Hospital Medical Coding in the Health IT Age

As the 2013 ICD-10 deadline looms, the need to abstract almost 20,000 diagnosis codes – a complex undertaking already - will soon become a labyrinth of over 150,000 codes. This would be significant enough to make any hospital take notice, without the additional squeeze on reimbursement from the Affordable Care Act. Thus, ICD-10 transition is a delicate undertaking that must be achieved with high precision, for there is no fat to waste in the revenue cycle. Indeed, many hospital CFOs and CIOs are busy applying technology and operational excellence solutions to this challenge. And as necessity is the mother of invention, this urgent evolution portends innovation in the future of coding best practices.

No innovation will replace the back-office coder, one of the fastest-growing occupations in the country. Even with the best, leanest and most integrated technology, a human being with a critical eye for error and common sense must be the final accountable party for claims batched for billing. But on top of the supply constraint of coders is the threat of a 30-50% reduction in productivity that other countries saw in the first months after adopting ICD-10. So this watershed event must herald changes for coding in the U.S. if providers are to hold onto control over their revenue cycle. Certainly at the point of care, much information critical for accurate coding is omitted in clinical notes by nurses and physicians, and still more is illegible or lost in paper charts. So it stands to reason that a good EMR enabling integrated clinical documentation goes a long way towards better back-end coding. Perhaps one day nurses using iPads with sleek technology, such as Nuance’s Dragon voice recognition, will capture exhaustive notes painlessly and pipe that information directly to coding software that submits claims in real time.

That day is inching closer, with many systems allowing for vendor-specific integration between clinical notes and billing. The challenge is that several intermediate steps, such as translating notes into diagnosis codes and procedure codes for reimbursement, remain a back-office slog-fest. Certainly, encoding lexicons like 3M’s Codefinder have been a boon for coders. Further, advanced computer-assisted coding software using natural language processing (NLP) – software that generates codes by finding patterns in the text of clinical notes – can further automate this process by linking to vendor-neutral clinical documentation. But, error rates in NLP engines are decreasing, they are not, however, negligible, especially when processing evaluation & management (E&M) codes, for example those that drive emergency department charges.

Conversely, other coding tools working from rules engines that are better at processing E&M codes have reduced error but generally require significant manual labor or are limited to feeds from other apps released by the same vendor. This means that hospitals using a best-of-breed portfolio instead of an enterprise solution have to touch records twice – once to code for diagnosis, and a second time to abstract procedure information into claims for billing. Workflows continue to be affected, and information management teams continue to be hampered by inefficiency. What coders want – and what hospitals need – is seamless integration capable of multiplying productivity.

Thus, the available technology begs the question: why can’t we put it all together and have the best of both worlds? What if clinical documentation could be fed through an NLP engine into a rules-based coder? If the NLP already automating diagnosis coding could learn from the human coders already entering structured data into E&M rules engines, a seamless, vendor-neutral feed between clinical documentation and billing codes might be enabled over time. The added automation would help transform coders from button monkeys into critical internal auditors, and give hospitals a significant solution against productivity loss due to ICD-10. Of course, all this is easier said than done. Even vendor-specific integration between clinical documentation and coding is a triumph. Still, vendors are busy developing technology potentially capable of these breakthroughs in automated coding. A decade from now, coders may look back and wonder how they were ever able to do their job by hand.

Stuart Kamin

MBA/MPH Candidate 2012
Haas School of Business
University of California, Berkeley