Relationship Between Physician and Hospital Coronary Angioplasty Volume and Outcome in Elderly Patients
Background With the expectation that physicians who perform larger numbers of coronary angioplasty procedures will have better outcomes, the American College of Cardiology/American Heart Association guidelines recommend minimum physician volumes of 75 procedures per year. However, there is little empirical data to support this recommendation.
Methods and Results We examined in-hospital bypass surgery and death after angioplasty according to 1992 physician and hospital Medicare procedure volume. In 1992, 6115 physicians performed angioplasty on 97 478 Medicare patients at 984 hospitals. The median numbers of procedures performed per physician and per hospital were 13 (interquartile range, 5 to 25) and 98 (interquartile range, 40 to 181), respectively. With the assumption that Medicare patients composed one half to one third of all patients undergoing angioplasty, these median values are consistent with an overall physician volume of 26 to 39 cases per year and an overall hospital volume of 196 to 294 cases per year. After adjusting for age, sex, race, acute myocardial infarction, and comorbidity, low-volume physicians were associated with higher rates of bypass surgery (P<.001) and low-volume hospitals were associated with higher rates of bypass surgery and death (P<.001). Improving outcomes were seen up to threshold values of 75 Medicare cases per physician and 200 Medicare cases per hospital.
Conclusions More than 50% of physicians and 25% of hospitals performing coronary angioplasty in 1992 were unlikely to have met the minimum volume guidelines first published in 1988, and these patients had worse outcomes. While more recent data are required to determine whether the same relationships persist after the introduction of newer technologies, this study suggests that adherence to minimum volume standards by physicians and hospitals will lead to better outcomes for elderly patients undergoing coronary angioplasty.
Previous studies have found that hospitals that perform higher numbers of coronary angioplasty procedures have lower complication and mortality rates.1 2 3 With the expectation that physicians and hospitals with more angioplasty experience should also have improved outcomes, the American College of Cardiology/American Heart Association (ACC/AHA) guidelines for angioplasty released in 1988 recommended minimum physician volumes of 50 coronary angioplasty procedures per year per physician and 200 procedures per hospital.4 In 1993, the number of recommended procedures for physicians was increased to 75 procedures.5 While there is growing evidence in the literature to support hospital volume requirements, there is little empirical support for physician volume standards.6 We examined the relationship between physician volume and outcome in Medicare patients undergoing angioplasty in 1992, the first year that physician volume could be calculated on the basis of the Unique Physician Identification Number. The Medicare population represented 95% of elderly Americans undergoing angioplasty during the time period.7
The Medicare National Claims History file consists of both Part A (hospital) and Part B (physician and outpatient) claims. Part A claims contain demographic and limited clinical information on all inpatient hospitalizations that were billed to Medicare, including the patient’s Health Insurance Claims (HIC) number, age, sex, race, discharge status including death, up to 10 discharge diagnoses and 6 procedures identified by International Classification of Diseases (ICD) codes, and attending physician according to Unique Physician Identification Number.8 Part B claims contain demographic and limited clinical information from physician and outpatient bills to Medicare, including HIC number, service or procedure provided according to Current Procedural Terminology (CPT) codes, and billing physician according to Unique Physician Identification Number.9 We excluded from the analyses patients eligible for Medicare for reasons other than age (end-stage renal disease, Railroad Retirement Board, and disability entitlements) and patients treated in Federal and non-US hospitals. Mortality after discharge was obtained from Medicare Provider Analysis and Review files maintained by the Health Care Financing Administration (HCFA) by matching HIC number.
Our study population consisted of all Medicare enrollees 65 years of age and over with physician and hospital claims involving coronary angioplasty in 1992. For each patient, a longitudinal record of hospitalizations, procedures, and death was created, linked by the patient’s HIC number. To avoid counting patients more than once, only the first hospitalization for coronary angioplasty in 1992 was analyzed in the bypass surgery and mortality comparisons.
Physician and Hospital Volume Calculations
We used Medicare Part B physician claims for coronary angioplasty to identify the physician associated with the procedure. Physician angioplasty volume was calculated by counting all claims for CPT code 92982 (percutaneous coronary angioplasty, initial vessel) submitted in 1992, including multiple procedures per day per patient. The procedure physician was identified according to the Unique Physician Identification Number submitted with the procedure claim. Claims associated with physician specialties other than cardiology or internal medicine were considered to represent coding errors and were excluded from the analyses. Hospital angioplasty volume was calculated by counting all Medicare Part A hospital claims containing an ICD code for coronary angioplasty (36.01, 36.02, or 36.05) in 1992 for each hospital according to the method of our previous hospital volume study.2 Kato and colleagues10 found that Medicare hospital volume was highly correlated with overall hospital volume (r2=.89, P<.001) for 109 hospitals in California, with Medicare cases representing 35% of overall hospital angioplasty volume.
Other Study Variables
Coronary bypass surgery performed from the day of angioplasty until discharge was counted as surgery after unsuccessful angioplasty. By linking hospital records, patients who transferred to other hospitals for coronary bypass surgery during the same admission were identified. Patients were designated as having an acute myocardial infarction admission if the first record of the initial hospital encounter contained a primary diagnosis of acute myocardial infarction (ICD code 410) or a secondary diagnosis of acute myocardial infarction and a primary diagnosis of an infarction complication such as papillary muscle rupture (ICD code 429.6), the last record for the encounter contained a 410 diagnosis code, and the length of stay for the encounter was at least 3 days if the patient was discharged alive.11 Claims identified as involving a subsequent hospitalization for the acute myocardial infarction according to the fifth digit of the ICD code (410.x2) were excluded from this classification. The medical school affiliations of hospitals were identified by Medicare provider-of-services files as “major,” “limited,” “graduate,” or “no affiliation.”
The primary outcome of the study was in-hospital bypass surgery or death. For the initial comparisons according to physician and hospital volume, patients were grouped by deciles according to the annual Medicare angioplasty volume of their associated physician or hospital. Tables comparing outcomes were generated for low-volume (<25 annual procedures), medium-volume (25 to 50 annual procedures), and high-volume (>50 annual procedures) physicians after stratifying for patient characteristics including sex, race, age, acute myocardial infarction, number of vessels undergoing angioplasty, and hospital characteristics including hospital volume and medical school affiliation. The boundaries for physician volumes were selected to represent distinct physician groups based on the decile plots and to match published volume guidelines.4 Similar comparisons were made for hospital volume strata.
Logistic regression models were developed to examine the relationship between angioplasty volume and the combined end point of bypass surgery or death after adjusting for patient-specific variables including age, sex, race, acute myocardial infarction, and comorbidity according to the Charlson index.12 Independent variables were transformed to maintain a linear relationship to the dependent variable where appropriate. Additional models examined the relationship between physician and hospital angioplasty volume and in-hospital death.
During 1992, 97 478 Medicare patients underwent 119 886 angioplasty procedures according to physician claims. The mean age of the patients was 72.6 years (interquartile range, 66 to 78), 44% were women, and the race distribution was 93.4% white, 3.4% African American, and 3.2% other or unknown. Patient characteristics by physician and hospital volume categories are displayed in Table 1⇓. Patients treated by low-volume providers were younger, had less comorbid illness, were less likely to undergo multivessel procedures, and were more likely to be hospitalized with an acute myocardial infarction.
According to Medicare Part B claims, there were 6115 physicians who performed coronary angioplasty in 1992, with annual volumes ranging from 1 to 494 Medicare procedures (Fig 1⇓). The median Medicare procedure volume per physician was 13 (interquartile range, 5 to 25). With the assumption that Medicare patients made up one half to one third of all patients undergoing angioplasty at the time, 13 Medicare procedures is consistent with an overall procedure volume of 26 to 39 cases per year.10 According to Medicare Part A claims, 984 hospitals performed angioplasty in 1992 compared with 934 hospitals performing angioplasty in 1990 (Fig 2⇓).2 The median number of Medicare procedures performed per hospital was 98 (interquartile range, 40 to 181), with a range of 1 to 1209 procedures. Assuming the same proportion of Medicare patients as above, a median of 98 Medicare cases is consistent with an overall hospital volume of 196 to 294 cases per year.
The rates of in-hospital bypass surgery, in-hospital death, and the combined end point of in-hospital bypass surgery or death were 3.3%, 2.5%, and 5.5%, respectively (Table 2⇓). Outcomes by physician and hospital angioplasty volume are plotted in Figs 3⇓ and 4⇓, with each point representing ≈9700 patients. In-hospital death declined with increasing hospital volume but did not change with physician volume. In-hospital bypass surgery and the combined end point of bypass surgery or death declined with both increasing physician and hospital angioplasty volumes. Stratifying by physician and hospital volume, the combined end point of bypass surgery or death was greatest for patients treated by low-volume physicians in low-volume hospitals and lowest for patients treated by high-volume physicians in high-volume hospitals (Table 3⇓).
After adjustment for age, sex, race, acute myocardial infarction, comorbid illness, and hospital volume to the extent possible with claims data, higher physician volume continued to be strongly associated with decreasing rates of in-hospital bypass surgery or death, mostly attributable to the decline inbypass surgery (P<.001). Improving outcomes were seen up to annual physician Medicare angioplasty volumes of 75 cases per year. This inflection point is consistent with an overall physician angioplasty volume of 150 to 225 cases per year. The logistic regression model with in-hospital death as the dependent variable did not detect a relationship between physician angioplasty volume and in-hospital death. If sufficient illness severity data were available to fully balance comparisons, the study had a 90% chance of detecting an absolute mortality increase of 0.4% for low-volume physicians (<25 annual procedures). Given the limitations of Medicare claims, a mortality difference was more likely to have gone undetected because of an inability to entirely account for illness severity rather than an insufficient number of patients.13
After adjustment for age, sex, race, acute myocardial infarction, comorbid illness, and physician volume, hospital volume was inversely associated with both in-hospital death (P<.001) and the combined end point of in-hospital bypass surgery or death (P<.001), with improving outcomes seen in up to 200 annual Medicare cases, attributable to both the decline in bypass surgery and death. This inflection point is consistent with an overall hospital volume of 400 to 600 cases per year, assuming the same proportion of patients <65 years of age as above. Interaction terms for physician and hospital volume were not significant.
The principal findings of this study are that more than 50% of physicians and 25% of hospitals performing coronary angioplasty in 1992 were unlikely to have met the ACC/AHA minimum volume guidelines published in 1988, and patients treated by these low-volume providers had worse outcomes.4 Low-volume physicians were associated with higher rates of bypass surgery, and low-volume hospitals were associated with higher rates of both bypass surgery and death. This is the largest study to date examining the relationship between physician volume and outcome and one of the first to provide empirical evidence in support of the expert consensus ACC/AHA guidelines for minimum physician volume standards.
The higher rate of bypass surgery associated with low-volume physicians observed in this study is consistent with the findings of Shook and colleagues6 regarding emergency surgery and resource use according to physician angioplasty volume. Examining 2350 angioplasty procedures at Good Samaritan Hospital in Los Angeles, Shook and colleagues found that low-volume physicians were associated with more complicated postprocedural courses, as reflected by more emergency bypass surgery (P=.005), higher hospital morbidity (P<.001), higher hospital costs (P=.03), and longer length of stay (P=.004). The finding of an inverse relationship between hospital angioplasty volume and the risk of death or bypass surgery is consistent with our previous observations regarding Medicare patients treated between 1987 and 1990.2 Other studies also have observed worse outcomes for low-volume hospitals.1 3
Prevalence of Low-Volume Providers
On the basis of the Medicare experience in 1992, it appears that physicians and hospitals widely disregarded the minimum hospital volume guidelines published in 1988. There were 3510 physicians who performed 15 or fewer Medicare procedures and 344 hospitals that performed 60 or fewer Medicare procedures, and these providers were unlikely to have attained overall angioplasty volumes of 50 cases per physician and 200 cases per hospital, as recommended by the coronary angioplasty guidelines. Physicians and hospitals may have chosen to disregard the minimum volume guidelines because of the lack of supporting evidence available in 1992. At the time, the only published data in support of minimum procedure volume involved procedures other than coronary angioplasty.14 15 16 Health system factors such as incentives regarding reimbursement for coronary angioplasty and an increase in the number of physicians trained to perform coronary angioplasty may have also contributed to the growth of low-volume hospitals as well as the large number of low-volume physicians.
Three hypotheses first proposed by Luft2 16 may explain the inverse relationship between volume and mortality. First, greater experience by high-volume providers led to better outcomes−the “practice makes perfect” theory. Angioplasty requires significant technical abilities, and it is likely that the increased skill of high-volume operators in successfully dilating lesions, as well as in “bailing out” procedures involving abrupt closure, led to the lower rates of bypass surgery for these providers. The second hypothesis involves selective referral to high-volume providers. That is, referring physicians knew which angioplasty providers had better outcomes, selectively referred more patients to such providers, and thus these providers performed the highest volume of procedures. In the absence of published provider outcome data such as are available in New York State, it was difficult for referring physicians to determine which physicians and hospitals had better outcomes. To the extent that angioplasty outcomes correlated with professional reputations, it is possible that selective referral led to better outcomes for high-volume providers.
The third possible explanation for the inverse relation between volume and outcome is that differences in patient characteristics led to worse outcomes for low-volume providers. Angioplasty volume continued to be inversely related to in-hospital bypass surgery or death after adjustment for the patient characteristics of age, sex, race, acute myocardial infarction, and comorbidity, according to the Medicare data. However, claims data are substantially limited in their ability to describe the severity of coronary disease as well as comorbid illnesses, and it is likely that the models did not entirely account for illness severity differences among patients.13
More detailed data from two studies that reported coronary disease severity according to physician procedure volume do not support the third hypothesis.6 17 Clinical characteristics, including ACC/AHA lesion severity classification by operator volume, were reported from the Society for Cardiac Angiography and Interventions (SCA&I) database for 7747 patients who underwent angioplasty at 18 hospitals in 1992.17 According to ACC/AHA classifications, type C lesions are most often associated with unsuccessful procedures because their characteristics include longer length, excessive tortuosity of the proximal segment, total occlusion for >3 months, or location on extremely angulated segments, in degenerated vein grafts, or by major side branches.4 In the SCA&I registry, 7.9% of the cases for low-volume (<50 annual cases) physicians involved type C lesions compared with 14.9% for higher-volume (≥50 annual cases) physicians (P<.001). The study by Shook involving 2350 angioplasty procedures between March 1991 and February 1994 also found that patients treated by high-volume operators had greater risk according to older age, more multivessel procedures, and more emergent/urgent procedures.6 These data imply that low-volume operators were actually performing angioplasty on patients with less complex coronary disease during the study period, making increased coronary disease severity an unlikely explanation for the worse outcomes for low-volume physicians. They also raise the possibility that an inverse relationship between physician volume and mortality was not detected in Medicare claims because of the lack of illness severity detail.
Impact of Newer Technologies
Since 1992, two treatments have been introduced, coronary stents and platelet glycoprotein inhibitors, which have significantly reduced the need for bypass surgery after angioplasty.18 19 20 21 The relationship between physician volume and outcome in this study was mainly attributable to higher rates of bypass surgery, whereas the relationship between hospital volume and outcome was attributable to both higher mortality and more surgery. With a decline in the surgical rate, the relationship between experience and bypass surgery may change in more contemporary data. However, it does not seem likely that factors related to worse outcomes by low-volume providers will be alleviated by platelet inhibitors or stents. The potential for complications related to the new technologies such as femoral artery hemorrhage or inability to adequately deploy a stent would be expected to be greater for inexperienced operators. Patients who experience abrupt closure and who previously would have been referred to surgery may now undergo “bail-out” stenting. More recent data are needed to determine whether the relationship between angioplasty volume and outcome continues to be observed, with attention to additional outcomes such as bleeding complications or emergent stenting.
Implications for Minimum Volume Standards
The observations of this study concerning angioplasty volume and outcome raise the important policy issue of where to set minimum volume guidelines. Using logistic regression, we found an inverse relationship between physician volume and outcome up to Medicare procedure volumes of 75 cases per year, or an estimated overall physician volume of 150 to 225 cases per year. For hospital volume, improving outcomes were seen up to 200 Medicare cases per year, or an estimated overall hospital volume of 400 to 600 cases per year. If low-volume physicians had achieved the experience and results of higher-volume physicians, there would have been 545 fewer surgeries at a minimum physician volume of 25 Medicare cases per year and 2742 fewer surgeries at a minimum physician volume of 75 Medicare cases per year. For hospitals, there would have been 124 fewer deaths and 145 fewer surgeries at a minimum hospital volume of 100 Medicare cases per year and 266 fewer deaths and 411 fewer surgeries at a minimum hospital volume of 200 Medicare cases per year. However, raising minimum volume requirements to the higher levels would have diminished access to angioplasty in a substantial number of settings. Minimum volume requirements for physicians and hospitals equivalent to 75 and 200 Medicare cases per year, respectively, would have restricted >95% of physicians and >75% of hospitals from performing angioplasty in 1992. In setting volume standards, maintaining reasonable geographic access to angioplasty will need to be balanced against the improved outcomes available from the most experienced operators and programs. Access is particularly important for patients suffering acute myocardial infarction who cannot be treated with thrombolytic therapy. On the basis of the Medicare experience of 1992, in which a substantial number of providers did not appear to adhere to the minimum volume standards, both an increase in the minimum volume standards and a stricter adherence to these standards are likely to lead to greatest improvement in outcome.
Quality Low-Volume Operators
A more basic issue concerning angioplasty guidelines is whether volume limits should be set at all. The experience of low-volume providers in this study was considered in the aggregate, yielding stable estimates of worse outcomes on average for low-volume operators. Among these low-volume physicians and hospitals, some may have had better outcomes while others had worse outcomes. As minimum volume standards would restrict low-volume providers with better outcomes from practice, some believe that guidelines should focus on the identification of quality low-volume providers rather than strictly on volume limits. Confidently identifying such low-volume providers with better outcomes is difficult because of the “low-volume operator paradox” suggested by Ellis and colleagues.22 Because of the small number of procedures that can be examined, outcome estimates concerning individual low-volume physicians or hospitals are subject to wide confidence limits. Also, most current observational angioplasty data are limited in the ability to reliably identify best practice because of limitations in measuring baseline risk as well as outcomes.23 Our study suggests that strategies that limit angioplasty to higher volume providers will improve outcomes. For occasional situations in which higher-volume providers are not available and low-volume providers with improved outcomes can be identified, volume standards may serve as a guideline rather than a strict cutoff. Such high-quality, low-volume physicians and hospitals may be identified by combining experience over a number of years to derive stable estimates of performance.
Incentive to Perform More Procedures
Another concern regarding volume standards is their potential to encourage unnecessary procedures. To meet the suggested volume targets, low-volume providers may perform additional procedures for marginal indications. While volume limits have the potential to encourage some operators to perform unneeded angioplasty, it is not clear whether the additional incentive will substantially lower the threshold for performing angioplasty below levels encouraged by reimbursement in the current health service system. As managed care and capitated payment lower reimbursement incentives, the role of volume guidelines in encouraging unnecessary procedures may become relatively more important.
In addition to the limitations outlined above, several other limitations to this study should be noted. First, the study involves patients >65 years of age and may be limited in its applicability to younger patients. Younger patients would be expected to be less technically challenging because of coronary disease characteristics such as a lower prevalence of lesion calcification and to have lower mortality because of factors such as less comorbid illness. On the basis of observations in this study and other work showing that younger patients have lower procedural mortality and higher rates of bypass surgery, the mortality relationship would be expected to be diminished in a younger cohort, whereas the bypass surgery relationship may persist.24 A second limitation involves our focus on the initial procedure in 1992. While this approach avoided confounding associated with double-counting the patients, it also led to underrepresentation of patients undergoing repeat angioplasty. Second procedures involving patients who had undergone prior angioplasty in 1991 or earlier were included, but repeat procedures in 1992 were excluded. Because there are no data demonstrating that repeat angioplasty carries risks different from initial angioplasty, the findings were unlikely to have been substantially altered by a study design that included additional patients undergoing repeat procedures. A third limitation is that we only identified one physician per procedure and thus were not able to examine the impact of multiple physicians per procedure on the volume outcome relationship.
More than 50% of physicians and 25% of hospitals performing coronary angioplasty in 1992 were unlikely to have met the minimum volume guidelines first published in 1988. Elderly patients treated by these low-volume providers had worse outcomes, manifested by higher bypass surgery rates for low-volume physicians and higher bypass surgery and death rates for low-volume hospitals. This represents the largest study to date providing empirical data to support minimum angioplasty volume standards for physicians. While more recent data are required to determine whether the same relationships persist after the introduction of coronary stents and glycoprotein IIb/IIIa receptor inhibitors, this study suggests that adherence to minimum volume standards for physicians and hospitals will result in better angioplasty outcomes for elderly Americans.
This study was supported by research grants HS-08805 and HS-06503 from the Agency for Health Care Policy and Research, Rockville, Md.
- Received February 28, 1997.
- Revision received April 17, 1997.
- Accepted April 18, 1997.
- Copyright © 1997 by American Heart Association
Ritchie JL, Phillips KA, Luft HS. Coronary angioplasty: statewide experience in California. Circulation. 1993;88:2735-2743.
Ryan TJ, Bauman WB, Kennedy JW, Kereiakes DJ, King SB, McCallister BD, Smith SC, Ullyot DJ. Guidelines for percutaneous transluminal coronary angioplasty. Circulation. 1993;88:2987-3007.
May DS, Kelly JJ, Mendlein JM, Garbe PL Surveillance of major causes of hospitalization among the elderly, 1988. MMWR CDC Surveillance Summaries. 1991;40:7-21.
International Classification of Diseases, 9th Revision, Clinical Modification. Los Angeles, Calif: Practice Management Information Corp; 1991.
Physicians’ Current Procedural Terminology 1994. Chicago, Ill: American Medical Association; 1994.
Kato NS, Ergun ME, Carter GM. Health policy implications of volume recommendations on percutaneous transluminal coronary angioplasty in the United States. Circulation 1996;94(suppl I):I-532. Abstract.
Annual Report of the California Hospital Outcomes Project. State of California, Office of Statewide Planning and Development; 1993.
Krone RJ, Vetrovec GW, Noto TJ, Johnson LW, and the Registry Committee. PTCA operator variability: procedure complexity and outcome from the registry of the Society for Cardiac Angiography and Interventions. Circulation. 1993;88(suppl I):I-300. Abstract.
Serruys PW, de Jaegere P, Kiemeneij F, Macaya C, Rutsch W, Heyndrickx G, Emanuelsson H, Marco J, Legrand V, Materne P, Belgardi J, Sigwart U, Colombo A, Goy JJ, Van Den Heuvel P, Delcan J, Morel MA, for the Benestent Study Group. A comparison of balloon-expandable-stent implantation with balloon angioplasty in patients with coronary artery disease. N Engl J Med. 1994;331:489-495.
Fischman DL, Leon MB, Baim DS, Schatz RA, Savage MP, Penn I, Detre K, Veltri L, Ricci D, Nobuyoshi M, Cleman M, Heuser R, Almond D, Teirstein PS, Fish RD, Colombo A, Brinker J, Moses J, Shaknovich A, Hirshfeld J, Bailey S, Ellis S, Rake R, Goldberg S, for the Stent Restenosis Study Investigators. A randomized comparison of coronary-stent placement and balloon angioplasty in the treatment of coronary artery disease. N Engl J Med. 1994;331:496-501.
Ellis SG, Omoigui N, Bittl JA, Lincoff M, Wolfe MW, Howell G, Topol EJ. Analysis and comparison of operator-specific outcomes in interventional cardiology: from a multicenter database of 4860 quality-controlled procedures. Circulation. 1996;93:431-439.
Califf RM, Jollis JG, Peterson ED. Operator-specific outcomes: a call to professional responsibility. Circulation. 1996;93:403-406.
Angioplasty in New York State 1994. State of New York, New York State Department of Health; 1996.