The healthcare industry is in the middle of a historic transition away from paper records toward a future where data and analytics drive treatment decisions. But messy data and siloed systems are slowing the transition.
Long after industries such as financial services, retail and marketing moved to electronic systems, healthcare continued to muddle through a largely paper-based record-keeping system. This started to change in 2009 with the HITECH Act, which incentivized the adoption of electronic health record (EHR) systems. More recently, elements of the Affordable Care Act that penalize hospital readmissions and incentivize keeping patients healthy over time encouraged providers to embrace analytics in healthcare to identify patients who are most likely to benefit from treatment.
But Mark Pitts, vice president of enterprise informatics, data and analytics at Highmark Health, said manipulating the data remains difficult. "The primary challenge is getting the data to where it needs to be," he said.
Pennsylvania-based Highmark Health functions as both insurer and care provider. Because all expenditures are internal, the network has an incentive to lower costs. It is trying to do this by identifying patients who have not received recommended preventive care or who have symptoms that suggest undiagnosed illnesses.
Pitts said most of the analytics is done ad hoc, which involves identifying necessary data, finding its source and running it through an analytics engine. Highmark does all its analytics in SAS Enterprise Miner and Enterprise Guide. But Pitts said he is looking to automate more of these tasks to speed things up. He said he is actively evaluating Hadoop-based systems, which could act as a centralized data store, something Highmark currently lacks. This could speed up the process of locating and moving data.
Mark Pittsvice president of enterprise informatics, data and analytics at Highmark Health
Highmark isn't alone in this struggle. Keith Blankenship, vice president of technology at accountable care organization consulting firm Lumeris, said many providers he works with have a hard time integrating data sources to speed up and simplify analytic operations. Today, most healthcare organizations have separate EHRs, picture archiving and communication systems, pharmacy record-keeping software, and billing tools. All these systems have data that would be useful for performing analytics in healthcare. But getting the data together isn't easy. "Because there's a lot of systems, making sure everyone is talking the same language" is difficult, he said.
There is no simple technology for the problem, Blankenship said. EHR vendors, which produce the technology behind the single greatest store of data in healthcare, have been slow to embrace interoperability, a problem that is expected to persist.
Blankenship said some of his clients hadn't implemented any type of analytics platform prior to working with him. He has been successful in helping organizations implement and integrate analytics software, so it can be done from a technical perspective. Resources are an issue he sees often. Today, the most successful integrations have happened at large medical centers. For smaller providers, it may be necessary to hire someone with expertise in this area or work with an outside expert.
Eventually the problems of non-interoperable systems and siloed data sets will have to be solved for the healthcare industry to embrace analytics at scale. Adam Kaufman, president and CEO of DPS Health, a Los Angeles-based company that helps providers keep patients engaged in efforts to improve their health, said healthcare has traditionally focused on collecting transactional data, such as health claims and treatment histories. But this data doesn't necessarily predict whether patients will continue taking their medication or participate in a workout program. That will require new ways of tracking and organizing data.
For example, he said, providers should be looking at patient data in the aggregate. Most today maintain patient histories individually. But to find meaningful correlations, they need to bring together data from multiple observations on many people over an extended period of time to truly understand how individuals engage with the healthcare system.
"The way we think about what people are doing ends up being different from the way we've traditionally structured our data sets," Kaufman said.
Learn how analytics benefits doctors
See how health data privacy challenges hold back analytics
Health data analytics: Is anyone really doing it?