Memorial Sloan Kettering project culls data to find insights
To support precision medicine and population health, Memorial Sloan Kettering Cancer Center in New York is working with analytics vendor Cota in a multi-year partnership.
The company will receive anonymized clinical and genomic data from which it will create new datasets that categorize patient factors, diseases and intended therapies by creating cohorts, or groups, of similar patients.
These cohorts create a “clinical story,” says Paul Sabbatini, MD, deputy physician-in-chief for clinical research at Memorial Sloan Kettering.
“The real challenge is that there is a lot of information on patients in electronic health record systems, but the data is in multiple formats and much of it is unstructured,” he explains. “If we can create the clinical story with mutations, various characteristics and other patient factors, we can start to find patterns.”
Cota has developed a process to explore the data in EHRs, extract information and display it so the data can be searched, analyzed and understood, he adds. The vendor takes characteristics and represents them by a number that is similar to a bar code, enabling a search for similar patients.
“There are ‘lanes’ in numbers, and each lane is a characteristic, such as the type of tumor, location, size and stage of the tumor. It’s a map and legend and we match characteristics to the legend,” Sabbatini says.
That means that Memorial Sloan Kettering can take a cohort of breast cancer patients and see who responded best to various therapies, and who did not. “It’s all about outcomes,” Sabbatini says.
The goal is to have cost-effective care while being able to more definitely be able to select the best treatment regimen. “We want to know that if a patient has this type of breast cancer, this is the optimal treatment course. Right now, all this information is just sitting around in the EHR.”
There are a variety of such datasets in other research facilities across the nation, and Memorial Sloan Kettering is working with others to combine clinical and genomic data, yet they are hindered because all the different EHRs don’t match with each other, so it’s difficult for the groups to put the datasets together.
Cota will structure data and give it back to researchers in raw form so each group can be doing their own analytics or cooperatively conducting joint analytics.
“This gives us more datasets from other large academic institutions working with Cota, which gives us the ability to combine in a larger dataset,” Sabbatini notes. “This is an unmet need in medicine; to more easily get structured and unstructured data.”