Forecastic requirements for physical therapists.

Key Words: Demand-based requirements; Health professions, forecasting; Need-based requirements; Physical therapist, requirements; Physical therapist, supply.

Findings point to a continued undersupply of physical therapists.(1) The American Physical Therapy Association (APTA) recently performed a study that produced evidence of a crisis of acute care personnel in the profession, with findings of months-long delays in filling physical therapy positions. Other studies have reported shortages in most settings in which physical therapists practice: hospitals, schools, nursing homes, and rehabilitation centers. The strong demand for physical therapy services that helped produce these shortages is expected to increase. The population over 65 years of age, which according to one estimate already consumes more than 20% of physical therapy services, should increasingly require physical therapy services as the US population ages. Increased reimbursement for home health care should add to the increased demand. The impact that health care reform will have, while unknown, should stimulate demand by emphasizing prevention of disability and greater attention to management of chronic conditions.

Having noted the existing imbalances between physical therapist supply and demand, the profession is contemplating the development of a training strategy to bring the supply in line with requirements. Although APTA estimates the supply of licensed physical therapists to be about 80,000, not all of whom are full-time or practicing, requirements for physical therapists have not been developed. Generating rational goals for training will depend upon forecasting requirements.

There is no best way to forecast requirements for physical therapists. Two approaches exist for developing health professions requirements, "need-based" and "demand-based," each of which provides an estimate but suffers from deficiencies. In models predicated on need for health care, experts calculate requirements using their knowledge of health professions standards, productivity, and expected morbidity patterns. The estimates are based on the experts' judgment of what health care ought to be. Requirements based on need provide a target level for health status and health services independent of those being delivered through the marketplace. Need-based requirements, therefore, stand as independent criteria on which to evaluate the efficacy of existing institutions. Because this method considers neither the preferences of health care consumers and providers nor the incentives, inefficiencies, and scarcities of the health care marketplace, need-based estimates tend to overestimate future consumption. The estimates can be useful for formulating goals, however, because they represent society's preferred level of service.

"Need-based" provider requirements is the quantity of health care services that expert opinion believes members of a society ought to consume over a relevant time period to remain or become as healthy as is permitted by existing knowledge.(2)

In contrast, market or demand-based requirements reflect the individual preferences of health care consumers and providers based on their observed behavior in health care markets and represent the value these persons place on the consumption and delivery of such services. As extrapolations from current utilizations, market or demand-based forecasts are logically and operationally inseparable from the existing patterns of care and distribution of resources. As such, they are not acceptable standards against which to evaluate the adequacy of existing institutional arrangements, nor do they take into account major changes in consumer or provider behavior.

Demand-based models are also more complex than those based on need for health care. They forecast requirements using econometric models that extrapolate from existing market conditions, institutional arrangements, access barriers, and individual preferences. The effectiveness of demand-based models depends upon having high-quality data on service utilization as a basis for making these forecasts. It also depends upon how well extrapolation from current conditions delineates future markets. That is, demand-based requirements projections indicate the level of service likely to exist in the future assuming the delivery system remains relatively unchanged. In the current situation, that means a continued bias toward acute, specialized care as opposed to chronic or preventive care. Given the changing demographics and the current climate of health care reform, market conditions are expected to change to favor more chronic and preventive care. The anticipated changes should increase demand for physical therapists.

"Demand-based" provider requirements are determined by a complex econometric relationship that is a function of many variables, including financial resources, price levels, size of the provider pool, and wants of the population (as reflected by consumer tastes and preferences) for all goods and services.(3)

[D.sub.PT]=f ($,...)

Comparing the results of the two models differentiates the quantity of services needed from the quantity of services demanded. Why do requirements estimates based on need-based models differ so much from those based on demand-based models? The answer derives from the disparity between health care need and realized demand. Realized demand results partly from health care need and partly from choices made by consumers and providers. In order for health care need to equal actual utilization, consumers would have to have accurate knowledge of their physical and mental needs, as well as knowledge of and confidence in the efficacy of appropriate forms of care. They would also have to be able and willing to forego other goods to pay for that care, as well as be able and willing to invest in the travel, inconvenience, and time needed to locate and receive care. Further, adequate resources would have to be available in the health system and providers would have to be willing to provide consumers with needed services.

This series of choices linking need and realized demand makes it unlikely that all health care need will be voluntarily transformed into realized market demand. For example, even if care were available, consumers completely informed, and fees nonexistent, access barriers such as travel and office waiting time alone would usually restrict demand to less than "need." So, although recent studies have documented instances in which realized demand exceeds need--such as when consumers succeed in engaging the services of a provider to care for a perceived need that does not exist, or when, for a variety of reasons, providers dispense services in excess of need--most often the health care market works to depress consumption of needed care. It therefore follows that need-based requirements usually exceed demand-based requirements.

Disparities between need and utilization in health care also arise because of market failure. In classical markets (in which price equilibrium is determined by the interaction of supply and demand), when demand exceeds the volume of delivered services, market forces encourage the price of services to rise until either demand is discouraged or provider supply is sufficiently increased to eliminate the unmet demand. In health care markets, the benefits and costs of service provision often do not fall directly or entirely on the providers or consumers who are actually making choices. Instead, a significant proportion of costs is covered by third-party payers. Consumers and providers are therefore often not influenced by the cost of the goods and services because third-party payers deal with these issues for them. As a result, the market may not adjust properly.

The critical challenge for reconciling the disparity between forecasts of "need" and "demand" is to discover ways to make consumers want and professionals provide only what consumers need, not more and not less. Referring to the consumer and provider choices described earlier, the means toward these ends include providing consumers with accurate and understandable information on their health status and options for care and on the likely outcomes of health-related behaviors; improving consumers' access to health care services; and enhancing incentives for health professionals to provide needed care. These goals form the basis for managed care programs.

The "need-demand gap" can measure how much effort must be exerted to remove information, resource, and access barriers. Together, the two approaches can provide complementary data on which to base policy formulation.

Methodological Issues in Measuring Requirements for Physical Therapists

Methods Based on Need

Requirements estimation based on need is a normative or externally imposed assessment of the type and quantity of health care the population should have in order to bring the full effect of relevant knowledge to bear on the target population. These forecasts are based on expectations of health status, are independent of actual provider utilization behavior (or expected behavior), and are determined by consensus. The requirements embody the judgments for health care delivery of the experts who establish the standards.

Because need-based forecasts are not predicated on market behavior, the forecaster must supply information on goals for resource allocation normally determined through the market. Specifically, need-based requirements forecasts must determine:

1. The target of health status (morbidity) for the population.

2. Services needed to attain and maintain this level of health.

3. The modes of care to be used (eg, inpatient, ambulatory, clinic, home).

4. The health professional resources to be used, including physiatrists, midlevel providers, technicians, and physical therapist assistants.

5. The social costs and expenditures of care.

In the application of the need-based methodologies currently available, the experts either specify or assume the needed information. Such answers are more often based on subjective judgment than on any rigorous analysis. The choice among particular need-based techniques becomes one of whose judgment is to be relied on in selecting assumptions regarding the linkages between health needs and service requirements.

The literature offers only one main option for forecasting requirements based on need.(4) The classic need-based requirements method developed to date has been requirements based on professional opinion. This methodology requires a panel of health experts to assess:

1. The productivity of individual providers.

2. The service intensity and mix for each morbidity.

3. The number of persons in the population who should receive the specific kinds of services from the practitioner, allowing for age, sex, or other demographic factors influencing the incidence and prevalence of the diseases or conditions requiring care (epidemiology).

Given these data, it is possible to calculate the total number of physical therapists required, now or in the future, by multiplying the population times the recommended utilization rates times the service time required and then dividing this product by the average work load (productivity) of the practitioner. To allow for the inevitable variations of opinion and for differences in volumes of care, incidence rates, and provider productivities, forecasts should be transformed into ranges of therapist requirements based on the varying judgments of experts on these factors and multiple delivery systems.

In this method, the health problems to be addressed serve as the basis for establishing how many health professionals should be available. It is logically an appealing method because it starts from what is physically or mentally in need of care, translates that into required services and finally into health care personnel, and becomes an easily understood and defensible method of establishing needs for specific types of health professionals. The method also allows for substantial disaggregation of both the health services being provided and the kinds of personnel being studied. That is, the method accounts visit by visit and service by service who the providers are and the purpose for the care. Similarly, the requirements may be calculated in a variety of service time measures--per episode, per person, per year--that a practitioner should spend in providing "good" care.

To expedite formulation of provider requirements without stating goals for health or functional status, expert judgment often sets goals for the process and not the outcome. For example, rather than key requirements to goals couched in terms of improved mobility rates following hip fractures, they are stated in terms of numbers of physical therapy visits and other similar process indicators. The general inability to define and measure health and functional status continually impedes policymakers. This issue is taken up later.

The development of professional opinion is costly and time-consuming. Consequently, the methodologies described in the following two sections will often be used, modified to quasi-need-based approaches.

Provider/population ratio techniques. Forecasting methods predicated on a subjectively defined ratio are of the form:

[R.sub.t]=[P.sub.t]r

where [R.sub.t] = providers required to serve a target population in time t, [P.sub.t] = size of the target population in time t, and r = normatively defined provider/population ratio.

Forecasting of requirements necessitates a forecast of the target population, which is then multiplied by the preselected therapist-to-population ratio from some reference group to yield total therapist requirements. Applications of this technique vary in terms of the choice of reference group and include national ratios, ratios from other countries, and ratios from well-served subnational systems (eg, health maintenance organizations [HMOs], preferred provider organizations).

Ratios can be modified to account for variations in the reference population across demographic groups (eg, age, gender, race, income). By calculating physical therapist requirements for each component of the reference area, the requirements for any target population can be calculated on the basis of actual or projected demographic composition. This process of disaggregation controls for disparities in requirements due to demographic differences between the reference and target populations.

Ratio methodologies, though simple and inexpensive to implement, imply a rigid relationship among population, health care need, service requirements, and provider productivity. The use of therapist-to-population ratios presumes all these factors to be unchanging over time, with no allowance for effects of new health care technology, patterns of task delegation, or service needs. If the ratio is assumed to change from year to year or from area to area, the ratio technique presumes that all component (but unspecified) factors change by similar amounts. Additionally, ratio methodologies presume that existing patterns of health services provider distribution in the reference location are ideal. It is obvious, however, that the ratios spring directly from existing utilization patterns in the reference area and, however ideal, are not reflections of need.

Health maintenance organization and other staffing patterns. Several studies have indicated that an alternative reference group for calculating provider requirements conformable to health care needs can be constructed from HMO staffing patterns vis-a-vis patient pool characteristics. Similar to the disaggregated ratio approach described earlier, the HMO patient pool is disaggregated by demographic characteristics noting the utilization rate per group. Based on these utilization patterns, the target population's need patterns can be approximated and controlled for demographic differences in the HMO and target populations.

The advantages of this approach are that marginal fees and access costs are often minimal in the HMO setting and that all aspects of health care quality, practice management, specialty allocations, and ancillary services are carefully monitored and controlled in these situations. Even within this tightly administered group of health care suppliers, however, there exists considerable variation in staffing patterns across HMOs. In order to select a model HMO staffing pattern to project as an "ideal" for the non-HMO populace, these disparities need to be explained and reconciled. Furthermore, studies have shown that access costs are often significant in these centralized clinics; as a result, some patients seek care outside the HMO. Finally, the population base served by HMOs is to some extent nonrepresentative of the nation. Studies of HMO patient pools have shown significantly lower ages, higher incomes, and generally better health than the national average.

Staffing patterns of HMOs may be useful in generating standards for non-HMO populations, but only if utilization and staffing patterns are disaggregated by sociodemographic characteristics, the access barrier effects quantified, and the standard staffing patterns adjusted to reflect the absence of these access barriers.

Analysis of need-based methodologies. Developing requirements based on health care need or professional judgment alone requires subjective assessment of what the level of health and distribution of health should be, what level of physical therapy services should be supplied to best achieve that level of health, and what specialty and geographic distribution of physical therapists and related personnel, such as physical therapist assistants, will be conformable with the preselected health status level and distribution. These complexities of determining requirements by fiat are usually truncated to provide estimates of service conformable with "accepted" standards of practice. Requirements formed in this fashion implicitly accept the goals of health status that such standards of care specify or imply as well as the "amount" of health they permit or lead to.

Need-based estimates can be both independent of consumer and provider preferences as revealed in the marketplace and abstract from the incentives and inefficiencies of the institutions of health care delivery. They do assume, however, a concomitant health/consumer information education effort to accompany a projected change in supply of physical therapists. In the presence of informational and access barriers to care, need-based estimates could serve as an ideal, generally upper limit, for future physical therapist requirements.

Need-based estimates are not without weaknesses. Requirement forecasts based on "need" are the product of a subjective perception of health care experts without reference to resource constraints or to rationing or allocation procedures. These judgments are grounded in best health practice techniques and require some "vision" of what health and health delivery ought to be. It is not at all certain that this vision includes, or can be made to include, an attainable and reasonable level of health status and care, given the scarcity of societal resources and the existence of other useful commodities. By not considering individual preference and access barriers, need-based requirements can fail to yield any insight into the modifications required of the delivery system to utilize all of the "needed" physical therapists if their services are in fact available. In other words, need-based requirements may bear little resemblance to employment opportunities. Without a mechanism to reveal alternate paths to transform "need" into "demand," it is not likely that all consequences of generating "needed" volumes of physical therapists can be foreseen and considered.

Methods Based on Market Demand

Requirements forecasts based on market demand are predicated on actual utilization or market behavior of the target population. By extrapolating from actual (observed) health care market phenomena, these forecasts embody the current interacting individual preferences for health care delivery of all consumers, providers, third-party payers, and regulators, given the current institutional structure and distribution of income and resources. Whether provision is made for health status variables, demographic characteristics, or economic factors, all of the market-based models relate some combination of these variables to actual utilization behavior. The literature is filled with variations on this theme.(5)(6) Conceptually, the boundaries on demand-based forecasting models are defined only by the availability of utilization data and the imagination of the investigator. A principal weakness of all market models is that their specification conforms to testing specific research hypotheses concerning utilization rather than to establishing provider requirements consistent with those utilization patterns. As a result, most models are incomplete and would have to be adapted to consider questions of physical therapy professional requirements.

To simplify discussion, consider three generic varieties of market-based techniques for estimating future physical therapist requirements.

Demographic models. Models have been developed that hypothesize that demand or utilization is dependent on a set of demographic variables such as age, education, income, gender, race, and marital status.(7) By specifying the apparent multivariate relationship between utilization and these factors, future utilization patterns can easily be obtained by substituting estimates of these variables for future years (from the Bureau of the Census) into that multivariate formula. These models contain implicit assumptions about unchanging health patterns, modes of health delivery, and provider productivity. These models also assume that the economic environment and accompanying access barriers that temper demand are either unimportant or similar to the present.

All absolute and relative prices, access costs, the structure of third-party payments, and general economic conditions are absent from this formulation. Not only does the forecast assume that these considerations are unimportant, but the model fails to specify any determinants of demand that are subject to policy modification.

Economic models. To remedy some of the deficiencies of the simple "demographic" models, economists have offered more fully specified models of utilization that provide for economic incentives, and thereby explicitly incorporate the realities of scarcity and economic choice as they affect utilization. By specifying the relative prices of various forms of care, account can be taken of economic incentives to substitute or switch from one type of care to another. Some multiequation models formally recognize that physical therapist services are only one facet of the health care system wherein utilization levels of all types of services are codetermined.(8)

Each model, however specified, delimits a finite network of casual relationships that interact with utilization. To the extent that alterations in physical therapist utilization are inextricably linked to alterations in utilization in other sectors (as, for example, nursing home beds), the model must provide for codetermination in order to be reliable. Lack of information about referral patterns among professionals, for instance, impedes reliable forecasts when separate methodologies are applied to individual professions.

The intricacy of the modeling effort is positively correlated with data needs and forces the researcher to strike a balance between reliability and the cost of the research effort. One notable model, for example, uses more than 100 equations and endogenous variables.(9) Although the model theoretically accounts for a wide variety of cause-effect links, its intricacy precludes data collection required for implementation. Less ambitious methods, however, do permit estimates of utilization based on economic and demographic data and might be of considerable policy assistance in measuring the effects of hypothetical policy changes such as alternate forms of national health insurance. Models may also include a variable for HMO patient pool size and thereby permit an evaluation of the physical therapist requirement stemming from alternative levels of HMO involvement in the target population.

The logic of the policy-making process suggests that a list of critical policy variables such as price, coinsurance rates, travel time, waiting time, HMO patient pool size, the other organizational factors as well as therapist productivity be specified prior to model formulation to achieve both maximum reliability and policy leverage in evaluating alternative policy strategies. Unfortunately, as most of these models were designed to answer particular research hypotheses and not to provide policy input, the alternative is to select from the menu of available models, each of which includes some, but not all, of the factors desired by policymakers.

Economy-based models. A final demand-based methodology finds its basis outside the specific health care marketplace. It argues that the demand for health professionals is more properly specified as a product of the general structure of the entire economy. This effort is supported by the Bureau of Labor Statistics (BLS).(10) It differs from other "demand models" in arguing that the demand for health professionals is more sensitive to movements in the general structure of the economy than to specific facets of the health care industry. It also differs in terms of the basis of the forecast. Other demand models forecast either by projecting trends in utilization from historical data or by projecting present utilization patterns on the basis of expected changes in the demographic structure of the population. The BLS technique bases the forecast of requirements on expected levels of economic activity and trends in the technological linkages between sectors of the economy.

The physical therapy sector is but one interactive component in this model, and, as expected levels of economic activity have an impact on the production level in sectors such as steel, automobiles, and textiles, so too is the production level of physical therapists expected to respond. The technical linkages between sectors of the economy are specified and projected from historical data and are the basis for creating impact projections for all sectors, including physical therapy.

The BLS projections do not afford health policy leverage in evaluating impacts of alternative strategies for altering levels of utlization. Rather, they are useful in evaluating the provider implications of general economic trends. Certainly, this technique is not suitable for setting requirements based on health care need, nor is it useful in evaluating alternative strategies of health policy. Demand forecasts from this model are based on economic linkages largely external to health markets. The basis for the forecast is a rich stock of economic and technological projections not conformable with other types of provider demand models. The value of this method is in serving as an independent check on other demand forecasts.

Analysis of demand-based methods. Demand-based methods yield estimates for provider requirements that reflect in part the preferences of individuals for health care and, as such, are a statement of the "value" society places on these services vis-a-vis the availability of other useful goods and services given the existing resource constraints. This statement of "value" is tempered by consumer ignorance, access barriers, and the existing physical therapist and income distributions. A critical shortcoming of the demand models is that most were designed to investigate specific hypotheses concerning utilization rather than to forecast numbers of providers required to service the predicted pattern of utilization.

Another very important shortcoming is that applying these models for inferential purposes requires massive databases and extensive knowledge concerning behavioral patterns. The complexities of such data needs make the application of demand-based methodologies very difficult. In addition, in order to adapt these methods to yield an estimate of physical therapist requirements, it is necessary to transform utilization (number of visits) into an equivalent number of physical therapists by means of assumptions of therapist productivity. Productivity is, of course, the product of economic incentives on the part of physical therapists regarding task delegation, hours of practice, skill, and available technology. Assumptions of productivity are also necessary in formulating need-based requirements, but in that application ideal productivity standards can be defined. On the other hand, in formulating demand-based requirements, which by definition attempt to account for market incentives and productivity measures, the forecaster must consider the actual effect of factors bearing on productivity such as task delegation, skill, technology, and preferences toward practice. Because many utilization models found in the literature are not focused on physical therapist requirements, the problem of productivity estimates has been largely ignored and little is known about behavior of physical therapists on which to specify assumptions about productivity.

All demand models face a potentially devastating, but only partially substantiated, charge that demand for health care is not truly measurable. Recent literature on health care utilization contains repeated references to instances in which utilization appears to be more responsive to the supply of services than to demand.(11) This creates some doubt about the relevance of forecasts based on characteristics of market demand. If it is true that demand is a passive force in the marketplace, then health care need is the only means by which goals can be set on which to base educational (supply) policy.

The primary advantage of the market demand models is their ability to measure objectively the response rates of utilization to various policy variables. Knowledge of how consumers and providers react to changes in fees or coinsurance rates, for example, affords policymakers a view of actual behavior that is valuable in selecting among strategies for achieving a health system goal.

Conclusion

As the preceding discussion suggests, neither the need-based nor demand-based approach is superior to the other. The need-based approach, focused on what health care ought to be, unrealistically assumes that all care given will be appropriate and necessary and that the services provided and demand for these services will rise to the ideal. Consumer and provider choices are absent from the equation. The demand-based approach, predicated on current utilization patterns, ignores the consensus that what exists needs to be changed, not emulated, and that health care reform virtually ensures that change will occur in all the variables that affect utilization, including the preferences and behaviors of health care consumers and providers, institutional arrangements, financial incentives, access barriers, and resource distributions. The impact of these changes will depend a great deal on the precise changes health reform brings.

For either approach to be used, extensive data must be available on the health status of the population to be served, provider productivity, utilization rates for the health care service being provided, and related factors. Unfortunately, no health profession has command of the requisite data. In physical therapy, as in many other health professions, there is a lack of definitive data on the number of physical therapists in active practice, the content of their practice, the use of physical therapist assistants and interactions with other health practitioners (including physicians), and the relation between the intensity of service provided and the outcome of interventions on patient well-being.

Given these constraints, the only intelligent approach is to capitalize on features of various models and approaches, using whatever data are available or can practically be generated to help produce a defensible health professions forecast. This pragmatic approach would borrow from the need-based approach the convening of a panel to define the contextual practice of physical therapists. That is, the panel--to be composed of physical therapists, physicians, and health services researchers and administrators--would describe how physical therapists function, the settings in which they practice, and their interactions with other health professionals, including delegation to physical therapist assistants. The panel would also review relevant data that relate to physical therapy practice, such as the number of beds in the hospital setting; the number and types of health professionals, such as physiatrists and orthopedists; and the number of patients in different demographic groups and their conditions. The panel would then link these data, such as the number of physicians, to the need for physical therapists.

The normative results of the panel deliberations need to be tempered with other available data that can be considered as a normative standard or that the profession is willing to adopt as a standard of appropriate use of services. Borrowing from the demand-based approach, this step anchors the recommendations of the panel in something known and acceptable. The needed comparison data on numbers of physical therapists practicing per population can be obtained from subpopulations such as HMO enrollees, the military, or any other well-defined health service organization that is seen as providing an appropriate level of services. Once definitive data are developed for a segment of the physical therapy community, requirements estimates can be extrapolated for larger areas or the nation as a whole.

This pragmatic strategy could be implemented relatively quickly and inexpensively. It has proved useful before in physician requirements forecasting when adequate data were not available for certain specialties.(12) Drawing from both need-based and demand-based models, it produces the normative targets of one while projecting from the actual health services data that form the basis for the other.

The physical therapy community urgently needs to develop a normative model for distribution of physical therapists among their various practice settings. This normative information would serve as a credible basis for garnering political support for the profession's goals at a time when long-term decisions are being made in the context of health care reform.

References

(1)Recruitment and Retention of Physical Therapists in Hospital-based Practice. Alexandria, Va: American Physical Therapy Association; 1992:4--5.

(2)Graduate Medical Education National Advisory Committee. Summary Report to the Secretary, Department of Health, Education, and Welfare, Vol 1. Hyattsville, Md: Health Resources Administration, Publication No. (HRA) 81--651, 1980.

(3)Graduate Medical Education National Advisory Committee. Physician Requirements Forecasting: Need-based Versus Demand-based Methodologies. Washington, DC: US Dept of Health, Education, and Welfare, Bureau of Health Manpower, Publication No. (HRA) 78--12, 1978.

(4)Lee RJ, Jones LW. The Fundamentals of Good Medical Care: Publication of the Committee on the Costs of Medical Care, No. 22. Chicago, Ill: University of Chicago Press; 1933.

(5)Mullan F, Rivo ML, Politzer RM. Doctors, dollars and determination: making physician work-force policy. Health Affairs. 1993; 12:138--151.

(6)Reuben DB, Zwanziger J, Bradley TB, et al. How many physicians will be needed to provide medical care for older persons? Physician manpower needs for the twenty-first century. J Am Geriatr Soc. 1993; 41:444--453.

(7)Kletke PR. The Demographics of Physician Supply: Trends and Projections. Chicago, Ill: American Medical Association; 1987.

(8)Reinhardt UE. Health manpower forecasting in a market context. In: Norman T, Bailey J, Thompson M, eds. Systems Aspects of Health Planning: Proceedings of the IIASA Conference; Baden, Austria; 1975.

(9)Reinhardt UE. Health manpower forecasting: the case of physician supply. In: Ginzberg E, ed. Health Services Research: Key to Health Policy. Cambridge, Mass: Harvard University Press; 1991:234--283.

(10)BLS Handbook of Methods. Washington, DC: US Dept of Labor, Bureau of Labor Statistics; 1992.

(11)Schwartz WB, Sloan FA, Mendelson DN. Why there will be little or no physician surplus between now and the year 2000. N Engl J Med. 1988:318:892--897.

(12)An Analysis of the Content of Specialty Practices and Their Service Capacities. Hyattsville, Md: US Dept of Health and Human Services, Public Health Service, Health Resources Administration, Publication No. (HRA) 81--649, 1981.

This paper was adapted from a presentation given by Dr Jacoby at the American Physical Therapy Association/Agency for Health Care Policy and Research Human Resources Conference; Reston, Va; August 26--27, 1993.

This article was submitted April 12, 1994, and was accepted August 30, 1994.