For more than two decades, the importance and unprofitable nature of outpatient-based postgraduate medical education and the difficulty of financing this education have been emphasized.1–7 Most of the contact that residents have with outpatients occurs in hospitals' clinics. The number and characteristics of their patients and the extent and nature of supervisory practices in teaching hospitals' clinics nationally have not been well quantified. Until 1996, the requirements governing Medicare Part B (Physician Services) reimbursement for outpatient services (IL-372) that include residents' contacts with patients were not clear to many teaching physicians.
In June 1996, the Office of the Inspector General of the Department of Health and Human Services announced a program to audit Medicare Part B payments to teaching physicians, and it made explicit the requirements for Part B reimbursement for patients seeing a resident.8 The reasonableness of the audit, which retrospectively applied previously ambiguous rules for reimbursing of the services residents provided to patients under the supervision of teaching physicians, has been the subject of debate in the medical literature and within teaching hospitals.9 As was made explicit in 1996, for most outpatient settings, Medicare Part B's reimbursement rules require a teaching physician to interact directly with patients seen by residents. This requirement does not apply to some primary care clinics. Medicare also requires that teaching physicians provide more detailed documentation of their roles in examining and formulating the treatment plans for patients seen by residents.
A decade ago, John M. Eisenberg, MD, subsequently a director of the U.S. Agency for Health Care Research and Quality, said “we should explore educational subsidies through Medicare Part B and other payers to finance ambulatory postgraduate medical education.”1 The strict Medicare Part B guidelines implemented in mid-1996 may have had the opposite effect, reducing the Medicare payments for patients' services that have been provided in hospital-based outpatient clinics and that use direct resident participation. This study quantifies the proportions of residents' patients who were also seen by staff physicians before and after June 1996 as well as the patients' and clinics' attributes associated with higher rates of direct staff physician contact with residents' patients.
For this study, I used data from five National Hospital Ambulatory Medical Care Surveys (NHAMCS) for Outpatient Departments, 1993 to 1997.10 These surveys, conducted annually by the National Center for Health Statistics (NCHS), are national probability sample surveys that quantify visits to hospitals' outpatient departments in non-federal short-stay hospitals in the United States. For each visit, the date of visit, diagnosis, and the type of practitioner seen are recorded. Based on these data, I quantified visits to various providers and the odds that hospital-based clinics' patients of given characteristics would be seen by a resident alone or by both a resident and a staff physician. To assess the extent of the change in practices after the new rules took effect, I compared the proportions of patients seen by a resident and also seen by a staff (or other non-resident) physician before and after June 30, 1996.
To test for statistical significance and calculate standard errors and 95% confidence intervals (CIs) for univariate estimates of the numbers of visits, I used formulas published by the NCHS.11 To calculate the proportion of visits of a given type, the odds ratios that compare these proportions before and after July 1, 1996, and 95% CIs of these odds ratios, I used the survey procedures of a standard statistical software package designed for the analysis of weighted survey data, the type of data collected in the NHAMCS, and to take into account sampling weights, clustering, and stratification of the sample. I used logistic regression techniques to calculate maximum likelihood estimates of odds ratios and their CIs. To obtain multivariate estimates of these odds, the models included all significant predictors of these odds in the univariate analyses. I calculated means and significant differences in means. Unless otherwise specified to reduce variance, all analyses were restricted to visits to medicine (and its subspecialties, including family practice), pediatrics, obstetrics—gynecology, and surgery clinics in sampled government (non-federal) hospitals (e.g., Veterans Administration and military hospitals are excluded) located in metropolitan statistical areas.
The five NHAMCS outpatient surveys sampled a total of 145,761 visits to hospital-based outpatient departments, including 33,612 visits involving a resident's direct contact with a patient. From 1993 to 1997, the surveys projected an average of 68 million visits each year to government (non-federal) hospital-based outpatient clinics. From 1993 to 1996, the estimated number of visits each year did not vary substantially (range = 63 to 67 million), but the estimated number of visits for the 1997 survey was significantly higher (77 million), with the differences primarily due to increases in visits to medical clinics and to specialty clinics not used in this study (drug abuse, psychiatry, etc.) Overall, more than 88% of all visits and 98% of visits to residents occurred in hospitals located in metropolitan statistical areas. Ninety percent of all clinic visits were made to medicine, surgery, pediatrics, or obstetrics—gynecology clinics, with the remainder made to a variety of other types of clinics, including psychiatry and drug abuse.
A total of 88% of patients attending hospital-based clinics were seen by at least one physician (resident or staff, MD or DO), while the remaining patients saw a nurse or other non-physician practitioner. Three fourths of all hospital-based clinic outpatients saw a staff (non-resident) physician, either alone or with a resident.
Each year, residents saw an average of 14,800,000 patients in hospital-based outpatient clinics, 22% of all the patients cared for at non-federal hospital-based outpatient clinics. Ninety-seven percent of visits to residents (14,300,000 per year) occurred in medicine, surgery, pediatrics, and obstetrics—gynecology hospital-based clinics located in metropolitan statistical areas (32,500 of 33,612 sampled visits). Therefore, residents saw 25.4% of all patients attending medicine (including family practice and medical subspecialties) pediatrics, surgery, or obstetrics— gynecology hospital clinics located in metropolitan statistical areas.
Over the five survey years, more than 1,000 hospitals located in metropolitan statistical areas had at least one of the types of clinic studied (medicine, surgery, pediatrics, or obstetrics—gynecology) in which residents were recorded as seeing sampled patients. Although the percentages of patients seen by a resident varied substantially by type of clinic (see Table 1), overall, they did not vary significantly among the study years (range = 24% to 27%). The percentages of hospital-based clinic visits from January 1993 to June 30, 1996, and from July 1, 1996, to January 1998 in which a resident saw a patient were nearly identical (25.3% and 25.5%, respectively).
For the univariate analysis, Table 2 presents the odds that various types of patients' visits to residents involved the patient's being seen by a staff physician. Perhaps most notable are the differences in the proportions of patients who attended different specialty clinics who saw a staff physician. There were also significant differences in the odds that a staff physician would also see a resident's patient based on the patient's age, patient's race, whether this was a first or return visit for the patient; the geographic location of the hospital; the type of ownership of the hospital [voluntary, (government) for-profit], and whether the patient was insured by Medicaid.
I used multivariate logistic regression for survey data to adjust for other significant predictor variables in determining the odds that a residents' patient with given characteristics would also be seen by another (non-resident) physician. I included in the model all significant univariate predictors of these odds (see Table 2). In the multivariate analysis, I also tested for interactions between the date of visit (before or after July 1, 1996) and other significant predictors of these odds. Significant interactions were found between date of visit (before or after June 30, 1996) and visits to hospitals located in the Midwest, visits to surgery clinics, first versus return visits, and patient's gender. However, the model that incorporated these interaction terms did not fit the data significantly better than did the simpler model that incorporated only all significant univariate predictors but no interaction terms, which was the final model I employed.
As detailed in Table 3, even after adjusting for all other significant predictors, after June 1996, the odds that a resident's patient would also be seen by a staff or other non-resident physician increased significantly (OR = 1.64, 95% CI = 1.11 to 2.41). In the multivariate analyses, patients who saw a resident and were 65 years old or older, attended surgery clinics, were cared for at hospitals in the Midwest, were white/non-Hispanic, were men, and had not been seen before in that clinic were more likely also to be seen by a staff physician than were residents' patients without these attributes.
Changes in Staff Physicians' Contacts with Residents' Patients
After June 30, 1996, a higher percentage of residents' patients were also seen by a staff physician than was the case from December 1992 to June 1996 (52% versus 42%, p <.05). For patients aged 65 or older, the increase in the percentage of residents' patients who also saw a staff physician was less than that for younger patients (13% versus 2%, p <.05). For specific patient groups, Table 4 presents the changes in the proportions of a resident's patients also being seen by a staff physician after June 30, 1996, compared with the prior three and a half years. As detailed in Table 4, the proportion of residents' patients seen in clinics located in non-federal government hospitals who also saw staff physicians did not increase. In contrast, the proportion of these patients in for-profit hospital clinics doubled. After June 30, 1996, the increase in the proportions of residents' patients seen by staff physicians were fairly comparable among most groups of patients (range 8% to 15%, see Table 4). However, residents' patients who were 65 and older and who had Medicare listed as their primary insurance were no more likely to see a staff physician after June 1996 than they had been before (change 1% or less for both groups).
Variation among Hospitals in Staff Physicians' Practices
There was great variability among the hospitals in the proportions of sampled visits to residents during which a patient's contact with a staff physician was recorded. In this analysis, to reduce variance in estimates, I considered only clinics that had at least 20 visits to residents sampled during the study. As noted in Table 5, for all four types of clinics studied, one fourth of sampled hospitals recorded 5% or less of sampled visits to residents during which a staff physician also saw the patient. In less than one fourth of hospitals' clinics were at least 75% of residents' patients also seen by a staff physician.
Annually, about 100,000 residents, including the 40,000 individuals enrolled in primary care specialties, participate in postgraduate programs accredited by the Accreditation Council for Graduate Medical Education in the United States.13 Each year, residents and interns see about 15,000,000 patient visits in hospital-based outpatient clinics, an average of fewer than three hospital outpatient visits per week for each resident. Even in pediatrics, a specialty that emphasizes ambulatory medicine, there is an average of fewer than six hospital clinic visits per resident per week. These small numbers reflect both the limited proportion of time devoted to training in outpatient settings and the limited volume of services trainees provide. Although ambulatory education also occurs at other sites, such as physicians' offices, federal hospitals, non-hospital clinics, and health maintenance organizations, for many residency programs hospital-based clinics are likely to be the primary site for ambulatory education.
The primary financial supporter of postgraduate medical education is Medicare. Until July 1996, Medicare rules governing staff physicians' contact with patients seeing a resident for Medicare Part B reimbursement were unclear to many teaching physicians. It is now clear that, with the exception of patients seeing residents in certain primary care practices, Medicare patients who see a resident in a hospital-based clinic also should see a staff (or other non-resident) physician if Part B reimbursement is to be sought.
Prior to July 1, 1996, data from the NHAMCS analyzed here document that staff physicians' contacts with residents' patients in hospital-based clinics were the exception rather than the rule. Overall, only 42% of patients seen by a resident were also seen by a staff (or other non-resident) physician. The proportions of residents' patients who also saw a staff physician varied greatly among specialties and for patients with different sociodemographic characteristics. Residents' patients in hospital-based obstetrics—gynecology clinics, patients insured by Medicaid, patients less than 65 years of age, and non-white or Hispanic patients (who, in most cases, were not insured by Medicare) were less likely than were patients with other attributes to see also a staff physician. Even after consideration of these factors, the odds that patients seeing residents in Midwestern hospitals would also see a staff physician were more than twice as high as were the odds for patients seen in other regions of the country.
Data from the NHAMCS did not permit me to determine the proportion of visits actually reimbursed under Medicare Part B that conform to the rules of IL-372 as explicitly restated in mid-1996. My data do not establish whether a Part B bill was submitted for a given Medicare patient seeing a resident in a hospital-based clinic. Also, in primary care clinics and certain other situations IL-372 does not always require that a staff physician actually see a resident's patient, but may require only that the billing physician supervise the care the resident provides.
Following the clear interpretation of IL-372 that became effective July 1996, the proportion of patients seeing a resident who also saw a staff physician increased significantly, to 52%. Surprisingly, the percentage of residents' patients with Medicare coverage or those 65 or over who also saw a staff physician did not increase after June 1996. For most other patients' groups, there was about a 10% absolute increase in the proportion of residents' patients also seen by staff physicians.
My analysis of this large, multi-year, national sample of patients involving more than 1,000 hospitals' outpatient departments demonstrates that a highly publicized government directive can affect physicians' behavior in the desired direction. Surprisingly, Medicare's action seems to have had the greatest effect on the process of care for non-Medicare patients. The greater increase in the rate of staff physicians' contact with residents' patients in for-profit hospital-based clinics, combined with no change in this proportion in government (non-federal) hospital-based clinics, suggests that for-profit hospitals paid greater attention to the strict interpretation of the 1996 regulations. The changes in staff physicians' supervisory practices after July 1996 are small compared with the great variability that continues among hospital-based clinics in the proportions of residents' patients who also see a staff physician.
Whether compliance with IL-372 will improve or degrade the quality of care provided to patients seen by residents and the education of residents is not clear or addressed by my data. However, explicit and strict interpretation of these rules did not substantially change staff physicians' contact with Medicare patients seen by a resident. This finding might reflect high compliance with Medicare Part B rules before July 1996, but this is unlikely given the variability in staff's practice patterns among clinics. Alternatively, strict and explicit rule setting alone may not be sufficient to have changed staff physicians' practices in many hospital-based clinics.