Author + information
- ↵*Reprint requests and correspondence: Dr. Paul T. Vaitkus, Cardiology Division, M/C 787, University of Illinois at Chicago, 840 South Wood St., Chicago, Illinois 60612 USA.
In manufacturing businesses and business schools throughout America, one of the prevailing tenets of management is that variance from a predefined norm is the source of avoidable costs, and its elimination is one of the keys to efficient operations. A widget that is larger or smaller than a defined standard is a source of consternation that must be eliminated. Leading businesses pride themselves in their fanatical devotion to the prevention and elimination of variance.
Angioplasties are not widgets
The same principles are now being examined and applied by those who manage health care provider and payer organizations. It has been well recognized that substantial variability exists in practice patterns both geographically and between individual practitioners and that this variability can translate into differences in cost (1,2). The elimination of variance is at the heart of utilization review as a management tool. Its goals are to identify and rectify practices that either exceed norms or go beyond the bounds of clinical guidelines. What has sometimes escaped the attention of
those who would stamp out variance in the practice of medicine is that the innate biologic variability among patients and the human dimensions of providing health care do not naturally lend themselves to efforts to achieve uniformity, much as in the manufacture of widgets.
The use of physician economic profiling has been examined in various clinical settings (3), and it comes as no surprise that these management practices have now reached into the cardiac catheterization laboratory. The conduct of percutaneous coronary interventions is capital-intensive. It involves expensive consumable items and personnel time. It is also readily apparent that much of this consumption comes in variable quantities at the hands of different interventional cardiologists. One interventionalist can, in a brief interval, directly consume large amounts of financial resources. Thus, it is natural that managers are interested in tracking and controlling behaviors in the catheterization laboratory.
Confronted with raw summary data of any aspect of clinical performance, be it clinical outcomes or economic costs, physicians are naturally and sensibly inclined to invoke the notion that their patients are more complicated than the patients of other physicians as an explanatory factor. What has been heretofore completely lacking in cost-accounting efforts in the catheterization laboratory is adequate control of case-mix complexity. Merely to enumerate the costs of cases or physicians, of which many hospital cost-accounting systems are capable, misses a basic point: complicated cases are more expensive (3–5). To be sure, there are commercially available cost-accounting systems that purport to be able to control for case-mix severity. With a single click of the mouse, a well-meaning administrator has the capability, ostensibly, to adjust physicians’ cost profiles for their case-mix severity. These systems’ clinical models, however, are not transparent. Neither the elements that are entered into the case-mix adjustments nor the validation data that support these models are readily available to would-be users; thus, from a physician’s perspective, they lack all credibility.
Upon this background, as presented in this issue of the Journal, Cowper et al. (6)at Duke University take the field a stride forward by showing us that appropriate profiling efforts must be based both on accurate cost-accounting and on appropriate clinical data. Over a two-year period, these investigators collected both economic and clinical data on a group of interventionalists. Their conclusion, which appeals to one’s common sense, is that adjusting for case-mix severity is important in correctly interpreting cost data. After adjustment, several operators who would initially have been identified as outliers no longer differed from the norm. Other operators, who in the unadjusted data-set looked like everyone else, subsequently emerged as outliers once the appropriate adjustments were performed.
Strengths and limitations of profiling and of the current model
The research methodologies of Cowper et al. (6)are sound. Without belaboring the minutia of cost-accounting and statistics, it will suffice to say that the cost data was appropriately tabulated and analyzed from clinical, managerial and statistical viewpoints.
The clinical data used for risk adjustment are fairly complete, though some potentially important elements may have been excluded. Angiographers will quickly recognize that lesion type can be an important driver of equipment utilization in the catheterization laboratory. Bifurcation lesions, for example, can be expected to be more expensive than simpler lesions. Though the issue of potentially incomplete or imperfect clinical modeling is a substantive one, we will not quibble about this. It is not likely that a more perfect model would change the overall results or conclusions reached by Cowper et al. (6).
The practice of the cardiologists at the institution studied was similar to national norms for such items as stents, other interventional modalities and adjunctive glycoprotein IIb-IIIa inhibitors. Thus, the cost dimensions of the Duke interventionalists may be expected to be representative of national patterns. Nevertheless, the cost structure of any one institution may not be applicable universally as different hospitals have different pricing and different fixed costs with which to deal.
One inherent limitation of all profiling tools is the issue of “temporal stability.” This is a relevant issue both for an individual operator’s performance and for the clinical model and the cost-accounting. To the former issue, an individual outlier can legitimately claim that his (or her) performance during a given time period is not representative as there may have been extraordinary cases or circumstances. Thus, to credibly label someone as a true outlier will take several periods of observations and a substantial numbers of cases. The reality is that it may take years of data before one can reach a firm conclusion that an individual is a true outlier. Like any measured phenomenon, catheterization-laboratory costs will likely demonstrate a regression toward the mean over time. Thus, even in the absence of any intervention to improve performance, an outlier during the time period studied by Cowper et al. (6)is likely to be within the normal range for the next time period and will expect a laudatory note from his or her administrator on the outlier’s improved performance.
Equally important is the idea that time marches on for the risk-adjustment models. As older technologies become less expensive in a competitive marketplace, and as newer, costly ones emerge to address specific clinical and angiographic problems, the predictors of what drives costs will change. Thus, the risk-adjustment models need continuous revalidation or recalibration. This will require much time and effort.
Overall, the present study (6)provides strong evidence to support risk adjustment for economic profiling, and physicians are within their rights in demanding it of any cost-accounting system that may be imposed upon them.
How can this risk-adjusted economic profiling be used?
For whom is this risk-adjustment tool intended? Payers have largely insulated themselves from the parochial issues of catheterization-laboratory costs by erecting diagnosis-related group (DRG)-based or capitated-payment systems. It is the minority of cardiology patients whose insurers reimburse hospitals on a per item basis. The worry of how efficiently the hospital runs is not the concern of Medicare or health maintenance organizations (HMOs).
Largely, the issue of containing costs at the level addressed by Cowper et al. (6)is of interest to hospitals. Self-contained systems, such as staff-model HMOs, that also own their hospitals, or the Veterans Affairs (VA) system, are uniquely focused on operational cost-containment, but in the aggregate these entities account for a minority of American medicine. For most hospitals, the revenues from an episode of care for a cardiovascular patient is fixed either by DRGs or capitation. Thus, cost containment becomes a matter of survival.
If it is the hospitals that have the greatest stake in controlling these sorts of operational costs, then it is the nature of their relationship to their physicians that will determine how they can put physician economic profiling to use. Hospitals that are in competition with nearby institutions to attract and retain private practice physicians and market share are not likely to strongly pressure their physicians on cost containment. Entities with greater control of their physicians, such as academic institutions with employee-physicians or those unusual institutions with geographic monopolies who do not need to worry about the loyalties of their catheterizing physicians, are positioned to more intensively monitor and manage their doctors.
Even a hospital with little leverage over its physician staff could find utility in a profiling tool. Conversely, even an institution that has potentially tight control of its physician staff may find its use of profiling limited.
Perhaps the most effective application of profiling is in physician-feedback efforts (7). Even in the absence of any administrative clout, these have been effective in reducing costs (7). In business schools, this is known as the “Hawthorne effect.” Shine a light on an issue and it will naturally improve. That which gets measured gets done. But whether efforts to control costs through the use of such profiling tools can go beyond physician feedback is highly dubious. It may, for example, be possible to ask an efficient colleague to coach an inefficient colleague, but one must account for the costs of the physician’s time in such an effort. More draconian initiatives, such as restricting physicians’ practices, are unlikely to succeed and will be far more costly in terms of intangibles such as loss of goodwill and possible legal redress on the part of the excluded physician.
In the final analysis, the usefulness of profiling may be very limited in terms of what is at stake. If the goal, like in manufacturing, is to reduce costs by reducing variance, the data of Cowper et al. (6)are illustrative of the limitations. Of the 50 operators, only 3 were outliers. We can estimate that their variance from the norm accounted for approximately $180,000 of the total variable catheterization-laboratory costs for the institution during the 18-month study period, or 1.5% of the catheterization-laboratory budget of $11.8 million. Examining the variability in total hospital costs, none of the providers were outliers, and thus there is no cost-savings opportunity in striving to eliminate variance.
Alternatively, if the goal is to prompt all the operators to become as efficient as the very efficient outliers, this may be an elusive goal, akin to goading the batters of a major-league baseball team to all hit 70 home runs like Mark McGwire. It may be that the two low-cost outliers are truly exceptional operators and all of the other highly trained, highly professional interventionalists will never equal their performance.
In conducting a program of physician profiling, the costs of appropriate risk adjustment and feedback must be balanced against the catheterization-laboratory costs to be saved. As Cowper et al. (6)underscore, the necessary clinical data to conduct appropriate case-mix adjustment does not exist in administrative data-sets. Someone has to gather, enter, and merge the clinical data with the administrative data. As more catheterization laboratories deploy automated systems for collecting clinical data (in particular to participate in regional or national databases) the merging of clinical and administrative data is simplified. But this raises issues about the accuracy of such efforts. For many clinical databases, the operators provide most of the key baseline information. If they perceive that this information is subsequently used to profile them, be it on clinical outcomes or economic dimensions, they will quickly learn to game the system by upgrading patients’ clinical or angiographic characteristics. This sort of “creep” in risk-factor reporting has been well recognized (8).
Other unintended consequences of profiling must also be considered. There are legitimate concerns that clinical-outcomes reporting can lead to refusal of operators to undertake complex cases (9). This could also apply to economic profiling.
The opportunities to contain costs in the catheterization laboratory are greater via approaches other than physician profiling. In contrast to profiling efforts, if the administrators could reduce costs by negotiating better contracts with vendors, the savings would apply to all operators’ work, not just the expensive outliers. In the present study, a 10% reduction in catheterization-laboratory hardware would have translated into a savings of $1.2 million, which dwarfs the $180,000 savings from “correcting” the outliers. The investigators admit, for example, that future decisions such as which glycoprotein IIb-IIIa inhibitor to use could have substantial impact on their costs. Also, concentrating their hardware inventory to one or several preferred vendors could increase the institution’s clout in purchasing contracts. Thus, seeking (and winning) the cooperation of the interventionalists in accepting a smaller inventory is likely to result in far greater rewards for the institution than are intensive efforts at physician profiling.
The goal the hospitals seek is cost-savings. Hospitals will not attain that goal (and will risk antagonizing their interventionalists) through profiling efforts, even efforts that are conducted with credibility and care such as the present Duke University effort. There are greater opportunities for the hospital via other cost-containment efforts in the catheterization laboratory, efforts achieved only through cooperation with the operating cardiologists.
↵* Editorials published in the Journal of the American College of Cardiologyreflect the views of the authors and do not necessarily represent the views of JACCor the American College of Cardiology.
- American College of Cardiology
- Cowper P.A.,
- Peterson E.D.,
- DeLong E.R.,
- et al.
- Omoigui N.A.,
- Miller D.P.,
- Brown K.J.,
- et al.