Turning Patient-Generated Data Into Meaningful Clinical Information

; Paul S. Teirstein, MD; Cheryl Pegus, MD, MPH; Joseph Wang, DSc


May 25, 2016

This feature requires the newest version of Flash. .

Editor's Note: In this segment from Medscape's Medicine 3.0: Patient-Generated Data event, held in March in San Diego, panelists discuss how massive volumes of patient-generated data can be distilled into meaningful information, both for patients and clinicians.

This abridged video and transcript have been edited for clarity.

Eric J. Topol, MD: LabCorp in recent months has set it up whereby any person can go order their own lab tests. That was a big deal, of course, because up until then you had to go to the doctor for the test you knew [you needed]. For example, if you knew you had to have a thyroid test or a lipid panel, you had to go to the doctor for an appointment. Then maybe you had to go back to the doctor for the interpretation.

Now you can order your test yourself and you can even get the LabCorp interpretation, or as you put it, Paul, if you get the on your smart phone, you can get it read with an algorithm. So you have this era of machine assistance or algorithms or software that helps interpret for people bypassing doctors. That, of course, is both a threat to physicians and a potential mechanism to avoid some of the unnecessary aspects of care for those who feel more autonomous. So these are some of the tensions that I think are there.

Paul S. Teirstein, MD: The other one would be the whole nurse practitioner–physician interface. There is coming to be a lot of conflict there now because the nurses want to take on a more independent role. And there is actually a bill before California right now to enhance their ability to take care of patients without having as much physician oversight. That fits very well with all of this, so it's probably going to end up going that way and the physicians will adapt.

Cheryl Pegus, MD, MPH: It's been going that way for a while. Many states have already passed that. You are right—California is a little later.

Dr Topol: As a non-physician, you are generating, through these sensors, all of this data. Does that bother you that you would have "TMI" and not be able to deal with this data? Or are you going to develop with your colleagues all kinds of algorithms so that people can interpret their data without the need for a professional?

Joseph Wang, DSc: We are concerned with the big picture: the sensor, the electronics, data fusion, and data security. So we are looking at the whole picture from the actual sensing to the data security and the communication. But, again, if you look at the glucose field, we do the test at home. This is amazing—[we can] take a drop of blood with 1 µL and fish out only the glucose with a sensor that costs 5 cents, and do it in 5 seconds. You can do the whole analysis because this is market driven—it is a $10 billion market, so it is driven by this. But now they can do it at home and do it in 5 seconds, fishing selectively and very fast for glucose. It's amazing.

Dr Topol: One thing you just touched on that is pretty interesting is that most engineers, they just want to get the data, to capture the data. They just want to get some ingenious sensor that accurately captures the data. But you are thinking much more broadly as to how you are going to process the data, how you are going to keep it secure, and that sort of thing. That is really important.


Dr Pegus: One of the things that you really want is to have a system that allows data to come in for a patient that they are aware is secure, and there is actually some information given back to the patient. So many of you remember the Institute of Medicine report on health literacy[1] in 2004 which, for me, remains one of those things that I always have to remember. For half of the people—90 million Americans—it was found that they had really poor health literacy, and it had nothing to do with economics, it had nothing to do with education. I tell this story all the time. I have a husband who is an MBA, and every time he goes to the doctor, he calls me from the doctor's office to say, "Hey, why don't you talk to my wife." It's an issue, so when we say that people are getting a lot of information, even for well-educated people, health information is complex.

What we really need is that when patients have patient-generated information—and by the way, this is my business idea and you guys can't steal it—it goes into a couple of companies or a company that is taking in the information and then giving back clear, health-literate, appropriate information to the patient. That company also has a two-way integration with the electronic medical systems that allows them to know it's secure data and that data are being put in, and that the patient has said, "Yes, I am okay with this being put in." And with the things that they are not okay with being sent in, they at least have it stored in a place where, if they had a question, they could ask someone.

Dr Teirstein: How is a company going to give feedback to a patient on an analysis of their data when it's often controversial?

Dr Pegus: It depends. Companies do it now, right? We heard that with LabCorp and Quest [Diagnostics], patients can order a test and they actually get a report back on what it means. There is a disclaimer at the bottom that says, "Do not do anything until you follow up with your doctor," but it is now actually pretty okay to do that. There are a lot of other companies that do it and, again, they are FDA cleared. So it is already occurring.

What is not occurring is someone giving some proactive guidance on how to help manage your health. There are newer companies—you guys may know of them: Rejuvenon and others—that are doing that right now. They are taking in patient health data, taking in patient self-reported data, and they are giving [patients] guidance.


Dr Topol: Do you think there is going to be good software help, such as algorithms, for the interaction with the individual patient to facilitate this, or do you still think it's going to be a mess?

Dr Pegus: Software algorithms will help. But I grew up on the Framingham Risk Score. And then I grew up on measures. And now I have an entirely new cardiac risk algorithm that has been developed. I think algorithms change—the way we look at how data helps us make good evidence-based decisions changes. And the algorithms by themselves, based on what I just said and the way we are really doing precision medicine, allows for you to have some individual interpretation as necessary.

Dr Wang: We are developing a multiple-parameter patch; we measure 24 analytes simultaneously and we can have 10 cardiac markers. With good software we can really generate different risk levels and warning levels and so on. But you need to fuse all of the data to combine all of these cardiac markers, for example, to get the picture.

Dr Topol: So if you are having chest pain and you get the readout of your troponin on your phone and it is negative, you don't go to the hospital. Is that right?

Dr Wang: It's too risky, yes.

Dr Teirstein: It would have to have it simplified, so it's a green light/red light for the patient.

Dr Wang: Exactly. A warning at least, from a combination of multiple parameters.

Dr Topol: One of the things that human beings are not too good at is processing a lot of data. And computing, actually, that's where we turn to, of course. We are not talking about just 20 more analytes; we are talking about sleep, activity, nutrition—all of these different labs that are changing dynamically over the course of time. You have got parts of the exam that you would be putting in, whether it's the official exam, if you will, from a doctor's office, or it's the kind where you are doing it with smart phone attachments, of a child or of yourself. You've got the sensor data. And now with this multilayered data, you have computing for getting you an avatar, a text or a voice, to talk to you—a virtual medical assistant. Do you see this—where you will be relying on your virtual medical assistant to help you navigate your health on a moment-to-moment basis, processing all of this data in real time? What do you think?

Dr Teirstein: I just can't imagine that working. I mean, it's going to be like talking to Siri, right? You'll get some funny answers.

Dr Topol: There was a Siri article last week and it said if you had an assault, if you were depressed, it didn't respond well. Siri wasn't really made for those things. But we are not talking about Siri; we are talking about a new era in which you have a machine learning about you.

Do you think this is possible or do you think this is science fiction? Cheryl, what do you think?

Dr Pegus: It's definitely possible. It's possible, and we already use some of this. An EKG gives you a simple readout; for more complex cases, someone over-reads to be able to do that. So depending on the type of data, I think we will risk-stratify data for "well populations," whom we have talked about—they are doing their BMIs, they may get their hip/waist measure, they may get some information on nutrition, prediabetes, . I think avatars can help with that. When you get to someone who may have had and has diabetes and maybe some CKD as well, there may be some simple day-to-day about how to manage your life, but the way we are going to be looking at that data, we will say, "When the trend analysis gets to blank point in the algorithm, someone else, a human, will be involved." And I think those things will happen. I think we can do it now.


3090D553-9492-4563-8681-AD288FA52ACE -
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: