Time is precious in the emergency department, where patients arrive unscheduled, the broad scope of medicine is fully encountered and rapid decisions are frequently made — in some cases with great impact on a patient's survival. Adding greater burdens on physicians in this setting is both unproductive and unwise.

Today, ED physicians may be using computer software intended to facilitate patient care, but which doesn't achieve that goal often enough. This may be related to poor user interface design, lack of good cognitive workflow or inefficient product ergonomics. ED physicians too frequently bear the burden of these deficiencies in a clinical setting that demands maximum efficiency.

The transition to ICD-10 may present yet another burden on the ED physician, depending upon how the EDIS, EHR or other documentation solution implements its ICD-10 support. Vendors will take different approaches for this process. Evaluating them is a prudent action to take well before the Oct. 1, 2014 date.

Generating adequate, appropriate ICD-10 codes is a key goal in any ICD-10 implementation plan. The patient’s diagnosis and associated ICD-10 code is necessary for claim submission, while also serving as a crucial data element in a host of different data collection processes, analysis and reporting.

Let the coder do it …

One of four unique approaches may be used by a vendor to generate ICD-10 codes. The first one is easiest to implement — it fully shifts the ICD-10 code assignment to the professional coder who reviews the physician’s chart.

The vendor simply allows for the recording of a patient’s diagnosis through free typing, handwriting recognition, voice input, selection from a list, etc. That textual representation is then assigned an ICD-10 code by the coding professional.

Such an approach is also easy for the physician since it passes the burden to the coder. The coder is then responsible for determining the appropriate ICD-10 code, which becomes more of a challenge with the increased number and complexity of ICD-10 codes. Coder productivity is estimated to be significantly impacted, with the potential to decrease by 20 percent or more.

Of considerable concern for the ED physician, this approach may significantly increase the assignment of "unspecified codes," which could have a negative impact on claims processing or reimbursement. Unspecified codes are those ICD-10 codes that contain the word "unspecified" as a component of their corresponding description. An example would be "unspecified asthma with exacerbation" (J45.901), contrasted with the fully specified code of “mild intermittent asthma with acute exacerbation” (J45.21).

Without prompting, no ED physician can document from memory all the necessary components or variables involved in any given category of diagnoses in ICD-10. Consider this example — a radius fracture of the forearm. There are approximately 400 seven-digit ICD-10 codes that describe uniquely different types of radius fractures — depending upon whether the fracture is open, closed, proximal, distal, displaced, segmental, comminuted, oblique, transverse, spiral, associated with a dislocation, etc.

Currently, the impact of unspecified diagnoses on claims processing and reimbursement with ICD-10 is not known. A variety of policies have been stated by payers, including "unlisted, unspecified and nonspecific codes should be avoided" and "nonspecific codes will not be accepted when a more specific code is available.”

Perhaps of greatest concern is that many payers have not stated their policy. Much will be learned in the first 120 days after the implementation of ICD-10. A better understanding of the financial impact will begin at that time — not a comforting thought for CFOs.

Natural language processing with autocoding

A second approach a vendor might use to generate ICD-10 codes could be via natural language processing (NLP) with autocoding, in which the computer identifies the textual diagnoses from a transcribed report or text-based computer generated chart and associates ICD-10 codes. Such an approach would be easy for the clinician to use and relatively easy for a vendor to implement.

Unfortunately, this approach also depends upon the physician’s memory to recall and appropriately designate all of the variables necessary to generate a “fully specified code.” This approach could also be associated with significant financial risk, depending upon how payers process unspecified codes.

ICD-10 search tool

A third approach would be for the vendor to provide a search tool that allows the physician user to enter a diagnosis by typing or use of voice or handwriting recognition. The application would then display to the physician a list of possible diagnoses with narrative descriptions and corresponding ICD-10 codes — unspecified and/or fully specified, depending upon the search terms entered.

For example, if the physician entered “right radius fracture” the search tool would provide approximately 700 uniquely different ICD-10 coded diagnoses, each unique in its description of the right radius fracture, depending upon variables such as the type of fracture and whether this was the first or subsequent visit for the fracture.

If the search tool was capable of limiting the results to just initial visits for the fracture — which is what one would expect in the ED most of the time — the number in the list could be decreased to about 132 choices. If the ED physician remembered to also include “closed” in his or her search entry, then the list could be narrowed further to only 44 choices. Still, that number of choices represents too much reading, scrolling and paging for the busy ED physician.

This third option would also be relatively easy to implement for the vendor but clearly would place an additional time burden on the physician user. Physicians would need to learn the variety of factors associated with the broad scope of common diagnoses encountered in the ED. On the up side, if the physician was able to take the time to read through the list of results presented, it could increase the likelihood of documenting a diagnosis with a fully specified ICD-10 code.

ICD-10 clinical content, feedback and code generation

Finally, a fourth option exists. While it’s more work for the vendor, it makes it easier for the physician as well as the professional coder. A vendor could modify the clinical content and provide a visual design that allows for easy visibility of the clinical variables associated with any given diagnosis.

That would make it simpler for the ED physician to document the maximum level of specificity for a given diagnosis without scrolling or changing screens. It would also eliminate the need for physicians to learn the terminology and categorization of diagnoses in ICD-10.

The corresponding text for the selected diagnoses could closely mirror the text found in ICD-10 making it easier for coders to assign codes — or the vendor could generate the ICD-10 codes based upon the selected information.

If the vendor also provided real-time ICD-10 related feedback to the user regarding specificity of diagnosis, it would further facilitate selection of fully specified ICD-10 codes, minimizing the potential risk of delayed or rejected claims.

Your choice

Different approaches will be used by various EDIS and EHR vendors to generate ICD-10 codes. Each approach will place differing burdens upon the vendor itself, the ED physician, the coders — and even the CFOs. Be sure to understand your vendor's approach so that it can be determined how well it fits with one's clinical and financial strategy and risk tolerance.