Coder vs. Machine—10 Rounds in the Ring

By Stephen C. Spain, MD, FAAFP, CPC, and Kathy Rowland, CPC, CEMC, MCS-P

The growing acceptance of electronic medical records (EMRs) continues to affect auditing services in new and significant ways. The number of our clients who use EMRs is on the rise, and recently we’ve encountered groups who increasingly rely on computerized selection of evaluation and management (E/M) level of service codes. Such audits provide an opportunity for a “head-to-head” comparison of conventional coding principles versus computerized coding systems that select level of service codes based on user input.

EMRs Promise Level of Service Code Accuracy

Like many providers and practices who have invested in EMRs, most of our clients were promised savings in processing and staffing, and improved reimbursement and charge capture. Within a year or two of adopting an EMR, some groups are secure enough to hand over the reins of code selection. This decision generally is made, in part, to allow the organization to realize fully the upside reimbursement potential of an EMR. Reliance upon the EMR’s selection process often is encouraged, if not urged, by EMR vendors.
Most EMRs can tabulate exam and history bullets to assign a level for these two key service elements. Software programs generally are incapable of calculating medical decision making (MDM) estimates, however. As part of the encounter documentation, most EMR systems require the provider select the MDM level. When the EMR has all the information, clients are told the software engine will assign the correct level of service code to the encounter.
So far, this sounds pretty good. In an ideal situation, the organization provides additional tutorials and coding advice to its providers—particularly, on the subject of calculating MDM correctly. In doing so, a provider group will be on track for a successful implementation of computerized code selection. In practice, unfortunately, problems quickly can ensue from this approach.

MDM Throws the First Punch

Some EMR coding engines can degrade accuracy of the coding process. In most cases, this occurs primarily because the engines often disregard MDM’s impact on code assignment, which creates code selection problems for established patient visits. Frequently, software gives numeric values to the history and exam key elements, and then assigns a numeric value to the MDM level inputted by the provider. The engine then assesses these three values.
This system may work well for new patients, where the lowest value of any of the three components would indicate the level of service. For established patients, however, where only two of the three key elements are required, we see many errors in code assignment by software systems.
When looking at three values, a computer algorithm simply has to pick either two values that match or, if none match, the program will pick the middle value. If MDM is a matching value, or is one of the two high-end values, then the code selection is accurate. If exam and history matches, or are the two highest components; however, then the MDM value has no bearing on code selection.
As an extreme example, an established patient with a pinky sprain could generate a level five visit (99215 Office or other outpatient visit for the evaluation and management of an established patient, which requires at least two of these three key components: A comprehensive history; A comprehensive examination; Medical decision making of high complexity) if the provider performs a comprehensive history and exam. Because an EMR allows the user to enter great amounts of often superfluous data, many visits are over-documented. Across the board, organizations and payers are seeing an alarming shift of level of service codes for established patients toward the comprehensive end of the spectrum. When considering use of an EMR to provide E/M level of service code selection, it is imperative to include coders in the decision-making process so they can test the EMR’s system of logic and understand fully its process for assigning codes.

Put Up Your Coding Defenses

To provide a better understanding of our approach to this coding issue, let’s look at our auditing practices and MDM. In a general retrospective audit, we typically apply the rules and regulations of the Centers for Medicare & Medicaid Services (CMS) and Medicare intermediaries because we find that these are the most published, most debated, and most vetted guidelines. Often, they are also the most restrictive. In our experience, if a provider is coding in compliance with the policies of CMS and its intermediaries, the encounter record and billed services generally are defensible across the board for all payers.
Coders know the CPT® manual states that only two of the three key elements are required for assessing a level of service for an established patient. Over the years, there has been debate as to whether MDM should carry additional weight for established patient visits as a marker of the overarching criteria of a service’s medical necessity. MDM is clearly a limiting factor in new patient code assignment.
In conversations with multiple Medicare medical directors, it has been affirmed to us that MDM is the best key element to base an assessment of medical necessity. One Medicare director wrote “when coding based on Med Necessity, then MDM is often the lynchpin.” Some Medicare intermediaries also include in their provider manuals the statement, “MDM is critical in determining the level of service.” CPT® clinical examples contain scenarios that can prove to be exceptions to MDM calculating. In our experience, however, if the level of service is supported by the MDM, the provider has a solid foot should any coding challenge arise.

Expose EMR Weaknesses with Audits

We approach our audits with this understanding of MDM and, as a result, we generally find errors with the EMR-assigned E/M level of service for established patients. Coders in a recently-audited organization reported that soon after making a switch to computerized coding, concerns arose that the EMRs were generating level of service codes with a high rate of over-coding errors. Providers who are aware that their EMR-assessed level of service seem too high are concerned. As we have counseled providers on their individual audit reports, many reported to us that they are perplexed to see a software system assign 99214 Office or other outpatient visit for the evaluation and management of an established patient, which requires at least two of these three key components: A detailed history; A detailed examination; Medical decision making of moderate complexity when straightforward or low MDM levels were entered.
Occasionally, our clients have their coding staff perform a parallel audit with us—both to affirm our results and to measure their own accuracy and reliability. In cases where an EMR’s coding engine was used, we have performed our audit twice: once applying our conventional rules for E/M code assessment, and another time applying the client’s EMR’s engine logic. These audits and methodologies generally expose computerized selection process inaccuracy. We have found, typically, the EMR engines generate approximately 15-30 percent more coding errors for established patients when compared to conventional auditing tools and coding practices.
At a hefty cost to our clients, claims with incorrectly-assigned service levels are re-submitted with corrections, and some reimbursement inevitably is returned to the payers. Fortunately, some of our clients with savvy coding staff identify errors quickly and convince their administration of the value of an outside audit to address their concerns.

Evaluation and Management – CEMC

Arm Yourself with Knowledge

EMR companies are gaining an understanding of the complexity of the MDM issue, and are working continually to improve the code selection process. It is common for software vendors to withdraw enthusiasm for their current engine and promise an imminent upgrade that will enhance accuracy.
As a result of our audits, and usually with the blessing of internal auditors, most of our client organizations have placed the computerized level of service code selection under intense review, or have restored that decision to the authority of providers—putting the organization back in the driver’s seat of the code selection process. Hopefully, the education and insight gained through careful coding oversight and auditing services will convince them to keep their hands firmly on the wheel until the promises of future versions and upgrades are proven to be true.
Dr. Spain has been engaged in the full-time practice of family medicine for over 25 years. In 1998 he founded Doc-U-Chart, a practice management consulting firm specializing in medical documentation. Dr. Spain can be reached at

Ms. Rowland is a specialist in the development and implementation of practice-based compliance plans. Her compliance experience also includes medical record auditing in many specialties. Ms. Rowland can be reached at

Medical coding books

Comments are closed.