Using data to drive population care management requires healthcare providers to adopt new processes, change behaviors and be mindful of new challenges, according to Earl Steinberg, MD, MPP, CEO of xG Health and executive vice president of innovation and dissemination at Geisinger Health System, based in Danville, Pa.
Dr. Steinberg highlighted the following six keys to transforming raw data into actionable information that supports population care management during a session May 7 at the Becker's Hospital Review 6th Annual Meeting in Chicago.
1. Access to relevant data. Organizations should standardize the collection and storage of several types of data including claims data, clinical EHR data, patient-reported data and data from other sources.
"There are challenges associated with the use of any of these types of data, it's not like one of these is the killer type of data, where if you don't have it you're dead, and the other types are marginal," said Dr. Steinberg. All the various types of data play a useful role in population care management and organizations need to know how to use them all, he added.
2. Acquisition, integration, cleaning and normalization of the data. No two data sets are alike. With 35 years of experience in claims data analysis, Dr. Steinberg said, "If you've seen one payer's data, you've seen one payer's data. If you've seen one payer's data for a month, you've seen one payer's data for a month."
This means that all raw data needs to be integrated, cleaned and normalized before it is stored. Claims data is notoriously inconsistent, according to Dr. Steinberg. His favorite example is from 10 years ago. He was analyzing the data of 30 million lives from different payers when he realized one payer sent every claim except those that occurred on Tuesdays.
He noted that the collection of EHR data also has plenty of challenges. It is not always easy to extract EHR data in a way that preserves the meaning of each piece of data.
3. Rigorous quality assurance. "The value of claims data in particular depends on your ability to interrogate and QA that data, and if you don't do that well, you will produce garbage in your results until you send a physician or to a patient that's dead wrong. If you're lucky you'll get away with that once, but if you do it more than once say goodbye to your credibility," Dr. Steinberg said.
EHR data is often ridden with false positives, false negatives and contradictory information as well, so organizations should build QA into their data collection, integration and normalization processes.
4. Analysis. According to Dr. Steinberg, the real challenge with data, is interpreting the results and understanding the implications of the findings are for population care management. Providers should not target patients solely because they have high-risk index scores, he said. Rather, providers should analyze the data in such a way that they are able to identify the most actionable opportunities with the biggest clinical and economic impact.
"Most of what goes on in the area of data analysis has room for improvement in terms of the ability to pinpoint the most impactful, actionable opportunities," he said.
5. Interpretation of actionable findings. There are three main uses of data in care management, according to Dr. Steinberg. These include targeting patient and provider opportunities, clinical decision support and performance evaluation and monitoring. Once the analysis step is complete, providers must be able to interpret the findings and pump that information to the source that needs it.
"I'm absolutely convinced that pumping results directly into the clinical workflow is extremely potent, though it's not only way to work with data," Dr. Steinberg said. "You can also send it to patients directly."
6. Appropriate data delivery. No matter the recipient of the data, it is important to make sure the results sent to each stakeholder are relevant, customized to their needs and understandable. Patient-specific information is important to those directly involved in patient care, Dr. Steinberg said, while provider-, specialty-, or facility-specific data is better suited to clinical department leaders, and population-wide clinical and economic performance data is good for executive leaders. Dr. Steinberg also stressed how important it is to tailor the way the results are provided because the medium of the message will impact the likelihood that it results in action.
"I assure you, you can go out and buy best tools and platforms there are, but if you don't have the know-how or support services surrounding them, you won't get results," said Dr. Steinberg.