Since the 1970s, the number of clinically active specialists in the United States per 100,000 population has more than doubled. For example, the number of orthopaedic surgeons increased from 3.6 in 1970 to 7.1 currently (Fig. 1). The increase in the supply of physicians has been associated with an increase in per capita costs and in utilization of specialist care. As a result of this escalation in health-care spending, the federal government has called for a dramatic change in the way in which they support graduate medical education. In this new scenario, institutions would be rewarded for reducing the number of residency trainees3.
However, it is necessary to determine how many physicians are needed. In order to do this prudently, the organizations that are responsible for the education and training of future orthopaedists commissioned a study by the RAND Corporation4. The results of this study, which appear in this issue of The Journal, indicate a current excess of 3546 orthopaedic surgeons, which will increase to 4122 by the year 2010.
The RAND researchers4 obtained these estimates of excess capacity by applying what they characterize as a demand-based method to estimate workforce requirements. According to their definition, demand is the same as the average utilization rate for the nation. They estimated this rate for services provided by orthopaedic surgeons by using several databases that, taken together, cover the major sectors in which orthopaedic care is provided: physicians' offices, ambulatory care centers, and hospitals. The number of full-time-equivalent orthopaedists required to provide services for the nation was then calculated on the basis of the rate of service delivery weighted by the average time spent providing care, as estimated with use of a national survey of practicing orthopaedic surgeons. In a final step, the researchers translated the person-hours required to accomplish the national orthopaedic workload into an estimate of the number of full-time-equivalent orthopaedists required to do the work. To take this final step, the researchers needed to estimate how many hours a full-time-equivalent orthopaedic surgeon ought to work in a year. Adopting the figure used in a study conducted by the Graduate Medical Education National Advisory Committee6, they estimated that a full-time-equivalent orthopaedist spends an average of 2220 hours providing direct patient care annually.
Projecting workforce requirements is hazardous duty, and RAND should get its fair share of recognition. By translating utilization into the time required to provide care and by calculating these estimates on a condition-specific basis, the study breaks new ground. Information on the proportion of workload allocated to specific tasks may be useful for a number of purposes. However, with regard to its strategy for determining the so-called right rate in order to arrive at the number of physicians that is required (and thereby end the filibuster regarding workforce policy), the study exhibits the same weakness as other workforce-planning efforts that have been based on projections of utilization rates. In the jargon of economics, the method assumes that demand in health-care markets is exogenous and that utilization is determined by illness rates and access to care, independent of the systems that provide the care, and therefore can be used to project supply needs.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. No funds were received in support of this commentary.
The findings of small-area-variation studies contradict the assumption that supply and utilization are explained on the basis of population demand. There are striking differences in the distribution of the workforce that are not explained by illness rates or the ability to pay for care. For example, the current number of orthopaedists in Ann Arbor, Michigan is 4.9 per 100,000 population, while in Seattle, Washington, it is 8.8. The average for the United States is 7.1 per 100,000 (Fig. 1). Variation in the rates of services provided to age and gender-adjusted communities with a similar number of orthopaedists is an example of this phenomenon. One wonders why the age-adjusted rates of operative procedures on the knee and the back (6.40 and 3.71 per 1000, respectively) in Sun City, Arizona are double those in Miami Beach, Florida (2.63 and 1.63 per 1000, respectively)7. (Sun City has 12.1 orthopaedists per 100,000 compared with 7.61 in Miami Beach7.) It may be an example of so-called supply-induced demand or it may simply be a difference in practice style: what some call the surgical signature of a community7 (Fig. 2).
The model that patient need determines the level of utilization in a community is true for only a few medical conditions. One example is fracture of the femoral neck or intertrochanteric fracture, for which rates of hospitalization vary relatively little among regions. The utilization data fit the assumption well that demand is exogenous, as determined by the incidence of fractures. However, this is so only because of the widely shared assumption (which is a good one) that all patients who have a fracture of the femoral neck or an intertrochanteric fracture should be hospitalized. Ninety-nine per cent (463,259) of 467,938 Medicare patients who had such a fracture between 1994 and 1995 were hospitalized, and the rate of hospitalization and the incidence were closely correlated (r2 = 0.99)7. In contrast, rates of hospitalization for fractures of the ankle or forearm vary much more among regions, as might be expected, simply because there is not a consensus in the profession that all patients who have a fracture of the ankle or forearm should be hospitalized (or need a procedure that is done in the hospital). Only 41 per cent (15,981) of 38,978 Medicare patients who had a fracture of the ankle and 35 per cent (8906) of 25,446 who had a fracture of the forearm between 1994 and 1995 were hospitalized; however, only one-third and one-quarter of the respective variation in the rates of hospitalization among regions was explained by the variation in incidence (ankle, r2 = 0.33; forearm, r2 = 0.27)7 (Fig. 3).
If the pattern of hospitalization for fracture of the femoral neck or intertrochanteric fracture was the rule rather than the exception for orthopaedic services—that is, if demand were truly exogenous—then demand-based projections of workforce needs would be more appropriate. However, this is not the case. The rates of utilization of hospitalization and procedures for most orthopaedic conditions vary strikingly among the nation's hospital service areas and referral regions.
The basic problems with the demand model become apparent when one considers operations on the back. Back pain can be treated in a number of ways, including several non-operative strategies. The flexibility in options is explained in part by the lack of information regarding a clear cause-and-effect relationship. For example, a diagnostic test like magnetic resonance imaging may demonstrate a herniated disc, but the herniation may not be causing the symptoms and, even if it is, the natural history is generally good. Similarly, there is a paucity of evidence-based outcomes data for many of the degenerative spinal conditions. This dilemma arises because clinical trials and other forms of outcomes research, although long overdue, have not been carried out because they are difficult to perform and the necessary research dollars have not been provided. However, the variation in utilization rates for operations on the back reflects a more fundamental problem concerning patient sovereignty and choice in health-care markets. The traditional model of delegating treatment decisions to physicians sets up a dynamic in which the opinions and preferences of providers of treatment can carry more weight than those of the patient. The evidence from studies of shared decision-making shows that utilization rates for operative procedures (demand) change dramatically when patients actively participate in the choice of care1,5.
It seems erroneous to project the need for orthopaedic surgeons on the basis of average rates of hospitalization for fractures of the ankle or of operative treatment of degenerative spinal conditions when those rates are the result of various professional practice styles. Such a strategy begs the more fundamental, normative questions concerning the costs, risks, and benefits of alternative ways of treating the fracture and, in the case of elective procedures, which form of treatment patients want.
If demand is not exogenous, workforce planning must be based on other assumptions. Some investigators have made workforce projections by assembling panels of experts to estimate need and what they consider to be the rate of orthopaedic surgeons that is necessary to provide care. However, such a strategy ignores the underlying problem of scientific uncertainty and the entanglement of provider (and health-plan) preferences with those of the patient. Our point of view is that the limitations of planning based on estimates of demand and need-based models are so severe that they probably are insurmountable. Our preference is to use benchmarking3 in workforce planning. This method involves more modest assumptions based on actual examples of how the workforce is deployed in regions where workforce strategies have appeared to succeed in the market (Table I). For example, Ann Arbor may provide a reasonable benchmark on the basis of the quality of care and the efficient size of its workforce (4.9 orthopaedists per 100,000 population).
We also recommend that workforce policy extend beyond the narrow concern of how many full-time-equivalent clinically active physicians are needed to the broader question of how physicians should spend their time. Patient care is not the only responsibility of the practicing physician. It is necessary to decide how much time physicians should spend in building the infrastructure for quality, in conducting outcomes research, in learning new skills and new ways of doing things, and so on. We think that at least 20 per cent of the work week is reasonable. If this assumption had been built into the RAND protocol, the estimate of excess capacity would have disappeared.
The future of the profession may depend on our opportunity to address the problems concerning quality and outcomes that plague the current system, and it may depend on our ability to adapt to change. This will require adjustment of the professional job description and financial incentives so that non-clinical tasks are considered part of the everyday practice of medicine. It also requires a new focus for educational policy that puts lifetime learning on an equal footing with entry-level training.
Workforce planning conducted with more modest assumptions about the limits on comprehensive rationality, but designed to help overcome some of these limits, may lead to new, more dynamic strategies for dealing with scientific uncertainty and technological change as well as uncertainty about so-called true demand in changing markets where patient preferences play a more important role in determining utilization. The idea that workforce planning should concentrate solely on the entry of new professionals denotes a deterministic economic model that no longer fits how the market works, particularly because it discounts the educational requirements to prevent the obsolescence of the workforce. Strategies that focus on the lifetime-learning requirements of the profession and on opportunities for dynamic reallocation of the workforce that are in synch with changing demand and technology offer a more realistic and optimistic scenario for dealing with the workforce planning. Freeing up orthopaedists to work on the complex problems of managing and improving the health-care system would greatly accelerate efforts by the orthopaedic societies to gain the high ground in the struggle over the future of the orthopaedic workforce. The outcomes agenda as represented by the The American Academy of Orthopaedic Surgeons Modems Project would be greatly enhanced by allocating a portion of the orthopaedic workforce toward this effort. This would suggest a strong commitment to outcomes research and would send an important message to the profession that such efforts are valued as much as clinical practice.
The outcomes efforts of The American Academy of Orthopaedic Surgeons should provide alternative opportunities for the orthopaedic workforce, the necessary data to conduct outcomes research, and an infrastructure for quality. Each step takes us closer to an answer to the question: Which rate is right?
James N. Weinstein, D.O., M.S.
David Goodman, M.D., M.S.
John E. Wennberg, M.D., M.P.H.
Center for the Evaluative Clinical Sciences
Dartmouth Medical School
7251 Strasenburgh Hall
Hanover, New Hampshire 03755