Tailoring treatments to the individual patient based on their predicted response or risk of disease is the next evolution of health care. Called personalized medicine, this field is rapidly moving from the experimental to the practical—and it’s readily apparent that we will need more data to feed the complex systems that can help predict a patient’s medication response or risk of disease.
At Partners, we’re working to satisfy this growing need by integrating previously isolated “islands” of patient data into a centralized repository called the Big Data Commons. Within the Commons classical clinical patient data is combined genomic and high-resolution imaging data and new types of information supplied by patients – through patient experience surveys that measure functioning before and after a procedures, and mobile and “wearable” tracking instruments like FitBit and Apple Watch.
“All of this aggregation is done under a strict veil of patient privacy and supervised by the Partners Institutional Review Board to ensure that individual patients never have their data inappropriately exposed,” said Dr. Shawn Murphy, Partners HealthCare Corporate Director of Research Information Systems and Computing and one of the leaders of the Big Data Commons.
Personalized medicine relies on comparing a patient’s unique characteristics against a large amount of data in order to tailor medical decisions and interventions to them.
The challenge with personalized medicine is to discover and refine methods that sort through the mountains of data in ways that ultimately impacts patient care. “Through the Big Data Commons, investigators are able to search for disease characteristics as well as genomic data, biobank samples, imaging data, and other attributes from many registries and data repositories within the Partners HealthCare system,” Dr. Murphy said.
This process, Dr. Murphy explained, allows researchers to create patient datasets that are relevant to specific questions. The goal is a better understanding of how diseases develop, and then how to manage, treat, or cure them in specific patients.
“As a first step, there is a lot of behind-the-scenes work to refine relatively low quality data into data that is meaningful for research and clinical care. We have developed sophisticated algorithms to achieve this,” Dr. Murphy said.
For example, depression diagnoses in billing data are only accurate half the time; this is because a depression billing code may be entered to cover a screening visit where a patient’s symptoms are discussed with a care provider prior to a true clinical diagnosis being made. However, the unique data processing methods within the Big Data Commons can refine the billing-code data by comparing it to additional patient data (such as subsequent visit notes and medications prescribed); patients who do not have depression are removed from the dataset, increasing the accuracy to about 87%. “Creating a high-quality dataset from these various data ‘islands’ is very powerful and can lead us to understand which differences in patients are important for determining disease,” Dr. Murphy said.
Further, by combining image data with a clinical finding of treatment-resistant depression, investigators were able to see changes in the growth of new nerve appendages between treatment-resistant and treatment-responsive patients. This suggests that differences in the microstructure of the brain underlie persistent, treatment-resistant depression. “Understanding if a patient has similar brain features may help with the medical intervention their care team develops to treat their depression,” Dr. Murphy said.
Another use of the Big Data Commons is to understand how gene variations impact different disease traits. Mining through the various data sources has allowed Partners researchers to predict treatment response to costly medications. For example, rheumatoid arthritis (RA) treatment was revolutionized by expensive biologic drugs that target inflammatory pathways. However, if physicians could predict a patient’s treatment-response, they could develop targeted therapies based on a patient’s individual genomic and biologic markers and substantially reduce the costs of RA treatment while also improving outcomes.
Ultimately, the Big Data Commons is a necessary response to the increasingly complicated world of medicine. As the work continues, Dr. Murphy and others at Partners believe it will result in a proven process for effectively managing complex treatment programs and the development of personalized care plans for patients.
“The Partners Big Data Commons is an advancement that few other health care systems have achieved,” Dr. Murphy said. “We’re forging a future that will apply advanced techniques to predict disease and improve outcomes for our patients.”
The Partners Big Data Commons is an advancement that few other health care systems have achieved. We’re forging a future that will apply advanced techniques to predict disease and improve outcomes for our patients.