Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
THE IMPACT OF CHANGES IN PAYMENT METHODS ON THE SUPPLY OF PHYSICIANS' SERVICES Robert H. Lee, Ph.D. Int roduct ion Change in the way that physicians are paid seems likely. Not only are businesses and governments actively seeking ways to save money by restructuring their insurance programs, but the market for physicians ' services is becoming increasingly competitive as the number of physicians grows. These two forces seem increasingly likely to lead to changes in the ways that physicians are paid, perhaps as a result of government policy or perhaps as a result of market pressures to adopt innovative payment methods. Physicians just beginning to practice are likely to respond in the most dramatic fashion, but changes in payment methods could aff ect the practice and income of every physic fan. How will the supply of services be affected by changes in the way that physicians are paid? An answer requires careful specific-a/ion of the changes that seem likely to occur and of what is meant by supply. Five kinds of changes seem likely, or at least possible. Three of these changes will be due primarily to policy changes and will directly affect the fee-for-service market: 1. Restructured schedules of allowed or reasonable fees. 2. Altered assignment or participation arrangements. 3. Experiments with per-case payments. The remaining changes will primarily be responses to a changed market, rather than explicit changes in policy. 4. Expanded numbers of physicians working in prepaid settings. 5. Increased numbers of physicians being salaried employees. All of these changes will alter the existing fee-for-service system. What is meant by the supply of services? A deceptively simple answer is The amount of a particular service that is available in a particular area at a given price. ~ An example would be the number of house calls 44
that physicians in Chapel Hill would be wi fling to provide for $30 per visit. Even this simple example becomes complex rather quickly, because the supply of house calls in Chapel Hill will depend on how many physicians have chosen to practice there, what their specialties are, and for how many of these physicians $30 is an adequate fee, which will depend on their values and on how much they can earn doing other things. Because the idea of supply is so complicated, it needs to be broken down into at least four separate components: 2. 3. What specialty physicians choose. What Locat ion physic fans choose . How much physicians choose to produce. What mix of output physicians choose to produce. This breakdown makes the question t ractable, even if the answers are not known. To assess precisely the effects of changes in the payment system on the various aspects of supply, two questions must be answered. First, how will a particular change affect the f inancial rewards of a particular specialty, location, or type of output? In the absence of specif ic proposals this question cannot really be addressed, but specif ic proposals can be evaluated if a second question can be answered. How much are physicians' decisions affected by financial considerations? A great deal, somewhat, or very little? This paper f ocuses on how much supply dec isions are at f ected by f inancial factors. show much will be expressed as the percent change i n the component of supply that is associated with a one percent change in the financial variable in question. For example, Berry et. al. estimate that if gross earnings increase one percent, net immigration increases by 3.4 percent. This way of describing results avoids any confusion due to how either gross earnings are measured (i.e., in dollars per year or thousands of dollars per year). It also offers a reminder that empirical results are historical in nature and can be extrapolated only with care. That is, if a change in policy increases gross incomes by 2 percent, it is probably reasonable to use the above est imate to calculate that net in-migration will increase by 6.8 percent. Because the change falls outside the range of historical experience however, it is not proper also to infer that a 50 percent rise in gross income would result in a 170.0 percent increase in net in-migration. Either a much larger response or a much smaller response could be possible. History serves as a very uncertain guide in the face of dramatic changes. Given these qualifications, what does existing research tell us about the impact of the payment system on physicians' specialty choices, location choices, productivity, and output mix? It tells us quite a lot ~ but data problems and the complexity of the existing payment system leave some important issues unresolved. -45-
The Payment System and Physicians' Specialty and Location Decisions There has been considerable analysis of physicians' career decisions. Much of this research, unfortunately, either has ignored the impact of f inancial factors on career choices or has relied on inadequate measures of the f inancial rewards of particular specialties or locations. Consequently, only a few studies merit careful attention. The fundamental problem conf ranting analyses of physicians' career choices is inadequate data. Data on prices and incomes are often not available, and the data that exist are usually of suspect quality. As a result, even for studies that have emphasized the impact of financial factors on career choices, data problems make the conclusions somewhat suspect. Location Choices Most established physicians are not likely to move. Consequently, it is the locat ion choices of young physicians that seem most apt to be If luenced by financial considerations. Although other factors seem likely to determine the location choices of most young physicians, an area that promises above average incomes seems likely to attract some additional physicians as a result. By the same token, an area that promises below average incomes seems likely to discourage some physicians f rom setting up practice there. The strength of this effect is the p r i no ipal research question. Because Canadian data on physicians' incomes are considerably better than U.S. data, two of the more reliable studies of the impact of income on location use Canadian data. Both find a relatively large impact. Berry et.al. ( 1978) examined the migration of physicians in and out of medical service areas in Quebec between 1971 and 1975. The authors concluded that net immigration (those entering minus those leaving) was quite strongly associated with average gross earnings in the service area during the previous year. Using multiple regression analysis, the authors estimated that a medical service area in which physicians average gross earnings were ten percent above average experienced net immigration of physicians who were not board certified that was 34 percent above average. In a study with comparable results, Hadley ( 1978 ) examined annual changes in the distribution of physicians in nine Canadian provinces between 1958 and 1976. Hadley estimated that a province in which physicians' net incomes were 10 percent above average could expect to attract 33 percent more new physicians than average. Because these data covered a relatively long period of time, Hadley was also able to examine how quickly physicians would respond to these differences in potential earnings. He estimated that it would take nearly seven years before the entry of physicians would be large enough to cause net income in a 46
high-income province to fall to average levels. So, even though practice location decisions appear to be significantly influenced by income opportunities, the adjustment process takes time. These two Canadian studies are of interest because they are based on much better data than comparable U.S. studies. To begin with, both had reliable income data. In addition, both had data on a number of areas for several years. This allows researchers to control for non-financial attributes of locations that may cloud the impact of financial factors. For example, one suspects that some physicians will be willing to accept a relatively low income to live in Montreal. Of primary interest, however, is the effect of a drop in potential income in Montreal relative to potential income elsewhere. To examine this question requires data for several years for each location. In addition, the Berry et . al. study focused on small areas. Since physicians choose to practice in communities, and since there can be considerable variation in a physician' s potential income among communities within a state or province, such an emphasis on small areas lends credence to the results. The quality of the data used and the comparability of the results of these two studies surpass those of any studies of the U.S. It may even be plausible to argue that Canada is enough like the U.S.--despite the differences in the two medical care systems--that the results may be seen as applying to the U.S. as well. That, however, has yet to be demonst rated. For the most part, studies of the impact of financial factors on the distribution of physicians in the U.S. have had to rely on unsatisfactory data. Information on the location of physicians is fairly good, but very little satisfactory data on physicians' prices or incomes are available. Where financial data exist, they often are too aggregate to be adequate. This may explain why U.S. studies generally have found a smaller effect of f inancial factors on location decisions than was found in the Canadian studies. A notable exception was Ramaswamy and Tokuhata's 1975 study of variations in the number of physicians per capita in the counties of Pennsylvania. The authors developed county-level price indices from Blue Shield claims data and so had rather good data on prices. The authors found that counties in which physicians could charge high prices had more physicians per capita. Unfortunately, their estimates are relatively imprecise. Ramaswamy and Tokuhata calculated that a county with a price index 10 percent higher than average had between 25 and 62 percent more physicians per capita. This imprecision may be due to unmeasured characteristics of the counties that also affect the supply of physicians for reasons unrelated to price. For example, the presence of a medical school might attract specialists who deliver care to patients f ram other counties. In any event, the results are valid only for Pennsylvania, so this study cannot supply a general conclusion about the impact of earnings potent ial on the supply of physic fans . 47
A study of a number of states by Fuchs and Kramer {1973) employed a fairly similar approach, but found smaller, if equally variable, effects. Fuchs and Kramer estimated -that a 10 percent increase in prices in a state was associated with an increase in the number of patient care physicians of between 4.2 to 11.4 percent. It does not appear, however, that the data on prices were as reliable as in the Ramaswamy-Tokuhata analysis. In addition' the Fuchs-Kramer study used states as the unit of analysis, which may or may not be satisfactory. Cantwell (1979) examined the relationship between prices and physician supply for a number of metropolitan areas and county groups. Unlike the other authors, Cantwell assumed that the impact of the supply of physicians on prices could safely be ignored. If not correct (most researchers would argue that it is not ) , this assumption would tend to bias downward the estimated impact of prices on supply. Cantwell found that areas with high prices had more physicians, but that the effect was smaller than in the two proceeding studies. This may be a result of the specification of the model. Several studies have investigated the relationship between physicians' incomes and the supply of physicians. All used the state as the unit of analysis, and only one examined the post-Medicare period. Harrison and Jud examined the interaction between physicians' incomes and the number of physicians in a number of states for 1967 and 1968. They found that states in which physicians' incomes were 10 percent above average had over 16 percent more physicians than otherwise comparable states. Held examined the impact of physicians' incomes on rates of in-migration and out-migration for some 1955-1965 graduates of U.S. medical schools. He f ound that high income in a state tended to increase in-migration and decrease out-migration, but the effect depended on the amount of contact physicians had had with the area to which they were migrating. An area in which physicians' expected earning potentials are high should be attractive as a result. Existing research is consistent with this perception, but gives no real guidance as to how strongly this affects physicians' location choices. For a number of reasons, data on prices and incomes are not good enough to permit much precision. Consequently, we really have very little idea of how long it would take for physicians to respond to a change in the payment system that made one area more attractive or less attractive financially. As the market for physicians' services becomes more competitive we would expect that financial concerns would play a larger role in physicians' location choices, so even good historical evidence might not accurately predict the response in the future. -48-
Specialty Choice Very little research on the impact of financial factors on specialty choice exists. Even if good data on prices or incomes were available, two problems would complicate analyses of specialty choice. To begin with, if financial factors affect students' specialty choices, it is expectations about potential earnings rather than actual earnings that matter. Presumably there is a link between reported income data and students' expectations, but very little is known about this relationship. In addition to this problem, it is likely that non-financial factors disguise the relationship between earning potential and specialty choice. For example, suppose that one specialty offered a more pleasant lifestyle than another. Economic theory predicts that students will enter the more attractive specialty until income in that specialty falls enough to make it comparable with the specialty with the less pleasant lifestyle. In principle, at least, information about these non-f inancial characteristics is needed to assess the impact of expected earnings on specialty choice. Only one study has directly tackled this problem. Hay (1981) uses a complex model that statistically adjusted for the effects of unobserved non-f inancial characteristics. Without the ad justment Hay found no significant effect of income on specialty choice. With the adjustment a higher income was found to significantly increase the probability that a specialty would be chosen. Hay concludes that failure to adjust for these non-financial effects should be seen as a significant limitation of other studies. Only three other examinations of financial factors and specialty choice have been published. The consensus of these studies is that there is a small impact of income on students' specialty choices. Sloan (1970) examined the impact of lifetime earnings on the number of residents in various specialties. He attempted to explain the decision to specialize as being partly due to the earnings in the several specialties relative to earnings in general practice. A more recent but more limited study by Lee (1980) looked at changes in the proportion of residents entering psychiatry between 1966 and 1973. Because it looked at changes over time, the Lee study was somewhat less vulnerable to problems due to inability to measure all the relevant non-financial factors. Despite their quite different approaches, the two studies reached nearly identical conclusions about psychiatry: an increase in psychiatrists ' incomes would lead to a statistically significant, but very small increase in the number of psychiatry residents. Sloan's results for other specialties also found that higher relative incomes would lead to more residents in the more financially rewarding specialty. A study by Hadley (1979) found relatively little support for the hypothesis that higher relative incomes increase the number of students choosing the more rewarding specialty. Only for internal medicine was there evidence of this effect. -49-
The conventional wisdom is that income expectations have some impact, but not too much, on specialty choice. Available data, however, permit only very tentative conclusions. In addition, Hay' s f indings suggest that earlier studies may have underestimated the effects of earning potential on specialty choice. In any event, a change in the payment system that significantly changed expectations about earnings in a particular specialty might well lead to a fairly dramatic shift in that specialty's popularity. Such changes in earnings have not been observed, however, so the response cannot be forecast accurately. Physicians' Productivity Economic research on physicians' productivity has two sides: theoretical and empirical. Its theoretical side makes a number of st ra ightf orward predict ions that have immediate policy relevance. I ts empirical side faces extremely difficult data problems and offers fewer unambiguous conclusions. From an empirical standpoint, the fundamental problem is that def ining a physician' s output in a satisfactory way is difficult. Consequently, there is always the lingering concern that empirical studies are measuring changes in the kind of output as well as the volume of output being produced. A brief overview of the main issues will help set the stage for the discussion to follow. First, physicians in solo, fee-for-serv~ce practice have incentives to be as efficient as possible, but it may not be possible for a physician in solo practice to be as productive as a physician in group practice due to the inherent advantages of group practice, which are termed economies of scale. Second, although the potential productivity of a physician in a fee-for-service group practice may exceed that of his or her solo counterpart, being in a group may dilute incentives enough to offset this advantage. Third, there is a strong suspicion that physicians in prepaid group practices--especially those who are salaried--will be less productive than their fee-for-service counterparts. Defining Efficiency and Productivity There are two ways of defining efficiency for physicians: narrowly, in terms of the cost of producing services, or broadly, in terms of the social cost of illness. The two concepts are related but not identical. The connection is that efficiency in the broad sense requires efficiency in the narrow sense, but efficiency in the narrow sense does not guarantee efficiency in the broad sense. Efficiency in the narrow sense entails producing a given output as inexpensively as possible. Alternatively, it could be described as producing as much as possible given the resources at hand. The two ways of characterizing efficiency are mirror images of one another; which is most useful depends on the situation at hand. 50
Eff iciency in the broad sense involves pushing the social cost of illness as low as possible. This concept may entail courses of action that are not purely medical in nature, such as prevention through educat ion or exercise. It also may entai 1 hard decisions about where resources should be focused, since what is possible depends on how resources are divided among individuals' and society's competing goals. St it 1, eff iciency in this broad sense requi res ef f iciency in the narrow sense. If it were possible to use fewer resources to produce the amount of medical care that is desired ~ that is, if we were ineff icient in the narrow sense), efficiency in the broad sense could not have been attained either. It is possible, of course, to produce medical care as inexpensively as possible, but to be producing the wrong mix of care, so efficiency in the narrow sense cannot ensure efficiency in the broad sense. For understandable reasons, the literature on productivity emphasizes the narrow sense of ef f iciency. Since what to produce and how to produce it as efficiently as possible are separate issues, this is an understandable approach. It limits, however, need to be remembered. Productivity in Solo, Fee-for-Service Practice Economic theory predicts that solo practitioners in fee-for-service markets should be as efficient as possible. Whatever the physician's goals--more income, more leisure, better patient care--increased efficiency lets the physician get more of it. In short, economic incentives encourage efficiency. Interestingly, despite this unambiguous prediction, the conventional wisdom is that physicians in solo, fee-for-service produce less than they profitably could because they have too few aides and too little equipment (Reinhardt, 1972~. Although widely believed, this conclusion seems likely not to be correct for two reasons. First, it hinges on the assumption that physicians are price takers, meaning that a physician's patients would seek care elsewhere if his or her prices were even slightly above the going rate. This is a fairly difficult assumption to accept, and it conf 1 icts with other evidence ~ Lee and Had fey, 1981 ) . Second, some technical criticisms of Reinhardt's work have been advanced, and a similar analysis of other data by Brown and Lapan (1980) finds results somewhat at variance with Reinhardt's. Further research on physicians' productivity is needed, but alternatives to the current approach should be considered. To challenge the theoretical presumption that physicians in solo, fee-for-service practice are efficient (given the constraints imposed by solo practice) requires more robust evidence. It is possible, of course, that even though as efficient as possible, physicians in solo practice are intrinsically less productive than physicians in groups. The primary reason, it appears, is-the size of the 51
typical solo practice: most solo practices are too small to permit the most efficient use of aides and equipment. There may be, in other words, economies of scale in medical practice, so that group practices have an inherent cost and productivity advantage. Although not all observers are convinced, several studies have f ound evidence of returns to scale {Reinhardt, 1972; Kimbell and Lorant, 1977~. The data in Table 1 are fairly typical of the evidence on scale economies. They suggest that physicians in solo practice are less productive and earn less than physicians in small groups. Part of the dif ference in product ivity is due to the greater use of aides and equipment by group physicians, but group physicians appear to be more Table 1 Output and Net Revenue per Hour by Practice Organization Total Patient Visits Net Income Per Type of Pract ice Per Hour Hour 1 MD 2.17 $33.18 2-3 MDS 2.48 35.03 4-7 MDS 2.52 37.32 8 + MDS 2.21 34. 65 Source: Profile of Medical Practice 1981, American Medical Association. productive even after this is taken into account. Part of the difference may be due to d if f erences in the type of output produced. Whatever the source of the differential, group practice appears to be more profitable than solo practice. Average net revenue per hour is between 4.4 and 12.5 percent higher for physicians in groups. Productivity in Group, Fee-for-Service Practices Two offsetting factors affect the productivity of physicians practicing in groups. First, there appear to be economies of scale that make physicians in groups more productive than solo practitioners, although the conventional wisdom is that the productivity gains of group practice are realized by groups that are relatively small (Kimbell and Lorant, 1977; Bailey, 1968; Golladay, Manser, and Smith, 1974~. As groups become larger, however, incentives for individual physicians to be as efficient as possible become diluted, and productivity may begin to fall (Sloan, 1974~. The data in Table 1 are consistent with this story. Solo practitioners see the fewest patients per hour and earn the least, 52
physicians in groups of eight or more see fewer patients and earn less than their counterparts in smaller groups. The incentive issue merits careful study, for increasing numbers of physicians are practicing in groups. If in fact the pattern in Table 1 ref lects variat ions in ef f iciency rather than variations in case mix or the characteristics of physicians who choose different practice styles, then increasing ef ficiency in groups may be one way for physicians to maintain net incomes even though the market has become increasingly compet it i vet The nature of the incentive problem is that in large groups the effect of a physician being inefficient can be spread throughout the group. For example, if a physician insists on overstaff ing his or her practice with aides or underutilizing aides, if costs are shared, every physician in the group will experience a rather small drop in net income. A solo practitioner would, on the other hand, suf fer the entire amount of the drop in net income. How costs are shared and how resources are allocated within groups may play a role in the efficiency of physicians in group settings. Compensation arrangements appear to affect productivity. In particular, salaried physicians generally work less, see fewer patients per hour, and earn less than physicians whose earnings depend more directly on their own ef forts (Bobula, 1979) . In part this may be due to self selection of physicians seeking lighter workloads into salaried practices, but it also seems likely that this is partly due to the reduced rewards to eff iciency and additional effort that salaried practice entails. As can be seen in Table 2, salaried physicians see an average of 11.7 percent fewer per hour than physicians paid on an incentive basis. Productivity differentials of this magnitude merit closer attention. Given the increasing importance of group practice, better evidence on the importance of incentives and organizational structure is needed. At the same time, care must be taken to control for self selection of physicians into types of practice organizations and differences in the types of patients being treated. Product ivity in Pre-Pa~d Group Practice Viewed narrowly, productivity in prepaid group practices is likely to be lower than productivity in fee-for-service groups or even in solo practices. Pre-paid groups are usually fairly large and usually have salaried physicians, two factors associated with reduced incentives for ef f iciency. Although the evidence is limited, physicians in prepaid groups do seem to be less productive than other physicians (Mechanic, lg75) . 53
Table 2 Visits per Hour by Specialty and Method of Remuneration for Non-Solo Physicians, 1973 Remuneration Method Specialty Shared Group Revenue from Salary Income own Patients General and Family Practice 4.29 5.42 5.12 Internal Medicine 3.14 3.45 3.19 Pediatrics Surgery Obstetrics- Gynecology 3.78 Source: Bobula, 1979. 4.20 5.17 5.38 2.88 3.16 2.83 3.97 3.79 A potentially important area of research, therefore, is the impact of alternative rules and financial incentives on the productivity of physicians in prepaid groups. The competitive position of HMOs would surely be enhanced were it possible to reduce costs through increased productivity, and compensation arrangements that combine salary with output-related pay might represent a promising mechanism. The def inition of productivity used above is a limited one. It refers only to ef f icient use of resources supplied by the physician. Medical care, however, is produced using many resources supplied by other sources, and fee-for-service physicians have an incentive to underestimate the cost of these resources and overuse them. This is particularly important regarding use of hospitals, because the social costs of emergency and inpatient care are so high. In a prepaid plan spending on hospitals usual ly reduces net income to the plan, so there are incentives to use hospital services eff iciently. Indeed, savings due to reduced use of hospital services appear to be the pr incipal way that successful prepaid plans lower costs ~ Luf t, 1978 ~ . Making sure that these incent Ives af feet the behavior of physicians in the group can be a difficult task, given the dilution of economic incentives that occurs in large groups. The problem is most acute in IPAs. Overuse of the hospital reduces the physicians' net income by a -54-
comparatively small amount because the cost is spread over all the physicians in the plan. To reduce use of hospital care, IPAs must rely on non-economic mechanisms. So, of course, must staff model HMOs, but these plans have the advantage of being able to recruit physicians that are inclined to use hospital services sparingly. This discussion of productivity is rather bloodless. An example, therefore, may clarify the issue somewhat. Compare the output of an internist in a group practice who shares income with his or her colleagues to the output of a salaried internist working in a staff model HMO. A reasonable estimate appears to be that the first physician would see 3.45 patients per hour, 52.3 hours per week, 47 weeks per year and that his or her salaried counterpart would see 3.14 patients per hour, 51.0 hours per week, 46.5 weeks per year. Were this the case, the salaried physician would provide over 1,000 fewer visits (12.2 percent fewer) than the non-salaried physicians. In short, even fairly modest changes in effort or productivity would have fairly dramatic effects on the supply of physicians' services. Participation in Health Insurance Programs In structuring a payment system insurers are often faced with three conflicting objectives: holding down the costs of the program, limiting beneficiaries' out-of-pocket costs, and setting rates that most physicians will accept. Although patients and physicians might initially welcome a system that paid a fixed percentage of charges, this arrangement does not appear to be satisfactory over the long run. Some screens are needed to catch clearly excessive charges. It also appears that, at least for insurers covering a substantial portion of the population, such a system leads to a rapid increase in billed charges. Limits on an insurer's allowed fees (often called reasonable fees) may expose patients to considerable out-of-pocket expense if physicians charge more than the allowed fee. If insurers attempt to limit allowed fees and to require that physicians accept the allowed fees as payment in full, beneficiaries' costs are held in check but physicians are likely to become increasingly unwilling to participate in the program. A good deal is known about the impact of the structure of the payment system on physicians' willingness to participate in insurance programs. The basic model was laid out by Sloan and Steinwald (1978) in their analysis of participation in the Blue Shield service benef it program. In fairly short order the approach was applied to Medicare and Medicaid by Sloan, Cromwell, and Mitchell ~1978~; Lee and Hadley ( 1978~; and Rogers and Musacchio (1983~. Some disagreements about fine points remain, but a consensus seems to have been reached about the general conclusions. Willingness to treat Medicaid patients is the simplest of the three cases, so examination of it allows us to see the outline of the standard approach. The basic rule is quite simple: The greater is the additional -55-
revenue from a private patient relative to the Medicaid fee, the less likely is a physician to treat Medicaid patients. In short, financial incentives depend on a comparison of revenues in the private market and revenues f ram Medicaid. For some physicians, revenues in the private market are high enough so that only a very significant increase in Medicaid fees would make treating Medicaid patients financially attractive. These physicians will, except for occasional acceptance of a Med icaid pat lent for humanitar iar~ reasons, assent tat ly opt out of the Medicaid program. For physicians with less profitable private practices however, Medicaid fees look more attractive, so an increase in Medicaid fees can significantly increase willingness to treat Medicaid patients. Paringer (1980) has shown that physicians who treat significant numbers of Medicaid patients charge their private patients an average of 5 percent less than physicians who do not. Sloan, Cromwell, and Mitchell (1978) estimate that for the 71 percent of physicians in their sample who treated Medicaid patients on a consistent basis, a 10 percent increase in Medicaid fees would lead to a 9.5 percent increase in the supply of services to Medicaid beneficiaries. At least for this group of physicians, a fairly strong response to changes in fees appears to exist. Even though the institutional arrangements are different for Medicare, similar results have emerged: The greater is the Medicare reasonable fee relative to the revenue expected from private patients, the more likely is the physician to accept the Medicare fee as payment in full. This finding has been replicated a number of times: Paringer, 1980; Rogers and Musacchio, 1983, Hadley and Lee, 1978. Two questions remain however. How much will an increase in the Medicare reasonable fee affect the proportion of cases in which the Medicare fee is accepted as payment in full? What impact will changes in Medicare fees have on the prices charged private patients? There are probably several different answers to the first question. Some physicians will find that fees are so low relative to the prices that they charge private patients that small changes will not affect their decisions not to assign claims. Some physicians will find that fairly modest increases in Medicare fees will make assignment an attractive option. Still other physicians will conclude that assignment was the best strategy before the Medicare f ee increase, and the increase will not affect their decisions. Empirical evidence of this diversity of responses is supplied by Paringer ( 1980 ~ . For physicians who treat substantial numbers of Medicaid patients (and thus accept fees less generous than Medicare), Medicare fees have little or no effect on assignment rates. For physicians who treat fewer Medicaid patients and whose pr ivate pr ices are higher, a 10 percent increase in Medicare f ees is estimated to result in a 13.5 percent increase in the assignment rate. Overall, it does not appear that an increase in reasonable fees would dramatically increase assignment rates. Rogers and Musacchio estimate that a 10 percent increase in reasonable fees would increase assignment -56-
rates by 3.8 percent. A weighted average of Paringer's estimates predicts a comparable 4.8 percent increase. An increase in program fees is likely to result in higher private fees as well as increased willingness to treat program beneficiaries. In essence, an increase in program fees increases the demand for physicians' services, and some physicians will raise prices as a consequence. No direct test of this hypothesis has been attempted, although the theory is generally accepted and there is evidence consistent with it (Lee and Hadley, 1981~. Two basic premises about payment systems are illustrated by research on participation in health insurance programs. First, the best strategy for a payment system depends on the strategies that others are following. So, whether a Medicaid fee is too high or too low depends on what patients and other insurers are willing to pay. Second, for large programs the impact of changes in the program on others in the system--providers, insurers, and patients--must be recognized. Considerable research on fee-for-service care and insurance programs has been undertaken, much of it at the behest of the Health Care Financing Administration. Very little is known about the role of prepaid plans and Medicare or Medicaid. To begin with, the public insurance programs were not designed to be compatible with prepaid groups. In addition to such legal and administrative barriers, Medicare and Medicaid benef iciaries have not been viewed as particularly good candidates for enrollment in prepaid plans. Medicaid benef iciaries tend to move off and on the rolls as their economic circumstances change. Such intermittent eligible lity makes them unsuitable candidates from the viewpoint of prepaid plans. Medicare beneficiaries could be candidates for prepaid plans, but many of the elderly have established physic~an-patient relationships, reducing the attractiveness of prepaid plans ~ especially in the absence of financial incentives to join). Recently, a number of prepayment demonstration projects have been funded by the Health Care Financing Administration. Only preliminary results are available, but it appears that it is possible to establish a capitation rate that will reduce Medicare costs, permit an extensive enough benefits package to attract Medicare eligibles and be consistent with the financial health of the HMO (Greenlick, et al., 1983~. It may be that prepaid plans will play a signif icant role in the payment system of the future. Conclusions In 1982 nearly 2 dollars of every 3 earned by physicians were paid by either public or private insurance (Gibson, et al., 1983~. As a result, changes in the structure of insurance programs have the potential to alter the f inancial rewards of different medical practices quite substantially. Growing pressure on insurers to control costs, combined 57
with a continuing rise in the number of physicians per capita, increases the likelihood that changes in the system of paying for physicians' services will emerge. These changes will probably result in lower fees for some services, hence lower incomes for physicians in some areas and in some specialties. How will physicians respond to such changes? A number of responses seem likely, including changes in specialty choices, changes In location cho ices, changes in practice set t ings, and changes in the mix of output produced. It appears that the impact of financial considerations is much greater for some career choices than for others. For example, the evidence thus far assembled suggests that differences in earnings have a much greater effect on physicians' location decisions than on their specialty choices. For several reasons, however, even this tentative conclusion should be viewed with caution. To begin with, no real consensus on the impact of earnings on location choices exists. Second, the data and methodological problems associated with analyses of specialty choices appear to be even more f ormidable than those associated with analyses of location choices. Consequently, relatively few studies of earnings and location choices have been attempted . Final ly, even If existing studies accurately portrayed the historical relationship between earnings expectations and location and specialty choices, those estimates might not predict the response to major shifts in relative incomes. In the face of significantly altered relative incomes, extrapolation from historical experience might well understate the response. In short, existing research does not allow us to predict with much confidence how much the pattern of location and specialty choices will shift if the payment system changes. Practice organization appears to have a ma jar impact on physicians' productivity and earnings. Physicians in small groups see more patients and earn more per hour than their counterparts in solo practice or in large groups. Compensation methods within groups also appear to af feet productivity, as salaried physicians work fewer hours and see fewer patients per hour than physicians paid on an incentive basis. The meaning of this evidence is far f ram clear, however. To begin with, physicians in different settings may be producing quite different products. Consequently, simple compar isons of visits may be misleading. Even so, differences in earnings reinforce the belief that there are real advantages to group practice and to incentive compensation methods. It remains to be shown to what extent it is the incentive structure per se, rather than the attributes of the physicians who choose different practice milieus, that lead to differences in productivity and earnings. Simple extrapolation of observed differences to physicians in general may prove to be misleading. 58
How much would physicians change their patterns of treatment if the relative prices of various services changed? We do not know. Studies of the impact of relative prices on the mix of patients that physicians choose to treat tend to f ind rather large effects, so the suspicion that changes in relative f ees might also lead to substantial changes in the mix of services provided cannot be dismissed out of hand. Careful analysis of this issue is long overdue. Primarily as a result of the resources that have been brought to bear on the question, the impact of relative fees on participation in insurance programs is the best understood of the issues discussed in this paper. Put simply, research consistently shows that physicians are more likely to participate in an insurance program the higher is the program fee compared to other fees. Analyses of participation in various insurance programs also show that the complex mix of private insurance, public insurance, and self payment forms a system, albeit an unplanned one. Changes in one component have repercussions for the rest of the system. By the same token, the impact of a policy change in one program depends on the policies being pursued by other insurers. For example, an increase in Medicaid fees may affect the prices that other payers are charged. Alternatively, the impact of a Medicaid fee increase depends in part on how it compares to others. This points out the advantages of coordinated public and private policy making. At the very least the interaction among insurers should be taken into account when strategies are being devised. This synopsis of research has focused on the most recent and most reliable evidence. Other, less satisfactory studies exist, but the studies cited here give a fair representation of the best evidence that we now have. It may be disheartening to see how significant are the gaps in our knowledge about the effects of the payment system on the supply of physicians' services. Still, even though the problems are difficult and resources are scarce, progress has been made. The limited good evidence that we have today would have seemed like a wealth of knowledge to the designers of payment systems 20 years ago. -59-
References Bailey, R.M., Economies of Scale in Outpatient Medical Practices, Group Practice {July, 1968~:24-33. Berry' C.J.' et al., A Study of the Responses of Canadian Physicians to the Introduction of Universal Medical Care Insurance: The First : Five Years in Quebec, ~ Princeton, NJ: Mathematica Policy Research, . 1978). Bobula, J.D., ~Physicians' Work Patterns under Alternative Methods of Reimbursement,. in Profile of Medical Practice, 1978, J.C. Gaffney, ed. (Chicago: American Medical Association, 1979~:17-27. Brown, D.M., and Lapan, H.E., Analysis of Physicians' Input Decisions, in Issues in Physician Reimbursement, N.T. Greenspan ea., (Washington, D.C.: Health Care Financing Administration, n. d. ): 135-145 . Cantwell, J.R., Implications of Reimbursement Policies for the Location of Physicians,. Agricultural Economics Research, 31 (April, 1979):25-35. Center for Health Services Research and Development, Prof lie of Medical Practice, 1981, D.L. Goldfarb, ed. ~ Chicago: American Medical Association, 19813. Fuchs, V.R., Kramer and M.J., Determinants of Expenditures for PhYsicianst Services in the United States 1946-1968, DREW Publication Number . (HSM) 73-3013, (Rockville, MD: National Center for Health Services Research and Development, 197 2 ~ . Gibson, R.M., Waldo, D.R. and Levit, K.~., National Health Expenditures, 1982,. Health Care Financing Review 5 (Fall 1983) :1-31. Golladay, F.L., Manser, M.E., Smith, K.R., Scale Economies in the Delivery of Medical Care: A Mixed Integer Programming Analysis of Efficient Manpower Utilization, Journal of Human Resources 9 (winter 1974~:50-62. Greenlick, M.R., et al., ~Kaiser-Permanente's Medicare PIUS Project: A Successful Medicare Prospective Payment Demonstration'. Health Care Financing Review 4 (Summer 1983~:85-97. -60-
Hadley, J., MA Disaggregated Model of Medical Specialty Choice, ~ in Research in Health Economics, Richard Scheffler, ed. {Greenwich, CT: JAI Press, 1979~. Harrison, J.L., Jud, G.D., MA Regional Analysis of Physician Availability,. paper presented at the Southern Economic Association Meetings, Houston, TX, November, 1973. Hay, J.W., Selectivity Bias in a Simultaneous Logit-OLS Model: Physician Specialty Choice and Specialty Income,. paper presented at the Harvard Conference on Econometric Modelling, Boston, MA, 1981. Held, P.J, The Migration of 1955-65 Graduates of American Medical Schools, unpublished Ph.~. dissertation, (Berkeley, CA: Department - of Economics, University of California, 1972~. Kimbell, L.J., Lorant, J.H., Physician Productivity and Returns to Scaled Health Services Research (Winter 1977) :367-379. Lee, R.H., Scholarship Programs and Medical Education Financing, ~ in Medical Education Financing, Jack Hadley, ea., (New York: Prodist, 1980 ): 128-148. Lee, R.H., Hadley, J., Supplying Physicians' Services to Public Medical Care Programs,. Urban Institute Working Paper No. 5925-4, (Washington, D.C.: The Urban Institute, 1979 ~ . Luft, H.S., show Do Health Maintenance Organizations Achieve Their 'Savings'?, New England Journal of Medicine 298, (June, 1978~:1336-1343. Mechanic, D., The Organization of Medical Practice and Practice Orientations Among Physicians in Prepaid and Nonprepaid Primary Care Settings Medical Care, 13 (March, 1975~:189-204. Paringer, L., Medicare Assignment Rates of Physicians: Their Responses to Changes in Reimbursement Policy, Health Care Financing Review 1 (1980~:75-91. Ramaswamy, K., Tokukata, G.K., Determinants of Expenditures for Physicians' Services in Pennsylvania, paper presented at the Joint Statistical Meetings, Atlanta, GA, August, 1975. Reinhardt, U., MA Production Function for Physician Services, The Review of Economics and Statistics, 54 (1971~:S5-66. - Rodgers, J.F., Musacchio, R.A., ~Physician Acceptance of Medicare Patients On Assignment,. Journal of Health Economics 2 (March 1983~:55-73. 61
Sloan, F.A., Lifetime Earnings and Physicians' Choice of Specialty, Industrial and Labor Relations Review, 24 (October, 1970) :47-56. - Sloan, F.A., Effects of Incentives on Physician Performance, ~ J. Rafferty, ea., in Health Manpower and Productivity ~ Lexington, MA: D.C. Heath, 1974) :53-84. Sloan, F.A., Cromwell, J., Mitchell, J.B. Physician Participation in - State Medicaid Programs, Journal of Human Resources 13 (1978), 211-246 . Sloan, F.A., Steinwald, B., Physician Participation in Health Insurance Plans: Evidence on Blue Shield, ~ Journal of Human Resources 13 ( 1978 ): 237-263. 62
