Home  |   Subscribe  |   Resources  |   Reprints  |   Writers' Guidelines

August 2014

CAC’s Effect on Accuracy Rates
By Elizabeth S. Roop
For The Record
Vol. 26 No. 8 P. 18

Several factors determine whether the technology produces the desired results.

When Children’s Medical Center Dallas implemented computer-assisted coding (CAC) in 2012, it did so with lofty expectations: a 30% increase in productivity while maintaining existing accuracy rates and increasing the case mix index by 5%.

“Our implementation of CAC was based on the need to mitigate productivity declines with ICD-10 and ensure accurate payment with the state of Texas moving to prospective payment for pediatric hospitals,” says Katherine Lusk, RHIA, chief HIM and exchange officer at Children’s Medical Center Dallas, the nation’s fifth largest pediatric facility with 591 beds.

To achieve those objectives, Children’s Dallas implemented 3M’s 360 Encompass System. It started with the annotation of clinical terms, which highlights key terms for easy recognition by coders, and culminated with the autoassignment of ICD-9 codes. The tool was integrated with the hospital’s HIM Coder workflow within Epic, enabling the population of ICD-9 and ICD-10 codes.

The results were almost immediate. Within the first month of launching the annotation features, coder productivity was up, and coders were capturing an average of one additional diagnosis per inpatient discharge. By the end of the first year, that figure had risen to an average of three additional diagnosis codes per discharged inpatient record. “Productivity increased with annotation [and] gradually increased with autoassignment of codes, culminating with a 30% increase in productivity overall,” Lusk says. “The coders moved into an editor role, freeing up time to use their extensive knowledge base for clinical documentation improvement [CDI] activities, resulting in an increase in queries.”

The productivity boost also enabled the facility to eliminate five contract coding positions and still maintain a four-day turnaround time. Case mix index rose by 7.5% over the previous year and 16% in the year following implementation. Lusk notes that the current case mix index more accurately represents the acuity of patients cared for in an academic tertiary care medical center the size and scope of Children’s Dallas, which offers more than 50 distinct services, including oncology, neonatal ICU, transplants, and a level 1 trauma center. “We continue to see the same very positive results with 2014 in all aspects,” she adds.

Validating Accuracy Expectations
But are the experiences of Children’s Dallas the norm or the exception? When it comes to measuring CAC’s impact on accuracy and timeliness, the answer can be difficult to pin down thanks to differing opinions on what constitutes accuracy. “It depends on your definition of accuracy,” says Keri Hunsaker, marketing manager for 3M’s 360 Encompass System. “Is it the accuracy of codes submitted for billing or is it the number of accurate codes autosuggested by the CAC system? What about the accuracy of documents that feed the CAC system and the accuracy level of the coder reviewing the codes? These are all critical factors in determining the overall accuracy of CAC systems.”

A study conducted by the AHIMA Foundation in collaboration with the Cleveland Clinic set out to clear up the confusion and examine CAC’s true impact on timeliness and data quality. Researchers focused on determining whether there is a measurable difference between traditional coding and the use of CAC in terms of timeliness and accuracy, and whether the use of credentialed coders in conjunction with CAC generates further improvements.

ICD-9 procedure and diagnostic codes were collected on 25 Cleveland Clinic cases and assigned by 12 credentialed coders and CAC. Six of the coders assigned codes without the assistance of CAC, while the others took advantage of CAC technology. The first phase was conducted shortly after CAC was implemented; the second took place six months later.

To assess accuracy, assigned codes were compared against the gold standard, which is the set of correct diagnosis and procedure codes for each medical record as established and validated by the Cleveland Clinic coding leadership and quality team. Accuracy rates were calculated through recall and precision.

The time it took coders to code inpatient records using CAC was significantly shorter, resulting in a 22% reduction in time per record. In terms of accuracy, the study validated that Cleveland Clinic, which requires coders to achieve a 95% accuracy rate before they can be assigned inpatient records, was able to reduce the time to code without decreasing quality.

However, CAC alone had a lower recall and precision rate. The precision and recall performance for CAC alone also was tested at implementation and then six months later. Over time, the recall rate improved for coding both diagnoses and procedures, an expected outcome because natural language processing (NLP) engines “learn” over time.

The findings, in particular the idea that NLP quality impacts CAC performance, align with the beliefs of Steve Bonney, executive vice president of business development and strategy at Records One, which recently partnered with Care Communications. “CAC accuracy, in large measure, is dependent upon quality NLP capturing both positive and negated concepts in the proper sections of the documentation,” he says. “If the physician isn’t using structured documentation methods, this is extremely challenging.”

An Integrated Solution
In fact, Bonney says, the Care Communications-Records One partnership emerged from a desire to close CAC accuracy gaps created when documentation and other systems aren’t properly utilized or integrated. Care Communications brings to the mix a depth of hands-on HIM expertise and coding quality improvement services, while Records One offers advanced CAC and CDI solutions. The partnership’s goal is to enable client organizations to improve coding and clinical documentation productivity and accuracy and also more efficiently transition to ICD-10.

The arrangement was driven by “creating additional value while bringing clarity to a confusing market space,” Bonney says, adding that client demands for “a concise, integrated, and balanced approach of HIM coding and documentation expertise with advanced technologies to solve critical HIM business issues” also were a factor.

Similarly, 3M has taken an integrated approach to its CAC-CDI offerings. According to Hunsaker, this allows coders and CDI specialists to communicate and collaborate from the start of a patient’s stay by making the CAC autosuggested codes visible to both groups. In turn, this allows them to more quickly identify the working diagnosis-related group. They also may be able to identify overlooked evidence or documentation gaps that should trigger a physician query.

“The foundation for success with CAC—whether it’s autosuggesting accurate codes, improving coder productivity, ensuring compliance, or promoting accurate claims—is complete and accurate clinical documentation,” Hunsaker says. “The most sophisticated NLP engines available won’t generate accurate codes if the EHR and other interfaced systems are feeding inaccurate and incomplete documentation. It’s a simple case of garbage in, garbage out.”

The experiences of Children’s Dallas appear to validate the effect integration—in particular the quality of data that feed into the system—can have on accuracy rates. If incomplete processes and actions upstream cause inaccurate or incorrect information to be fed into the system, it will be reflected in the CAC results, Lusk notes.

Mary Beth Haugen, MS, RHIA, president and CEO of the Haugen Consulting Group, a health care consulting firm and educational services provider, adds that effective integration requires proper planning. “Do not overlook how the system addresses late documentation,” she says. “The system can only look at the documentation identified, so if you have key information the coder is reviewing, you need to make sure it is included in your integration plan.”

Integration does not have to stop at coding and documentation. For example, 3M integrates across functions and systems, combining CAC and CDI with quality indicators and patient data into one integrated system. Doing so promotes accuracy by improving communication among coders, CDI specialists, and physicians as well as quality and performance management, Hunsaker says.

Greater integration has had “a profound impact on the efficiency and accuracy of coding due to the system reinforcing collaboration rather than disagreement between coders and CDI specialists,” she says. “Some clients have seen the agreement rate between coders and CDI reviewers grow to 88%. With both the coder and CDI specialist working from the same view of the record and a common workflow, any remaining issues can be easily addressed. System integration also provides real-time data that can be seen and reviewed by all stakeholders, providing a clear opportunity to understand and/or impact patient outcomes.”

Outside Forces
Other clinical information systems aren’t the only outside forces that can mess with CAC accuracy rates. Other departments and individuals also can wreak havoc on the system’s ability to deliver on expectations. Hunsaker ticks off a list that includes the following:

• physicians, who must provide the most complete clinical story;

• quality, which is responsible for tracking the correct patient indicators; and

• CDI, which must ensure that the correct information is available for the coding process.

Coders must be able to communicate with all three areas to deliver accurate and compliant coding, Hunsaker says, adding that CAC accuracy can impact other departments as much as coding. “By integrating it with other functions such as CDI and by incorporating quality metrics such as key patient safety indicators, core measures, severity of illness, and risk of mortality, CAC technology can contribute to impacting patient care in real time,” she says. “We have clients that are using CAC in real time to identify patients at risk for preventable readmissions or hospital-acquired conditions, making it possible to undertake care interventions while the patient is still in the hospital. Accurate CAC coding integrated with CDI and quality metrics has contributed to declines in PSIs [patient safety indicators] and HACs [hospital-acquired conditions] for our clients—for some, as much as 63%.”

Hunsaker says the key is to make sure that the appropriate documents are interfaced with the CAC system and determine how hybrid records are handled. Other considerations include knowing what CDI specialists review to ensure complete documentation, what coders review, how queries are handled, and how responses are tracked.

Haugen emphasizes the importance of continuous review and validation of any coding information that is implemented within the CAC system, noting that while “many codes are picked up, additional codes may need to be added as well as the sequencing of codes may need to be changed.”

Start From a Solid Foundation
As with any technology, a major factor in how well CAC achieves expected accuracy rates depends on the strength of its foundation. Key components of a solid base include comprehensive training and education and continuous refinement.

Because CAC implementation impacts productivity, accounts receivable also feels its effects. Training, combined with a comprehensive understanding of just how much productivity may decrease at implementation, will help offset the impact. “I would also recommend having a coder educator or lead reviewing the coding to ensure accuracy of both the system and the original coder. The coder truly becomes the auditor, and the CAC will require feedback to improve on the accuracy of the coding,” Haugen says. “I have heard repeatedly that the implementation of the CAC had a greater impact on coder productivity than anticipated, so anything that can be done ahead of time to mitigate the impact should be considered.”

At Children’s Dallas, coders and CDI specialists received four hours of classroom training that focused on understanding CAC and providing feedback to allow the system to learn. Internal policies that placed responsibility for accurate coding on coders were not altered with the introduction of CAC.

It also is important to include postimplementation refinement in any CAC strategy. For example, the volume of documentation being sent through its system forced Children’s Dallas to make key adjustments. Initially, all documentation was being fed into the CAC system, causing information overload as the tool annotated nursing and allied health notes. As a result, it “minimized the productivity gains we expected, so we reassessed and determined that we would send only documents that a coder or clinical documentation improvement specialist would use to assign a code,” Lusk says.

If adjustments are indeed necessary, ensure coders have a voice in any changes. Often, this means reassuring them that the goal of CAC is not to eliminate staff, but rather to improve documentation quality and decrease the turnaround time to final bill.

“There is always a little bit of worry that a machine will replace them, but the tool is an adjunct adding to the tool kit,” Lusk says. “The more time your coders can spend on the implementation initially to help the system learn, the faster you will receive the benefits. They, of course, will be more inclined to help if they understand the tool is meant to augment, not replace.”

— Elizabeth S. Roop is a Tampa, Florida–based freelance writer specializing in health care and HIT.