CHICAGO (GenomeWeb) – As the much-hyped Precision Medicine Initiative works toward its initial goal of collecting collect medical, environmental, lifestyle, and genetic data on 1 million people with millions of taxpayer dollars, it may be failing to address two fundamental questions: what to do with all that data and how to translate knowledge into clinical practice.
"The PMI's emphasis on genetic and molecular discovery overlooks important aspects of patient-centered care delivery," according to a recent commentary in the New England Journal of Medicine.
"The PMI's scientific advances may add further complexity to delivering high-quality, cost-effective care in keeping with patients' values. If we don’t grapple with that complexity, patients may wait years to reap the diagnostic and therapeutic benefits of the PMI," wrote Ravi Parikh, a resident physician in internal medicine at Brigham and Women's Hospital in Boston, and University of Pennsylvania physicians and health economists Sanford Schwartz and Amol Navathe.
"But though a better understanding of cancer biology could yield insights into innovative diagnostics and treatment, an exclusive focus on laboratory-based genotype–phenotype–therapy links ignores the real-world challenges of delivering expensive technology, assessing preferences and trade-offs, and undertaking shared decision making," they wrote.
Enter the Precision Delivery Initiative, their proposal to augment the Precision Medicine Initiative with a dedicated effort on marrying predictive analytics, new data sources, and the power of engaged patients with scientific advances from the PMI.
"There is more to personalized medicine than genes and small molecules. Enhanced digitization, transparency, and application of clinical data may put precision medicine into the hands of today’s patients — and make the grand conception of the PMI a reality," the researchers explained.
"A PDI research agenda could spur collaborations among disciplines such as epidemiology, biostatistics, computer science, and systems and human factors science to develop and evaluate novel analytics and implement interventions," they continued. "PDI research and investment could improve point-of-care and longitudinal clinical decision support, realizing benefits from decades of electronic data entry that has often frustrated clinicians. Potential applications include enhanced risk prediction and early-warning systems for acute events, as well as targeted discharge planning for patients at risk for readmission."
Corresponding author Parikh spoke with GenomeWeb this month about his proposal.
Below is a transcript of the conversation, which has been edited for clarity.
What made you believe there was a piece missing from the Precision Medicine Initiative?
There was a lot of sense even before the Precision Medicine Initiative even came about that doctors have trouble personalizing care for patients. We tend to learn things in broad strokes. We tend to learn things by evidence bases from large populations for treatments. When it comes to actually seeing individual patients, they are very different than large population aggregates. Sometimes, individual patients need different treatments than what were conventionally taught.
This notion of personalizing medicine and the angst of doctors to try to individualize the right treatment to the right person has been there long before precision medicine ever came about.
Evidence-based medicine has not been easy, either. It was derided as "cookbook medicine." How would a Precision Delivery Initiative address that skepticism?
That's a fundamental tension: How do you differentiate doctors' practice between what's evidence-based and what doctors feel is really right for that individual patient? I don't know any doctor who intentionally disregards evidence when thinking about treatments for an individual patient.
I would venture that most doctors are trying to do right by patients even when they pursue non-evidence-based or off-label therapies because something might be right for an individual patient. Just because a treatment guideline suggests ciprofloxacin for a UTI, for example, maybe ciprofloxacin isn't right for that particular patient because of a side-effect profile or prior reaction. That's an example of where personalizing treatments can be really useful.
What the Precision Medicine Initiative has tried to do is develop and implement personalized diagnostics and therapeutics for individuals, which I think is certainly essential. We have the capacity to measure people's genetic risk predispositions or molecular/biological predispositions for a disease. It makes sense to try to develop targeted therapies based on that.
What prompted us to write the article was that there's equally as much innovation on the informatics and information technology side that can be used to personalize treatment that isn't necessarily covered under traditional conceptions of precision medicine.
You wrote that advances in precision medicine would make delivering care more complex, which makes sense because you're dealing with so much more data. Could you discuss some specific areas where this might apply?
I'll give the example of oncology because I'm going into oncology training.
One of the goals in precision oncology is trying to match patients' genomic and biologic profiles for targeted cancer therapeutics. That's useful, but, to the point of complexity, oncologists today who have been in practice 20 years learned during their fellowships that the available amount of treatment for any given malignancy was far different than what they're treating patients with now and what they will have available in their armamentarium in the future. That's a good thing. We want more therapeutic options for our patients. But the complexity inherent in learning a side-effect profile, in learning an indication, in learning a delivery method for another new drug is just so high.
The point we try to make is that, one, precision medicine might not necessarily make things easier. It might make things more complex for the individual physician. And, two, we can leverage information technology to really try to integrate some of this data that's been there and try to match treatments and diagnostics to the right patient, incorporating patient wishes.
Patient empowerment is a big change, too. Doctors need to learn how to listen, but patients and doctors also need help deciphering this new type of data. Not every patient who's newly diagnosed with cancer or some other serious, life-threatening disease knows what a genotype or a phenotype is. Do you agree?
Absolutely. One of the examples of our proposed Precision Delivery Initiative — ways to leverage novel information technology and data sources — is encouraging patient ownership of data. I think, among other things, that might give some more agency for patients to get a sense of what's actually in their data, what's actually in their medical file, what's actually in their biologic and genomic electronic profiles. Ownership of data, while certainly posing a lot of challenges, could indeed encourage a lot more patient agency and autonomy in understanding their own health data. That's an important initiative here.
The other thing, we're not talking about algorithms and informatics replacing doctor-patient interaction. Certainly, any algorithm that we even think about coming out of such an initiative would be interpreted in context of the patient-doctor relationship. I think that extra piece of information that the algorithm and our information sources can offer can only be helpful in promoting more informed patient-doctor discussions, because right now, so many decisions are made with a lack of knowledge in the gray zones of medicine. That can be challenging.
HIPAA has been used as an excuse to deny people access to their own data, whereas the regulation says the exact opposite, that patients have a right to their own health information. Where you are, at Partners HealthCare, many of the primary care clinics have joined the OpenNotes movement, where clinicians let patients see unedited progress notes. Where do you foresee this fitting with a Precision Data Initiative?
[Misunderstanding about HIPAA] doesn't negate the fact that proper measures for security and privacy need to be out there, but with the advent of personalized health records and the means to put health data in the hands of patients, I think that there's certainly an appetite for this out there. While HIPAA and information security are important, we shouldn't let that be the barrier to getting this essential information in patients' hands.
It does have to be in a way that's understandable. Just putting a series of radiology images and unintelligible doctors' information in patients' hands won't actually be of much use for helping patients understand their data.
OpenNotes is really an amazing initiative. The next step is to take an analogous version of OpenNotes and put this information in the hands of patients and allow patients to use some of that shared data in research, getting patients to engage in their own investigation. Imagine how many better-powered trials or more comprehensive health dashboards we could get if we just put patient data in the hands of patients and allowed them to do what they wanted with it, as opposed to "vaulting" patient data into the confines of obscure academic databases.
That's really the next step here. How can we make patient ownership of data more than just an information-sharing tool and engage them in the data-generating process?
There is so much unstructured data sitting in electronic medical records, but not a whole lot of use of natural language processing. Do you believe it's essential that health systems find a way to extract this data?
There is so much data inherent in how doctors write their discharge summaries and their history-and-physicals. There is so much that's locked up in there, and yet, almost all of our observational studies in health systems are based on structured databases that we've collected. So much of our health data exists in text form — doctors' notes, imaging, echocardiography reports, notes to patients — all sorts of things.
In the article, you cited the Department of Veterans Affairs' Homeless Program Hotspotter Initiative that looks at innovative data sources and population data to find homeless veterans at risk for mental illness and high healthcare utilization rates. That's one example of NLP?
Not only are academic researchers realizing this, entrepreneurs are beginning to realize this as well. There are a lot of different mental health apps out there using text-based natural language processing analytics to identify people who are at risk of mental health acute conditions. That's the kind of thing that's really interesting: mining things like Twitter and Facebook.
Your vision for the Precision Data Initiative would be to create massive databases that can be shared, like what the VA does; one of your co-authors, Amol Navathe, works at the Philadelphia VA. These deidentified databases include information about social services and data from wearable devices, but are you also talking about genetic registries, similar to disease registries?
There are examples of genetic registries that have been created. We cite the Million Veterans Program. The National Cancer Institute has its own genetic registry. That's freely available data that genomics researchers can tap into. Just imagine an informatics boom for delivery system-focused researchers. What if there was something similar that they could tap into? That's what we're trying to target here.
As we curate all this electronic data, from EMRs, from wearables, from social media sources — whatever it might be — to have that data exist solely within limited-size institutions would be somewhat of a waste I think. That data, arguably even more than the genetic data, is meant to be shared broadly to generate more insights because it can improve delivery systems today.
You want to democratize research so community hospitals can participate as much as academic medical centers?
Yeah. There's obviously a tremendous need for innovation at the community hospital and community practice level as well. This data can't exist only within institutions. It needs to be democratized, with adequate security, of course.
Was this paper intended to inform policy? How much money might a data initiative take in comparison to the Precision Medicine Initiative?
It depends on what aspects of the initiative people choose to implement. We deliberately strayed away from advocating specific funding sources because research funding is so unpredictable and up in the air. Honestly, that's one of the things that I worry about. Even with the Precision Medicine Initiative, we see research subject to the vagaries of congressional funding decisions and public funding decisions. I worry that an initiative like the Precision Medicine Initiative, which has its output so far down the line, would be subject to some of these vagaries.
One advantage of the Precision Delivery Initiative is that, theoretically, its rewards could be realized much quicker than the Precision Medicine Initiative. The insights that are generated based on existing electronic medical data could be theoretically generated a lot faster [now that EMRs are ubiquitous at US hospitals]. It's time to leverage some of that data.
The point about where the funding is going to come from and how much money it is going to take is going to take is a good one that we deliberately left for a later article.