Small County Hospital Uses Expert System Software to Increase Reimbursements While Cutting Code Time in Half

By Marilyn James, Medical Records Supervisor
Kathy McLeish, Medical Records Clerk
Lane County Hospital, Kansas

With one person doing DRG (diagnosis related group) coding by hand, Lane County hospital was experiencing a problem common to many hospitals. Inaccurate and missed diagnoses were causing the hospital to receive less reimbursement than it was entitled to from Medicare, Medicaid, and private payers using the DRG system. A random sample of charts revealed that 30% were inaccurately coded and in most cases the errors resulted in underpayment of approximately $500. By implementing Clinical Coding Expert software from IRP Systems, Wilmington, Massachusetts, the hospital has increased its reimbursements significantly because the software alerts the coder to complications and alternate diagnoses. And because the software simplifies the coding process, DRG coding now takes one-half the time it was taking by hand.

Lane County Hospital, established in 1962, is the smallest hospital in Kansas. It has 10 acute care beds and 21 long-term care beds. Its one full-time physician heads a clinic which is run by the hospital. A part-time physician also manages a satellite rural health clinic which is staffed with a PA (physicians assistant). The hospital averages 10 in-patients per month whose care is submitted for reimbursement. Since the hospital doesn't perform any major surgical procedures or obstetrics, most of the reimbursements are relatively small, but these funds are important to the hospital nonetheless.

Prior to automating the coding process, one person was doing all the coding by hand. She would go to the index in the ICD-9-CM manual and look up a diagnosis. If the terminology she was using didn't match the manual's, she was referred to another section. Many times, she would have to go through several different pages and categories, often being directed to additional references. To determine the appropriate DRG, she had to work through a tree. It was easy to make mistakes and not get the right codes for a diagnosis, or to miss the sign that a fifth digit for a more specific diagnosis was needed. The result was that the hospital did not always get as much reimbursement as it was entitled to, or it had to send money back when payments were made in error.

In Kansas, hospitals are monitored through PRO, a foundation for medical care. Among other things, this group checks charts for DRG coding errors. Random samples from Lane County hospital showed that codes were being missed in nearly one-third of the cases. The hospital staff purchased Clinical Coding Expert, a DRG encoder which incorporates the entire ICD-9-CM and CPT-4 coding book on a computer, in hopes of improving its coding accuracy.  A secondary goal was to decrease the time spent coding since this is a small hospital and the person doing the coding had other responsibilities as well.

The software has now been with us for a year and a half. It proved to be very easy to learn because it prompts the user as she goes through the coding process. The process begins by entering identifying information such as the patient's age, gender, date of admission, date of discharge, etc. Then the user simply types in the principal diagnosis from the chart. For example, for a rib fracture she types in "fracture" and the software immediately takes her to the alphabetical listing of fractures and their codes in the ICD-9-CM code book. At that screen she types in "rib" and pushes the enter key. The computer automatically provides the code for a rib fracture, 807.0. Compared to stumbling through a manual, this process is nearly instantaneous.

One of the main benefits of this system for Lane County Hospital is that it guides the coder to select a more precise diagnosis as substantiated by the medical record documentation. Frequently, this results in a higher reimbursement level than if the simpler alternative was selected manually. For example, the coding manual uses a little sign to indicate whether a diagnosis needs a fifth digit. This digit is a more specific code that speeds up and often increases reimbursements. However, the sign in the manual is easy to miss. With Clinical Coding Expert, when the coder hits the enter key, the software immediately gives the fifth digit if it is needed. In the example of the rib fracture, once the code for rib fracture is displayed, the screen also lists the fifth digit for all possible variations of rib fracture such as one rib broken, two ribs broken, sternum also broken, and so on.

The real power of the software comes in its ability to help identify complications. Working by hand, the principal diagnosis itself is so difficult to determine that there is a great temptation to stop there. The software greatly simplifies the process of selecting complications by reminding the user that no complications have yet been entered, determining whether each item entered qualifies as a complication, and suggesting possibilities. In many cases, these complications are items the person doing the coding manually may not consider.

Because Clinical Coding Expert clearly demonstrates on the screen other possible, alternative DRG assignments, this helps the coder review records to see if a secondary diagnosis, with a higher DRG reimbursement, could actually be the principal diagnosis. In the case of a patient who was hypoglycemic and also a newly diagnosed diabetic, for example, the software indicated the higher DRG which made a $200-$300 difference in reimbursement.

Clinical Coding Expert is an expert system containing over 6,000 rules that were derived and verified from analysis of 20 million case abstracts. These rules are applied against the case currently being coded. The current case is analyzed for clinical conditions that indicate whether a more detailed review of the chart is warranted to improve DRG reimbursement or to improve adherence to coding guidelines.

If additional review is determined, the system identifies specific complications, co-morbidities, secondary diagnoses and related procedures which may be present in the chart but have not been coded. As a result, the coder is prompted to review only those charts which have been screened for likely coding improvement instead of all charts. The list of suggested diagnoses and procedures is tailored to be clinically relevant to the chart being coded. Factors such as patient gender, age, diagnoses, procedures and discharge status are used to fine tune the data. The addition of a missing element is a simple matter of selecting it from the list shown. Any time the current chart is modified, a new set of rules is calculated. If clinically appropriate, new selection lists are presented and the process is repeated.

Since Lane County Hospital has been using Clinical Coding Expert, subsequent reviews of its coding have revealed almost no errors. Coding now requires only 10 hours a week instead of 20. Clinical Coding Expert met the hospital's expectations, improving accuracy and reducing coding time, and as a result the hospital is getting proper reimbursement for its services.

<< back to Articles