There’s no question that coding is on the minds of everyone involved in ICD-10, and the two biggest questions are:

One: Will the coders be productive enough to complete their coding assignments?

Two: Will the coding be accurate enough?

The ICD-10 Coding Test

To answer these questions, we recommend an ICD-10 practice coding exercise. Here are the steps:

• Begin with the charts you used in the top 25 DRGs analysis we recommended here.
• Pass these charts out among your ICD-10 trained coders,  and ask them to code the charts.
• Give each coder four to eight hours of work, so that you will have sufficient data for analysis of each coder’s performance.
• Allow each coder to complete their assignment, and then analyze.

Productivity

For each coder, analyze the time required for him or her to code the selected charts in ICD-10. If you have data on the same charts when coded in ICD-9 from your coding system, you can easily make a direct comparison about productivity. If not, your coders average ICD-9 productivity would make a reasonable baseline.

Don’t be surprised if the ICD-10 coding takes much longer than ICD-9 coding. There have been a number of studies from Canada’s implementation (which did not include procedure coding) and early test exercises conducted with the United States’ implementation of ICD-10, that all suggest a 50 percent drop in productivity can be expected.

Ask your coders to generate physician queries as they normally would when coding production charts. If you wish, you can have the queries routed to a single source in the department for responses, but the query process will be a significant negative driver on coding productivity and must be included for valid productivity comparisons. The queries will also be useful as a driver of your physician education effort, identifying common areas where documentation does not currently meet ICD-10 standards.

Accuracy

The ICD-10 National Pilot Program, a joint project carried out by HIMSS and the Workgroup for Electronic Data Interchange (WEDI), recently reported the results of a preliminary ICD-10 test and found that the average accuracy of the (properly trained) coders included in the study was 63 percent.

Given that most hospitals strive for 95 percent accuracy or better, that 63 percent number is a sobering concern, with significant implications for compliance, continuity of care, quality measures, and reimbursement. Having an understanding of coding accuracy is critical for directing your ongoing training efforts, predicting denials, and managing your HIM shop.

If you used an outside consultant to review charts for documentation, as discussed in the previous post, it’s a simple matter to compare your coder’s results to the consultant’s and calculate accuracy. If you did not, one good strategy is to have multiple coders code the same chart and compare results. Where coding differs, reviewing the differences with all the coders involved will provide an educational experience, and likely a reasonable consensus on the expected codes.

At this point, you should also review the results to see how many codes were coded using non-specific codes, i.e. codes that are not the most specific codes of a diagnosis set, but serve as a useful placeholder when full specificity is not available. An excess of non-specific codes is a sign of trouble, indicating that either your documentation isn’t as specific as it should be, or that your coders aren’t recognizing the complexity of the codes.

Putting it all together

Combining the productivity and accuracy data from this test should provide you with a solid understanding of your coder’s abilities to complete ICD-10 coding, and an understanding of where mistakes were made. This will provide you with the information you need to customize each coders ongoing training over the next few months.

You should consider repeating the exercise at that time to assess your progress, and refine your calculations about the need for additional coding support in light of the additional training and expected improvements in documentation

Sufficient coding support is required, if you want to avoid substantial financial impacts. For more information about the financial impacts of ICD-10, download our Guide to the Financial Impacts of ICD-10.