- NEW ORLEANS - Healthcare big data analytics and precision medicine won’t get very far without the knowledge, expertise, and unique skills of health information management professionals, said John Showalter, MD, MSIS during a standing-room-only session at the 2015 AHIMA Convention yesterday.
In order to provide maximum value to healthcare organizations - as well as to their own careers - HIM professionals must navigate the transition from traditional medical librarian to data curator, crafting a meaningful portrayal of a patient, event, community, or health system from disparate data sources.
Showalter, who serves as Chief Health Information Officer at the University of Mississippi Medical Center (UMMC), introduced the concept of the “electronic phenotype,” which he defined as “the appearance of an individual or organization based on their data from EHRs and other systems.”
This portrait would ideally contain data from every available source: in-house electronic health records, encounter data from external care providers, patient-generated health data, pharmacy and medical claims, socioeconomic information, outcomes or complications, and genomic data.
Taken all together, this wealth of big data could enable rich and meaningful predictive analytics and precision medicine. But the healthcare system continues to face familiar challenges for HIM professionals and data scientists who are trying to assemble their treasure hoard of information.
A lack of interoperability, intractable data siloes, disappointing documentation quality, and data integrity roadblocks are leaving providers with nothing more than a “Swiss cheese” version of a patient’s true health status. Healthcare organizations may suffer from poor reputations and skewed consumer perceptions about quality and outcomes if the data being used for reporting is inaccurate or incomplete. After all, statistics can be sneakily unreliable, and different methods of sorting and slicing the same data can produce drastically different results.
Similar problems arise when attempting to use patchwork data for decision-making purposes. “We have people pushing analytics and precision medicine, but how are we going to do precision medicine on the Swiss cheese version of the patient?” Showalter asked.
“In ten years, when we’re ready for precision medicine, we can’t have Swiss cheese people in the EHR. In ten years, we’re going to have the genetic fingerprint for to improve decision making through precision medicine. The problem is that if the EHR doesn’t know everything it needs about the patient, it’s going to give the wrong directions to the clinician. We can’t wait ten years. Even right now, a Swiss cheese person makes a Swiss cheese provider, which makes a Swiss cheese health system.”
As value-based reimbursement puts pressure on healthcare providers to harness the power of big data analytics for population health management programs, these gaps in the patient record are going to become increasingly costly - not to mention worrisome for care coordination and patient safety.
Healthcare providers will have to look at the possibility of integrating datasets that fall outside of the traditional healthcare domain, such as socioeconomic factors, geographical data, and community information if they wish to understand why they are seeing so many emergency department visits or failing to achieve optimal medication adherence.
“You can get into some thorny privacy and health disparity issues with non-clinical data. But if your patients live on a block that has no public transportation, and they are fifteen miles from the nearest pharmacy, and they are almost universally uninsured because their state didn’t expand Medicaid…guess what? They’re not going to fill their prescriptions.”
“There are a lot of issues to work through, but it’s so critical to bring in those outside determinants to care that we don’t like to talk about. We’re going to have to talk about them.”
The health information management department must keep that conversation going, and bring a strong focus on data integrity, information governance, accuracy, and informatics with them. “There is no other group in the health system prepared to do this,” Showalter declared. “The only people who can do this are the HIM professionals.”
“HIM are the intermediaries between what the doctors write and what [public reporting data and consumer-facing resources] say about your health system,” he added. “I don’t know that your training has prepared you for that role. But that’s the world we live in, where you’re responsible for the phenotype of your patient and your medical center.”
Fortunately, HIM professionals have an array of different health IT tools at their disposal to ensure that their organizations are creating, collecting, and leveraging their big data appropriately. At UMMC, the health information management team keeps a close eye on physician documentation with real-time reminders, alerts, and query features.
UMMC’s computer-assisted physician documentation program uses natural language processing to fire an alert if a clinician includes a term that is too vague, such as “anemia” or “congestive heart failure” that prompts the user to further clarify his statements. If a physician copies and pastes a note from one day to the next without making adjustments for temporal phrases like “today” or “yesterday,” a real-time reminder will flag the issue.
If the physician ignores the alert, the HIM department will get a ticket stating that the issue is unresolved. They can then contact the offending provider through a text message or phone call with a personalized reminder to correct the problem and avoid further action.
“Computer-assisted physician documentation is about getting the right clinical context in place,” Showalter explained. “Don’t just say chronic heart failure. It matters in a clinical context if it’s systolic or diastolic. I need that information so I know what to do when I’m responding to a patient who can’t breathe in the middle of the night.”
“You’re not able to do the physician’s job, but you need them to do their job better so you can do yours,” told members of the audience, which included many HIM educators and students.
The opportunities for health information management professionals and healthcare data scientists are set for significant growth in the near future, said Showalter, as organizations start seeking qualified talent to get their big data analytics programs off the ground.
That is easier said than done, he noted. An ongoing shortage of credentialed data management applicants - especially those with a strong working knowledge of the healthcare industry - means that attractive applicants with real-world experience on their resumes are few and far between. UMMC has turned to teamwork to fill some of its informatics gaps, Showalter said.
“We have a PhD physicist working on some of our data analytics projects,” he said. “He’s brilliant and can do all the data science, but he doesn’t know that ‘RN’ stands for ‘registered nurse.’” Showalter paired his data scientist with an experienced Registered Health Information Administrator (RHIA) to provide the missing piece of healthcare-specific education, and the team has thrived.
“I don’t expect the RHIA to know PhD physics-level algorithms, but the RHIA understands all about the EHR, and she has the ability to bring the right ontology to bear for whatever problem they’re tackling,” he said.
That is a key competency for health information management experts who want to make themselves useful in an era of big data analytics, detailed population health management, and innovative medical research.
“The number one skill for an HIM professional is critical thinking and the ability to do self-learning,” Showalter stated. “The industry is moving so much and so fast that the skills you have today are not the ones you’ll need tomorrow. I work with my people to keep them moving forward with professional certifications. The other is the ability to dig into technology. You don’t need one particular coding language or analytics skill, but you need to be able to dig into the technology and figure out how to use it.”
“Right now, industry is really good at computer assisted coding, for example,” he continued. “You’re getting 70 percent of the codes right, but that still doesn’t give you the type of big data that will help you to think about your electronic phenotype. We need between 80 and 90 percent accuracy for that, and that’s a huge HIM opportunity.”
“Understand how your CAC makes its decisions and how things get into the EHR,” he suggested. “We’re not going to get to 80 or 90 percent accuracy because natural language processing algorithms are going to get that much better. But you can make them work a whole lot more effectively if you get data and documentation captured properly on the front end.”
ICD-10 and its focus on detailed and complete documentation may provide the impetus for healthcare organizations to raise the bar on their clinical documentation. The new code set may provide enough information within the clinical note to support clinical care and decision-making while also enabling big data analytics on a broader scale.
“The specificity coming with ICD-10 is going to give us more information in the next 6 months than we’ve had in the last ten years. The detailed information we’re going to have from ICD-10 is going to completely revolutionize administrative data in the United States. ICD-10 will be painful for the next 3 to 6 months or a year. But in the next 3 to 5 years, you and your children will have much better healthcare as a result of putting ICD-10 in place,” Showalter said, garnering a round of applause.
The new code set is poised to be a disruptive force for health information managers, but a commitment to ongoing education, independent learning, and gaining the technical competencies necessary to thrive in a big data analytics environment will help HIM professionals stay well equipped for the future of precision medicine, population health management, and quality patient care.