February 2008
Feature Story

Karen Titus

Getting a fix on lab test utilization is as simple as pinning down a hologram. You see it—or at least you think you do. But move around a bit, and you’re no longer sure. Take a step back—maybe it’s nothing. Reach out and try to touch its contours, and you’ll come away empty–handed.

But what laboratory doesn’t want to stop the flow of unneeded, unwanted, sometimes costly, occasionally downright silly test requests? Who doesn’t want to make sure the right tests are being ordered? The idea of better test utilization is so beguiling, it’s hard to give up the ghost completely.

Kent Lewandrowski, MD, has been unraveling utilization for years. About eight years ago, he and his colleagues at Massachusetts General Hospital knew exactly which threads to yank: those leading to specific tests that they deemed to be of little or no clinical value, whose use could be linked to specific subspecialties. Bleeding times. Five-hour oral glucose tolerance. Certain tumor markers. Back then, Dr. Lewandrowski would approach the hospital clinical practice management committee, tee up his proposals, gain approval, and put them into practice.

“That was very much going from one foxhole to the next, picking off easy opportunities,” recalls Dr. Lewandrowski, associate chief of pathology and director of clinical services (anatomic and clinical), MGH, and associate professor, Harvard Medical School.

Soon other threads emerged. These stretched a little longer and were linked to systemic problems. So Dr. Lewandrowski and colleagues began to establish algorithms for test ordering, the typical thyroid screening algorithm being the classic example. Other algorithms took aim at selected situations, particularly in coagulation and specific microbiological areas of the laboratory.

Eventually, however, Dr. Lewandrowski found himself tugging on a big knot. “We ultimately realized that there was another fundamental problem,” he says. The first cases were lower-volume tests that lent themselves to easy cost accounting—he and his colleagues could identify a test they had halted, then point to the dollars the hospital was saving. “You keep rolling up all the dollars that you saved—$100,000, $200,000, $300,000,” he says.

Meanwhile, total test volumes were expanding, chiefly because of rising orders for routine tests such as basic metabolic panels, calciums-phosphorus-magnesiums, and the like. But eliminating these tests, despite their large numbers, wouldn’t necessarily save much money. “There was one study by Winkelman years ago [Winkelman JW. Clin Lab Med. 1985;5:635–651] that said to get a 10 percent reduction in cost in the laboratory with automated testing, you need a 50 percent reduction in volume,” Dr. Lewandrowski says. Pulling tests off at the margins saves only incremental, variable costs; fixed costs—obviously a significant component of the laboratory—remain steady. So Dr. Lewandrowski never gave much thought to cutting back on the routine tests.

Slowly their thinking began to change. Maybe money wasn’t the only reason to change test utilization. “Maybe it’s also about the volume of testing that is flooding the laboratory,” he says.

No doubt it is. But as Dr. Lewandrowski and others spend more time untangling the knot, they’re starting to fathom how big and how snarled it is. Looking into test utilization has brought them face to face with medical and payment practices that stretch back 40 years. Do labs really want to pull at those threads?

Dr. Lewandrowski and his colleagues, inside and outside the laboratory, remain unfazed. Their focus, for now, is on inpatient testing, which accounts for 60 to 65 percent of the laboratory’s volume. This lets Dr. Lewandrowski “exercise command and control,” as he puts it, versus juggling multiple off-site clinics and practice managers, with whom he doesn’t have regular contact. Reimbursement nuances are another reason for tackling inpatient testing—under DRGs, the lab gets paid a fixed amount for a given hospital event, regardless of the amount of testing done; outpatient testing, on the other hand, has a revenue stream associated with its costs.

Their guiding hypothesis is simple: Doing useless tests takes time away from useful ones. “If you’re flooded with basic metabolic panels and calcium-phosph-magnesiums and CBCs, that detracts from your ability to provide rapid services that are necessary for patient management,” Dr. Lewandrowski says.

The problem mushrooms from there, thanks to the statistical method of determining normal reference ranges. With most high-volume automated tests, the normal range is defined as the mean plus or minus two standard deviations. That means statistically five percent of all tests will be abnormal, simply because of how the normal range is defined. By reducing overall tests, the number of false abnormal results would also drop, reducing subsequent workups, or “clinical adventures,” as Dr. Lewandrowski puts it. “This excess testing affects the entire system, from making the lab busy to making the doctors rat race around chasing after laboratory values.”

True, not every test will launch exhaustive workups. Many physicians will discount abnormal results, or at least those few slightly elevated results, says Mark Lifshitz, MD, director of clinical laboratories, New York University Medical Center. “But they may not. Everyone likes to err on the side of completeness with diagnostic workups, because everyone’s afraid they’re going to get sued.”

And so the issue moves beyond simple cost savings. “The real question is, what’s the impact of an inappropriate test?” Dr. Lifshitz says. Direct financial impact—the obvious cost of doing the test in the lab—is often the smallest one. The indirect impact can be much greater, especially in the aforementioned scenario. Add in the burdens of repeat testing, referrals to specialists, other interventions (potentially with associated morbidity), and patient angst, and suddenly that first, inexpensive test doesn’t look so cheap and easy anymore.

The solution can be summed up quite easily in words if not in practice. “If you don’t need the test, maybe you shouldn’t order the test,” Dr. Lifshitz says.

Granted, one person’s necessary test is another person’s unnecessary test. Labs are caught in another crosscurrent as well: While labs want physicians to think more, physicians want labs to help them think less.

From the busy physician’s perspective, less thinking makes sense, Dr. Lewandrowski says. “They try to set as many things on autopilot as they can from the start. So they say, ‘Look, I don’t want to have to go through all my patients every day and figure out whether I need a calcium or a basic metabolic—I’ll just get them.’”

And why not? For the clinician, it’s easy to order plenty of tests—they’re showing up in the system automatically, often as part of panels compiled three or four decades ago, when vendors were developing automated chemistry equipment. Though these panels have earned the approbation of groups like the American Medical Association and Medicare, they may bear little resemblance to the actual testing needs of individual patients. This gives rise to the not uncommon scenario of physicians ordering, say, blood glucose tests on nonfasting patients. “Why is that test even in the panel? I don’t know,” says Dr. Lifshitz, who then provides a morsel of Zen clarity: “It’s in the panel because it’s in the panel.”

One box on a computer screen puts 12, 14, 20 tests at a clinician’s fingertips, which can be an irresistible temptation. “You don’t have to think about anything,” Dr. Lifshitz says. “When you realize how many patients physicians churn through their practices, they don’t have a lot of time to think. It’s almost easier to order it than not to order it.”

This has forged some deeply rooted habits, says Dr. Lewandrowski. “There are certain departments where house officers have a voracious appetite for testing, and the culture of the house officers is to order daily labs,” typically a basic metabolic panel with calcium-phosphorus-magnesium; a CBC; and not infrequently a PT/PTT. “Those are the ones they just seem to order on every patient, every single day.

“If I were a patient, I would be distraught if the phlebotomist showed up and I had to get stuck every morning. That’s another reason I’m passionate about this—for the poor patient who’s being stuck,” Dr. Lewandrowski continues, adding that in long-term patients this can lead to therapeutic anemia.

Recalling the early days in his career when he cared for patients directly, he doesn’t necessarily fault time-strapped clinicians. What looks selfish or careless to the lab can be a survival mechanism for exhausted physicians, he says—to a doctor who hasn’t slept in 36 hours and is waiting to clear patients for handover to the next shift, it makes sense to order everything on everyone, stat.

In pursuing utilization problems, Dr. Lewandrowski and his colleagues have also traced some of their difficulties to the order entry system, which at MGH is not connected to the laboratory information system. That makes physicians dependent on paper requisitions. The disconnect, says Dr. Lewandrowski, leads to faulty interpretations and more tests being ordered. When a clinician orders electrolytes (which do not appear on the requisition), the order might be translated by a nurse or operations assistant into a request for a basic metabolic panel, which does appear on the requisition and includes electrolytes but also BUN, creatinine, glucose, and calcium.

Another disconnect occurs in the intensive care units, where nurses are accustomed to ordering tests preemptively, knowing that patients coming into their units usually get blood gases, electrolytes, CBCs, and PT/PTTs. While the nurses’ intentions are good, Dr. Lewandrowski says, he estimates that in some ICUs, 30 to 40 percent of the tests the laboratory receives are never ordered by a physician.

Having identified problem areas, it was now time for Dr. Lewandrowski to go after them.

Starting with the order entry system, he and his colleagues selected tests “that we felt were ridiculously over-ordered,” largely because they greeted clinicians on the ordering screen. “They’re going ‘click, click, click, click, click’ with their little mouse, and they click on the tests you present.” Case in point: CPK-MB popped up in the provider order entry system, although the lab was trying to persuade physicians to rely on a troponin-based rule-out MI protocol. Taking the unnecessary test off the screen cut utilization by 50 percent. Numbers fell more when clinical physician guidelines and scenarios were added, says Arjun C. Rao, MBBS, senior consultant at MGH’s Center for Quality & Safety, and assistant, Health Decision Sciences, General Internal Medicine Unit.

The next step was to build rules into the order entry system, making it harder for a clinician to order, for example, a daily basic metabolic panel. Though physicians could still order an approved set of daily lab tests in specific situations, they were limited to q. day ∞ 3, versus an indefinite order. That too produced a significant decrease in volume, Dr. Lewandrowski says.

The net result of their maneuvers was a 16 to 18 percent reduction in inpatient tests per discharge. The numbers plateaued, however, as the house officers developed workarounds, such as reverting to write-in orders to duck the q. day limitation.

The lab then did what laboratories do in these cases: held seminars, worked through patient cases, and provided practice guidelines. The clinicians responded the way clinicians respond: They changed their behavior briefly, then returned to their old ways.

“We made a dent, but it was not a sustainable dent,” says Dr. Rao. He says it’s partly the nature of the beast, that it’s not easy to disseminate information among residents. “We have some attendings trying to educate them, and they believe it in theory. But because of the sheer volume of work on the floors, it’s hard for them to keep track of the residents and make sure they don’t order these tests on an ongoing basis.” Plus, the medical service has a high turnover of residents. MGH has 100 to 120 new interns a year, making it even harder to reach everyone. “We are moving in the right direction. It is very, very slow,” Dr. Rao says.

Now the lab is taking another run at it. One tactic is to work with the medicine juniors, meeting with them regularly to show them their test volumes. The goal is to budge the culture a bit, influencing the interns who, of course, will be next year’s juniors.

Later this year, a pilot program managed by Anand Dighe, MD, PhD, director of the MGH core lab, will link the provider order entry system with the LIS. If successful, it could end the interpretive adventures launched by paper requisitions. And since only physicians will be able to routinely order tests, it should curb the preemptive thoroughness of nurses. Finally, the lab is going after the q. days again, placing blocks on orders and forcing house officers to explain their requests. Inconvenience and auditability may be the twin slayers of a beast that has stalked the lab for years.

The redesigned computerized lab order entry process will also provide clinicians with information about lab test utility, turnaround time, cost, past results, and alternatives. All the information the lab knows about the test and the patient will be brought to bear on the provider’s decision whether to order a particular test, Dr. Dighe says. “We consider the two minutes that a provider sits down and orders lab tests to be a key educational moment for us.”

Information technology may be the best tool pathologists have for controlling utilization, says Dr. Lifshitz. It can provide real-time feedback to physicians when they order tests. It can propose ordering pathways for specific diagnoses and spot duplicate test orders. It can also restrict ordering based on specialty. At NYU, for example, BNP ordering is limited to emergency room physicians and cardiologists.

And, labs can simply keep tests off the ordering screen. At NYU, the lab offers only basic metabolic and hepatic panels for inpatients. “If I offered the comprehensive instead of the basic, everyone would be ordering that instead,” Dr. Lifshitz says with a laugh. “They can order a panel, but at least it’s a smaller one.”

At MGH, the ICUs have already been a success story. The first stop was the surgical ICU. The lab and SICU worked together to create new ordering guidelines (see box, page 84), trying to limit preemptive orders. While RNs can still order stats, they are now required to get the physician’s order retroactively. The result has been a 38 percent reduction in SICU test volume—some 10,000 tests a month, Dr. Lewandrowski estimates (see chart, page 78). He hopes to replicate this in the medical ICU, where a similar project was set to launch in January. From there he hopes to extend the effort to the cardiac ICU and other units.

Gone are the days of easy pickings. Now Dr. Lewandrowski sees high-volume automated testing as the big game to go after; the big gun is the order entry system, which should change the culture by identifying test ordering patterns of individual physicians. He anticipates a 15 to 20 percent improvement, and he expects to get about half that within the next 18 months.

There’s nothing sneaky about what he and his colleagues are doing. “We don’t try to be tricky,” Dr. Lewandrowski says. “Because if you have one situation where you’ve committed some egregious crime, they’ll never trust you again.” He tries to involve everyone, from nurses and lab colleagues to surgeons and house staff, and he and Dr. Rao lead the clinical labs advisory committee, from which they regularly seek advice. “If you try to just do it from your office, and put out decrees, you’ll fail.”

This is tricky work, nonetheless. In essence, the lab is telling fellow physicians that they’re doing something wrong. Or maybe they’re doing something wrong. Frankly, it’s not always easy to tell.

Well-founded hunches can start with a simple review of monthly lab stats. That’s how Dr. Lifshitz discovered that a cardiology practice was suddenly ordering huge numbers of C-reactive proteins when they meant to order hs-CRP. That’s also how he flattened a bump in troponin T send-outs. His lab offers troponin I; the physician placing the orders for T was interested in troponin, but only knew to ask for cTnT; the clerks evaluating the orders didn’t realize cTnI was an acceptable substitute.

Dr. Lifshitz has spotted other problems, however, only to find, when he picks up the phone to talk to the ordering physician, that a new protocol has been put in place, or that a spike in tests is the result of a temporary research protocol, and the “problem” will soon slip away as quietly as it arrived, like the Loch Ness monster.

If the problem is indeed a problem, the solutions can be just as elusive. In trying to reduce the daily labs, says Dr. Rao, “we know the impact on pathology, but we have not yet cracked into the system to put our finger exactly on what’s happening on the other side.”

In fact, Dr. Lewandrowski says, one of the surprises for him, in tackling utilization, is realizing how little he actually knows, despite his years of training and experience. “It’s true. You think you know a lot, because you talk to nurses and physicians and others. Suddenly you realize you don’t know as much as you think you do. I’ve found this even in areas where I think I have some subspecialty knowledge, how often it comes up that I really don’t know what I’m talking about. Some of my ideas are wacko. And they’ll tell me, ‘That’s stupid, Kent.’”

More often his clinical colleagues agree with his ideas—but say they don’t have time to take it on as a project. But the lab does, says Dr. Lewandrowski, and it should. After all, the lab knows what’s being ordered, in what volumes, and where.

It also knows when improving utilization will help. Someone with less savvy may cook up a scheme to drop chloride out of the electrolyte panel, knowing it’s the least useful test on the panel and would mean a 25 percent drop in test volumes. But, as Dr. Lewandrowski points out, the savings—in time or money—would be nil, since the blood is still drawn and the sample accessioned, processed, and loaded on the instrument. “All you save is the incremental cost of the chloride.” Better, he says, to eliminate tubes, not tests—that’s the real way to take cost and effort out of the system.

Dr. Lifshitz agrees. Knocking tests off a panel may save only cents per sample. Logic might suggest pitching panels, which by now seem to have all the attributes of a weed. But that’s a battle no sane laboratory would choose.

“I have no doubt in my mind that there are far more panels ordered than there need to be,” Dr. Lifshitz says. But other than looking at their inappropriate placement in specific clinical pathways, or determining whether they’re used in standing orders by specific groups of physicians, it’s difficult to get the best of panels. Nor is it worth the effort. “You want to pick and choose the battle of the $20 test or the $50 test, not the five-cent test,” he says, “unless there’s some generic way where you can eliminate huge amounts of panels, tens of thousands or hundreds of thousands of tests—and I’m not sure I know how to do that. I’d be the first person in the country to do it if I could figure out how.”

An obvious source of help—evidence-based guidelines—won’t help with panels either. They work well for specific tests and specific conditions—a cardiac marker pathway, say, or an algorithm for evaluating celiac disease. In such cases, Dr. Lewandrowski would meet with the specialists involved and start sounding out ideas for streamlining diagnostic workups. “But there really is no evidence on daily labs, and there are no papers on how often you should order a CBC on your average inpatient medicine case.”

All of which leaves labs back where they started, pulling at threads, with no certainty of where they’ll lead. “In the end, all you know for sure is the order came into the lab,” says Dr. Lifshitz.