Here's how to root out internal variations in care
Before you can reduce variation, you need an understanding of your current practice. Janet Niles, RN, director of utilization and quality management at Carilion Health Plans in Roanoke, VA, uses a cost-based method to spot variation and then looks deeply into medical records to determine the causes. This occurs in a spirit of education, not as "economic profiling," Niles stresses.
Here are the steps she takes:
1. Identify your high-volume, high-cost procedures or diagnostic codes.
Clearly, you want to put your greatest effort where you can expect to make the most impact. To begin, you may want to list the top five or 10 diagnoses by volume and cost. By reviewing that list, you can choose the diagnosis that you think will reap the most immediate results, perhaps because of physician leadership or opportunity for change, Niles says.
2. Review the accuracy of your data.
You may be surprised to find aberrations in your data. But until you clean them up, you can't go on to the next step, says Niles.
"What we found was that there were claims processing issues," she says. "Location codes [showing where care occurred] were not being put in the way we thought they were put in."
Physicians are sometimes lax about coding procedures and secondary diagnoses. "They need to make sure they're putting all of the diagnosis codes on the bill," says Niles. "If you say your patient is sicker, but you're not [including] the code, then how can you prove it?"
You can check your data by comparing the claims information with the actual medical record on a sample number of cases. If you find problems with coding or claims processing, you will need to educate physicians and staff, Niles says.
3. Analyze cost per episode of care for a select diagnosis.
Niles calculates her average cost per episode by medical specialty and location of care. (See sample charts, p. 80.) "Most of the time physicians are surprised where the dollars actually are going," she says.
Next, she calculates the cost per episode for each physician with a substantial volume in that procedure or diagnostic area. She uses a severity adjustment tool built into the software to recalculate the numbers and discover physicians who differ significantly in cost from their colleagues.
Don't stop at the obvious
When you find a physician whose care involves significantly higher costs, even when the data are adjusted for severity of illness, your work has just begun, says Niles. You can't jump to conclusions.
4. Look for "unexpected services" as indicators of care differences.
One way to begin to analyze variation is to calculate the physicians' rate of "unexpected services." In other words, if a congestive heart failure patient usually has two chest X-rays, but Dr. A's patient has four, then he would have two "unexpected services."
Of course, there may be valid reasons for physicians to veer from the average. That will simply fuel your analysis. "Most of the time this leads you to ask more questions and more questions until you finally come to the answers," says Niles.
When you examine individual cases, you may discover that the data were skewed by certain unusual cases and your variation isn't as significant as you believed, says Niles.
"You have to look at what the physicians' patients look like and how they are moving through the continuum of care," she says. "What are their outcomes? A lot of times you will find someone who appears to be high cost because they're doing the more expensive test first to cut to the chase. Other patients may not do as well."
5. Design interventions to improve care based on best practices.
National benchmarks are readily available from the medical literature and national or international quality improvement projects. For example, the Institute for Healthcare Improvement in Boston conducts a "Breakthrough Series" with meetings in which clinical leaders discuss their route to best practices.
(For more on the Breakthrough Series, see Patient Satisfaction & Outcomes Management, June 1997, p. 65.)
Internal and external data go hand-in-hand, says Niles. "You can't learn anything from best practices unless you know what you're doing," she says. "You may already be a best practice and not be aware of it."