A Love-Hate EHR Relationship
When it comes to EHRs, Adele Towers, M.D., sees the good and bad.
When it comes to EHRs, Adele Towers, M.D., sees the good and bad. Towers serves as medical director of health information management at the University of Pittsburgh Medical Center’s Melwood hospital, a role in which she supervises coding and documentation quality efforts.
She also practices geriatrics at the Benedum Geriatric Center. At HIMSS13 in New Orleans, Towers described how UPMC’s EHR, from Cerner, has expedited physician documentation, improved note legibility and removed guesswork around orders. At the same time, the technology has spawned a multitude of notes with little meaningful information, as they contain an abundance of material cut and pasted from other notes. “You read the same note over and over,” Towers said. And in many instances, the information that coders really need to assign the proper diagnosis code is missing or unclear, leading to missed billing opportunities.
In 2011, UPMC hired an outside auditor to analyze its charts. The report was not pleasant. Over a one-year period, UPMC was leaving behind some $18 million in legitimate reimbursement due to poor documentation.
The medical center decided to take a look at a technology called natural language processing to help sort through the sea of information. It’s working as a development partner with Optum Insight, a firm that provides the medical center’s computer-assisted coding software. The NLP system is designed to comb through patient charts—both narrative information and discrete data from labs—and identify possible instances where a more complex diagnosis is warranted. Charts flagged for possible deficiencies may then be routed back to physicians for clarification.
Towers described a tough case of a patient who spent nearly two weeks in the hospital after being beaten with a baseball bat. The case was coded at a lower level of severity than warranted. The NLP system flagged a couple of findings in the radiology report that suggested “cerebral edema,” a more serious condition which would be reimbursed more highly, was the possible diagnosis. Coding staff went back to the attending and determined that indeed, cerebral edema was the right choice.
Towers cautioned, however, that even though NLP technology can flag likely chart deficiencies, providers must be cautious about how they word queries to physicians about a particular case, lest they run afoul of billing regulations. “Physicians say you are only doing this for money, and we say ‘yes we have to, to keep the hospitals open,’” she said.
She also practices geriatrics at the Benedum Geriatric Center. At HIMSS13 in New Orleans, Towers described how UPMC’s EHR, from Cerner, has expedited physician documentation, improved note legibility and removed guesswork around orders. At the same time, the technology has spawned a multitude of notes with little meaningful information, as they contain an abundance of material cut and pasted from other notes. “You read the same note over and over,” Towers said. And in many instances, the information that coders really need to assign the proper diagnosis code is missing or unclear, leading to missed billing opportunities.
In 2011, UPMC hired an outside auditor to analyze its charts. The report was not pleasant. Over a one-year period, UPMC was leaving behind some $18 million in legitimate reimbursement due to poor documentation.
The medical center decided to take a look at a technology called natural language processing to help sort through the sea of information. It’s working as a development partner with Optum Insight, a firm that provides the medical center’s computer-assisted coding software. The NLP system is designed to comb through patient charts—both narrative information and discrete data from labs—and identify possible instances where a more complex diagnosis is warranted. Charts flagged for possible deficiencies may then be routed back to physicians for clarification.
Towers described a tough case of a patient who spent nearly two weeks in the hospital after being beaten with a baseball bat. The case was coded at a lower level of severity than warranted. The NLP system flagged a couple of findings in the radiology report that suggested “cerebral edema,” a more serious condition which would be reimbursed more highly, was the possible diagnosis. Coding staff went back to the attending and determined that indeed, cerebral edema was the right choice.
Towers cautioned, however, that even though NLP technology can flag likely chart deficiencies, providers must be cautious about how they word queries to physicians about a particular case, lest they run afoul of billing regulations. “Physicians say you are only doing this for money, and we say ‘yes we have to, to keep the hospitals open,’” she said.