BOSTON—The cost of mapping an individual human genome is dropping logarithmically, from $100 million just 12 years ago to $5,000 today. Silicon Valley entrepreneurs hope to drive the price below $1,000, the cost of an MRI test, and within a decade it very well may be possible to conduct a whole genome sequence for every newborn at birth.
Cynics point out that genome sequencing is the only healthcare cost actually going down, at a time when U.S. healthcare spending is projected to approach $3 trillion in 2013. But the research and analytics that results from such data is poised to change the way healthcare providers, insurance companies and pharmaceutical companies do business—if only everyone who has the information is willing and able to share it.
Crowdsourcing One Possibility for Genomic Research
A whole genome sequence can be cost effective, says Sandy Aronson, executive director for IT at the Partners HealthCare Center for Personalized Genetic Medicine, since it can be run once and then used across multiple "episodes of care." Aronson and others spoke at the recent Medical Informatics World conference.
Today, clinicians can run 2,900 tests against a patient's genome, Aronson notes. The challenge is twofold, he adds: Knowing how to interpret the results of a test and, with a genome containing as many as 5 million variants, making sure nothing is missed.
To address this, organizations must be prepared to "pre-position" IT infrastructure to take advantage of ever-changing genomics research and incorporate it into mainstream clinical care, Aronson says. On top of that, the industry needs to change regulatory and reimbursement frameworks, provide training for healthcare providers, payers and patients, and lean on society's resources.
The latter could involve "highly structured crowdsourcing," he says, which places tests—of diseases, variants, pharmacological effects and others—in the context of patient phenotypes and family history. This augments a patient's record and can add further value with, for example, alerts that are triggered by certain test results.
Such information is also of interest to insurance companies, says Dr. Lonny Reisman, senior vice president and chief medical officer for Aetna, as it offers the opportunity use "phenotypic manifestations" for predictive analysis of patient populations. This, in turn, can be applied to "value-based insurance design," which Aetna has used to waive co-pays for certain procedures or medications that offer proven long-term benefits to patients with heart conditions, Reisman say.
For Patients to See Value, Data Must Flow Both Ways
Even with waived co-pays, the patient compliance rate remains less than 50 percent, Reisman says. This points to a larger concern in the healthcare industry: Improving patient engagement. For Aronson, Reisman and others speaking at Medical Informatics World, better information sharing will lead to better patient engagement.
Dr. Mark Davies, executive medical director of the Health & Social Care Information Centre within Britain's National Health Service, says physicians should have an "adult" relationship with patients—one that makes them feel like they're part of an equal partnership. This, in turn, must be coupled with a "bidirectional flow of insight" among patients, providers and patients, Reisman says. The benefit is bidirectional, too. Patients have better access to more robust personal health information, while patient-reported outcome measures can be used for quality, accountability and transparency improvement initiatives, Davies says.
For this to succeed, though, there must be a clear value for patients. Right now, unfortunately, that isn't the case, says Dr. John Halamka, CIO at Boston's Beth Israel Deaconess Medical Center. While the U.S. government's meaningful use incentive program does require healthcare providers to offer technology that lets patients download, request and transmit data, there is little "value add" for personal health record or disease management applications, Halamka notes.
In most cases, patients visit these apps once but don't come back. Poor usability and functionality are often to blame, Halamka says. "Go build apps that provide value."
Predictive Modeling Is Where the Value Is
For Julie Meek, clinical associate professor in the Indiana University School of Nursing, that value is in predictive modeling. Bringing together demographics, billing and pharmacy claims, lab test results, patient-supplied data and genomic research—and then incorporating it all into the clinical workflow via an EHR system—gives patients a much better sense of the health indicators than the height, weight, blood test and urine test of the annual physical ever could.
The key is making sure that no data sets are missed. Meek's predictive modeling—which is more than just an exercise in data mining, she says, because it incorporates logistic regression and model validation—considers 39 separate variables. Many stick to age and gender data, as both are readily available, but, as Meek puts it, "Cheap data is no substitute for legitimate inquiry."
She advocates such a comprehensive approach to population health management because the status quo isn't cutting it. Twenty percent of Medicare patients who are hospitalized are subsequently readmitted within 30 days—and many, for whatever reason, don't follow up with a physician in between hospital visits. This is costly and inefficient.
Determining who will come back isn't easy— John D'Amore, founder of clinical analytics software vendor Clinfometrics, says this analysis must take into account 60 variables—but it can be done. Take a group of 15 patients being discharged from the hospital and, D'Amore says, you can identify the five at the highest risk of being readmitted. That's important because, in that group of 15, 74 percent of the readmissions come are one of those five patients, he says.
Genomics Research Allows for 'Precision Medicine'—If Data's Available
The data that's gleaned from genomics research could play an increasing role in this type of modeling, whether it's reducing readmission rates or researching cancer in the name of "precision medicine" that's tailored to individual patients' needs.
It's could and not will because, while Davies says "we have some fantastic technology out there" to first conduct and then share genomic research, a mix of professional, personal and cultural factors combine to make data dissemination difficult. Patients fear that data will be sold to pharmaceutical or life sciences firms, while researchers and providers persist in creating data silos.
What's needed, Aronson says, is a more detailed regulatory framework that can address data privacy as well as genomic data use case standards. From a care-coordination and knowledge-sharing standpoint, primary care physicians, specialists and genetic researchers have to be connected. (However, Halamka points outs, health information exchange is no easy task, as each U.S. state and territory has different data sharing standards; what's legal in Massachusetts may be illegal in neighboring New Hampshire.)
There's also an educational component at the caregiver level, Aronson adds. Families need to understand the importance of sharing genomic information.
Or, as Davies says, the industry needs to know that "failure to share data kills people."