Optimizing the Monitoring of ICD-10 Coding Productivity

With the help of hindsight and data, we can now more accurately predict coding productivity and staffing needs. The run-up to ICD-10 had most of us very concerned, expecting to experience a decline in productivity of as much as 40 percent or more.  

Early productivity reports, based on perceptions and/or small sample sizes, confirmed that productivity impacts were not nearly as great as anticipated. The results of these reports, however, had significant variation, making it difficult to determine solid facts about trends.

The Backdrop

As Ciox Health implemented ICD-10 to support its clients, it also decided to gather productivity facts by undertaking two large coding productivity research studies using its own inpatient database of discharge cases arising from October 2015 through July 2016.

“Ciox’s goal for conducting these studies was to create realistic and dynamic productivity expectations to support the management of its coding services, as well as provide coding productivity insights to the industry for use in setting ICD-10 inpatient coding standards,” said Patty Sheridan, MBA, RHIA, FAHIMA, senior vice president of health information management (HIM) services at Ciox Health.  Ciox partnered with the University of Pittsburgh’s Department of Health Information Management to examine inpatient coding times specifically. The University of Pittsburgh provided the research design and statistical analysis leadership required for a study of this size and complexity.

The overall goals of the study were to:

1. Understand the impact of ICD-10 on inpatient coding productivity;
2. Understand the impact of variables such as length of stay and case mix index on productivity;
3. Create a formula that allows organizations to factor these variables into productivity expectations;
4. Provide recommendations to the industry in order to update aging standards that focus on ICD-9; and
5. Scrutinize large data sets rather than perceptions.

As part of the study, coder demographics were collected. These included credentials, training, and previous ICD-10 coding experience. Ninety percent of the coders in the study have a credential, 57 percent did not perform dual-coding, and 95 percent are employed full-time. 

All of the coders had previous inpatient coding experience, and 80 percent had more than 60 hours of ICD-10 training. Coding productivity times were calculated automatically during the coding process by Ciox’s Prism Coding workflow system.    

Coders used common industry encoders 100 percent of the time and used computer-assisted coding 24 percent of the time. The amount of time that most coders spent doing activities other than coding, such as abstraction, was in the range of 1 to 15 minutes per hour.   

The Results are In

After examining 10 months’ worth of post ICD-10 data using a sample size of over 320,000 records, coding times were shown to have decreased from approximately 44 minutes per record in October 2015 to approximately 37.5 minutes by July 2016.  When we compared this to an ICD-9 data set of over 80,000 records, we found that productivity decreased about 22 percent for the first five-month data set, from October to February. Then, for the next five-month data set, from March to July it was cut in half – productivity decreased by 11 percent.

Again, variables that may influence coding times, such as length of stay (LOS) and case mix index (CMI), were included in the analysis of productivity. Not surprisingly, as LOS increased, so did coding times. For example, for a length of stay of one to two days, coding times were approximately 28 minutes per record, compared to LOS greater than 10 days, wherein coding times were recorded at more than one hour per record.  

This also held true for CMI; as CMI increased, so did coding times. For example, a CMI of less than or equal to 1.0 was recorded at 34 minutes per case, compared to 44 minutes per case for a CMI greater than or equal to 2.11. The data revealed a statistically significant correlation between LOS and coding times as well as CMI and coding times. While there are other variables that influence productivity time, the focus of this study was on CMI and LOS. Additional variables, such as the use of computer-assisted coding, extent of abstraction, physician query, etc. can be examined in future studies.

Predicating Coding Productivity

Multiple linear regression analysis was conducted to predict coding time based on CMI and LOS. As a predictive analysis, multiple linear regression is used to explain the relationship between one continuous dependent variable, which in this case is coding time, from two or more independent variables, such as LOS and CMI. 

“A significant regression equation was found when running the regression analysis, which is significant in the statistical sense, since the results we are seeing are real and not just due to chance,” said Valerie Watzlaf, PhD, MPH, RHIA, FAHIMA, an epidemiologist and associate professor at the University of Pittsburgh. “To help in developing staffing plans, a formula from the regression analysis was created, which can be used to predict coding times based on LOS and CMI.”

The formula is as follows (coding time is noted in the formula as minutes per chart):  

Coding Time = 19.166+ + 6.650(CMI) + 1.743 (ALOS)

The formula considers that for each unit increase of the CMI, the coding time increases by approximately seven minutes on average, controlling for all other effects in the model.

And for each additional day in the hospital, the coding time increases by approximately two minutes on average, controlling for all other effects in the regression model.

For example, if you have a CMI of 1.3 and an ALOS of 3.0, the formula predicts that you should be coding at a rate of 33.04 minutes per chart.

Productivity Optimization and Monitoring

At this point in the implementation of ICD-10, the industry is focused on optimization and monitoring of coding productivity. Measurement is a critical first step on the path to optimization, and some steps to follow include:

Step 1: Identify the baseline productivity for your facility.

Step 2: Calculate the target productivity for your facility.

If your baseline is greater than your target, things are looking good. Keep it up. But if your baseline is less than your target, identify challenges and solutions and work to achieve your goals. Typical challenges include excessive abstraction, system interoperability and/or response time, and cumbersome physician query processes.

Knowing your expected productivity levels enables you to focus on areas that may need to improve. As the industry tackles productivity, it’s also time to optimize and monitor coding quality. Stay tuned to future studies on coding quality. 

The complete coding productivity studies were published in the March 2017 and August 2016 issues of the Journal of the American Health Information Management Association.