Executive summary
More hospitals are taking a serious look at computer-assisted coding (CAC). While CAC still requires human input, the latest technology - natural language processing that extracts code-related terms from electronic text-has helped boost productivity. Both Eastern Maine Medical Center and Massachusetts General have documented shorter turnaround times in their coding tasks.
Computer-assisted coding (CAC), after making substantial inroads in outpatient departments, is starting to emerge on the inpatient side as well. While only a few hospitals around the country have rolled out CAC across the board, more facilities are beginning to look at it. That number is expected to rise because of the coding challenge posed by the changeover to ICD-10 diagnostic codes.
Due to the complexity of inpatient care, the vast number of medical terms used, and the prevalence of hybrid paper and electronic records, hospitals don't use CAC without human intervention, as ambulatory-care providers sometimes do. But the latest CAC technology - which employs natural language processing to extract code-related terms from electronic text - has been shown to raise productivity by automating portions of the coding process.
As a result, CAC is breaking out of the radiology/pathology/ER ghetto and is crossing over to the inpatient arena. Among the institutions that are beginning to use it are Yale-New Haven (Conn.) Hospital, Baltimore-based Johns Hopkins Hospital, University of Pittsburgh (Pa.) Medical Center and Eastern (Bangor) Maine Medical Center. Vendors of CAC systems include 3M, St. Paul, Minn.; CodeRyte, Bethesda, Md.; A-Life Medical, Inc., San Diego, Calif.; and Artificial Medical Intelligence, Eatontown, N.J., now partnered with Medquist, based in Mt. Laurel, N.J. Health information services vendor QuadraMed has also just added a CAC module.
Another approach to CAC is to use electronic health record (EHR) documentation to generate some of the codes. Rick Schooler, vice president and CIO at Orlando (Fla.) Health, which uses Sunrise Clinical Manager, supplied by Atlanta-based Eclipsys Corp., on the inpatient side, says that the seven-hospital system has decided to move forward with this kind of CAC. As it rolls out its EHR, Orlando Health will use a term-mapping solution from Intelligent Medical Objects, Inc., Northbrook, Ill., that will be integrated with Eclipsys, he says. The structured input from the EHR will be combined with natural language processing.
“To the degree that the IMO product does natural language processing and electronically reads through the documentation looking for key words, we'll probably have EHR templates that generate the key words,” says Schooler. “It will help us move from ICD-9 to ICD-10, which is going to be a monster.”
Productivity gain at EMMC
CAC has already improved the productivity of coders at Eastern Maine Medical Center, says Mandy Reid, coding manager at the flagship of a seven-hospital system. After the hospital introduced the 3M product in August 2009, she says, the number of records processed per hour increased 15 percent in the first 45 days and 30 percent in the first 90 days.
Eastern Maine Medical Center also decreased its average turnaround time for coding records from five to four days. Its average days in accounts receivable dropped 7 percent, and it was able to reduce its coding staff by one full-time employee, says Reid.
When the hospital decided to adopt 3M's inpatient CAC software, Reid says, the goal was to “take the page turning out of the coders’ job.” By combining natural language processing with 3M's coding decision support tool, she explains, the CAC system enables coders to click on an annotated term, such as chronic obstructive pulmonary disease (COPD), and follow the decision tree logic as far as it can take them with a high degree of confidence. “Then it lets the coders make the final few decisions to get to the correct code. So compliance-wise, it only takes them as far as it accurately can without someone intervening.”
CodeRyte, which supplies 3M's natural language processing, has added a quality assurance module that enables the program to learn from its mistakes. Essentially, the software registers how often coders follow its recommendations on a particular code and adjusts its approach accordingly. For example, it can figure out when it's misinterpreting unfamiliar terms that physicians sometimes use. Over time, “training” the software in a particular hospital should make it more reliable.
Schooler compares the evolution of CAC to that of voice recognition software, which is widely used to transcribe medical dictation with some editing by transcriptionists. “Coding will go down the same path. There will be certain technologies that will allow the coders to transition to more of a review function, as opposed to bottom-up coding on every patient. And there will be technologies that capture the codes 100 percent upon documentation. I think it's going to be a combination of the two. Will we ever eliminate the need for coders? I doubt it.”
Kelly Taylor, an Atlanta-based consultant with CSC in Falls Church, Va., agrees that, though CAC can increase productivity, it requires human intervention in the complex hospital environment. “According to the American Health Information Management Association (Chicago), CAC technology should be viewed as a tool to assist coding staff rather than as a replacement for coding staff,” she says. “I don't think you can ever take the human coder out of the equation.”
Getting the staff aboard
Eastern Maine Medical Center overcame the staff's apprehension about CAC, says Reid, by getting the coders involved at an early stage in the implementation process. “We made sure they saw the product and knew how it functioned.” Now, she says, the staff loves it. Aside from sparing them the routine scut-work of coding, she says, it also does auto-abstracting of physician names and dates of service.
From a technical point of view, CAC required the hospital to add a few extra servers to feed transcribed reports from its EHR to the CAC program. Interfaces also had to be written to the billing, lab and radiology systems, but the admissions-discharge-transfer (ADT) system was already linked to the 3M coding decision support tool.
The more electronic text a hospital has, the more effective CAC is, Reid points out. While a hospital without an EHR could apply CAC to transcribed reports alone, Eastern Maine Medical Center's EHR also includes electronic forms and nursing notes (but not yet physician notes) that the CAC software can parse. It can't, however, search scanned documents.
Eastern Maine Medical Center is now rolling out CAC to its radiology and pathology departments and its outpatient physician offices, says Reid. She expects greater productivity gains in those areas because of auto-coding.
Radiology department experience
Outpatient departments have much more experience with CAC than inpatient facilities do. For example, the radiology department of Massachusetts General Hospital in Boston has been using the technology since 2002. At that time, it took two to three weeks to code a case and get it out the door, says Christie James, lead radiology billing manager at Mass General. Today, all cases are coded the same day that tests are performed, and 30 percent of those are auto-coded. James says her staff is working toward raising that number to 40 percent; some radiology departments, she says, auto-code as much as 60 percent of their claims.
CAC is easy to use with X-rays, mammograms and other tests that always require the same set of codes, James notes. In those areas, she says, the accuracy rate of Mass General's CodeRyte program is above 95 percent. She knows this because the department's claims denial rate has dropped significantly. Coders review high-tech procedures like computed tomography and magnetic resonance imaging scans, she says, “because that's where all the money is and where there's more scrutiny.”
CodeRyte has figured out how to program each health plan's coding rules into its CAC program, James points out, although there are still glitches. “We'Re still not where we want to be with that, but we'Re getting closer.”
Overall productivity has increased, notes James. Although the department's volume increased 17 percent from 2002 to 2009, she didn't have to add any coders. Moreover, all coders now work from home, saving office space that is very expensive in Boston.
James admits that implementation of CAC was not easy. “It took a while to get there; it was very painful. But now it's working really well.”
Healthcare Informatics 2010 July;27(7):22-25
