One year of ICD-10: First half 2016 data shows coding trends and impacts

by Eileen Dano Tkacik

One year following the official implementation of ICD-10, the coding industry is beginning to report valid results regarding accuracy, productivity, and denial trends. While some of these facts and figures are self-reported by HIM directors and anecdotal in nature, other findings are grounded in hard, fast coding performance data. Such is the case with the results from Central Learning (www.centrallearning.com), a web-based system that electronically assesses coder knowledge using real medical record cases and expert-verified answer keys.

This article summarizes coder performance data as measured across 50 health systems and 300 coders as of June 30, 2016. It compares these findings with other industry reports and extrapolates key findings for HIM directors and revenue cycle executives. Since coding and diagnosis-related group (DRG) assignment are the major drivers behind health system revenue streams, consistent data analysis helps to ensure accurate coding and reimbursement.

ICD-10 coding accuracy on the rise

According to Central Learning data, coding accuracy is slightly increasing after nine months under ICD-10 for both experienced coders and coders-in-training. While the industry overall still lags behind the 95% accuracy benchmark achieved in ICD-9, we’re getting closer in all three major patient types: inpatient, outpatient, and emergency services.

Fifty health systems are represented in the data, providing a broad-based assessment. We compared coder accuracy from Q1 (January 1?March 31) with Q2 (April 1?June 30) to identify recent, timely trends in code quality. The figure on p. 13 lists the most current benchmark of our status through June 30, 2016.

We expect the uptick in coder accuracy to continue as coders and health systems engage in more targeted education and training for ICD-10. Actively monitoring code quality through monthly coding audits, combined with coder knowledge assessments, helps prevent denials and mitigates compliance risk. This two-pronged approach to coding management is critical, as payer denials and Recovery Auditor audits are expected to increase later in 2016.

Education and communication are the keys to making continued improvements in coder accuracy over time. This is especially true for the five identified areas of coding accuracy concern.

Five areas of coding accuracy concern

A closer look at Central Learning data from June 30, 2016, identified five coding categories where accuracy remains below acceptable levels (ranging between 65%?75% accuracy). With concentrated training efforts in these specific areas, coders can see a marked improvement in the quality of their work. For example, the coding of injury, poisoning, and other external causes ranked third lowest in Q1, but only sixth worst for coding quality in Q2 with a 12.6% improvement in coding quality following focused educational efforts.

HIM directors and coding managers can use these nationwide benchmarks to compare coding quality across internal teams and identify specific areas for coding risk.

Other ICD-10 coding quality data results

Alongside these specific Central Learning data reports, AHIMA recently conducted interviews to benchmark coding accuracy and productivity in ICD-10. The survey was conducted by the AHIMA Foundation in May 2016 (http://journal.ahima.org/2016/06/13/survey-coding-productivity-dipped-after-icd-10-implementation). In the self-reported survey results, "respondents noted they experienced a 14.15% decrease in productivity, yet only a 0.65% decrease in accuracy."

With such a dramatic variation between AHIMA’s self-reported results and system-generated data from Central Learning, it is evident that continued coder assessments and monitoring are essential. Solid coding data drives performance transparency?a critical component of revenue cycle preparation and denial prevention in ICD-10.

Putting ICD-10 coding data to work

Practices and health systems are smart to flag specific areas of coding concern based on their areas of challenge to introduce targeted education and training. This is especially true for specific diagnoses and procedures frequently used by providers.

Coding performance data analysis is also a first step to develop follow-up performance measures for coding teams, coding audits, and coding compliance programs. Positive trends in high coding quality should be accelerated, while poor performance areas should be targeted for risk mitigation.

A great starting place is to identify your coding team’s top five most and least accurate code categories during coding audits or coder knowledge assessments. This will bring your strengths and weaknesses to the forefront so you can conduct appropriate training.

Once you’ve identified these categories, use your data to answer these five important questions:

1.How accurate is the code assignment methodology used for high-risk service lines within these code categories?

2.What are the specific coder knowledge gaps by diagnosis, procedure, and coder?

3.Is clinical documentation accurate, complete, and as specific as possible?

4.Are payers paying high-risk service lines correctly?

5.How much revenue, if any, is lost due to incorrect coding?

Coders are using training tools, such as AHIMA seminars and the Central Learning training tool, to enhance their knowledge and experience with ICD-10. The result should be continual improvement in coding accuracy. The impact of accurate coding on the revenue cycle, compliance, and accurate reimbursement becomes more obvious as we get further down the ICD-10 road.

Correlation between coding and revenue stream

High levels of coding accuracy are essential in both fee-for-service and value-based reimbursement models. So far, the payment trends under ICD-10, to the surprise of many, have been positive. There has been a steady decrease in claims processing and payment velocity. Also, the deluge of claims denials has not yet happened?but may occur after October 2016 when the one-year grace period for code specificity concludes.

The following data was collected in a recent year-over-year six-month period (October 1, 2015?March 31, 2016), compared to the same six-month period from a year ago (October 1, 2014?March 31, 2015) according to RemitDATA, a healthcare claims clearinghouse company:

• Average staff processing time has shown a steady decrease during the year, with average staff processing time of 17 days in January to an average of eight days in May
• Average payer processing time has decreased throughout the year, with an average of 15 days in January to 12 days in May
• Total claims processing time was reduced by nearly 60%, with total processing time of 32 days in January to 12 days in June

Coding certainly has played a big role in this trend. Precise coder accuracy measurement and analysis of coding data are the first steps to making this transition.

Editor’s note

With over 30 years of combined expertise in audit, information technology, and revenue cycle operations, Tkacik is the director of operations and information of Aviance Suite, Inc. She last served as the Interim Director, Revenue Cycle at Lehigh Valley Physicians Group (LVPG). Prior to LVPG, she served as the Vice President of Information Technology and Patient Accounting. It was Tkacik’s combination of revenue cycle operations and information technology that led her to Aviance Suite. Aviance Suite is an integrated platform of web-based software applications that helps hospitals and health systems make better revenue cycle and clinical coding decisions.

HCPro.com – HIM Briefings