Volume–outcome relationships in coronary artery
bypass graft surgery patients: 5-year major
cardiovascular event outcomes
Herng-Ching Lin, PhD,aSudha Xirasagar, PhD, MBBS,bNai-Wen Tsao, MD,cYi-Ting Hwang, PhD,d Nai-Wen Kuo, PhD,aand Hsin-Chien Lee, MD, MPHe
Objective: Using nationwide population-based data from Taiwan’s National Health Insurance database, we examined the association between hospitals’ coronary artery bypass grafting surgery volume and 5-year major adverse cardiovascular events. Methods: We used Taiwan’s National Health Insurance claims data linked to the Cause of Death file for the years approximately 1997 to 2004. All 5718 patients who underwent nonemergency coronary artery bypass grafting operations during 1997 through 1999 were classified into one of 4 hospital volume groups: 282 cases or less (low volume, n 5 1584 patients), 283 to 517 cases (medium volume, n 5 1317), 518 to 725 cases (high volume, n 5 1437), and 726 cases or more (very high volume, n 5 1380).
Results: Increasing hospital volume is associated with increasing 5-year major adverse cardiovascular event–free survival (72.0%, 75.5%, 76.9%, and 79.4% in low-volume, medium-volume, high-volume, and very high-volume hospitals, respec-tively). Cox regression analysis shows that increasing hospital volume predicts a systematic decrease in adjusted major adverse cardiovascular event hazard at 5 years. The 5-year major adverse cardiovascular event hazard ratios for high-volume and very high-volume hospitals were 0.884 (95% confidence interval, 0.809–0.965) and 0.811 (95% confidence interval, 0.728–0.904) relative to low-volume hospitals after adjusting for patient demographics and economic status, initial case severity, coronary artery bypass grafting procedure attributes, and hospital characteristics. Conclusions: The findings suggest that high-volume hospitals have some processes, infrastructure/personnel factors, or both that seem to produce not only better short-term outcomes but also better long-short-term outcomes.
T
here is a growing recognition of the relatively better postoperative outcomes for high-risk procedures with increasing provider volumes. Policymakers in-ternationally have responded with regionalization policies and incentives to concentrate the volumes of high-risk procedures in a few hospitals. In Canada and the United Kingdom, a policy of regionalization of high-risk procedures was imple-mented.1Coronary artery bypass grafting (CABG) surgery is one high-risk procedure showing a consistent and comprehensive body of evidence,2-9leading to a widespread belief that if all patients were treated at high-volume hospitals, thousands of lives would be saved each year. In the United States, in 2003, the Leapfrog hospital quality initiative established an annual hospital volume of 450 CABG procedures to qualify for premium bonus payments over and above the established fee schedule.10Currently, the literature on the CABG volume-outcome relationship is limited to in-patient or 30-day mortality,11with little documentation on long-term outcomes. Using nationwide population-based data from Taiwan’s National Health Insurance (NHI) database, we examined the association between hospitals’ CABG surgery volumes and long-term (5-year) patient outcomes. Under NHI, all citizens (.96% of the island’s
From the School of Health Care Administra-tion,a Taipei Medical University, Taipei,
Taiwan; the Department of Health Services Policy and Management,b Arnold School
of Public Health, University of South Caro-lina, Columbia, SC; the Division of Cardio-vascular Surgery,c Taipei Medical University Hospital, Taipei, Taiwan; the Department of Statistics,d National Taipei
University, Taipei, Taiwan; and Department of Psychiatry,eTaipei Medical University
and Hospital, Taipei, Taiwan.
This study was supported partially by a grant from the National Science Council (NSC 95-2416-H-038-001) in Taiwan. This study is based in part on data from the National Health Insurance Research Database pro-vided by the Bureau of National Health In-surance, Department of Health, Taiwan and managed by the National Health Research Institutes. The interpretations and conclu-sions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health, or the National Health Research Institutes.
Received for publication Aug 6, 2007; accepted for publication Oct 4, 2007. Address for reprints: Hsin-Chien Lee, MD, Department of Psychiatry, Taipei Medical University and Hospital, 252 Wu-Hsing St., Taipei 110, Taiwan (E-mail:ellalee@ tmu.edu.tw).
J Thorac Cardiovasc Surg 2008;135:923-30 0022-5223/$34.00
CopyrightÓ 2008 by The American Asso-ciation for Thoracic Surgery
doi:10.1016/j.jtcvs.2007.10.005
Abbreviations and Acronyms
CABG 5 coronary artery bypass grafting ICD-9-CM 5 International Classification of Diseases–
ninth revision–Clinical Modification MACE 5 major adverse cardiovascular event NHI 5 National Health Insurance
PTCA 5 percutaneous transluminal coronary angioplasty
population of about 23 million) have full choice of health care provider under a single health benefit plan, a comprehensive benefit package, and very low copayments. This structure en-ables virtually full access to all citizens when they perceive the need. Furthermore, all care, outpatient or inpatient, is captured in the claims database, permitting every long-term outcome to be tracked. This is unlike health systems, such as that of the United States, where patient segmentation by insurer, health plan, and provider panel disrupt the complete-ness of information on postdischarge events and outcomes. Furthermore, many other health care systems use gatekeeper or referral systems that limit provider choice, confounding volume-outcome studies. Taiwan’s NHI claims data provide an opportunity to examine long-term outcomes without the above-mentioned sources of selection bias or follow-up bias. In this study we use major adverse cardiovascular events (MACEs) as our outcome measure to examine the long-term effects of hospital volume. MACE outcomes include death, myocardial infarction, stroke, and repeat revasculari-zation procedures (percutaneous transluminal coronary angioplasty [PTCA] or repeated CABG). Rather than limit to mortality alone, MACEs have been widely used to evalu-ate the outcomes of cardiac procedures in the current litera-ture.12-14 To date, MACEs have been used to evaluate short-term outcomes of comparable/alternative care proce-dures. Our study is innovative in presenting evidence on long–term (rather than short-term) MACE outcomes (rather than mortality alone) over a 5-year follow-up period relative to hospital procedure volume.
Materials and Methods
Database
We linked NHI claims data with Taiwan’s Cause of Death file for the years 1997 through 2004. These data cover all inpatient and outpa-tient medical benefit claims for all citizens (.96% of the Taiwanese population [23 million]). Each claim has International Classification of Diseases–ninth revision–Clinical Modification (ICD-9-CM) codes for 1 principal operative procedure, 1 principal diagnosis, and up to 4 secondary diagnoses, along with details of the care provided, patient demographics, and provider characteristics.
The Cause of Death file provides data on the date of death and underlying cause of death (ICD-9-CM). Because of mandatory death registration, these data are accurate and comprehensive. Because these were deidentified secondary data released for public
access for research purposes, the study was exempt from full review by the internal review board.
Study Sample
All in-patient claims for CABG surgery between January 1, 1997, and December 31, 1999, were screened for the following: (1) ICD-9 code 36.10-36.20 (bypass anastomosis for heart revasculari-zation), (2) first-time CABG surgery for the patient, and (3) not being an emergency department admission. NHI was implemented in 1995, and paying for CABG operations out of pocket was beyond the reach of most Taiwanese before 1995. NHI claims since 1995 were scrutinized to verify that the sample cases were first-time CABG recipients. We excluded emergency CABG recipients (admitted through the emergency department) because these patients are documented to have disproportionately poor out-comes.15,16 Based on the above criteria, we generated a study sample of 5718 patients undergoing CABG surgery.
Hospital CABG Volume Groups
Based on unique hospital identifiers, we calculated each hospital’s total CABG volume over the study period. Hospitals were then sorted, in ascending order, according to their total CABG volume, and cutoff points were determined by the volume that most closely sorted the sample patients into 4 quartile groups of roughly equiva-lent size. This is the standard documented method of classifying patients by the provider’s procedure volume to obtain balanced cell sizes for the volume groups.6The 4 hospital volume groups were 282 cases or less (low volume, n 5 1584 patients), 283 to 517 cases (medium volume, n 5 1317 patients), 518 to 725 cases (high volume, n 5 1437 patients), and 726 cases or more (very high volume, n 5 1380 patients).
Study end points. The study objective was to examine post-CABG MACE-free survival at 30 days and 5 years (the latter excluding patients with a MACE outcome within the first 30 days). This is because, relative to 5-year follow-up, 30-day MACE incidence is more likely driven by initial disease severity and by patient-specific clinical instability during the preoperative and postoperative phases.
We identified 4 clinical end points (MACE qualifying outcome) examined at 30 days and 5 years after each patient’s operative date: death, myocardial infarction, stroke, and repeat revascularization procedure (PTCA or repeat CABG). During each patient’s post-CABG 30-day and 5-year periods, death (from Taiwan’s Cause of Death file) or any subsequent inpatient claim or claims with a record-ing of a MACE event was logged into the study database. In addi-tion, a composite MACE outcome was defined: any patient with 1 or more of the 4 events/outcomes above was defined as having a MACE outcome. The remaining patients were classified as MACE-free survivors at 30 days and 5 years (after 30 days). The composite definition of MACE-free survival in cardiovascular medicine has been widely used in the current literature.12-14
Variables of interest. The patient was the unit of analysis. The key independent variable was hospital volume. The dependent vari-ables were dichotomous: having (or not having) a MACE within 30 days or 5 years (the denominator for the latter excluded patients with any MACEs within 30 days). Both outcomes were evaluated as hazard ratios by hospital volume.
We adjusted for patient and hospital characteristics. Patient variables included age, sex, income, and clinical severity at first
admission, as well as CABG procedure attributes. Income was categorized as NT$0 (not working), NT$1 to NT$15,840 (income less than minimum wage), NT$15841 to NT$25,000, and NT$25,001 or greater. NT$15,840 is Taiwan’s minimum-wage level stipulated for a full-time employee. Clinical severity was captured by using 8 dichotomous (yes/no) variables: myocardial infarction (ICD 410), any other coronary artery disease (ICD 411–414), diabetes (ICD 250), chronic obstructive pulmonary disease (ICD 490–496), hypertension (ICD 401–405), renal dysfunction (ICD 580–586), congestive heart failure (ICD 4280), and stroke (ICD 430–438). We controlled for 2 CABG procedure attributes (partly representing unmeasured clinical severity and partly representing MACE propensity caused by operative complex-ity): concomitant valve repair (ICD procedure code 35) and use of an internal thoracic graft (ICD procedure code 36.15 and 36.16).
We adjusted for hospital and patient variables. Hospital vari-ables included ownership (public, private not-for-profit [NFP], and for-profit [FP]), hospital level (medical center, $500 beds; re-gional hospital, 250–499 beds; and district hospital, #20 beds), and geographic location (north, south, east, and central Taiwan). Teaching status is excluded because all medical centers and regional hospitals are teaching hospitals, which would cause colin-earity. Medical centers and regional hospitals, as well as not-for-profit status, are generally associated with state-of-the-art clinical technologic infrastructure. Medical centers and regional hospitals also have higher, round-the-clock, high-intensity moni-toring of patients by residents, attending physicians, and teaching faculty, as well as nursing students of various levels. Therefore be-tween the 2 variables, hospital size and ownership, the regression accounts for the facilities’ clinical technology infrastructure and personnel support that might affect short-term and long-term outcomes.
Statistical Analysis
The SAS statistical package (Version 8.2; SAS Institute, Inc, Cary, NC) was used. Kaplan–Meier methods were used to estimate 30-day and 5-year (after 30 days) MACE-free survival, and their associa-tions with hospital volume were assessed by using the log-rank test. Survival time was computed from the surgical date to the date of MACE within the 30-day or (post-30-day) 5-year follow-up periods. Cox proportional hazard regressions yielded hazard ratios that esti-mate the contribution of hospital volume to MACE-free survival adjusted for hospital and patient characteristics.
Results
Table 1describes the distribution of the sample patients by using severity measures at the time of CABG and CABG pro-cedure attributes. Of 5718 first-time CABG hospitalizations, the majority (76.8%) of patients was male, the mean age was 69.4 years (SD, 9.9), 99.2% had either myocardial infarction or other coronary artery disease, and 29.1%, 38.4%, and 6.5% had diabetes, hypertension, and stroke, respectively.
Hospital and patient characteristics by hospital CABG volume groups are presented in Table 2. Mean hospital CABG volume for the study period was 214 operations. The majority of low-volume hospitals were regional and
not-for-profit hospitals, and all very high-volume hospitals were teaching hospitals (medical centers or regional hospi-tals).
The in-hospital mortality rates were 3.5%, 3.9%, 2.9%, and 3.1% for low-, medium-, high- and very high-volume hospital groups, respectively. The 30-day mortality rates were 4.2%, 4.3%, 4.2%, and 3.2% for low-, medium-, high- and very TABLE 1. Distribution of first-time CABG patients in Taiwan, 1997–1999
Totals*
Variables No. %
Hospital volume group
Low volume 1584 27.7
Medium volume 1317 23.0
High volume 1437 25.1
Very high volume 1380 24.1
Patient sex
Male 4389 76.8
Female 1329 23.2
Patient age (y)
,65 1534 26.8 65–74 2278 39.8 .74 1906 33.3 Coronary disease MI as primary diagnosis or secondary diagnosis 92 1.6
Other coronary artery disease 5582 97.6
No coronary artery disease 44 0.8
Diabetes Yes 1662 29.1 No 4056 70.9 Hypertension Yes 2196 38.4 No 3522 61.6 COPD Yes 228 4.0 No 5490 96.0 Renal disease Yes 416 7.3 No 5302 92.7
Congestive heart failure
Yes 411 7.2
No 5307 92.8
Stroke
Yes 371 6.5
No 5347 93.5
Concomitant valve repair
Yes 401 7.0
No 5317 93.0
Internal thoracic artery graft
Yes 1141 20.0
No 4577 80.0
MI, Myocardial infarction; COPD, chronic obstructive pulmonary disease. *Total patient sample 5 5718.
TABLE 2. Hospital and patient characteristics in Taiwan by hospital CABG volume groups, 1997–1999
Hospital CABG volume group
Low (#282) Medium (283–517) High (518–725) Very high ($726)
Variable No. % Mean (SD) No. % Mean (SD) No. % Mean (SD) No. % Mean (SD)
Hospital characteristics*
No. of Hospitals 32 8 4 3
Mean of hospital CABG volume 95 (87) 445 (66) 698 (28) 922 (136)
Hospital level Medical center 7 21.9 6 75.0 2 50.0 2 66.7 Regional hospital 21 65.6 2 25.0 1 25.0 1 33.3 District hospital 4 12.5 — — 1 25.0 — Hospital ownership Public 7 21.9 2 25.0 2 50.0 2 66.7 Private (not-for-profit) 16 50.0 5 62.5 2 50.0 1 33.3 Private (for-profit) 9 28.1 1 12.5 — — — — Hospital Location Northern 11 34.4 3 37.5 2 50.0 3 100 Central 11 34.4 1 12.5 1 25.0 — — Southern 9 28.1 4 50.0 1 25.0 — — Eastern 1 3.1 — — — — — — Patient characteristicsy
Total no. of Patients 1584 27.7 1317 23.0 1437 25.1 1380 24.1
Mean age of patients (y) 68.8 (9.8) 69.3 (9.7) 68.4 (10.4) 71.0 (9.3)
Patient sex Male 1155 72.9 1006 76.4 1086 75.6 1142 82.8 Female 429 27.1 311 23.6 351 24.4 238 17.3 Patient age ,65 y 453 28.6 349 26.5 445 31.0 287 20.8 65–74 y 643 40.6 547 41.5 582 40.5 506 36.7 .74 y 488 30.8 421 32.0 410 28.5 587 42.5 Coronary disease MI as primary or secondary diagnosis 39 2.5 12 0.9 14 1.0 27 2.0 Other coronary artery disease 1530 96.5 1301 98.8 1411 98.2 1340 97.1 No coronary artery disease 15 1.0 4 0.3 12 0.8 13 0.9 Diabetes Yes 540 34.1 318 24.2 364 25.3 440 31.9 No 1044 65.9 999 75.8 1073 74.7 940 68.1 Hypertension Yes 645 40.7 391 29.7 548 38.1 612 44.4 No 939 59.3 926 70.3 869 61.9 768 55.6 COPD Yes 64 4.0 25 1.9 61 4.2 78 5.7 No 1520 96.0 1292 98.1 1376 95.8 1302 94.3 Renal disease Yes 145 9.2 82 6.2 88 6.1 101 7.3 No 1439 90.8 1235 93.8 1349 93.9 1279 92.7
Congestive heart failure
Yes 167 10.5 38 2.9 152 10.6 54 3.9 No 1417 89.5 1279 97.1 1285 89.4 1326 96.1 Stroke Yes 120 7.6 58 4.4 74 5.2 119 8.6 No 1464 92.4 1259 95.6 1363 94.8 1261 91.4
ACD
high-volume hospital groups, respectively (data not shown). Examined on the basis of individual MACE type, hospital CABG volume is inversely associated with the 4 end points and the composite end point, MACE, as shown in
Table 3. During the 5 years after CABG, low-volume hospi-tals had the highest stroke and repeat CABG/PTCA rates, which systematically decreased with increasing volume category, although the differences did not attain statistical significance. However, mortality accounted for a small pro-portion (9%) of all MACE events (range, 6% to 12% across volume groups). When the cause of mortality was investi-gated, about one third of deaths were due to cardiovascular causes, with the distribution of cardiovascular and noncardio-vascular causes being similar across the 4 volume groups (data not shown).
The crude 30-day and 5-year MACE-free survival rates, as well as adjusted hazard ratios by hospital volume groups, are provided inTable 4, showing increasing 30-day and 5-year MACE-free survival rates with increasing hospital volume.
Crude hazard ratio for a 30-day and 5-year MACE event decreased with increasing hospital volume (30-day crude hazard ratio for low-volume hospitals of 1.585 [reciprocal of 0.631] relative to very high-volume hospitals). After adjusting for patient demographics, initial case severity, CABG procedure attributes, and hospital characteristics, the negative association between hospital volume and 30-day and 5-year MACE hazard is sustained. Increasing hospital volume predicts a systematic decrease in adjusted MACE hazard at 5 years. The 30-day MACE hazard ratios for patients in low-volume hospitals was 1.502 (P , .05) relative to very high-volume hospitals. The 5-year MACE hazard ratios for patients in low-volume hospitals were 1.131 (P , .01) and 1.233 (P , .001) relative to those of patients in high-volume and very high-volume hospitals.
Discussion
This study makes a new contribution to the CABG volume-outcome literature by investigating MACE hazard rather TABLE 2. Continued
Hospital CABG volume group
Low (#282) Medium (283–517) High (518–725) Very high ($726)
Variable No. % Mean (SD) No. % Mean (SD) No. % Mean (SD) No. % Mean (SD)
Concomitant valve repair
Yes 107 6.8 78 5.9 100 7.0 116 8.4 No 1477 93.2 1239 94.1 1337 93.0 1264 91.6 Internal thoracic artery graft Yes 134 8.5 260 19.7 514 35.8 233 16.9 No 1450 91.5 1057 80.3 923 64.2 1147 83.1 Income-related insured amount 0 557 35.2 432 32.8 552 38.4 350 25.4 NT$115,840 400 25.3 478 36.3 275 19.1 719 52.1 NT$15,84125,000 498 31.4 296 22.5 468 32.6 208 15.1 $NT$25,001 129 8.1 111 8.4 142 9.9 103 7.5
CABG, Coronary artery bypass grafting; SD, standard deviation; MI, myocardial infarction; COPD, chronic obstructive pulmonary disease. *Total number of hospitals 5 47.yTotal patient sample 5 5718.
TABLE 3. Five-year follow-up results (excluding those with a MACE within 30 days) by hospital CABG volumes, 1997–1999
Mortality Myocardial infarction Stroke Repeat CABG or PTCA All MACE
Yes No Yes No Yes No Yes No Yes No
Variables No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) Hospital CABG volume
#282 50 (3.3) 1456 (96.7) 28 (1.9) 1478 (98.1) 165 (11.0) 1341 (89.0) 179 (11.9) 1327 (88.1) 422 (28.0) 1084 (72.0) 283–517 20 (1.6) 1235 (98.4) 13 (1.0) 1242 (99.0) 131 (10.4) 1124 (89.6) 143 (11.4) 1112 (88.6) 307 (24.5) 948 (75.5) 518–725 25 (1.8) 1347 (98.2) 20 (1.5) 1352 (98.5) 134 (9.8) 1238 (90.2) 138 (10.0) 1234 (90.0) 317 (23.1) 1055 (76.9) $726 26 (1.9) 1308 (98.1) 14 (1.0) 1320 (99.0) 129 (9.7) 1205 (90.3) 106 (7.9) 1228 (92.1) 275 (20.6) 1059 (79.4) Total patient sample 5 5467. MACE, Major adverse cardiovascular event; CABG, coronary artery bypass grafting; PTCA, percutaneous transluminal coronary angioplasty.
than mortality alone and by examining long-term MACE-free survival rather than 30-day or inpatient mortality, which has been the norm in these types of studies. It also distin-guishes between 30-day and 5-year MACEs (after 30-days) to sequester the high and variable vulnerability of the first 30 days from the relatively nonvolatile, post–30-day period. Other key features that qualify its unique contribution are the following: (1) the data’s ability to capture every subsequent MACE event treated anywhere in Taiwan and linkages to the Cause of Death data file and (2) coverage of all CABG age groups throughout the country, unlike most documented literature from the United States that focuses exclusively on the Medicare population because of the inability to follow-up younger age grofollow-ups because of periodic churning in insur-ance status and plan type. Although biases from these several sources have confounded past efforts to investigate the true relationship of provider volume with long-term outcomes, our study is able to show that unequivocally, hospital CABG volume affects MACE outcomes. Although our study might not be readily replicable in many countries because of a lack of such a data source, it makes a signal contribution by providing empiric evidence for policymakers to benefit the populace and to reduce health care costs arising from care-intensive MACE outcomes.
Our findings indicate that the favorable association between provider volume and in-hospital/30-day mortal-ity5-10 extends beyond the immediate postoperative period to 5 years after the operation. Furthermore, our study also indicates that not only mortality but also intermediate events and repeat revascularization in the following 5 years are less frequent among patients treated by higher-volume providers after adjusting for patient severity and other clinical
charac-teristics, demographics, and income, as well as hospital char-acteristics.
Our study also confirms that mortality represents the tip of the iceberg of adverse outcomes (only 9% of all MACEs). It demonstrates that limiting CABG outcome studies to mortality alone would greatly underestimate the effect of volume or, worse, fail to detect long-term outcome differ-ences across volume groups. This is one likely reason why the only documented study of long-term mortality versus provider volume11contradicts our finding. They reported no difference in risk-adjusted 3-year survival at a low-volume hospital compared with a high-volume hospital, when the same high-volume surgeons performed the procedure. Other reasons for their finding could be the following. First, they compared 1 low-volume hospital in Ohio with 1 high-volume hospital for a 5-surgeon team. Therefore lack of sta-tistical power to compare hospital groups would preclude statistically relevant conclusions. Second, in the United States high-volume hospitals in metropolitan areas are very likely teaching hospitals that are also safety net providers to treat the uninsured, Medicaid, and African American populations. Their patient profile confirms the significantly higher proportions for these groups among the high-volume hospital’s patients. Poorer risk-adjusted outcomes of CABG operations (as well as most high-risk and chronic conditions) among these groups is well documented. There-fore their finding of no difference between the low-volume and high-volume hospital could actually represent better average performance by the high-volume hospital when the generally poor risk-adjusted outcomes of the African American, uninsured, and Medicaid population is taken into account.
TABLE 4. Thirty-day and 5-year MACE-free survival (excluding those with a MACE within 30 days) and hazard ratios by hospital CABG volume groups
Variables 30-d MACE event-free survival rate (%) Crude hazard ratio (95% CI) Adjusted hazard ratio* (95% CI) Hospital CABG volumey
#282 95.1 1.000 1.000
283–517 95.3 1.019 (0.708–1.465) 0.952 (0.671–1.349)
518–725 95.5 0.780 (0.547–1.113) 0.922 (0.659–1.288)
$726 96.7 0.631 (0.428–0.932)z 0.666 (0.459–0.967)z
5-year MACE event-free survival (%)
Crude hazard ratio (95% CI)
Adjusted hazard ratio* (95% CI) Hospital CABG volume{
#282 72.0 1.000 1.000
283–517 75.5 0.882 (0.797–0.976)z 0.912 (0.812–1.081)
518–725 76.9 0.855 (0.769–0.949)x 0.884 (0.809–0.965)x
$726 79.4 0.765 (0.689–0.850)k 0.811 (0.728–0.904)k
MACE, Major adverse cardiovascular event; CABG, coronary artery bypass grafting; CI, confidence interval. *Hazard ratios are adjusted for the patient's age, sex, income-linked premium category, comorbidities, procedure attributes, and hospital characteristics, including hospital ownership, hospital level, and geographic location.yTotal sample size 5 5718. zP , .05, xP , .01, and kP , .001. {Total sample size 5 5467.
To compare our study with theirs, we evaluated mortality alone, and yet we find that high volume predicts lower adjusted mortality hazard. Because Taiwan has a universal coverage, single-payer system with low copayments and an ethnically homogeneous population, differences in treatment by insurance status is not a source of confounding, rendering our finding robust to extraneous confounding.
To explain our finding, 2 hypotheses are documented in the literature. One is that ‘‘practice makes perfect,’’ causing high-volume providers to have in place better care proce-dures, recognition of potential complications before they su-pervene, and better care organizations right up to discharge.17 The second documented hypothesis is that ‘‘self-referral’’ might cause more patients to attend hospitals with a reputa-tion for better outcomes, causing such hospitals to become high-volume hospitals.18 Although the former is possible, the latter might also play a role in Taiwan19,20 because of full provider choice, a considerable density of hospitals of all types in the populated areas, and good transportation systems.
Additionally, some researchers have suggested that the vol-ume-outcome relationship in CABG surgery might result from systematic differences in patient severity between low- and high-volume providers.21,22Many opponents of regionaliza-tion policies have argued that administrative databases might not permit adequate risk adjustment. We submit that our ad-justment for a very comprehensive array of clinical risk factors and the CABG attributes of internal thoracic artery graft use and concomitant valve repair should have accounted for most of the patient-specific vulnerabilities to MACEs. More-over, we excluded all CABG cases admitted through the emer-gency department, which also precludes a major source of variance in clinical severity. Furthermore, it must be noted that, in general, severely ill but not emergency patients are most likely to choose medical centers or regional hospitals (which are also large and noted for their state-of-the-art tech-nologic infrastructure). Therefore it is unlikely that systemat-ically higher unmeasured severity at low-volume hospitals is driving our findings. Consistent with this reasoning, patients treated at very high-volume hospitals had higher rates of con-comitant valve repair and the use of internal thoracic artery grafts than other volume groups. Other studies have also shown that low-risk patients for cardiac interventions were more likely to be treated by low-volume providers.23Another question could be that patients of lower socioeconomic status could be disproportionately represented among the low-vol-ume hospitals’ patient panels. Our adjustment for patient’s in-come-linked premium deduction category takes care of this potential source of confounding.
Finally, it could be argued that 5-year MACE differences between low- and high-volume hospitals could be due to lower all-cause mortality among the high-volume hospital patients, in turn because of generally better medical and surgical inpatient care for noncardiovascular morbidities
subsequent to their CABG surgery. Our finding that the distribution of cardiovascular versus other cause of death is similar across the volume groups rebuts this potential expla-nation for our findings. Moreover, as explained earlier, mor-tality contributes to less than 10% of total MACE incidence; the sentinel cardiovascular MACEs independently show a systematic (adjusted) decrease with increasing hospital volume.
There are some study limitations. First, the claims data-base lacks information on the clinical history of the patient, such as cigarette smoking, alcohol consumption, left ventric-ular ejection fraction, level of block and number of arteries blocked, body mass index, creatinine levels, and clinical severity scores. Yet with the large sample used, it is unlikely that systematic differences in these factors are driving our findings. A second limitation is that because of overall lower CABG volumes in Taiwan compared with the United States or other developed countries, we were unable to use the con-ventional hospital volume thresholds documented in the liter-ature for comparability of our findings across countries.
Notwithstanding the above limitations, the findings sug-gest that high-volume hospitals have some processes, infra-structure/personnel factors, or both that seem to produce not only better short-term outcomes but also better long-term outcomes. All high-volume hospitals are teaching hos-pitals. Detailed comparative studies are necessary of the care processes, medical and nursing monitoring, technical-support infrastructure, and care provider interactions with patients to understand what makes for better long-term outcomes.
Theoretically, short-term outcomes could be attributed to better surgical skills because of higher surgeon volumes. Yet analysis of the above data by surgeon volume instead of hos-pital volume does not show an association (data not shown). Our finding that hospital volume, but not surgeon volume, predicts adverse long-term outcomes is plausible for a risk procedure such as CABG performed on high-risk patients. The speedy and appropriate response of a hos-pital’s preoperative and postoperative technical support and skilled medical/nursing manpower can make or break a patient’s MACE vulnerability.
Intuitively, one would expect that the above attributes would affect short-term MACE survival, but one might ques-tion why these instituques-tional attributes would affect long-term outcome. One possibility is that the high intensity of qualified manpower in a teaching institution (eg, residents, medical students, nursing students, and dieticians) might result in a greater probability of appropriate advice and health educa-tion support to patients to attend regular and full-scale check-up visits, as well as to change their lifestyle and risk factors after leaving the hospital.
Although the above explanations are clearly speculative, they are presented as possible mediating variables underlying our finding. The differences in long-term outcomes in
addition to short-term outcomes amounts to 50% greater odds of a MACE within 30 days (reciprocal of 0.666) among patients in low-volume hospitals (adjusted for severity, comorbidities, patient demographics, and other factors) and an additional 23% greater (adjusted) MACE odds after 30 days but within 5 years (reciprocal of 0.811). This magnitude of difference alone justifies detailed process studies by peer task forces to leverage our findings into suitable professional and policymaker interventions. The answer might not neces-sarily be regionalization of high-risk procedures because some low-volume hospitals can produce excellent long-term outcomes and some high-volume hospitals can provide poor outcomes. Only a detailed study of care processes and services common to high-volume hospitals can provide guid-ance to address this issue.
References
1. Nallamothu BK, Eagle KA, Ferraris VA, Sade RM. Should coronary artery bypass grafting be regionalized? Ann Thorac Surg. 2005;80: 1572-81.
2. Wu C, Hannan EL, Ryan TJ, Bennett E, Culliford AT, Gold JP, et al. Is the impact of hospital and surgeon volumes on the in-hospital mortality rate for coronary artery bypass graft surgery limited to patients at high risk?Circulation. 2004;110:784-9.
3. Hannan EL, Kilburn H, Bernard H, O’Donnell JF, Lukacik G, Shields EP. Coronary artery bypass surgery: the relationship between in-hospital mortality rate and surgical volume after controlling for clin-ical risk factors.Med Care. 1991;29:1094-107.
4. Hannan EL, O’Donnell JF, Kilburn H, Bernard HR, Yazici A. Investiga-tion of the relaInvestiga-tionship between volume and mortality for surgical pro-cedures performed in New York State hospitals. JAMA. 1989;262: 503-10.
5. Rathore SS, Epstein AJ, Volpp KG, Krumholz HM. Hospital coronary artery bypass graft surgery volume and patient mortality, 1998-2000. Ann Surg. 2004;39:110-7.
6. Birkmeyer JD, Siewers AE, Finlayson EV, Stukel TA, Lucas FL, Batista I, et al. Hospital volume and surgical mortality in the United States.N Engl J Med. 2002;346:1128-37.
7. Hannan EL, Wu C, Ryan TJ, Bennett E, Culliford AT, Gold JP, et al. Do hospitals and surgeons with higher coronary artery bypass graft surgery volumes still have lower risk-adjusted mortality rates? Circulation. 2003;108:795-801.
8. Hannan EL, Siu AL, Kumar D, Kilburn H, Chassin MR. The decline in coronary artery bypass graft surgery mortality in New York State. The role of surgeon volume.JAMA. 1995;273:209-13.
9. Showstack JA, Rosenfeld KE, Garnick DW, Luft HS, Schaffarzick RW, Fowles J. Association of volume with outcome of coronary artery bypass
graft surgery. Scheduled vs nonscheduled operations.JAMA. 1987;257: 785-9.
10. Shahian DM. Improving cardiac surgery quality—volume, outcome, process?JAMA. 2004;291:246-8.
11. Zacharias A, Schwann TA, Riordan CJ, Durham SJ, Shah A, Papadimos TJ, et al. Is hospital procedure volume a reliable marker of quality for coronary artery bypass surgery? A comparison of risk and propensity adjusted operative and midterm outcomes. Ann Thorac Surg. 2005;79:1961-9.
12. Youssef AA, Chang LT, Hang CL, Wu CJ, Cheng CI, Yang CH, et al. Level and value of interleukin-18 in patients with acute myocardial infarction undergoing primary coronary angioplasty.Circ J. 2007;71: 703-8.
13. Wang TY, Hasselblad V, Peterson JL, Wijns W, Parhizgar A, Kereiakes DJ, et al. The Cobalt chromium STent with Antiproliferative for Restenosis II (COSTAR II) trial study design: advancing the active-control evaluation of second-generation drug-eluting stents.Am Heart J. 2007;153:743-8.
14. Rodriguez AE, Baldi J, Ferna´ndez Pereira C, Navia J, Rodriguez Alemparte M, et al. Five-year follow-up of the Argentine randomized trial of coronary angioplasty with stenting versus coronary bypass surgery in patients with multiple vessel disease (ERACI II).J Am Coll Cardiol. 2005;46:582-8.
15. Tomasco B, Cappiello A, Fiorilli R, Leccese A, Lupino R, Romiti A, et al. Surgical revascularization for acute coronary insufficiency: analysis of risk factors for hospital mortality.Ann Thorac Surg. 1997;64:678-83. 16. Chen Y, Almeida AA, Goldstein J, Shardey GC, Pick AW,
Moshinsky R, et al. Urgent and emergency coronary artery bypass graft-ing for acute coronary syndromes.ANZ J Surg. 2006;76:769-73. 17. Luft HS, Bunker JP, Enthoven AC. Should operations be regionalized?
The empirical relation between surgical volume and mortality.N Engl J Med. 1979;301:1364-9.
18. Jollis JG, Peterson ED, Nelson CL, Stafford JA, DeLong ER, Muhlbaier LH, et al. Relationship between surgeon and hospital coro-nary angioplasty volume and outcome in elderly patients.Circulation. 1997;95:2485-91.
19. Cheng SH, Song HY. Surgeon performance information and consumer choice: a survey of subjects with the freedom to choose between doctors. Qual Saf Health Care. 2004;13:98-101.
20. Wen HC, Tang CH, Lin HC, Tsai CS, Chen CS, Li CY. Association between surgeon and hospital volume in coronary artery bypass graft surgery outcomes: a population-based study.Ann Thorac Surg. 2006; 81:835-42.
21. Shahian DM, Normand SL. The volume-outcome relationship: from Luft to Leapfrog.Ann Thorac Surg. 2003;75:1048-58.
22. Halm EA, Lee C, Chassin MR. Is volume related to outcome in health care? A systematic review and methodologic critique of the literature. Ann Intern Med. 2002;137:511-20.
23. Shook TL, Sun GW, Burstein S, Eisenhauer AC, Matthews RV. Com-parison of percutaneous transluminal coronary angioplasty outcome and hospital costs for low-volume and high-volume operators.Am J Cardiol. 1996;77:331-6.
