August 20, 2007

Digging for Data: Use the Right Tools
By Judy Sturgeon, CCS
For The Record
Vol. 19 No. 17 P. 6

How many times does someone call the coding office and ask, “What’s the DRG (diagnosis-related group) code for total hip replacement?” or “What’s the code for cancer of the whatever?” In most cases, the coder will simply answer exactly what is asked, and that
can sometimes be a point of concern. The coder’s job is to carefully and precisely translate diagnoses and procedures and provide the correct codes that result. Consequently, for most
cases, this is the perfect response.

Let’s review, however, what can happen if the caller is someone trying to set up a data request for research purposes. These requesters are generally not coders. They may be scientists, programmers, or data hounds from marketing and pharmaceutical research. What they have in common is that they think they know what they want, and that’s what they will ask of the coder. The researcher gets what he or she asked for, is happy, and heads off to start reporting data generated by the code. The coder goes back to work, happy for having been able to help the caller.

Return to the part where the requester thinks they know what they want when asking for a code or a DRG. If the coder stops to ask, “For what reason will you be using the code(s)? I’d like to make sure you get everything you need,” it may expand the need to a whole range of information. The inquiring mind on the other end of the line may not need a DRG at all but a CPT or ICD-9 code. They could even be making erroneous assumptions about the use and format of those codes and subsequently report data or arrive at decisions that are equally incorrect.

For example, imagine that a caller asks the coding office for the DRG for total hip replacement. Let’s say the coder gives the most likely DRG. Suppose the caller is planning to research total hip arthroplasties for the orthopedics department. Armed with the coder’s response (DRG 544 as of July 2007), he or she gets a report of the total number of patients with DRG 544 in the past five years. Then he or she gets the total charges for each, adds them up to calculate an average cost, and plans the budget for next year’s surgeries based on the results. Would the figures be correct?

The answer is no because this DRG changed in 2005 and, along with DRG 545, would have been reported as DRG 209 prior to discharges as of or before October 1, 2007. If any patient had bilateral hip replacement in one admission, then the DRG for all five years would be DRG 471. But it gets worse: The DRGs in question also include items such as reattachment of a foot, total ankle replacement, and resurfacing of the femoral head. Since a DRG is based primarily on the reason for admission, if a patient was admitted for other problems but still had a hip replacement while in the hospital, it could result in numerous and altogether different DRGs.

The researcher’s data is therefore based on too many patients in some DRGs, incorrect DRGs for some years, and other possible missed DRGs. And it’s not only the population that’s in error: The average costs were based on erroneous data, making the financial expectations well out of range, too.

If the coder had known initially what the requester was trying to accomplish and for what period of time the information was needed, the response should have been quite different. Running the report based on the specific procedure code for total hip arthroplasty instead of the most common DRG would isolate the exact type of surgery and include correct populations and costs. Unlike DRGs, the codes had not changed for the previous five years, and subsequent reports would have included the correct patients while excluding ankle repairs, amputees, and revisions. This is not always the solution; sometimes both ICD-9 and CPT codes change for the times needed. Sometimes, it is the DRG that is the only constant and all will vary by the years in question.

If the caller is someone who needs a code for diabetes to put on an insurance form, the coder will know to keep it simple. If the requester needs to determine life expectancy, readmission statistics, or surgical histories of patients with a particular diagnosis or procedure, the answer can get increasingly complex. A code for “cancer of the whatever” and a patient who “had a biopsy” will not be covered with a single diagnosis code or one ICD-9 or CPT code.

Consider the following questions that may need to be answered in a scenario such as the one previously stated in order to provide the researcher with the appropriate data:

1. Is the patient in question likely to be inpatient, day surgery, emergency department, or any of the above? Will research be done on hospital or physician data or both?

2. Do you only want patients whose primary reason for care was cancer of (specify the type here)?

3. Do you want to include the patient if he or she was admitted for something besides cancer but is still having the cancer treated?

4. Do you want to include patients who have had the cancer surgically removed but may be getting treatment for metastatic sites?

5. If you need codes for procedures, will they be done as inpatient or outpatient?

6. Do you need a specific range of ages or dates of service?

Why does this matter just to get codes for people with a “cancer of the whatever” who had a biopsy?

1. Coding rules for the various patient statuses vary. Some may only code the symptom, while others will code the most probable cause. Some will get codes for everything that affected the stay, and some will get only codes for what the physician tended to personally during that encounter.

2. The primary reason for care is usually the first diagnosis listed. Sometimes, however, a physician specialty may list a primary cancer as the first diagnosis, but the hospital may code the metastatic site as primary if the patient has both and most of the care was devoted to the secondary site.

3. and 4. A patient with the surgical removal of a breast tumor may be admitted for gallstones, but during the cholecystectomy, a biopsy of a suspicious colon lesion shows recurrence of the cancer. The patient will not have a code for current breast cancer. The biopsy may be a simple sampling, an excision of a tumor, or a resection of part of the colon. Both procedures may be laparoscopic or open. Not only does the diagnosis have to be well-defined, so does the definition of biopsy. The principal diagnosis will be gallstones, and the DRG will be determined by the types of procedures performed.

5. and 6. Codes, and even types of codes, will vary by the status and dates of service.

How will the diligent researcher get the correct information from the coding office? Communication is the key for the coder and the requester. If you are the person taking the call, ask for more information so you can best provide the correct codes—the six questions listed are a good sampling. The office can have a designated research request coder if there’s a specific talent and interest. Perhaps it would be more beneficial to have everyone be adept at handling this type of customer.

When asking for codes or DRGs, start by summarizing what you want to do with the information. Be patient when the coder asks what seems like 20 questions before you get an answer that you can take to the data bank. Your responses will further refine their responses so you’ll get exactly and only what you need.

In either case, working together can create an important partnership for both groups. The coding office gains credibility when the data reported from “their” codes is accurate and trustworthy, and the person reporting the data gets exactly the same benefit. Take the extra time, ask the extra questions, and understand that a skilled coder is one of the most valuable tools any researcher can have in a data-mining operation.

— Judy Sturgeon, CCS, is the hospital coding senior manager at The University of Texas Medical Branch in Galveston and a contributing editor at For The Record.