Discovery Stories

Health Leaders Magazine March 2015

Cloud Adoption Gains Traction Health Leaders Magazine March 2015

Web-based services are now finding hospital-based and even industrywide applications.

The following is an extract from the March 2015 Health Leaders Magazine. In this article, Shawn N. Murphy, MD, PhD, the director of Partners Research Information Services & Computing is featured. The full article can be viewed:  http://www.healthleadersmagazine-digital.com/healthleadersmagazine/march_2015#pg50


Cloud computing can also extend what existing hospital data centers can do to improve healthcare. At Boston- based Partners HealthCare, a new initia- tive, the Partners Data Lake, leverages cloud resources provided by storage provider EMC to permit an agile data and analytics platform.

The purpose of a data lake, as opposed to a data warehouse, is to cull a subset of a population for research- ers, says Shawn N. Murphy, MD, PhD, corporate director of research information systems and computing at Boston-based Partners, which employs 60,000.

“Let’s say they want to find patients who are diabetics who are on insulin but who have a very high hemoglobin A1C, which means that the diabetics are out of control,” he says. “Well, you might say, ‘I’ll just give them all the

diabetics, and let them sift through the medical records and find the ones that they want’; but considering that a good 10% to 20% of our population might be classified as having diabetes, that means they’re giving researchers access to a large number of patients who really don’t even fit their definitions, so we don’t do that.”

Instead, Partners has built a deiden- tified patient data registry, which gives researchers data and encrypted patient identifiers. “Only when those research- ers have homed in on precisely those patients that they need to know about for their research, and that question has been approved by our institutional review board, or IRB, which carefully looks over each question to make sure that the scientists are qualified and they’ve got a good study, a good study design, only then do we let those lim- ited numbers of records” be released to researchers, Murphy says.

The cloud’s role is present in the intermediate stages of research, when data sets can grow so large that Partners’ own data center may not be the appro- priately large scale to hold such big data sets. Using EMC cloud infrastructure, “we know that when we need to, we can expand externally,” Murphy says.

As another example of how it might work, Murphy offers the example of a brain MRI study Partners has already conducted on a smaller scale. In a cloud use case, such a study may start

with 100,000 images, each of which is between 10 and 50 gigabytes in size.

“If you want to see if the white matter tracks are bigger or smaller in patients with treatment-resistant depression, the 100,000 images you start with quickly becomes a million, because of all the steps that you go through in your com- putations. Then when the process is finally done, and you’ve gotten your final results, you usually delete all those intermediate copies.”

The data lake’s elastic storage pays another dividend: allowing researchers the kind of continuous backup they need to avoid a data disaster. “Although the intermediate data might seem to be tem- porary, if you lose it, you’ve just lost a year, maybe a year and a half of your life, not to mention that if you lose it, the chances are that you’re just never going to be able to get quite back to it, because of the way things work in science,” Murphy says.

One of the first applications Partners will host at its data lake will be the i2b2 clinical trials genomics research data- base. Informatics for Integrating Biol- ogy and the Bedside is a National Insti- tutes of Health–funded National Center for Biomedical Computing, which was based at Partners HealthCare System.

“This is an infrastructure to encour- age the talented folks at Partners to engage in the Partners data lake,” says Murphy. “We’ve purposely not lim- ited it to a specific disease, but rather we’ve kind of focused around various modalities—imaging, genetics, and data from the healthcare workflow in the clinics, and so forth.”

Since i2b2 has now outlived its orig- inal NIH grant funding and become a separately governed resource for researchers nationwide—not just at Partners—it offers growing evidence that the cloud’s benefits have outlived their original utility at individual orga- nizations, and are becoming a national asset in healthcare.