This Article Abstract Full Text (PDF) Free Letters to the Editor: Submit a response Alert me when this article is cited Alert me when Letters to the Editor are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My File Cabinet Download to citation manager Rights and Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by LEE, P. P. Articles by RELLES, D. A. Search for Related Content PubMed PubMed Citation Articles by LEE, P. P. Articles by RELLES, D. A. Social Bookmarking                   What's this?

Demand-Based Assessment of Workforce Requirements for Orthopaedic Services*

PAUL P. LEE, M.D., J.D., CATHERINE A. JACKSON, PH.D. and DANIEL A. RELLES, PH.D., SANTA MONICA, CALIFORNIA

Investigation performed at RAND, Health Program, Santa Monica

	Abstract

Top Abstract Introduction Methods Results Discussion References

 
On the basis of an analysis of the supply of and demand for orthopaedic surgeons, we projected that there will be 21,134 full-time-equivalent orthopaedists in the year 2010 if training continues at current levels. We estimated a demand-based requirement of 17,012 full-time-equivalent orthopaedic surgeons, indicating a surplus of 4122 full-time equivalents. In terms of orthopaedist-to-population ratios, we estimated that there will be 7.5 full-time-equivalent orthopaedists per 100,000 population in 2010 compared with a demand-based requirement of 6.0 full-time equivalents. However, we did not include estimates of the demand for orthopaedic surgeons as assistants in the operating room in our model. If an assistant orthopaedic surgeon is required for all procedures, an additional 3906 full-time-equivalent orthopaedists would be demanded, thus eliminating the surplus. The demand for an assistant orthopaedic surgeon in only half of the procedures would still lead to a sizable reduction in the surplus.

	Introduction

Top Abstract Introduction Methods Results Discussion References

 
The debate about the physician workforce in the United States has centered on two related but distinct questions: whether there is an overall surplus of physicians and whether there are too many specialists relative to primary-care physicians. When these issues were addressed in the 1960s, policy makers concluded that there were too few physicians generally and too few specialists specifically. This began a period of substantial increases in the level of financial support for medical education and postgraduate medical training. Currently, however, the overwhelming opinion of policy analysts, health services researchers, and workforce specialists is that there is an overall surplus of physicians and that there are too few generalists relative to specialists in the United States^7,8,11,16,17. Few policy analysts or authors have challenged either of these consensus views, although some believe either that there is only a small overall surplus, with an imbalance in the specialist-to-generalist ratio, or that there are enough generalists but too many physicians overall.

The methods used to derive physician workforce estimates have, for the most part, focused on comparing the physician workforce (either as a whole or by specialty) with the population of the United States to arrive at physician-to-population ratios. Forecasts of physician workforce requirements are made by applying these ratios to census estimates of the future population^12,14,17. To place such ratios in an interpretable context, similarly constructed ratios from health maintenance organizations (staff, group, or network models) or specific geographic markets are used as benchmarks of feasible workforce configurations^14,17. However, even among such benchmarks there are often twofold differences in staffing levels.

The use of such ratios and comparative benchmarks provides the opportunity to place bounds on extant staffing ratios and can yield insight into the status and potential future shape of the physician workforce. These ratios are easily understandable and can provide information for those involved in discussions of physician workforce policy in the private and public sectors.

However, because workforce ratios are simply the number of all physicians (sometimes by specialty) divided by the total population (without distinction of need for care or other proxies such as age or gender), they are not very precise measures of the requirement for physicians. When such ratios are developed for given health-care plans or regional areas, several assumptions are made. The first assumption is that demand is being met appropriately, with regard to quality and quantity. The second assumption is that the populations from which the ratios were derived are and will remain similar. Finally, the relationship between the population's demand for care and the ability of physicians to provide the care is projected to be static.

In the current report, we present the results of a more detailed approach to the estimation of the workforce needed for orthopaedic care. This method explicitly takes into account aspects of health care that are only implicitly included in the workforce ratio approach. Specifically, we estimated the demand for orthopaedic services with use of publicly available, nationally representative survey data to project the demand (utilization) for specific orthopaedic services by age and gender. We also conducted a survey to find out how much time orthopaedists spend in providing care for specific orthopaedic conditions. We then combined the detailed utilization data with the work-time data to estimate the amount of orthopaedic services required to satisfy current demand. We used various sensitivity analyses to assess the robustness of the estimates and to construct confidence intervals around our workforce estimates. Unlike other workforce studies that are based on ratios, this more detailed approach provides the flexibility to estimate the probable effects of changes in health care, such as changes in the intensity of the use of services, shifts in care between specialties, different physician practice patterns, efficiencies arising from new technologies, and demographic changes within the population^9,13.

	Methods

Top Abstract Introduction Methods Results Discussion References

 
In order to conduct a demand-based assessment of orthopaedic workforce requirements, we needed to obtain estimates of the supply of orthopaedists and the demand for orthopaedic services and we required a method to reconcile supply with demand. In order to analyze and reconcile supply with demand, we used time as the base unit of analysis. Thus, we had to convert the utilization data from the secondary data sets into work times. The use of time as the base unit meant that we could maintain maximum flexibility to adapt our modeling to changes in work hours per year, intensity of treatment (the amount of time needed to care for conditions), technological efficiencies, and other factors.

Supply Component
We used the membership files of The American Academy of Orthopaedic Surgeons¹ for 1994 as a foundation for our supply estimates of the number of active orthopaedists. The data set is thought to contain at least 98 per cent of all active orthopaedists in the United States and identifies the current status of each member (active member, in training, or retired). We compared the numbers from The Academy with those from the American Medical Association as reported in the Bureau of Health Professions Area Resources File⁵ for the calendar year 1993. The total number of orthopaedic surgeons is fairly comparable in the two sources (membership in The Academy, 20,259; Area Resources File supply, 20,413). Thus, we used the data from The Academy as the source of our supply data because it included physician-specific demographic information that could be included in our projection models as well as physician-location information to assist us in administering our work-time survey.

In comparing orthopaedic supply with demand, we wanted to ensure that equivalent units were being analyzed. From our survey of orthopaedists, we determined both the total time spent in direct patient care, excluding overhead activities such as completing insurance forms, and the time required to provide specific types of orthopaedic services. Hence, we wanted to express the supply of orthopaedists in terms of orthopaedic time available to perform direct patient care rather than in terms of the total amount of time that an orthopaedist spends in his or her practice. From our survey, we determined that orthopaedists worked an average of 2600 hours per year. As this included direct patient care as well as administrative time, it was too high. We compared this number with the orthopaedic-specific figure of 2200 annual hours of direct patient care used in the Graduate Medical Education National Advisory Committee study¹⁷ and the surgeon-specific figure of 2374 annual hours of direct patient-care activities from the American Medical Association (Socioeconomic Characteristics of Medical Practice 1997)³. We discussed all of these alternatives with an advisory panel made up of members of The Academy and representatives of the subspecialty societies, which ultimately settled on the number 2200 as the low end of what seemed reasonable.

We converted the number of orthopaedic surgeons into full-time equivalents. Assuming that an individual orthopaedic surgeon spent 2200 hours a year in direct patient care, we adjusted for training status and gender with use of information from previous workforce studies. Orthopaedists in their first year of postgraduate training were considered 0.35 of a full-time equivalent; those in their second or third postgraduate year, 0.5; those in their fourth postgraduate year, 0.75; and those in their fifth postgraduate year or higher, 1.0. Female orthopaedists were considered 0.85 of a full-time equivalent¹⁶. There were 19,502 orthopaedic surgeons (active or in training) in The Academy's membership database for 1994, which corresponded to 18,296 full-time-equivalent orthopaedic surgeons when reductions were made for orthopaedists in training and for female orthopaedists. This current supply translates into 7.6 orthopaedic surgeons (or 7.1 full-time equivalents) per 100,000 population, and it can be compared with health maintenance organization-based adjusted ratios of 3.1 to 7.0 orthopaedic surgeons per 100,000 member-population¹⁸.

A population projection model for the supply of orthopaedic surgeons was developed with use of the entry rates of residents from the National Residency Match Program as inputs to supply and the retirement and mortality rates from the Bureau of Health Professions model as exits from supply⁹. We assumed that orthopaedic training would continue at current levels—that is, 602 physicians would enter orthopaedic residencies annually. Accordingly, we projected that there will be 22,413 orthopaedic surgeons or 21,134 full-time-equivalent orthopaedic surgeons in the year 2010 (Fig. 1). This translates into 7.5 full-time-equivalent orthopaedic surgeons per 100,000 population in the year 2010, an increase of 0.4 full-time equivalent per 100,000 population compared with the current level (Fig. 2).

View larger version (16K): [in this window] [in a new window]   Fig. 1 Graph showing the trends with regard to the number of full-time-equivalent orthopaedists under two levels of residency training: current (602 physicians a year) and a 50 per cent reduction (300 physicians a year).

View larger version (13K): [in this window] [in a new window]   Fig. 2 Graph showing the trends with regard to the ratio of the number of full-time-equivalent orthopaedists to the estimated population of the United States under two levels of residency training: current (602 physicians a year) and a 50 per cent reduction (300 physicians a year).

Demand Component 1: Utilization Data

Sources of Data
To estimate the current levels of utilization of orthopaedic services, we analyzed several national data sets generated by the National Center for Health Statistics (Table I). These data sets are helpful because they reflect national utilization patterns in a wide range of treatment settings, including physicians' outpatient offices (National Ambulatory Medical Care Survey), hospital inpatient services (National Hospital Discharge Survey), hospital-based emergency rooms (National Hospital Ambulatory Medical Care Survey), hospital-based outpatient clinics (National Hospital Ambulatory Medical Care Survey), and ambulatory surgery centers (National Survey of Ambulatory Surgery). From these data sets, we calculated age and gender-specific utilization rates.

Organization of Data
The International Classification of Diseases, Ninth Revision⁹ (ICD-9), was used for diagnosis and procedure coding in all of the national utilization surveys. However, for our analysis we needed to develop a typology of orthopaedic-related diagnoses to reduce the thousands of diagnosis codes into a manageable number of groups specifically relevant to orthopaedics. Whereas the ICD-9 is universally used for diagnostic coding, often the Physicians' Current Procedural Terminology² is used for procedure-related physician billing. Because these two coding systems are different, we had to develop a link between diagnoses and their related procedures. This typology was developed by project staff, with the assistance of consultant orthopaedic surgeons and a member of the coding committee of The Academy.

Identification of the Universe of Orthopaedic-Related Diagnosis Codes
We began the process by reviewing the full listing of diagnosis codes, which was displayed in code-numerical order, and identifying the ranges of likely orthopaedic-related codes. We supplemented this list with the diagnosis codes (primary, secondary, and tertiary) reported by orthopaedic surgeons in the National Ambulatory Medical Care Survey. From this combined list, we identified key word components (such as arthro and osteo) related to orthopaedic care. We then used these key word components to conduct a computer search of the full ICD-9 code listing. We identified additional diagnosis codes and added them to the composite list.

In a parallel effort, staff at The Academy had compiled a list of orthopaedic-related diagnosis codes. We compared our completed list with The Academy's list and augmented our list as needed.

For the final refinement, we reviewed the entire listing of diagnosis codes and indicated those that were included in our composite list. In this way, we could ascertain which codes were identified as orthopaedic related and which were not. Our final list consisted of 2407 orthopaedic-related diagnosis codes.

Grouping of the Orthopaedic-Related ICD-9 Diagnosis Codes
We reviewed all of the orthopaedic-related diagnosis codes and developed a typology to describe the current practice of orthopaedics as well as the orthopaedic subspecialty structure. The typology consisted of two interrelated groupings: anatomical location and disease manifestation. These groups formed the building blocks of the workforce model.

Seven specific anatomical locations were identified as relevant to orthopaedic practice: upper extremity and shoulder, foot and ankle, hand, hip, knee, lower extremity, and spine. To these seven, we added two general anatomical areas, systemic and unspecified, that appeared often among the orthopaedic-related diagnosis codes. We reviewed these nine anatomical groupings, identified key words related to each anatomical location, and sorted the orthopaedic diagnosis list so that codes that used similar key words were displayed consecutively. The list was reviewed and refined by staff at The Academy and by our orthopaedic consultants.

We then reviewed the diagnosis codes to identify disease and condition clusters that were frequently used in the nine anatomical location groups. We identified ten specific diseases or conditions among the codes: inflammatory arthritis, arthrosis, congenital, infections, mass or tumor, neurological, sprain or strain, orthopaedic trauma, other trauma, and vascular. We also identified two general groups: systemic and unspecified. We used an algorithmic key-word sorting strategy to organize the codes into groups that shared similar disease entities or characteristics; for example, we sorted the disease codes so that all codes that included mass in their description were listed together.

We next assigned each orthopaedic-related diagnosis code to one, and only one, anatomical location-condition grouping. However, since not all diseases could occur within each anatomical location, there were ninety-five final groups rather than the 108 that would be created from nine anatomical locations and twelve disease-condition clusters.

Organization of Procedure Codes
To identify the universe of orthopaedic procedure codes, we started with all procedures (as coded in the ICD-9) used in association with an orthopaedic diagnosis as reported in the various secondary data sources (the National Ambulatory Medical Care Survey, National Hospital Discharge Survey, National Hospital Ambulatory Medical Care Survey, and National Survey of Ambulatory Surgery). We then supplemented this empirically-derived list with codes identified through a manual review of all of the ICD-9 procedure codes. This review ensured that all codes potentially used by orthopaedists or reflecting orthopaedic care were included.

Unlike the diagnosis codes, each of which was assigned to only one location-condition pairing, a given procedure code could be used for more than one location-condition pairing. Initially, we assigned procedures to location-condition pairs with use of the ICD-9 codes with which they were associated in the national data sets. We then reviewed the assignment manually and revised it as needed, being particularly careful to identify the procedure codes, such as internal fixation, that should be assigned to multiple location-condition pairings. This process resulted in the assignment of all orthopaedic-related procedures to the location-condition categories for which they could potentially be used as treatment.

Demand Component 2: Work-Time Data
To translate the location-condition utilization data into work times, we conducted a survey of practicing orthopaedic surgeons to ascertain the amount of time that they spent in the care of patients during various diagnosis and procedure-specific encounters. In the survey, we used scenarios for the location-condition pairs that were fairly common and representative of orthopaedic practice. An example of a medical-visit scenario, which was included on all surveys, was a patient of variable age having aspiration or injection of a joint (a non-incisional procedure). An example of an operative scenario, which was included on the adult hand survey, was a forty-eight-year-old woman with carpal tunnel syndrome having a release of the transverse carpal ligament at the wrist. We asked orthopaedists about specific components of their work time, including the duration and annual frequency of medical visits and evaluations and the duration of procedures (including preoperative, intraoperative, and postoperative care).

Design of Sampling Strategy
As a first step in the design of the primary data-collection effort, we considered two sampling strategies: simple random samples and stratified random samples. A simple random sample would be appropriate if orthopaedic surgeons' practices did not cluster by patient characteristics, such as age or location-condition. Stratified random samples would be appropriate if such clustering did occur. To determine which was more appropriate, we used data from The Academy's biannual demographic census and examined three variables: the self-designation of orthopaedic generalist, orthopaedic generalist with a specialty, or orthopaedic specialist; the age distribution of the physician's patients; and the distribution of the physician's practice by anatomical location (for example, spine, foot and ankle, or hand).

Because we wanted to elicit information from orthopaedists about their particular practice and about the time required to manage patients with various conditions, we needed to determine whether or not orthopaedists' practices were concentrated with respect to the patients' age—that is, whether or not an individual orthopaedist managed predominantly pediatric or geriatric patients—or with respect to the patients' conditions. With use of The Academy membership file, we reviewed the patient-age distribution of physicians' practices by their self-designated specialty interests. We found a cluster of orthopaedists, especially specialists, whose practices were predominantly pediatric. We then cross-classified practices with use of the distribution of anatomical locations of conditions. When we attempted to determine the patient age-patient condition categories that represented a large fraction of an individual orthopaedist's practice, we found that certain specialists focused on adult conditions of the hand and spine and other specialists focused on pediatric conditions of the foot and ankle or the spine and hip. From this analysis, we decided that orthopaedists' practices were sufficiently specialized that, in order for us to target the content of the survey appropriately, we would need to develop separate surveys for pediatric and adult orthopaedics, with specialist and generalist distinctions in each group.

With use of The Academy's membership file, we assigned each orthopaedist to one of six mutually exclusive categories: pediatric foot and ankle, pediatric spine and hip, pediatric general, adult hand, adult spine, and adult general (Table II). Because the adult general category was so large, we randomly divided it into five subcategories. From each of the ten mutually exclusive categories or sample population stratum, we randomly drew samples. We wanted to draw fairly equal samples across the various strata, but some of the pediatric strata were small. One hundred and twenty-five orthopaedists were selected from each of the two adult specialty strata; 150, from the pediatric spine and hip stratum; seventy-five, from the pediatric foot and ankle stratum; and 175, from each of the six (five adult and one pediatric) general strata. Thus, more than half of the surveys were sent to orthopaedists in the adult general category. Overall, 1525 orthopaedic surgeons were sampled, and 45 per cent (686) responded (Table II). The characteristics of the respondents were similar to those of the full Academy membership (Table III).

Design of the Surveys
To address the specific orthopaedic conditions for each stratum, we concluded that, at a minimum, we would need different surveys for each of the six strata. To balance our concerns about the time required for completion of the survey (and thus acceptable rates of response) and about sufficient coverage of orthopaedic care, we determined that each survey could include as many as approximately fifteen diagnoses and fifteen procedures. With use of the secondary data sets, we identified conditions and procedures that had high levels of frequency or use. This list was reviewed by our orthopaedic consultants and the advisory panel, who identified the diagnoses and procedures of importance to the orthopaedic community. We augmented the list with procedures that had been included in the orthopaedic practice Medicare Resource-Based Relative-Value-Scale development effort¹⁰.

Components of the Surveys
We divided the work-time segment of the survey into two parts. The first segment concerned outpatient physician visits, including in-office procedures. The second segment concerned orthopaedic operations. In both segments, we asked respondents to provide information about various characteristics of the care that they provided.

All estimates of time requirements included the time required to provide direct patient care, including the time needed to review medical records as well as laboratory and radiology reports, to document care in the charts, and to consult with other physicians. The time estimates explicitly did not include the time required to obtain preauthorization or approval for payment, as we wanted to focus on patient care regardless of the type of health-care organization, payment, or reimbursement mechanism.

To determine the time required for office visits, we asked the duration of an initial visit as well as the number and duration of follow-up visits. These data were used in conjunction with the outpatient utilization data that also identified individual visits as new (a new patient or a new condition in a previously seen patient) or continuing care for a previously diagnosed condition. In addition, respondents estimated the proportion of patients who had an operation for each of the medical conditions surveyed. This information was used to provide an independent verification check of the procedure utilization numbers reported in the two national surgery data sets (National Hospital Discharge Survey and National Survey of Ambulatory Surgery).

To ascertain the time requirements for operative procedures, we asked for estimates of the work time needed for preoperative assessment, the operative procedure, same-day postoperative care, and global ninety-day operative care. We also asked how frequently the primary surgeon was assisted by a practice assistant (either a physician or a non-physician assistant during the operation). However, there was considerable variability among the responses for each procedure. Because we wanted to determine the impact that operative assistants would have on the overall workforce estimation, we simulated the impact of such assistants.

Reconciliation Component
Once utilization data (the numbers of office visits and procedures) and time data (the duration of the office visit or procedure) were available, the total time required to provide orthopaedic care for each of the ninety-five location-condition categories could be calculated. In this final reconciliation phase, we refined this seemingly simple multiplication in several ways to avoid overcalculation or undercalculation of the time estimates.

Basic Calculation
We derived, from the National Ambulatory Medical Care Survey, the annualized numbers of new and follow-up office visits for each of the ninety-five categories. Using the identification of physician specialists providing care for each category as reported in the data set, we adjusted total utilization by the proportion of visits provided by self-designated orthopaedists (either Doctors of Medicine or Doctors of Osteopathy) (Table IV). The disease and condition categories were drawn with as high a degree of specificity as could be described within the limitations of the ICD-9 coding system. In many cases, a given ICD-9 code was not specific in its anatomical description, so all diagnoses with that code had to be included in the unspecified location category. In other cases, an anatomical description was somewhat more specific but was not sufficient to place that diagnosis into a specific joint category. For example, arthrosis is a disease of joints; however, the ICD-9 system is often no more specific than upper extremity or lower extremity. Thus, there is not a specific category for arthrosis of the knee; all such cases are included in the category of arthrosis of the lower extremity. The proportion of cases in the ninety-five location-condition groups seen by Doctors of Osteopathy overall was small. We therefore did not make any distinction between so-called regularly trained orthopaedists (Doctors of Medicine) or osteopathically trained orthopaedists (Doctors of Osteopathy) in the analysis; both were included. Using the time estimates for new and follow-up visits derived from the survey, we calculated the total amount of time required by orthopaedic surgeons to meet the current demand for new and continuing-care medical visits.

To determine the utilization of operative and inpatient orthopaedic care, we identified the patients who had both an orthopaedic diagnosis and an orthopaedic procedure, as well as the patients who had only an orthopaedic diagnosis, during an inpatient stay (as recorded in the National Hospital Discharge Survey) or during a visit to an ambulatory surgery center (as recorded in the National Survey of Ambulatory Surgery). The time required to manage the patients who had both an orthopaedic diagnosis and a procedure was attributed solely to orthopaedists. For each inpatient procedure, we recorded the orthopaedist-provided operative time as well as the time for the follow-up visit. For the visits to the ambulatory surgery center, we considered only the operative time, since any follow-up care would be reflected in the outpatient data sets.

Care of inpatients (as recorded in the National Hospital Discharge Survey) who had only an orthopaedic diagnosis could have been provided by non-orthopaedic physicians. Since the inpatient data sets did not provide information concerning the specialty of the treating physician, we used the proportion of orthopaedist-provided care as given in the National Ambulatory Medical Care Survey to adjust the level of demand. We then assigned one patient visit to each inpatient stay and used the duration of new patient visits to capture inhospital orthopaedic consultations. Thus, from the inpatient utilization data set, we approximated the amount of both orthopaedic medical inpatient care and operative care.

In summary, our estimates of the demand for orthopaedic office visits and related work-time requirements derived from various nationally representative survey data sets reflect care provided in physicians' offices, hospital outpatient clinics, and emergency rooms. New and continuing-care visits were identified in these data sets so that the appropriate work times could be applied.

Adjustment for Multiple Conditions
We assumed that the amount of time required to manage patients who had multiple orthopaedic conditions would be overestimated by the simple addition of each condition-specific time. In order to determine what adjustment should be made, we used data from the National Ambulatory Medical Care Survey to calculate the duration of the average visit for patients who had two orthopaedic diagnoses and for those who had three orthopaedic diagnoses or more. We then compared these averages with the simple average of the times for each of the individual diagnosis-specific visits. We found that the duration of a visit was 6 per cent higher than the simple average for each condition for patients who had two orthopaedic diagnoses and 26 per cent higher for patients who had three diagnoses or more. Thus, in our utilization calculations, we monitored the number of orthopaedic conditions reported for each patient in the secondary data and adjusted the work times when multiple conditions were present.

Full-Time-Equivalent Calculations
We defined a full-time-equivalent orthopaedic surgeon as one who performed 2200 hours of direct patient care annually. This figure was used in the Graduate Medical Education National Advisory Committee study¹⁷ and was accepted by the advisory panel of The Academy as an appropriate estimate. The estimate corresponds to the work times reported in the survey—that is, the estimate of all time required to provide medical care for the patient, including reviews of laboratory tests and documentation in the chart but excluding work that is uniquely required by a particular insurance company or payment mechanism.

Future Projection
For estimates of future demand, we applied the current gender and age-specific utilization rates for each of the ninety-five categories to population projections to the year 2010 developed by the census bureau. The age categories were zero to twelve years, thirteen to eighteen years, nineteen to forty-four years, forty-five to sixty-four years, and sixty-five years or more. Our projection of demand explicitly adjusts for age and gender-related differences in utilization, which may be especially important as the so-called baby-boom generation ages.

Dealing with Uncertainty and Variation
The methodology that we used to estimate the current and future demand for orthopaedic services involved several different types and sources of data and numerous assumptions. In order to calculate confidence intervals for the demand estimates, we used a statistical technique called the bootstrap, which samples the input data and repeats the model calculations many times. We ran the sample-repeat cycle 1000 times to generate a 90 per cent confidence interval around the final results for medical, operative, and total demand.

We also assessed the robustness of our results through various sensitivity analyses. For these, we ran the model using different assumptions about the data.

	Results

Top Abstract Introduction Methods Results Discussion References

 
We estimate the current demand to be 5655 (90 per cent confidence interval, 5519 to 5812) full-time-equivalent orthopaedic surgeons for the medical component of orthopaedic services and 9096 (90 per cent confidence interval, 8909 to 9292) full-time equivalents for the operative component (Table V). The estimated total current demand for orthopaedic services in the United States is 14,750 (90 per cent confidence interval, 14,521 to 15,027) full-time equivalents. With a current supply of 18,296 full-time-equivalent orthopaedic surgeons, there is a surplus of 3546 (90 per cent confidence interval, 3269 to 3775) full-time equivalents. The confidence intervals reflect the statistical variability in the estimates and were quite narrow, reflecting the large sample sizes in the national data sets. The estimates for future demand and supply indicate that there will be a surplus of 4122 (90 per cent confidence interval, 3804 to 4383) full-time-equivalent orthopaedic surgeons in the United States in the year 2010.

We divided the total demand according to the nine anatomical locations (Table VI) and the twelve disease-condition categories (Table VII) that were the basis for the analysis and that parallel the subspecialty classifications used in orthopaedics. Overall, the operative component of orthopaedics accounts for 62 per cent of the overall demand, but this varies by anatomical location and condition. This demand, however, reflects the current behavior of patients seeking care as well as the referral patterns in health-care organizations and in communities. These estimates do not indicate what the best practice may be; they merely reflect what is currently being done. Moreover, because orthopaedists tend to work across anatomical location-condition groupings, we did not estimate the current or future supply according to specific anatomical locations or conditions. Thus, these location and condition-specific demand estimates are provided as a point of interest rather than as evidence of particular trends.

	Discussion

Top Abstract Introduction Methods Results Discussion References

 
Our analysis indicates that there is currently a surplus of 3546 full-time-equivalent orthopaedic surgeons. Even if we use the lower bound of the confidence interval, we estimate that there is a current excess of 3269 orthopaedic surgeons in the United States. Moreover, we estimate that the surplus will persist until 2010 if current levels of training and utilization remain constant.

Central to the results of this study are the survey-based estimates of the time that is required to perform orthopaedic services. One weakness of self-reported information, such as that obtained through surveys, is that the reported data can be biased. An alternative approach to obtaining work-time estimates is the time-motion study. However, the logistics and related costs of such an effort were beyond the scope of this project. To the extent that the survey respondents underestimated (or overestimated) work times, we too underestimated (or overestimated) the demand for orthopaedic services. In order to minimize such bias, we assured all respondents that their answers would be held in strict confidence and we underscored the importance of accuracy in reporting. Finally, we asked the advisory panel to review the responses to make sure that they were reasonable; in no instance did the advisory panel indicate that the mean or median response was not in keeping with their experience.

The finding that there is a current surplus of orthopaedic surgeons was based on certain assumptions. First, we assumed that the average full-time-equivalent orthopaedic surgeon devotes 2200 hours to direct patient care each year. The current surplus could be eliminated if practice hours changed. For example, if orthopaedic surgeons reduced their work hours by 19 per cent and if demand remained constant, there would be a corresponding need for 19 per cent more full-time-equivalent orthopaedic surgeons and the current surplus would disappear. However, if it was assumed that a full-time-equivalent orthopaedic surgeon devotes 2600 hours per year to direct patient care, the total demand for full-time equivalents would be reduced by nearly 20 per cent, thereby increasing the potential surplus of orthopaedists.

Second, the surplus is based on current patterns of utilization among different physician specialties. If orthopaedic surgeons were to become the preferred physicians for more conditions and procedures, then the demand for orthopaedists would grow correspondingly. However, many reallocations are the result of the fiscal pressures of managed care, which drive patients away from specialists and toward generalists. A push toward primary-care providers could result in a decrease in the demand for orthopaedic surgeons to treat a particular condition.

Third, most of the national data sets that we used to estimate demand do not distinguish between Doctors of Medicine and Doctors of Osteopathy. Thus, our estimates of demand for orthopaedic surgeons included the demand for Doctors of Osteopathy who practice orthopaedics. Our estimates of supply were based on the members of The American Academy of Orthopaedic Surgeons, most of whom are regularly trained orthopaedic surgeons. Only Doctors of Osteopathy who met the criteria for membership in The Academy were included in the supply estimates. However, Doctors of Osteopathy do not provide a large proportion of care for conditions treated by regularly trained orthopaedic surgeons in doctors' offices (Table IV). Thus, we are confident that our supply-and-demand estimates refer principally to regularly trained orthopaedic surgeons. In any case, since we included the demand for Doctors of Osteopathy but may have underestimated the supply of these physicians, the bias would result in an underestimation of the surplus.

Fourth, a perfect distribution of orthopaedic surgeons is unlikely; thus, some surplus or redundancy is expected and probably needed.

Finally, it is common for a second orthopaedic surgeon to assist in an operative procedure. However, there are no reliable data to indicate the degree to which this practice occurs, and no data are available for the age-specific groups of procedures used in our model. Moreover, there are no data with which to determine whether such assistants need to be orthopaedic surgeons or whether other types of providers, such as surgical scrub nurses or technicians, could be used effectively. As previously mentioned, the data from our survey concerning the procedure-specific need for operative assistants varied to such a degree that we chose not to use it in the modeling. However, if every orthopaedic operation required an assistant orthopaedist during the portion from the initial incision to the skin closure, an additional 3906 full-time-equivalent orthopaedists would be demanded. Such a practice would eliminate the current surplus.

In the future, quality-of-care and outcomes studies could provide information that will reshape specialty-specific allocation of physicians. Without such studies, patients' selection of physicians may be influenced by concerns other than the quality of care.

Advantages of Detailed Modeling
The workforce model that we developed is flexible in several ways and thus has advantages compared with the more static physician-to-population ratio approach. For example, we can estimate the impact of new technologies on procedure-specific durations of operations. The model can also incorporate changes in treatment—such as those that increase the duration of the visit, operative procedures, and the like—resulting from condition-specific best-practice treatment protocols.

More interesting, perhaps, is the fact that the model allows us to change several factors simultaneously and to ascertain the potential impact. For example, we can explore the overall effect on demand that could result from a shift in the treatment of arthritis of the cervical spine to neurologists and of pain in the lumbar spine to orthopaedists and a simultaneous reduction in the time required for repair of a joint ligament. Such simulations could be enormously helpful in understanding how the workforce balance may look in the near and distant futures.

Since the late 1970s, there has been a major shift away from hospital-based care as a result of both technology and the reorganization of our health-care system. More recently, there has been a shift from specialists to primary-care providers. Because the model presented here distinguishes between the medical and operative components of care, we can explore additional changes in health-care delivery. For example, the parameters of the model can be modified to examine the impact of changes in health-care delivery, such as limiting orthopaedists to the operative or procedural aspects of care or requiring that all patients who have an orthopaedic-related condition be managed by an orthopaedist.

Unlike the detailed demand-modeling done here, physician-to-population ratios cannot incorporate explicit changes in health-care delivery. Because they are aggregate measures, the structure of the delivery of care is treated as a so-called black box. Changes within that black box cannot be examined. Thus, extension of the physician-to-population ratios into the future assumes that nothing changes: that interrelationships extant in health care remain static. Therefore, changes in utilization due to new procedures or innovations cannot be readily incorporated into the ratios. Adjustments for overuse or underuse cannot be made, despite the well documented variation in the use rates of many services and procedures.

Limitations of Detailed Modeling
Detailed modeling has several limitations. The most important is the need for large amounts of data, some of which are not currently available. Thus, the detailed modeling presented here requires primary data collection in addition to multiple secondary data sets. The model is generic in concept but not in structure, which means that a specialty-specific analytic framework has to be developed in order to generate these more detailed workforce estimates.

Second, there is a risk that detailed modeling provides a false sense of confidence in the accuracy of generated estimates of demand, supply, and resulting balance^4,15. As far as possible, we have explicitly identified the assumptions that we made and we conducted sensitivity analyses to determine the robustness of the results.

Third, since we used time and full-time equivalents as our common measure for both supply and demand, the results are vulnerable to changes in health-care practice or structure that affect the number of hours that are worked in a year. We have defined one full-time equivalent as 2200 hours of direct patient care a year. If changes in the health-care environment make that number unreasonable, the workforce estimates could change. Such changes could include parity in the gender-specific number of hours of patient care that are provided; different contributions to patient care by those in training; and increases in the hours devoted to activities not related to patient care, leading to reductions in the currently defined 2200 annual hours of patient care. However, our model can estimate the impact of such changes.

Finally, our use of a full-time-equivalent measure means that we cannot address distributional issues concerning the supply of orthopaedic surgeons. However, this limitation is not unique to our model; there will always be situations where less than a full provider, no matter how defined, will be needed and only a full provider can be supplied. More work needs to be done to determine the overall effect of these varied needs on physician workforce balance.

Future Needs in Orthopaedic Workforce Modeling and Studies
The results presented here provide a more detailed, clearer picture of the demand for orthopaedic services and the workforce balance. However, research is still needed in several areas. We have used the demand for services to reflect the need for services. Population-based estimates of the prevalence and incidence of many common musculoskeletal disorders would be helpful if we wanted to estimate the workforce requirements to satisfy the need—not just the demand—for services. The distinction between need and demand may not be too great since many orthopaedic conditions, such as fractures, necessitate immediate care. However, there are other conditions that may benefit from treatment but currently are not being treated. If such conditions were to be treated, more providers would be needed, but the question of how many more remains unanswered.

If universal health insurance becomes a reality, nearly forty million Americans will have better access to health care, and we would expect that the demand for orthopaedic services would increase. However, we would not expect the demand to increase uniformly across all types of orthopaedic use. We do not believe that it would increase for conditions such as fractures (orthopaedic trauma) of the long bones, as we have assumed that the demand for care nearly equals the actual prevalence or need for treatment of these conditions. In contrast, we would expect an increased demand for care for conditions that may have gone untreated previously, such as early forms of degenerative arthrosis.

Finally, we need to understand and describe the allocation of patients across the array of providers of musculoskeletal care, including non-orthopaedists, non-physicians, and providers of alternative medicine. To the extent that these providers treat the same conditions that orthopaedists treat, the demand estimates from our model will be affected. Currently, the secondary data sets that are available to us focus only on physician providers even though non-physician providers, such as chiropractors and podiatrists, satisfy the demand for certain musculoskeletal treatment. Indeed, Cooper estimated that there will be considerable growth in the number of non-physician providers during the next twenty-five years. While Medicare Part-B files identify non-physician providers, these records are restricted to patients who are at least sixty-five years old or disabled. We need more nationally representative data on the utilization of these types of providers before we can construct a complete picture of the orthopaedic workforce.

Implications for Training Programs
We have estimated a current surplus of orthopaedists, and this surplus is expected to continue if orthopaedic training continues at current levels. If 50 per cent fewer residents were trained, the surplus would be eliminated in 2010. However, the surplus will be eliminated only if training levels are decreased immediately. Adjustment of the overall supply of orthopaedic surgeons occurs very slowly. Because of limitations in data concerning the allocation of patients among different orthopaedic subspecialties, we cannot address the issue of where reductions should occur.

On the supply side, we can simulate changes in the number of orthopaedists who are trained, but we cannot predict the amount of work-time effort that will be provided or say much about the geographic distribution of orthopaedists and related concerns regarding access. On the demand side, our models, like many others, assume that current utilization patterns and rates will continue. If new knowledge or technology leads to major advances in orthopaedic treatment, then the demand for orthopaedists will grow commensurately as new patients are brought into the system. Similarly, if the allocation of care among different providers is changed because of new data on quality of care, patient satisfaction, or cost-effectiveness, then the relative balance among providers and specialists (and thus for any one specialty such as orthopaedics) will be affected.

Most importantly, even with a large (50 per cent) reduction in the number of residents who are trained and the resultant near balance between supply and demand in the future, the orthopaedist-to-population ratio (6.0 full-time equivalents per 100,000 population) remains greater than some of the ratios reported in the literature^14,17. Consequently, reliance on health maintenance organization or managed-care staffing ratios alone may result in a tight supply of orthopaedists at best or in a future shortage at worst. In fact, our results clearly indicate that more orthopaedists are required than would be estimated with the use of health maintenance organization staffing ratios.

Since World War II, workforce policy for health-care providers in the United States has been based on flawed assumptions and often inadequate data. Our detailed model of the orthopaedic workforce suggests that the use of physician-to-population ratios could misguide policies regarding graduate medical education. We counsel caution—and maintaining a bit of reserve in the workforce—when changing policy to reduce the overall supply of any one type of provider.

NOTE: This work was funded by The American Academy of Orthopaedic Surgeons, The American Orthopaedic Association, the Academic Orthopaedic Society, the American Association of Hip and Knee Surgeons, the American Orthopaedic Foot and Ankle Society, the American Orthopaedic Society for Sports Medicine, the American Society for Surgery of the Hand, and the Pediatric Orthopaedic Society of North America. The authors thank Dr. William Tipton for supporting their efforts. They thank Dr. Blair Filler, Dr. Bernard Morrey, Mr. Alan Praemer, Dr. Michael Simon, Dr. Paul Tsou, Dr. James Urbaniak, and the advisory panel—Dr. Arthur Boland, Dr. Richard Brand, Dr. Bruce Browner, Dr. Cecil Christensen, Dr. Robert D'Ambrosia, Dr. Harold Dick, Dr. Jeffrey Eckardt, Dr. James Kleinert, Dr. Mehrdad Malek, Dr. Ron Norris, Dr. Kent Reinker, Dr. Ronald Smith, Dr. Vernon Tolo, and Dr. David Wong—for their critiques and suggestions.

	Footnotes

 
*Copyright, 1998, The American Academy of Orthopaedic Surgeons. Published by permission.

Duke University, School of Medicine, Box 3802, Durham, North Carolina 27710.

RAND, Health Program, 1700 Main Street, Santa Monica, California 90407-2138. E-mail address for Dr. Relles:relles@rand.org.

	References

Top Abstract Introduction Methods Results Discussion References

 

1. American Academy of Orthopaedic Surgeons: Membership masterfile. Rosemont, Illinois, The American Academy of Orthopaedic Surgeons, 1997.

2. American Medical Association: CPT 1993. Physicians' Current Procedural Terminology. Chicago, American Medical Association, 1993.

3. American Medical Association: Socioeconomic Characteristics of Medical Practice 1997. Edited by M. L. Gonzales. Chicago, Center for Health Policy Research, 1997.

4. Anderson, G. F.; Han, K. C.; Miller, R. H.; and Johns, M. E.: A comparison of three methods for estimating the requirements for medical specialists: the case of otolaryngologists. Health Services Res., 32: 139-153, 1997.

5. Bureau of Health Professions: Area Resources File, prepared by Office of Data Analysis and Management. Rockville, Maryland, Bureau of Health Professions, 1994.

6. Cooper, R. A.: Seeking a balanced physician workforce for the 21st century. J. Am. Med. Assn., 272: 680-687, 1994.[Abstract/Free Full Text]

7. Feil, E. C.; Welch, H. G.; and Fisher, E. S.: Why estimates of physician supply and requirements disagree. J. Am. Med. Assn., 269: 2659-2663, 1993.[Abstract/Free Full Text]

8. Greenberg, L.: Forecasting the Future Supply of Physicians: Logic and Operation of the BHPr Physician Supply Model. Department of Health and Human Services Office of Health Professions Analysis and Research Publication 3-93. Washington, D.C., United States Department of Health and Human Services, 1992.

9. Health Care Financing Administration: International Classification of Diseases, Ninth Revision (ICD-9). Department of Health and Human Services Publication 91-1260. Washington, D.C., United States Department of Health and Human Services, 1991.

10. Hsiao, W. C.; Braun, P.; Kelly, N. L.; and Becker, E. R.: Results, potential effects, and implementation issues of the Resource-Based Relative Value Scale. J. Am. Med. Assn., 260: 2429-2438, 1988.[Abstract/Free Full Text]

11. Kindig, D. A.; Cultice, J. M.; and Mullan, F.: The elusive generalist physician. Can we reach a 50% goal?. J. Am. Med. Assn., 270: 1069-1073, 1993.[Abstract/Free Full Text]

12. Kronick, R.; Goodman, D. C.; Wennberg, J.; and Wagner, E.: The marketplace in health care reform. The demographic limitations of managed competition. New England J. Med., 328: 148-152, 1993.[Abstract/Free Full Text]

13. Lee, P. P.; Jackson, C. A.; and Relles, D. A.: Estimating Eye Care Provider Supply and Workforce Requirements. RAND Publication RAND-MR-516-AAO. Santa Monica, California, RAND, 1995.

14. Mulhausen, R., and McGee, J.: Physician need. An alternative projection from a study of large, prepaid group practices. J. Am. Med. Assn., 261: 1930-1934, 1989.[Abstract/Free Full Text]

15. Sanderson, C. F.; Hunter, D. J.; McKee, C. M.; and Black, N. A.: Limitations of epidemiologically based needs assessment. The case of prostatectomy. Med. Care, 35: 669-685, 1997.[Medline]

16. Schwartz, W. B.; Sloan, F. A.; and Mendelson, D. N.: Why there will be little or no physician surplus between now and the year 2000. New England J. Med., 318: 892-897, 1988.[Abstract]

17. United States Department of Health and Human Services Public Health Service: Summary Report of the Graduate Medical Education National Advisory Committee. To the Secretary, Department of Health and Human Services. Department of Health and Human Services Publication 81-651-657, vols. I-VII. Washington, D.C., United States Department of Health and Human Services, 1980.

18. Weiner, J. P.: Forecasting the effects of health reform on US physician workforce requirement. Evidence from HMO staffing patterns. J. Am. Med. Assn., 272: 222-230, 1994.[Abstract/Free Full Text]

CiteULike

Connotea

Del.icio.us

Facebook

Technorati

Twitter

This article has been cited by other articles:

				R. F. Kyle Workforce Analysis J. Am. Acad. Ortho. Surg., May 1, 2007; 15(5): 264 - 265. [Full Text] [PDF]

				F. A. Farley, J. N. Weinstein, G. M. Aamoth, M. S. Shapiro, J. Jacobs, J. C. McCarthy, J. Kramer, and Orthopaedic Workforce Taskforce to the Board of Di Workforce Analysis in Orthopaedic Surgery: How Can We Improve the Accuracy of Our Predictions? J. Am. Acad. Ortho. Surg., May 1, 2007; 15(5): 268 - 273. [Abstract] [Full Text] [PDF]

				G. W. Woods, D. P. O'Connor, and P. Pierce Orthopaedic Surgeons Do Not Increase Surgical Volume After Investing in a Specialty Hospital J. Bone Joint Surg. Am., June 1, 2005; 87(6): 1185 - 1190. [Abstract] [Full Text] [PDF]

				D. Shipton, E. M. Badley, and N. N. Mahomed Critical Shortage of Orthopaedic Services in Ontario, Canada J. Bone Joint Surg. Am., September 1, 2003; 85(9): 1710 - 1715. [Abstract] [Full Text] [PDF]

				D. W. Jackson The Orthopaedic Clinician-Scientist J. Bone Joint Surg. Am., January 1, 2001; 83(1): 131 - 131. [Full Text]

				J. D. Mabrey The Emerging Impact of the Information Age on Orthopaedic Surgery* : Electronic Resources for the Orthopaedic Surgeon J. Bone Joint Surg. Am., October 1, 2000; 82(10): 1494 - 1494. [Full Text] [PDF]

				R. A. Meals, H. L. Bassewitz, and F. J. Dorey Academic Longevity and Attrition of Full-Time Orthopaedic Faculty Members J. Bone Joint Surg. Am., July 1, 2000; 82(7): 1042 - 1042. [Full Text]

				A. H. Morris and A. Sarmiento Correspondence J. Bone Joint Surg. Am., September 1, 1999; 81(9): 1347 - 8. [Full Text]

				J. M. Melhorn Rediscovering Occupational Orthopaedics for the Next Millennium J. Bone Joint Surg. Am., April 1, 1999; 81(4): 587 - 91. [Full Text]

				J. D. HECKMAN, P. P. LEE, C. A. JACKSON, D. RELLES, J. N. WEINSTEIN, M. C. GEBHARDT, M. A. SIMON, J. J. CALLAGHAN, and R. D. D'AMBROSIA Symposium - Orthopaedic Workforce in the Next Millennium J. Bone Joint Surg. Am., October 1, 1998; 80(10): 1534 - 51. [Full Text]

				J. D. HECKMAN Let's Keep Our Eye on the Sparrow J. Bone Joint Surg. Am., June 1, 1998; 80(6): 775 - 81. [Full Text]

				J. N. Weinstein, D. Goodman, and J. E. Wennberg Commentary - The Orthopaedic Workforce: Which Rate is Right? J. Bone Joint Surg. Am., March 1, 1998; 80(3): 327 - 30. [Full Text]

This Article Abstract Full Text (PDF) Free Letters to the Editor: Submit a response Alert me when this article is cited Alert me when Letters to the Editor are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My File Cabinet Download to citation manager Rights and Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by LEE, P. P. Articles by RELLES, D. A. Search for Related Content PubMed PubMed Citation Articles by LEE, P. P. Articles by RELLES, D. A. Social Bookmarking                   What's this?

Stanford University Libraries' HighWire Press (TM) assists in the publication of JBJS Online.
Copyright © 1998 by The Journal of Bone and Joint Surgery, Inc. Online ISSN: 1535-1386.
The Journal of Bone and Joint Surgery®, JBJS®, JB&JS®, and eJBJS® are registered in the U.S. Patent and Trademark Office.