III.  Physician Requirements

Physician “requirements” refers to an assessment of the total number of physicians needed to provide a specified level of services to a given population. Estimating physician requirements and projecting future requirements is often the most difficult and controversial component of assessing the adequacy of the physician supply.  There exists no consensus on

• what constitutes an adequate level of services (given the needs of a population and its ability to pay for services),
• the relationship between requirements and certain key determinants, and
• how the characteristics of the health care system and other key determinants of physician requirements will change over time.

In the remainder of this Chapter we discuss approaches to modeling physician requirements, provide an overview of the PRM, describe the major determinants of physician requirements, and present projections from the PRM.

A.   Approaches to Model Physician Requirements

Approaches to estimate physician requirements range from estimating some socially optimal number (a needs-based analysis), to estimating the number that society will likely employ (a demand-based analysis), to some variation of these approaches.  In addition to simply extrapolating current physician-to-population ratios, four major approaches have been used to estimate physician requirements:

1. Needs-based approach.  The Committee on the Costs of Medical Care (CCMC) might be considered the first attempt to apply scientific principles to determine the adequacy of physician supply in the United  States.   In 1933, the CCMC published its finding that the Nation  needed 140.5 physicians per 100,000 population (an estimate that exceeded existing supply by 10 percent), and that 82 percent of the physician workforce should be generalists. CCMC reached its conclusion by estimating (1) the incidence of disease and other health problems, (2) the expected number of patient-physician encounters per incidence of disease, (3) the average amount of physician time per encounter with a patient, and (4) the average amount of physician time per year spent in patient care activities.  A major criticism of this approach is that it ignores the economic realities of the health care system. ^[16] Schroeder (1994) and others have criticized this approach as being open to bias because it relies heavily on the subjective assessments of expert panels.  A further complication of using this needs-base approach is that it fails to account for the technological changes in the practice of medicine which increase the ability to treat complex clinical conditions; these technological advances are also heavily concentrated in specialty care, thereby having a larger impact on the need or demand for specialists than for primary care.  GMENAC (1981) estimated physician requirements using an “adjusted” needs-based approach, similar to the CCMC study, but adjusted downward their initial requirements estimates to reflect “realistic” physician and patient behavior.
2. Demand/utilization-based approach.  The demand-based approach extrapolates current patterns of utilization of physician services taking into account changing demographics and trends in key determinants of the demand for physician services.  This approach relies primarily on empirical analysis to estimate the relationship between utilization of health care services and its determinants, and relies less on subjective assessments from a panel of experts.  This approach forms the basis for HRSA’s requirements models, as well as for numerous studies on individual clinical specialties.  A major criticism of the demand-based approach is that because it extrapolates current health care utilization and service delivery patterns, inequities in the current system are carried into future requirements projections.
3. Benchmarking approach.  Benchmarking involves the identification of a certain standard of care, and then extrapolating that standard to a different population.  Examples of benchmarks include: (1) physician staffing patterns in HMOs (e.g., Weiner, 1994; Weiner 2004), and (2) physician-to-population ratios in other countries.  Some utilization-based projections of physician requirements could be considered benchmarking, where the physician staffing patterns in future years are compared to physician staffing patterns in the reference year.  The implicit assumption of benchmarking is that the benchmark (e.g., HMO, country, time period, etc.) reflects an efficient (or at least adequate) mix and number of physicians for the population served.   A challenge with using this approach is that health care delivery systems might be very different between the benchmark entity and the population of interest such that a comparison of physician-to-population ratios requires substantial adjustments.  For example, the role of primary care physicians in the United States  might be quite different than the role of primary care physicians in other countries, which complicates the comparison of simple statistics such as physician-to-population ratios. Weiner (2004) studied three prepaid group plans and found that compared to the U.S. population as a whole, physician-to-population ratios at the three prepaid group plans were about 25 percent lower for primary care physicians and 32 percent lower for specialists.
4. Trend analysis approach.  Cooper (2000) and colleagues (2002) use a “Trend Model” that estimates the correlation between a proxy for the demand for physician services (they use physician-to-population ratios under the assumption that historically, supply has equaled demand) and factors hypothesized to be major determinants of demand.  Cooper uses aggregate-level, time-series data to estimate the relationship between physicians-per-population and its hypothesized determinants: per capita GDP and demographics to control for population growth and aging.  Cooper concludes that there is a trend to desiring a higher level of care (specialist care, in particular) that is limited primarily by our ability and willingness to pay.

B.  Physician Requirements Model Overview

The PRM uses a utilization-based approach to project future physician requirements.  The PRM projects requirements for 18 medical specialties through 2020.  Projections are based on current use patterns of physician services and expected trends in U.S. demographics, insurance coverage, and patterns of care delivery.  These use patterns are expressed as physician-to-population ratios for each specialty and population segment defined by age, sex, metropolitan/non-metropolitan location, and insurance type.  The baseline ratios are established using 2000 data.  Thus, the major components of the model are:

1. Population projections by age, ^[17] sex, and metropolitan/non-metropolitan location;
2. Projected insurance distribution by insurance type, age, sex, metropolitan/non-metropolitan location; and
3. Detailed physician-to-population ratios (Exhibit 27).

Exhibit 27. Overview of the Physician Requirements Model

The model’s base year is 2000, which means that the model’s baseline scenario projects growth in demand for physician services based on the level of care provided to the U.S. population in 2000.  Population growth and aging, as projected by the U.S. Census Bureau, are the main drivers of growth in demand for physician services in the baseline projections.  Alternative scenarios are projected using different assumptions regarding changes in insurance coverage and type, economic growth, and the increased use of NPCs.  We explore trends in major determinants of physician requirements and their implications for future physician requirements.

C.  Determinants of Physician Demand

Physician requirements derive from the demand for physician-related services.  The demand for such services is the outcome of countless decisions made by consumers, physicians providing services, and other entities involved in the health care system such as insurers.  The PRM is a simplified model of a complex health care system and tries to capture the major trends that affect the demand for physician services, and thus the number of physicians needed to provide that level of service.  The major determinants of physician requirements are population growth and aging, changes in medical insurance coverage and type, economic growth, the growing role of NPCs, advances in science and technology, changing public expectations, the price of physician services, and government policy.  We discuss each of these determinants in turn.

1.   Population Growth and Aging

The United States Census Bureau projects a rapid increase in the elderly population beginning in 2010 when the leading edge of the baby boom generation approaches age 65 (Exhibit 28). Between 2005 and 2020, the population younger than age 65 is expected to grow by about 9 percent, while the population age 65 to 74 is projected to grow by about 71 percent and the age 74 and older population is projected to grow by about 26%.

Exhibit 28. Projected Percent Growth in Population, by Age: 2005 to 2020

[D]

Source: United States Census Bureau population projections (April 2005 release).

The elderly, especially those over age 85, use much greater levels of physician services relative to the non-elderly, so the rapid growth of the elderly population portends a significant increase in demand for physician services.  To estimate differences in use of physician services by different demographic groups, for each physician specialty we estimated per capita encounters for segments of the United States population categorized by age, sex, and insurance status.  We analyzed health care use data from the National Ambulatory Medical Care Survey (NAMCS), National Hospital Ambulatory Medical Care Survey (NHAMCS), National Inpatient Sample (NIS), National Nursing Home Survey (NNHS), and National Home and Health Survey (NHHS) (BHPr, 2003).  After determining what portion of physicians’ time is spent with each segment of the population, we calculated physician-per-population ratios that reflect current use patterns and current patterns of care.

For presentation purposes, these ratios are summarized in estimates of physician requirements per 100,000 population for four categories of physicians and six age groups (Exhibit 29).  In 2000, for the United States population as a whole, approximately 253 active physicians (MDs and DOs) were engaged primarily in patient care per 100,000 population.^[18] The aggregate estimates ranged from a low of 149 for the population ages 0 to 17, to a high of 781 for the population ages 75 and older.  The ratios vary substantially by medical specialty.  These data suggest that the aging of the population will contribute to faster growth, in percentage terms, for specialist services relative to the growth in demand for primary care services.

Exhibit 29. Estimated Requirements for Patient Care Physicians per 100,000 Population, by Patient Age and Physician Specialty, 2000

Source: PRM. ¹ Includes general and family practice, general internal medicine, and pediatrics. ² Includes cardiology and other internal medicine subspecialties. ³ Includes general surgery, obstetrics/ gynecology, ophthalmology, orthopedic surgery, otolaryngology, urology and other surgical specialties. ⁴ Includes anesthesiology, emergency medicine, pathology, psychiatry, radiology, and other specialties.

The PRM segments the U.S. population into 176 mutually exclusive categories based on age, gender, metropolitan/non-metropolitan location, medical insurance status, and insurer type. The PRM considers differences in per capita health care utilization in each of these 176 categories and uses this information to estimate physician-to-population ratios in the base year for each population category.  Combining these physician-to-population ratios with population projections creates the baseline demand projections in the PRM.

2.   Medical Insurance Coverage and Type

Whether a person has medical insurance and type of insurance plan are important determinants of the amount and type of physician services utilized.  Insured persons have greater access to physician services, relative to the uninsured, because insurers typically negotiate discounts with providers and cover much of the cost of services.   Modeling insurance coverage and type is especially important for projecting how demand for physician services will likely change under alternative insurance scenarios (e.g., implementing Federal policies or programs that expand medical coverage), and in response to trends in the health care system (e.g., such as changes in managed care enrollment rates).

Managed care plans attempt to control health care costs through the use of gatekeepers, preferred providers, utilization review and other managed care practices.  The lower physician-to-population ratios among managed care plans form the basis for Weiner’s (2004) analysis that suggests that the Nation could get by with substantially fewer physicians—especially specialists.

The PRM models 10 insurance categories.  These include four payer categories: private insurance; Medicare (defined for modeling purposes as all government-sponsored insurance for the age 65 and older ^[19] ); government-sponsored insurance for the age < 65 (which is primarily Medicaid); and uninsured.  The three insured categories are each divided into three subcategories: traditional fee-for-service; exclusive network HMO (i.e., group-, staff-, network- or mixed-model HMO); and all other managed care plans (to include preferred provider organizations [PPOs], point-of-service [POS] plans organized as open-ended HMO, non-HMO POS, and other HMO/managed care plans).

The PRM starts with the insurance distribution in 2000 (Exhibit 30), with the probability of being in a particular insurance category differing by age and sex.  The baseline projections in the PRM assume that, controlling for age and sex, the probability of being in a particular insurance category remains constant over time.

The PRM uses an index to scale the physician-to-population ratios used for each demographic group in different insurance categories (Exhibit 31). ^[20] In this index, the fee-for-service (FFS) setting is used as the comparison group and has an index value of 1.  Consider the index values for anesthesiology.  The value of 0.86 for the exclusive network HMO indicates that a person enrolled in an HMO will tend to utilize only 86 percent of anesthesiologist time, on average, compared to a similar person insured under a fee-for-service plan.  Enrollees in other types of managed care plans will use approximately 98 percent and the uninsured will use only 29 percent of anesthesiologist time, on average, relative to a comparable population insured under a fee-for-service plan.

Exhibit 30. Estimated U.S. Population, by Insurance Status: 2000

Source: Analysis of the 1999, 2000 and 2001 NHIS.

Exhibit 31. Per Capita Index for Use of Physician Services (relative to a fee-for-service)

Source: BHPr (2003). ¹ Estimates capped.

3.   Economic Growth

Economic theory and empirical research suggest a positive correlation between ability to pay for physician services and demand for such services. ^[21] At the micro level, the ability of an individual or a household to afford medical insurance and out-of-pocket expenses influences whether a person seeks needed medical services.  At the macro level, the Nation’s ability to pay determines the number of persons who receive medical insurance and the generosity of such insurance in terms of services covered and out-of-pocket costs to beneficiaries. ^[22]   For example, during an economic expansion, employers might provide more generous medical benefits to attract and retain employees (Christianson and Trude, 2003).  Economic growth also affects tax revenues, which in turn affect the ability of the Federal and State governments to fund programs such as Medicaid, Medicare, and the Children’s Health Insurance Program (CHIP).

Income elasticity, the economic term for a measure that quantifies the relationship between ability to pay and demand, is defined as the percent increase in demand for physician services for each 1 percent increase in ability to pay.  While the direction of the relationship between ability to pay and overall demand for physician services is clear, there exists no consensus regarding the size of this relationship.  Obtaining precise estimates of this relationship is complicated by several factors:

1. The relationship likely varies by medical specialty (e.g., elective and cosmetic procedures being among the most sensitive to ability to pay).  Cooper et al. (2002) find a positive correlation across States between the number of active physicians per capita (which they use as a proxy for demand) and personal income per capita, with the relationship being stronger for specialists compared to generalists.
2. The relationship reflects decisions made at the household level (e.g., whether or not to visit the doctor), at the employer level (e.g., whether to offer medical insurance), and at the societal level (e.g., whether or not to expand a government-sponsored medical insurance program).
3. The relationship is distorted by the nature of a three-party health care system—patients, physicians and insurers.  Once insured, most patients are relatively shielded from the costs of physician services and physicians are less constrained by patients’ ability to pay. This tends to be true regardless of whether the insurance is public or private.
4. The relationship is confounded by variables that are correlated with both ability to pay and utilization of physician services (e.g., health status, adequacy of physician supply, and implementation of new technology).

The following is a brief summary of key and recent studies of the relationship between ability to pay and demand for health care or physician services, as well as our own empirical analysis.

Cooper et al. (2002, p. 143) state that “the major trend affecting the [per capita] demand for physician services is the economy.” Assuming that historical, national rates of physicians per capita reflect demand for physician services, the authors estimate the relationship between physicians per capita and per capita GDP using annual data from 1929 to 2000.  The authors conclude that each 10 percent increase in per capita GDP results in a 7.5 percent increase in demand for physician services (i.e., income elasticity [e]=0.75).  They view changes in economic growth over time as both an indicator of increased ability to pay and a proxy for technological advances, and argue that because of increased ability to pay for health care services consumers will demand a higher level of services in the future than is provided under the current health care system.  Cooper et al. project the future supply of and demand for physicians and conclude that we face a looming shortage of specialists.

Other researchers have expressed concerns with Cooper et al.’s assumptions and conclusions (e.g., Barer, 2002; Grumbach, 2002; Reinhardt, 2002; Weiner, 2002).  One critique is that the authors do not establish a causal relationship between economic well-being and demand for physicians despite the finding of a statistical correlation.  Another critique is the assumption that physician supply and demand were in equilibrium during the 70 year period included in the analysis, which assumption is necessary when using physicians per capita (a supply measure) as a proxy for demand.  Cooper et al.  assume that in the past market (and other) forces helped keep a balance of supply and demand, but in the future supply and demand will diverge.

Exhibit 32 illustrates one of the problems with using a simple correlation between physicians per capita and measures of economic well-being to estimate the relationship between demand and ability to pay.  This exhibit shows a positive correlation across States between personal income per capita (controlling for cross-State differences in cost of living) and physicians per 100,000 population.  Two series are plotted: (1) a series that adjusts for out-of-State consumption of health care services, and (2) a series that does not adjust for out-of-State consumption of health care services.   The adjusted series is computed based on work by MEDPAC (2002), which estimates Medicare payments per beneficiary with and without adjusting for out-of-State service use and health status.  The unadjusted series suggests that each $1,000 increase in personal income per capita increases the number of physicians per 100,000 population by approximately 9.7.  Evaluated at the mean personal income per capita, this translates to an elasticity of 1.12, meaning a 10 percent increase in personal income per capita is correlated with an 11.2 percent increase in physicians per capita.  The series that adjusts for out-of-State consumption of health care services suggests that each $1,000 increase in personal income per capita results in an increase of 3.8 physicians per 100,000 population.  Evaluated at the mean personal income per capita, a 10 percent increase in personal income per capita is correlated with a 4.7 percent increase in the supply of physicians.  The 95 percent confidence interval for this estimate is quite large, though, ranging from 0.06 to 0.87.

This comparison of the adjusted and unadjusted series suggests that States with higher per capita income are net exporters of physician services.  Controlling for where patients receive services explains away approximately 60 percent of the observed cross-State relationship between physicians per capita and personal income per capita.  While the adjusted series still shows a positive correlation between personal income per capita and physicians per capita, the estimated relationship is substantially smaller than Cooper et al.’s estimate, which in turn is substantially smaller than the estimate from the unadjusted series.

Exhibit 32. Relationship across States between Real Per Capita Personal Income and Physicians per 100,000 Population: 2001

[D]

Sources: Income estimates from the U.S. Bureau of Economic Analysis.  Population estimates from the U.S. Census Bureau.  State-level estimates of physician supply from AMA’s Physician Characteristics and Distribution in the U.S. (AMA, 2003).  Note: The District of Columbia, which is omitted from this graph, has a physician-to-100,000 population ratio of 698.

Anderson et al. (2003) report statistics that show a positive correlation between a country’s per capita GDP and the percentage of GDP spent on health care for the  Organization for Economic Cooperation and Development (OECD) countries (Exhibit 33). ^[23]   Although the United States  spends a disproportionate amount of GDP on health care relative to other countries, Anderson et al. attribute this phenomenon largely to higher prices for health care goods and services in the United States rather than higher utilization of such goods and services.

Exhibit 33 Relationship Between GDP Per Capita and Total Health Spending as a Percent of GDP in 2000: OECD Countries

[D]

Source: Anderson et al (2003).

A visual inspection of Exhibit 33 suggests that the relationship between per capita GDP and percent of GDP spent on health care is strongest among countries with per capita GDP less than $25,000.  Close to 60 percent of the OECD countries have per capita GDP in the $25,000 to $35,000 range, and among this subset of countries there is little relationship between per capita GDP and the percent of GDP spent on health care.

Using data on physicians per capita from Anderson et al., we estimated the relationship between physicians per capita and per capita GDP by estimating a log-log model using a bivariate regression analysis.  The estimated income elasticity is 0.4, which suggests that each 10 percent increase in per capita GDP is associated with a 4 percent increase in physicians per capita.  The estimated 0.4 elasticity is roughly half the 0.75 estimate found by Cooper et al., but the 95 percent confidence interval is large, ranging from 0.10 to 0.70.

Exhibit 34 plots the relationship between per capita GDP and supply of physicians per 100,000 population for OECD countries. ^[24] The number of physicians per capita in the United States falls below the trend line, suggesting that demand for physicians exceeds supply if economic well-being is the major determinant of demand for physicians.  A caution against drawing causal conclusions from observed statistical correlations, such as presented here, is the failure to control for differences across countries in the health care system, demographics, and indicators of health care needs.  For example, the United States has a much greater supply of other trained health workers (e.g., NPCs, nurses, technicians, etc.) compared to other countries, which distorts the observed simple relationship between physicians per capita and measures of economic well-being. Demographics, lifestyle, the public health infrastructure, and other important determinants of physician demand vary substantially by country.

Exhibit 34 Relationship Between GDP Per Capita and Physicians Per 100,000 Population in 2000: OECD Countries

[D]

Source: Anderson et al (2003).

Koenig et al. (2003) estimate the relationship between income and expenditures for physician services both using aggregate-level data and using beneficiary-level data.  The authors simultaneously control for nine categories of factors hypothesized to affect expenditures for physician services: demographics, health status, insurance product and benefit design, provider supply and organization, provider payment, practice operating costs, health care regulation, medical technology, and economic activity.  One analysis uses State-level data from 1990 to 1998 to estimate the relationship between per capita expenditures for physician services and various explanatory variables.  This analysis produces an income elasticity estimate of 0.76 with a 95 percent confidence interval of 0.57 to 0.95.  The second analysis uses data on beneficiaries enrolled in a national preferred provider organization from May 1998 to May 2000.  This second analysis produced an income elasticity estimate of 0.31 with a 95 percent confidence interval of 0.10 to 0.53.  This lower estimate, however, is based on a subset of the population that is insured and thus excludes the income impact on ability to purchase medical insurance.

Cookson and Reilly (1994) model health care consumption using a trend analysis, where national health care expenditures is modeled as a function of numerous factors including measures of real personal income.  These authors find a significant lagged “wealth effect,” with change in real income being a statistically significant predictor of change in health care expenditures 3, 4 and 5 years into the future.  The estimated income elasticities for 3, 4 and 5 years are, respectively, 0.17, 0.39 and 0.33.  The cumulative effect is approximately 0.88, which implies that a 10 percent increase in real per capita personal income eventually translates into an 8.8 percent increase in health care expenditures.

The findings and methods used in these studies are summarized below (Exhibit 35).  The estimated elasticities range from 0.31 to 0.88, with relatively large standard errors on these estimates.

Exhibit 35. Income Elasticity Estimates

The extant research sheds little light on how to incorporate national economic growth into the PRM projections.  Empirical questions raised include:

• How does the relationship between national economic growth and demand for physician services differ by medical specialty?  Theory would suggest that elective services (e.g., some plastic surgery services) are more responsive to ability to pay than are less elective services (e.g., cardiologist services).  Indeed, when Cooper et al. look at the relationship between States’ physician per population ratio and income per capita they find a stronger relationship for specialists compared to generalists.  Therefore, separate estimates of income elasticity are needed for each specialty.
• To what extent is the relationship between national economic growth and demand for physician services already built into the PRM via the insurance distribution assumptions?  If the main effect of national economic growth is to move people into more generous insurance status (e.g., from uninsured to insured, from managed care into fee-for-service), then the research needed to incorporate economic growth into the PRM is very different from the research needed to estimate health care utilization changes within insurance status.
• Does the size of the relationship between ability to pay and demand for physician services diminish at higher income levels?  As discussed earlier, when using country-level data the relationship between per capita GDP and health care measures such as percent of GDP spent on health care and physicians per capita seems to diminish as a country’s per capita wealth increases.

The argument that economic growth should be considered in the PRM projections raises a philosophical question regarding how the PRM projections should be used.  Incorporating economic growth into the model will result in higher demand projections because the hypothesized impact is that consumers will demand higher-quality care as the Nation’s wealth increases.  Historically, projections of an impending physician shortage have had large policy implications and have been instrumental in securing additional Federal funding for training new doctors.  Should the government role be to help ensure sufficient supply to meet demand (even if society wants the “sport-utility vehicle [SUV] version of health care” as discussed by Grumbach [2002])?

For this study we project the future demand for physicians under alternative scenarios.  The baseline scenario omits economic growth, and in essence projects future demand for physicians assuming the same level of care that is currently supplied is provided in the future.  An alternative scenario incorporates economic growth.  Projections under this alternative scenario are presented in a later section, but these alternative projections assume the following:

• Economic growth of 2 percent annually between 2000 and 2020.  The CBO projects a 3 percent annual growth rate in real GDP between 2003 to 2013, which is approximately equal to about a 2 percent average annual growth in real per capita GDP.
• Income elasticity of 0.25, 0.5, and 0.75 varies by specialty.  There is no consensus on what the income elasticity of demand is for physician services.  Physician requirements are projected under the assumption that demand for some specialties is relatively insensitive (elasticity=0.25) ^[25], modestly sensitive (elasticity=0.50) ^[26], or more sensitive (elasticity=0.75) ^[27] to economic growth.

Real, per capita economic growth occurs through increased productivity.  The CBO’s economic growth projections, therefore, imply that productivity will increase by approximately 2 percent annually, on average, throughout the economy.  Productivity growth will differ by industry and occupation, and because physician services are labor intensive it is reasonable to assume that growth in physician productivity (however measured) will lag behind overall productivity growth in the economy.  An increase in physician productivity of 1 percent annually would offset the increased demand for physician services under the scenario and accompanying assumptions described above.  Trends in physician productivity are discussed in more detail in a later section, but the limited available data suggest physician productivity appears to be growing at about 1 percent annually.

4.   Role of Nonphysician Clinicians

Three trends during the past decade have increased the proportion of health care services being provided by NPCs (Cooper, Laud and Dietrich, 1998; Druss et al., 2003).  First, the number of NCPs has grown substantially.  Second, State legislatures have expanded the legal scope of practice for NCPs (Cooper, Henderson and Dietrich, 1998).  Third, pressure to contain rising health care costs has fueled demand for NPCs.

The size of the NPC workforce has grown significantly in recent years and is projected to continue growing rapidly.  Assessments of the adequacy of the future physician supply should take into consideration changes in the growth and use of NPCs.  The BLS projects that between 2000 and 2010 employment of PAs will increase 53 percent, employment of chiropractors will increase 24 percent, employment of optometrists will increase 19 percent, and employment of podiatrists will increase 11 percent.  Cooper et al. (2002) report that by 2015 there could be as many as 525,000 NPCs—including 275,000 nurse practitioners, physician assistants and nurse –midwives; 150,000 acupuncturists and chiropractors; and 100,000 other NPCs engaged in various medical specialties.  These authors calculate that by 2015 these NPCs could be providing services that are the equivalent of 40 physicians per 100,000 population (with most of the services displacing services that historically have been provided by primary care physicians).

Physicians view NPCs in their role as both a complement and a competitor in the provision of care (Grumbach, 1998).

1. NPCs as complements to physicians.  The view of NPCs as complements to physicians is one in which NPCs allow physicians to leverage their expertise.  In this model of care, there is a division of labor such that NPCs provide those services within the scope of their training, with physicians handling the more complex cases.  Druss et al. (2003) find evidence of increased collaboration between NPCs and physicians over the period 1987 to 1997, and Trude (2003) reports that between 1997 and 2001 the proportion of physicians in non-institutional practice settings who worked with NPCs increased from 40 percent to 48 percent.  The increase in working with NPCs was most noticeable for group practices of three or more physicians where the percentage of physicians working with NPCs increased from 53 percent to 66 percent between 1997 and 2001.  Part of the increased collaboration between physicians and NPCs might be considered voluntary on the part of physicians (e.g., physicians hiring NPCs), while part of the increased collaboration might be considered imposed (e.g., in the case of a managed care company that hires both physicians and NPCs).  A recent survey by Farber and Murray (2001) found that 19 percent of responding physicians indicated that to counter lagging income they are considering hiring nurse practitioners or physician assistants.
2. NPCs as competitors to physicians.  As noted by the American Academy of Physician Assistants (AAPA, 1999) and others, the increasing financial uncertainties in the health care system has increased concerns by physicians about encroachment into their practice territory by physicians in other specialties and by NPCs.  Cooper, Henderson and Dietrich (1998) document legislation passed by States giving NPCs greater business and clinical autonomy.  Because NPCs can provide some services currently offered by physicians but at a lower cost, there exists an economic incentive to use NPCs to provide services within their legal scope of practice.

Whether viewed as competitors or complements in delivering care, the increasing size of the NPC workforce will likely reduce demand for physicians.  The growing role of NPCs raises several questions pertinent to modeling future demand for physicians in specific specialties:

1. How large is the overlap in services provided by NPCs and physicians.  That is, what proportion of physician workload can legally and safely be transferred to NPCs?
2. Although the supply of NPCs is growing rapidly, is there a saturation point at which there will be a NPC surplus?  Will all NPCs trained displace physicians, or will the United States reach a point at which there is no longer the ability to partially substitute NPCs for physicians?
3. To what extent are NPCs practicing in markets that are left unfilled by physicians? Physician assistants and nurse practitioners are disproportionately employed in rural areas that have difficulty attracting physicians.

Cooper, Laud, and Dietrich (1998) analyzed State distributions of physicians and NPCs per 100,000 population and found that for most NPC specialties, there were higher NPC-to-population ratios in States that also had high physician-to-population ratios.  This finding suggests that economic growth is associated with higher demand for health care services, with little economic-related preference for physicians over NPCs.  This analysis, though, does not control for cross-State variation in factors that would be correlated with demand for both NPC and physician services (e.g., health care needs of the population).

The baseline projections in the PRM assume that current patterns of health care delivery will continue over the projection horizon.  Physicians and NPCs will continue to have overlapping scopes of practice, and each will maintain their share of total services provided.  An alternative scenario assumes that the increased supply of NPCs will retard growth in demand for physicians by capturing a growing percentage of total patient volume.  Projections for this scenario are produced for physicians in the aggregate, not by specialty.Specific assumptions are that (1) the number of active NPCs will double between 2000 and 2020, (2) all NPCs that are trained will become employed ^[28] and will provide services that otherwise would have been provided by physicians, and (3) on average each NPC will provide 40 percent of the work currently provided by a physician.  The projections under this scenario are presented later.

5.   Science and Technology

Advances in science and technology have great potential to affect the demand for physician services and thus physician requirements.  Most health workforce projection models, however, do not include trends in science and technology in the set of determinants because there is too much uncertainty regarding how new innovations will impact the supply of or demand for certain health care services. Furthermore, technological advances will likely have differing impacts on individual physician specialties.  Advances in science and technology have the capability to affect the demand for physicians by:

1. Creating additional demand for physician services so that consumers seek treatments that previously did not exist (e.g., fertility treatment) or that can now be provided at lower cost (e.g., minimally invasive surgery);
2. Increasing physicians' productivity  (which increases the amount of services that physicians can provide and thus reduces the number of physicians needed to adequately serve a given population); and
3. Eliminating certain illnesses or otherwise reducing the amount of time needed to treat certain health problems, thus reducing demand for physician services.
4. Increasing longevity that may eventually increase the demand for physician services as patients age and seek care for other health problems.

The following are areas of scientific and technological advancement with the potential to affect physician supply and demand.

• Information technology: Physician reliance on electronic medical records is increasing rapidly and will only continue to do so in the next decade.  The goals of these systems are to “enhance patient safety and reduce the amount of paperwork for clinicians, giving them more time to devote to patient care” (Morrissey, 2004).  Efforts to expand electronic data collection and allow for interoperability (exchange of data within a facility) and portability (exchange of data across facilities) are underway.  Significant hurdles, such as privacy concerns and system compatibility, remain.  Improvements in physician productivity likely would reduce the demand for physicians because fewer physicians would be needed to provide the same level of services.
• Emerging forms of communication: The Internet, telemedicine, video conferencing, and telesurgery are emerging innovations with the potential to transform the way some physician services are delivered.  Currently, most physician services are provided through face-to-face encounters with patients.  The potential for technology to reduce the importance of geography in providing physician services has mixed implications for the amount of physician services utilized.  There is the potential to increase utilization by improving access to care by rural and other underserved populations.  There is also the potential for outsourcing some services (e.g., interpreting scans and test results) to physicians in foreign countries.  Telemedicine could potentially affect all medical specialties, but the greatest current applications are found in providing diagnostic-related services in radiology, pathology, and cardiology.  Advances are also being made in systems that use robotics, minimally invasive surgical gear, and video equipment to provide surgical services remotely.

Miller and Derse (2002, p. 168) opine that “the emergence of the Internet portends a dramatic shift for health care and the relationships of patients and physicians,” potentially improving the quality, timeliness, and efficacy with which physician services are provided.  Miller and Derse document that physician services currently provided over the Internet include: (1) proscribing medications, (2) responding to consumers’ medical questions, (3) providing psychotherapy, (4) reviewing biopsies and medical records, and (5) providing second opinions to a patient’s physician.  The overall impact of patient Internet access to physicians is uncertain.  Such access will likely increase the level of correspondence between patients and physicians, but this increased demand for physician services might be offset by increased efficiency in the delivery of services.

While face-to-face encounters between physician and patient will continue in importance, these other forms of communication have great potential to alleviate the geographic misdistribution of physicians.  For example, email may facilitate more productive interactions between patients and physicians without necessitating an office visit.  Email, while not having the urgency of a phone call, allows both the patient and the physician to communicate efficiently.  Physicians might find email useful for sending additional medical or condition-related information, thereby potentially reducing the need for an office visit.

Although the technology exists to provide a larger proportion of physician services remotely, third-party payers for medical services have been slow to change their reimbursement structure to compensate physicians for services provided remotely.  Currently, only capitation allows physicians to fully realize the benefits of communicating with patients via nontraditional means such as email.  Thus, reimbursement system and regulatory changes are needed before these new forms of communication with patients can be fully implemented (Terry, 2000).

• Minimally invasive surgery:  Most surgical specialties increasingly use minimally invasive surgery (MIS) in place of open surgery.  The benefits of MIS over open surgery include reducing patient recovery time, risk of infection, pain, and scaring.  MIS requires greater surgical skill, and sometimes longer operating times are required compared to open surgery.  Because MIS greatly reduces the direct medical costs, patient indirect costs (e.g., lost time from work), and patient suffering, advances in MIS have increased demand for such procedures.  For example, between 1990 and 1994 the number of patients that elected to have a cholecystectomy increased 84 percent (AHA TrendWatch, 2002).  Advances in MIS, therefore, will likely increase demand for surgeons because of increased time to perform surgeries and increased demand for such surgeries.  These advances might decrease demand for hospitalists and primary care physicians because MIS reduces patient recovery time and risk of complications.
• Diagnostic equipment: Advances in diagnostic equipment such as X-rays, ultrasound, CT and MRI scans allow physicians to identify health problems earlier and with greater accuracy.  Technological advances that increase the ability to diagnosis health problems and reduce the cost of such equipment will likely lead to increased demand for physicians (e.g., radiologists, neurologists) and technicians who provide these diagnostic services.  The ability to more quickly identify health problems has a mixed effect on demand for physician services.  There will likely be increased demand for services to correct health problems, while the ability to identify health problems at an earlier stage could reduce the need for more extensive physician services that would occur if a health problem becomes more advanced.
• Pharmaceutical and vaccination advancement:  Rational Drug Design, the development of new chemical and molecular entities by looking at the physical structure and chemical composition, continues to shorten the drug discovery process.  The discovery of new and more potent pharmaceuticals with fewer side effects will enable physicians to treat patients quickly and effectively, reducing the need for time consuming treatments and surgeries.  In addition, pharmaceutical advancement will bring with it an increase in prophylactic drugs and vaccines.   These in turn will reduce the overall burden of illness and help alleviate certain chronic conditions.  Prophylactic and therapeutic pharmaceuticals and vaccines will reduce the demand for physicians by both reducing the number of face-to-face visits and surgical procedures required to treat a patients.  (RWJF, Health and Health Care 2010). Pharmaceutical and vaccination advancement, therefore, is likely to reduce overall demand for physician services.
• Genetic mapping, genetic testing, gene therapy, and transplantation: Genetic mapping, genetic testing, gene therapy, and transplantation are technologies which will come to play an increasingly important role in medicine in the future.  New innovations can increase demand for specific tests and therapies, but the ability to cure diseases and other ailments creates the potential to reduce overall, long-term demand for physician services.  Therefore, the likely long-term impact on demand for physician services is unknown and will likely differ substantially by medical specialty.

6.   Public Expectations

Public expectations of medicine are different today than they were 100 years ago, or even 20 years ago.  New medicines have improved the ability to care for chronic conditions and have improved the quality of life for many individuals.  The Institute of Medicine (2000) has highlighted the prevalence of medical errors which has led to increased scrutiny of quality of care by the public and by policymakers.  The elderly baby boom population will not have experienced the same hardships as their grandparents which may also affect their expectations of the health care system.  Rising public expectations will increase physician requirements.

7.   Price of Physician Services

In most economic models, the price of goods and services is a major determinant of the quantity of such goods and services demanded.  Likewise, theory would suggest that as the price of physician services rises (falls), utilization of such services would fall (rise).  The scenarios modeled in the PRM assume no changes in the relative price of physician services over the projection horizon.

The responsiveness of physician requirements projections to changes in the cost of services is an area for future research.  There are many challenges to quantifying the price of physician services, projecting how these prices will change over time, and determining the likely impact of price changes on utilization of physician services.  The following are a few such challenges:

1. The quality of services changes over time due to technological advances and increased skill and sophistication of the services provided, thus increasing the difficulty of obtaining historical price data on a give set of physician services.
2. It is difficult to predict how the price of physician services will change over time relative to the price of other goods and services.  Because physician services are labor intensive and thus less likely to benefit from technological advances that increase productivity, it is reasonable to assume that physician services will become relatively more expensive over time.  That is, the cost per unit of physician services is likely to become more expensive relative to the cost per unit of food, transportation, etc.
3. Additional research is needed to determine the price elasticity of various goods and services—that is, how responsive health care utilization is to changes in price.  Price elasticities are particularly difficult to estimate for health care services because our third-party payment system reduces the cost-consciousness of patients and physicians making health care utilization decisions.  Unlike the purchase of most goods and services, the majority of patients are shielded from the true cost of health care services.  Employers and insurers, however, are relatively sensitive to the cost of health care services and have a strong financial incentive to keep physician payments to a minimum.  Employers demonstrate their sensitivity to health care costs through the selection of insurers.  Many insurers demonstrate their sensitivity to the cost of physician services through the use of selective contracting with physicians and the implementation of other managed care techniques.  
4. Additional research is needed to quantify trends in the total price of medical services—including physician charges for services and indirect costs (e.g., wait times) associated with seeking services.

A final note regarding the price of physician services is that market pressures (e.g., competition among physicians, bargaining clout of employers and insurers) will help to correct imbalances in physician supply.  A shortage (surplus) of physicians will tend to drive prices up (down), thus reducing (increasing) per capita utilization of services.

8.   Government Policy

The changing role of government, which is closely linked to public expectations, may also have a significant impact on the demand for physician services.  This includes the impact of regulation as well as payment policies.  Policies that might increase demand for physician services include more generous Medicare and Medicaid benefits, while policies that might reduce physician requirements include giving NPCs greater clinical and business autonomy, and efforts to ration or otherwise limit access to certain services.

Government Medicare projections take into account expected changes in physician behavior resulting from changes in program policies and practices.  For example, Nguyen (1994) finds that legislation that reduced physician fees for providing Medicare services in 1989 and 1990 had the unintended consequences of increasing the volume of services provided.  Nguyen found that that each planned dollar decrease in physician payments due to fee reductions was offset by a $0.40 increase in payments attributed to higher volume.  Nguyen attributes this volume offset to physician behavior motivated by a desire to maintain earnings.  The size of the volume offset differs by medical specialty, with surgical specialties showing a larger volume offset, on average, compared to non-surgical specialties.  Nguyen cites other studies with similar findings, as well as studies that find no evidence of a volume offset.

Although the PRM can be used to model the demand implications of policy changes, this report contains no projections modeling changes in government programs or policies.

D.  Physician Requirements Projections

The baseline projections take into account the growth and aging of the population, but are calculated on the assumption that the United States will provide the same level of care in the future that is currently provided.  Essentially, the baseline projections assume that the future will use today’s health care system.  Alternative projections are based on different assumptions of how the health care system will evolve over time.

The baseline projections suggest that between 2005 and 2020 overall requirements for physicians engaged primarily in patient care increase 22 percent, from approximately 757,300 to 921,500 (Exhibits 36, 38, and 39).  In percentage terms, growth is lower for primary care (20 percent) than for non-primary care (23 percent).  If it is assumed that requirements for physicians engaged primarily in non-patient care activities (e.g., administration, teaching, and research) remain relatively constant at approximately 6 percent of total physicians, then total requirements for physicians will increase from about 802,100 to 976,000 during this period. ^[29]

On a per capita basis, demand for physicians is increasing as a result of an aging population (Exhibits 37, 40, and 41).  For example, under the baseline scenario, requirements for physicians engaged in patient care increases from approximately 256 to 274 (7 percent) per 100,000 population between 2005 and 2020.  In percentage terms, the increase is greater for non-primary care (8 percent) than for primary care (5 percent).

Projected growth in requirements between 2005 and 2020 varies substantially by specialty (Exhibit 42).  Between 2005 and 2020, specialties with the highest percentage growth are cardiology (33 percent) and urology (30 percent).  Specialties with the lowest percentage growth are pediatrics (9 percent) and obstetrics/gynecology (10 percent).

Exhibit 36. Baseline Projections of Physician Requirements

* Base year assumes that physician supply and demand are balanced.

Exhibit 37. Baseline Physician Requirements per 100,000 Population

* Base year assumes that physician supply and demand are balanced.

Exhibit 38. Physician Requirements

[D]

Exhibit 39. % Growth in Physician Requirements

[D]

Exhibit 40. Requirements per 100,000 Population

[D]

Exhibit 41. % Growth in Requirements per Capita

[D]

Exhibit 42. Baseline Physician Requirements Projections

Note: Totals might not equal sum of subtotals due to rounding.

The baseline projections assume that patterns of health care use and delivery of care remain unchanged over the projection horizon and that changing demographics are the primary driver of changes in physician requirements.  To better understand the implications of possible changes in utilization and delivery patterns, physician requirements are projected from 2005 to 2020 under alternative scenarios (Exhibits 43 and 44).

• Growing role of NPCs.  This scenario assumes that (1) the number of active NPCs will increase 60 percent between 2005 and 2020, (2) all NPCs that are trained will become employed and will provide services that otherwise would have been provided by physicians, and (3) on average each NPC will provide 40 percent of the work currently provided by a physician.  Under this scenario, by 2020 physician requirements would be approximately 90,000 physicians less than the baseline projections.  NPCs will have a disproportionate impact by specialty, with NPCs having a greater impact on reducing demand for generalists.
• Economic growth.  This scenario assumes that economic growth will allow the Nation to afford a higher-quality health care system.  This new health care system will require more physician and, in particular, more specialists.  Physician requirements are projected under the assumption that per capita income will grow by 2 percent annually, and that demand for some specialties is relatively insensitive (elasticity=0.25), modestly sensitive (elasticity=0.50), or more sensitive (elasticity=0.75) to economic growth (Exhibit 34).  This scenario produces the highest projections, with requirements growing to 1.1 million physicians in 2020 (136,000 higher than the baseline projection).
• Physician productivity increase.  Requirements are projected under the assumption that physician productivity increases 1 percent annually (i.e., each physician can see one percent more patients per year through improved use of staff and technology).  Projected physician requirements remain relatively constant through 2020 under this scenario, with the 2020 projection suggesting 137,000 fewer physicians than projected under the baseline scenario.
• Economic growth offset by physician productivity increase.  Combining the previous two scenarios, the growth in demand for physician services due to economic growth is offset by the increased productivity of physicians resulting in projected requirements of 956,000 in 2020 (20,000 fewer than under the baseline scenario).

Exhibit 43. Alternative Requirements Projections

[D]

The scenarios described here produce a large range in projected future demand for physicians. The sensitivity of the projections to key assumptions regarding the impact of economic growth and increases in physician productivity illustrate why researchers arrive at such different conclusions regarding the future requirements for physicians.  These national projections use current patterns in the utilization and delivery of physician services as a starting point. Throughout the Nation there remain pockets of under service, especially in poor, rural and urban areas.  These geographic disparities are discussed in Chapter IV .

Exhibit 44. Physician Requirements by Medical Specialty:  High Economic Growth Series

Note: Totals might not equal sum of subtotals due to rounding.