The Limits of Computer-Assisted Coding


Over the past several years, there has been much angst over the conversion from ICD-9 to ICD-10. Although there was plenty of warning – ICD-10 was first made available to World Health Organization (WHO) member states in 1994 – putting it in place still came down to the wire for many organizations.

Then the Centers for Medicare and Medicaid Services (CMS) announced in July a 12-month grace period on mistakes as long as the coder used a code from the correct ICD-10 family.

Of course, having technology in place and using it properly are often two different things. One way organizations are overcoming some of the ongoing difficulties is with Computer-assisted Coding (CAC), which provides a way to improve productivity as well as accuracy among coders by removing the need for human coders to page through 68,000 ICD-10 codes to make an entry.

Instead, the CAC technology scans medical entries in electronic health records (EHRs) looking for keywords and using its algorithms to suggest what it determines are the most appropriate codes. Human reviewers then determine whether the suggestions are correct and make adjustments as-needed. In effect, they become more like editors than researchers, allowing machines to do what they do best (review large amounts of data) so humans have a better starting point.

SEE ALSO: Staying the Course on ICD-10 Will Bring Immense Benefit

In the current context, using CAC makes sense – it solves the immediate problem of complying with the transition. Yet it also begs the question as to whether we are treating the health IT symptom, rather than the disease.

The Root Problem

From a technical standpoint, migrating from ICD-9 to ICD-10 should not have been as difficult as it has turned out to be. Yes, there is a learning curve, and some challenges in mapping the 16,000 old codes to the 68,000 new ones, but there were organizations that managed to be prepared well ahead of the deadline. So what is the root cause that necessitates measures such as CAC?

Much has to do with the age of the infrastructure and applications being used in healthcare, both on the payer and provider sides. At a meeting this summer hosted by Oracle, CEO Mark Hurd stated that the average age of enterprise applications (regardless of industry) is 23 years. Many applications in healthcare, especially on the payer side, are actually closer 30 or 35 years old; think about how the world has changed since 1985.

Today, it is difficult to find the tools that were used to build those ancient applications, much less people who know how to use them. Yet healthcare organizations have continued to patch these old systems because replacing them with new technologies built for today’s world can be hugely expensive and disruptive.

It’s going to have to happen, however, because the difficulties we’re seeing with the transition to ICD-10 are merely symptoms. How can organizations that have had this much trouble migrating to the new codes expect to move into far more complex areas such as bundled payments, where they are being asked to aggregate billing from sometimes dozens of individual providers and submit it all as a single invoice? Or include patient satisfaction scores and other measures?

The level of tracking and management they will need to add to their operations will be far more complex and detailed than what was required to migrate to ICD-10, not to mention the analytics they will need to perform in order to determine the ROI of current contracts and guide them when negotiating new ones.

To understand the impact, suppose a bundled payment contract for a knee replacement includes a provision for physical therapy, but the physical therapist mistakenly submits an invoice directly to the insurance company. It is unlikely that the outdated technology will be able to catch it because the therapy may be included in one contract and not another. It may turn up eventually in an audit, but in the meantime there is a double payment. These issues are likely to be fairly commonplace until the technology catches up to changes in the industry.

Technology Transplant

Many healthcare organizations need a technology transplant to operate effectively in the 21st Century, but like a patient who needs a heart transplant, they’re going to put it off for as long as they can before submitting to technology’s knife.

Upgrading those core applications and systems will be extremely disruptive. Implementing technology such as CAC allows organizations to survive a while longer with the status quo, but time will eventually run out on even the most meticulously maintained systems. Then, organizations will have some difficult decisions to make.

Chief among these is whether to continue implementing technology internally or move it to the cloud. Healthcare organizations have been resistant to going all-in to the cloud due to concerns about security and the privacy protected health information (PHI), largely driven by their responsibilities under the Health Insurance Portability and Accountability Act (HIPAA). They are reluctant to give up complete control over their data and infrastructure, but that may be inevitable.

Oracle’s Hurd believes all enterprise computing will eventually be performed on technology delivered by three or four major cloud providers – and when a change such as the migration to ICD-10 occurs, it will be on the cloud provider to perform the necessary mapping and upgrades to bring the organization into compliance. This will allow healthcare organizations to focus their technology efforts on using data to benefit the health and wellness of their patients.

Preparing for the Future

How do we get there? The first step is to take it one piece at a time. Take and test an action and once there is success, use the same methods for other parts of the business.

Another important step is to start thinking long-term. Healthcare organizations tend to think in terms of 12-18 months (CAC is an example of this). Taking a longer-term view will help guide better decisions, placing less focus on immediate functionality and more on future adaptability. A “solution” that works today but won’t in five years isn’t really a solution; it’s just a temporary patch.

If a healthcare organization decides to get out of the IT business and work with partners for its technology needs, it will be important to ensure those partners are thinking long-term and have adaptable systems as well. The organizations should share a vision for where the industry is headed and how they will react to changes to avoid becoming trapped in outdated technologies.

Healthcare organizations need to evaluate their approach to technology overall and ensure they have a plan that allows them to adapt to whatever the world throws at them so they can reach a worthwhile destination.

Craig Kasten is Chairman of SKYGEN USA, a collection of benefit solution companies that help healthcare organizations be market-leading and reform-ready. He can be reached at craig.kasten@skygenusa.com.

About The Author