A cooperative strategy for curbing test abuse.

AS far back as 1976--before we ever heard of DRGs--I was asked to look into laboratory utilization at our 725-bed hospital.

Looking at published studies about lab overuse, and seeing lab workload grow by 10 to 15 per cent a year while patient days stayed level, made us concerned that clinicians might be overordering tests. Did the house staff lack information that the lab could provide? Could we help quell rising public unrest about health care costs?

Studying laboratory utilization proved more difficult than I imagined. Several years of investigation now have taught us, at the very least, that problems don't always respond to seemingly obvious cures; that the effects of fixes should be carefully considered in advance; and that the lab and clinical staff must work together when assessing potential problems and attempting change. HEre then, are some of our hard-learned lessons.

Cost containment pressures and changes in reimbursement practices have made it easier to promote and implement modifications in test ordering procedures. But there is a great danger of overusing the very tactics that combat lab overuse. Interventions may not always be in patients' best interests. We are at a cross-roads in health care today. Before attempting major change, we should listen, look, and perhaps stop as the DRG express moves determinedly towards us.

Listen. The laboratory staff, clinicians, and administrators bring vital viewpoint to the issue of utilization. They also have unique and sometimes conflicting needs: The laboratory wants clinical information that isn't always obtainable; clinicians request unresonable turnaround from the lab; and administration tries to curb overall resource use. The first step is to hear out the other sides, understand their perspectives, and make goals complementary, not divisive. Avoid narrow-minded thinking at all costs.

Why is listening so important? Medical decisions must usually be made in the face of incomplete knowledge, irreducible uncertainty, and an unmeasurable personal value system applied by the physician on the patient's behalf. Consequently, criteria for determining "good" or "bad" laboratory use arise, at least in part, from personal value systems, and therefore are somewhat subjective. Judge others in their practice of medicine by your subjective criteria, and they will probably reject you--unless your guidelines are either ironclad, which is impossible because they are value-laden, or identical to theirs.

In other words, changes in lab utilization should arise from consensus. Both clinical and laboratory reasoning must be heard if subjective criteria are to be agreed on.

If the clinical laboratory test were like a major surgical procedure, it would be easier to establish criteria for its use. But the lab is only one cog in the wheel of patient care, even though it contributes so much. Other cogs--nursing service, surgery, clinical staff, pharmacy, radiology, and so forth--may control patients' outcomes much more directly.

So it's usually difficult to link patient outcome to laboratory use (Figure I). Tests used to exclude disease, for example, have no apparent effect on the patient even though they strengthen the clinician's pre-existing judgment.

In addition, the usual patient care setting is not conductive to experimental investigation of proper and improper use. Such experiments can't be piggbybacked onto the daily patient care process. Each patient presents with a unique constellation of problems and responses. Ethically, we cannot exert experimental control and probe the key question in utilization decisions: What will happen if we don't run this test?

The elusiveness of clinical reasoning has been well studied, but it still isn't well understood. We know clinicians don't always employ lab tests optimally. They don't take into account important probabilistic concepts such as prevalence of disease, diagnostic sensitivities, specificity, and positive or negative predictive values.

But thinking in statistical terms is not a nautral or intuitive process. Research suggests that the very structure of our mental machinery often prevents us from solving such problems optimally. Studies have shown health professionals to be unreliable probabilistic thinkers.

Human limitations in thinking can therefore lead clinicians to order redundant, wrong, or unlikely tests and not make the best use of available information. In spite of inherent cognitive limitations, and resulting suboptimal thinking, the physician problem-solver generally performs quite well on behalf of the patient. Those who set criteria with optimality as the target--and fail to recognize human thinking constraints or the difficulty of pragmatically applying probabilistic concepts to lab testing--do not serve anyone well.

Except when frankly misapplied, most tests lie on a spectrum of possible payback, even if very small. They're ordered in good faith and with reasonably valid incentive.

Screening tests are a good example. Their value to the general patient population is small, but they are essential to the few in whom they signal disease. Likewise, normally unremarkable chemistry profiles ordered during routine physical examinations occasionally hint at underlying health problems.

In the past, it was reasonable to order a test' as long as it had a shadow of probable worth. Today, in the face of a mounting bill for health care, it appears that society may have reached a limit in its willingness to pay. Clinicians, accustomed to acting and thinking solely on behalf of their patients, now face the dilemma of making judgments under societal pressures, too. Providing expensive or excessive technology for one patient may mean that another's needs are ignored. Clinicians are being forced to weigh more carefully the potential value of a test against its cost, thereby moving up the probable worth scale.

The tradeoff hinges on the value of the information a test might provide and the value of the benefit we may forgo. The threshold is not uniform for all tests or for all patients. In some settings it may be appropriate to eliminate the expensive, time-consuming WBC differential as an admission screening test in the absence of clinical indications or suspicions, because the loss of useful information is negligibel. On the other hand, the potential loss to the patient caused by not applying some tests, such as the neonatal screen for hypothyroidism, is too great even though it has a low yield of positive results.

In summary, decisions about lab use are based on irreducible uncertainty, are applied to a complex but limited world, and ware value-laden. By listening we can consider all the uncertainty, subjectivity, and limitations that constitute the real world of medical practice. It allows us to set utilization criteria by consensus and thereby gain support for and commitment to improving laboratory use.

Look. The hospital is an institution of complex interrelationships. A prudent systems analyst would hesitate to tinker with its complexisty without first understanding it--just as a technician would not interfere with a delicate mechanical system before fully comprehending its workings.

We have to exercise the same restraint in modifying laboratory testing practices. We must judge as best we can the effects--negative and positive--that change will hae on a system. This requires careful examination of the situation.

Looking isn't always easy. Nor is it always welcome, especially when consensus is lacking. Sharp systems analysis spots weaknesses in department organization and administration, such as slow turnaround time, and in the structure of health care delivery. It may step on the toes of highly territorial managers or threaten the incomes of the hospital or its feefor-service staff members.

Looking can be costly and time-consuming. To look at patterns of laboratory use, we need to compare many cases by disesae severity, coincidental disease, and physcian's specialty. furthermore, fragmented paper record systems still prevail at most institutions and make this investigation difficult. ECG reports, laboratory results, and chart notes, to name a few, must first be gathered before we can abstrat valuable conclusions from them.

Our hospital has tackled this problem by using computers to amass scattered data. The hospital has no overall integrated information system, but relies on several modular systems. We're using four of them for a current study of platelet tranfusion practices: the blood bank computer for information about platelet transfusions; the chemistry and hematology sections' computer for data, such as platelt counts and BUN results, that explain why platelet transfusions are requested; the administration computer for demographic data such as admissions, lengths of stay, diagnoses, and surgical procedures; and a fourth, general purpose scientific computer that has been used to aggregate the other information into a coherent data base.

AT this exploratory stage, new questions arise for every answer we obtain. With data funneled to a central source, we're discovering how many platelet units each service orders, which ones transfuse, at higher pretransfusion platelet counts than others, which diseases demand the most platelets, and so forth. For example, we've found a surprisingly high number of single-unit transfusions. Eventually, we'll have the clear patterns of data needed to attempt transfusion-practice changes.

Computers are no panacea for the hard job of looking carefully at laboratory utilization. It's difficult to extract the desired data from laboratory information systems that aren't designed to answer our multifaceted questions. We also have the usual troubles that arise when trying to combine data across different types of computers. As well integrated hospitalwide information systems become more the norm, these problems should be much less of a hindrance.

If we establish a beachhead of reasonable, well accepted, consensus-formed criteria for transfusion practices, implementing reform will put significant new burdens on laboratory and hospitalwide information systems. Ordering must be monitored as it happens, so that attention can be promptly directed to practices that are not in agreement with the criteria.

Stop. When you've identified problems and assessed the impact of change, you may want to draw up a plan to modify practice. Imposed changes in clinical practice have never been simple or long-lasting, but today's incentives to cut medical costs are stronger than ever.

Implementation usually presents a choice between education and enforcement, and the former hasn't been very successful. (See "A Clinciian's View of Laboratory Utilization" on page 51.)

One California hospital trained an eye on thyroid function and cardiac enzyme tests because they were clearly being overused. The investigating team simply changed the order form for thyroid function tests by offering clinically relevalnt test groups rather than a wide open menu. This step discouraged clinicians from ordering comprhensive but inefficient panels. REquests for thyrotropin (TSH) and T4 (RIA) tests dropped dramatically. On the other hand, when the team used education, such as house staff meetings and individual conferences; to alter ordering patterns for CK and LDH isoenzyme tests, chance was minimal.

The lesson: While education may facilitate other measures directed to improving lab use, it has little effect by itself.

It's difficult to replace old informal policies with new formla policies. Here's an example of the tenacity of old policies: One neighboring hospital had reported no positive CFS cultures for tuberculosis in the last 10 years, yet 70 per cent of CSF specimens were cultured for tuberculosis. In this geographical area, tuberculous meningitis has a very low prevalence.

We searched for the source of this ordering practice. When we asked the house staff, "Do you order TB cultures on CSF routinely?" all of the neurology staff and two-thirds of the medicine staff said yes. Lab records supported their comments.

This overordering, however, didn't correlate at all with medical faculty opinions; only 21 per cent of the faculty advocated the practice. The pattern of high use suggested that routine TB culture on CSF specimens was a long-standing informatl policy among the house staff. Note that this practice has some underlying validity: tuberculosis can elude diagnosis by clinical findings alone, and the cost of missing a case is very high. Clinicians may want to spare patients the need for a second lumbar puncture if tuberculous meningitis later becomes part of the differentical diagnosis.

Those proposing to replace old informal policies with more carefully developed formal policies may find old practices hard to dislodge because they have underlying validity. Physicians must be convinced that the new approach is better before a change in practice can be expected.

Once change is indicated, don't stop listening. Consensus remains important in developing strategies to change practices. Don't stop looking, either; attempts to modify practice can fail. When they do, change or scrap them.

Listen, look, and stop is not a magic formula for effective change. It doesn't cover many of the variables you'll encounter in investigating improper lab utilization. But it does offer a starting point and underlines the importance of consensus and laboratory-medical staff cooperation.

Working with these guidelines, we've documented for the first time a commitment to better lab usage by all services.