turning up the dial on “engagement.” Personal science is focused on observations. Registries remain oriented toward outcomes, which mostly means comparative study of interventions. But what happens when diagnoses are contested or ambiguous, or interventions are not yet known? If our data is collected for purposes remote from the clinical context, there is no reason for us to connect to a registry as a “patient.”

Connecting biomedicine to the everyday tools of health discovery requires a revised concept of the registry that begins with the individuals ability to securely collect, store and access their own data, and then builds around this data store a system for granting and revoking access to academic and clinical researchers. Let’s call this system not a “patient registry” but a “self-registry.” Since they focus on individual benefit, self-registries must accept any data type, including photos, videos, subjective self-assessments and free text. A self-registry, by definition, allows it’s individual participants to review, join, and edit their own records. Self-registries allow individuals to associate themselves with the patient communities where they believe they belong, and to grant access to their medical records on a per-project, rather than a one-time, basis.

What if individual people with an interest in their own health don’t have the time, skill, support, or power to work with the resources of a self-registry? I care about this question a lot. It’s a question at the threshold of a new approach to public health, and I’ll intentionally not answer it here.