- © 2003 by American Society of Clinical Oncology
Role of Surgeon Volume in Radical Prostatectomy Outcomes
- From the Departments of Urology and Health Services, University of California, Los Angeles David Geffen School of Medicine and School of Public Health, Los Angeles, CA; Abt Associates Clinical Trials, Bethesda, MD; and TAP Pharmaceutical Products Inc., Lake Forest, IL.
- Address reprint requests to Mark S. Litwin, MD, Department of Urology, UCLA School of Medicine, Box 951738, Los Angeles, CA 90095-1738; email: mlitwin{at}ucla.edu.
Abstract
Purpose: To examine the effect of hospital and surgeon volume on postoperative outcomes and to determine whether hospital or surgeon volume is the stronger predictor.
Patients and Methods: Using 1997 to 1998 claims data from a national 5% random sample of Medicare beneficiaries, we identified 2,292 men who underwent radical prostatectomy at 1,210 hospitals by 1,788 surgeons. Hospitals were classified as high (≥ 60 per year) or low (< 60 per year) volume according to radical prostatectomy experience over the 2-year period. Surgeons were classified as high (≥ 40 per year) or low (< 40 per year) volume. Multivariate logistic regression was performed to control for patient demographics and comorbidities when assessing the association of hospital and surgeon volume with in-hospital complications, length of stay, and anastomotic stricture rates. In-hospital complications included cardiac, respiratory, vascular, wound, genitourinary, and miscellaneous surgical and medical conditions.
Results: High-volume surgeons had half the complication risk (odds ratio [OR] = 0.53; 95% confidence interval [CI], 0.32 to 0.89) and shorter lengths of stay (4.1 v 5.2 days, P = .03) compared with low-volume surgeons. High-volume hospital patients tended to have fewer anastomotic strictures (OR = 0.72; 95% CI, 0.49 to 1.04). Patient age (≥ 75 years) was associated with more complications (OR = 1.9; 95% CI, 1.39 to 2.70), more anastomotic strictures (OR = 2.2; 95% CI, 1.54 to 3.15), and longer hospital stays (parameter estimate = 2.26; 95% CI, 1.75 to 2.77).
Conclusion: Surgeon volume is inversely related to in-hospital complications and length of stay in men undergoing radical prostatectomy. Hospital volume is not significantly associated with outcomes after adjusting for physician volume. Further study is necessary to elucidate the mechanism of the volume-outcome effect.
IN ACCORDANCE with the widespread adoption of cost containment measures to control increasing healthcare expenditures, outcomes research has focused increasing attention on assessing the quality of medical care. In the classic Donabedian framework, quality of care comprises the components of structure, process, and outcomes.1 Studies analyzing the process of surgery and the outcomes of mortality and complications have shown that high-volume hospitals offer better outcomes for vascular surgery, liver transplantation, and pancreatic, lung, esophageal, breast, kidney, and bladder cancer surgery.2–,7 In addition, Dudley8 suggested that 26% of deaths resulting from high-risk surgery, AIDS or human immunodeficiency virus at low-volume hospitals were preventable had these patients received care at high-volume hospitals. These findings have prompted employers to use volume as a surrogate for quality and to request that insurers refer patients to high-volume centers for high-risk procedures.9 In addition to rewarding high-volume providers with more patients, employers may increase reimbursement accordingly for superior performance.10
Prostate cancer is the most common malignancy in men and the second most common cause of cancer death in men older than 60 years.11 The US federal government is the largest payer for the treatment of prostate cancer, with Medicare costs exceeding $1.4 billion in 1994.12 Although controversy exists regarding which therapy is optimal for patients with organ-confined prostate cancer, radical prostatectomy remains the most common treatment in the United States.13 Higher hospital radical prostatectomy volume is associated with lower mortality rates, shortened length of stay,14 lower costs,15 and lower complication rates.15–,17 In addition, high surgeon volume is associated with lower complications rates.16 To determine whether surgeon and hospital volumes are independent predictors of radical prostatectomy outcomes, we examined the association of both volume elements with in-hospital complications, anastomotic strictures, and length of stay in a national sample of Medicare beneficiaries. We also investigated whether physician or hospital volume was more strongly associated with these outcomes.
PATIENTS AND METHODS
Patients undergoing radical prostatectomy were identified from a 5% national random sample of 1997 and 1998 claims data from the Center for Medicare and Medicaid Services (CMS), formerly known as the Health Care Financing Administration. Men who underwent radical prostatectomy over this 24-month period (N = 2,292) were identified on the basis of the presence of either International Classification of Diseases (9th revision [ICD-9]), code 60.5 or Physicians Current Procedural Terminology Coding System (4th edition [CPT-4]), codes 55810, 55812, 55815, 55840, 55842, or 55845. Patient demographics, complications, length of stay, hospital characteristics, and surgeon and hospital radical prostatectomy volume were obtained from an analytic file created by linking encrypted beneficiary identification numbers from three Medicare Standard Analytic Files representing hospital inpatient, hospital outpatient, and physician-supplier part B care. Previous studies provided internal validation for similar research using Medicare data.18–,21
Patient comorbidity was assessed using the Charlson index, and was constructed using ICD-9 diagnosis codes during the year before radical prostatectomy.22 The Charlson score represents the sum of weighted diagnosis codes for each comorbid condition. A higher score indicates greater comorbidity.
Geographic region classification was consistent with the methodology used by the US Census Bureau (northeast, midwest, south, west).
Each patient was longitudinally tracked for 12 months after radical prostatectomy. Key outcomes were identified using all available ICD-9 or CPT-4 codes for relevant procedures and diagnosis according to the protocol used by Yao et al.17 Outcomes of interest were in-hospital mortality and complications (cardiac, respiratory, vascular, wound-bleeding, genitourinary, miscellaneous medical, miscellaneous surgical), hospital length of stay, and anastomotic strictures. Perioperative complications and mortality were assessed in the hospital setting using the hospital inpatient file. Anastomotic strictures in the 12 months after surgery were assessed in both inpatient and outpatient settings using the inpatient, outpatient, and physician-supplier files.
Hospitals were classified as either high- or low-volume providers on the basis of their cumulative experience in 1997 and 1998. High-volume hospitals performed 60 or more radical prostatectomies per year (corrected for the 5% sampling). Our rationale for setting this volume cutoff point was based on our subjective clinical assessment that a high-volume hospital should perform more than one radical prostatectomy per week. This approximates Ellison’s high-volume hospital definition of more than 54 radical prostatectomies per year15 and is somewhat higher than Yao’s threshold of 35 radical prostatectomies per year.17
Physicians were classified as high or low volume on the basis of their cumulative operative experience in 1997 and 1998. High-volume surgeons performed 40 or more radical prostatectomies per year (corrected for the 5% sampling). Our rationale for setting this volume cutoff point was also subjective, and was based on doubling the national mean radical prostatectomy surgeon volume, which in 1997 was 18.23
To analyze possible interaction effects between surgeon and hospital volume, the outcomes of interest were stratified by four volume categories: low-volume surgeons in low-volume hospitals, low-volume surgeons in high-volume hospitals, high-volume surgeons in low-volume hospitals, and high-volume surgeons in high-volume hospitals.
To ensure that our results were both statistically sound and clinically meaningful, we considered provider volume as a categorical, rather than continuous, variable. An odds ratio reported from a continuous measure would have indicated the incremental improvement in outcome of one additional surgery, which we judged to have limited clinical interpretability. Our approach is also consistent with prior similar research.15–,17
Statistical Analysis
A descriptive summary of patient and hospital characteristics was derived for the high- and low-volume groups. The outcomes of interest were compared between groups with volume as the independent variable. Categorical variables were analyzed using the χ2 statistic with an alpha level of 0.05. Length of stay was compared using one-way analysis of variance. The distribution of the length of stay data was unimodal and symmetric about the median with a slight tail containing larger valued outliers. Given the robustness of normal theory with respect to symmetric distributions, we determined that a linear regression without any data transformations was appropriate.
The multivariate analysis to determine the association between volume and length of stay was conducted using a linear regression model. Perioperative complications and anastomotic stricture outcomes analyses were conducted using multivariate logistic regression.
R2 and C (rank correlation index) statistics are reported to provide overall model predictive value for the linear and logistic models, respectively. Both statistics have values that range between zero and one. Zero indicates no predictive value and one indicates perfect predictive value of model.
RESULTS
We identified 2,292 patients who underwent radical prostatectomy in 1997 or 1998. Two hundred twenty patients had missing physician information and were excluded from the surgeon and multivariate analyses. Patient and provider characteristics stratified by volume and geographical region are shown in Table 1⇓. Patient demographics were similar with the exception that low-volume hospitals performed radical prostatectomy on older patients (P = .001). In addition, high-volume providers were more often academically affiliated (P < .0001).
In bivariate analysis, significant volume-outcome effects were evident for overall complications, mean length of stay, and anastomotic stricture rates (Table 2⇓). Low-volume surgeons had twice the in-hospital complication rate of high-volume surgeons (21.9% v 11.8%, P < .01). Furthermore, low-volume surgeons hospitalized radical prostatectomy patients 1 day longer than did high-volume surgeons (5.2 v 4.1 days, P < .01). Patients at high-volume hospitals experienced fewer anastomotic strictures (19.8% v 26.8%, P = .01) and shorter lengths of stay (4.4 days v 5.2 days, P < .01); however, hospital volume was not significantly associated with these outcomes in the multivariate analysis.
Simultaneous estimation of the effects of hospital volume and physician volume is presented in Table 3⇓. This model was used to determine whether the interaction between hospital and surgeon volume should be included in the multivariate model. The only significant interaction effect is for the length of stay outcome. The interaction is ordinal. There is a larger difference between lengths of stay attributable to surgeon volume (4.7 v 3.7 days) in patients of high-volume hospitals than in low-volume hospitals (5.2 v 4.8 days). However, when all main effects in the final multivariate analysis are controlled for, the interaction between surgeon and hospital volume became nonsignificant (data not shown).
Table 4⇓ summarizes the characteristics associated with in-hospital complications, anastomotic strictures, and length of stay in multivariate analysis. After simultaneously adjusting for hospital volume and patient characteristics, patients of high-volume surgeons were half as likely to experience in-hospital complications (odds ratio [OR] = 0.53; 95% confidence interval [CI], 0.32 to 0.89) and had shorter lengths of stay compared with patients of low-volume surgeons (parameter estimate [PE] = −0.66; 95% confidence limit, −1.26 to −0.06). Length of stay was greater for nonwhites (PE = 0.60; 95% CI, 0.15 to 1.05) and those in the northeast region (see table). Finally, greater age was significantly associated with worse outcomes for in-hospital complications (OR = 1.94; 95% CI, 1.39 to 2.70), anastomotic strictures (OR = 2.21; 95% CI, 1.54 to 3.15) and length of stay (PE = 2.26; 95% CI, 1.75 to 2.77).
The R2 value for the mean length of stay model is 0.08, indicating that 8% of the variation in length of stay is explained by the predictors in the model. The corresponding C statistics for the logistic regressions were 0.6, demonstrating good model fit.
DISCUSSION
Our study suggests that higher surgeon volumes are associated with lower in-hospital complication rates and shorter lengths of stay after radical prostatectomy. After adjusting for physician volume, hospital volume was not significantly associated with either outcome. However, patients at high-volume hospitals demonstrated a tendency toward lesser risk for anastomotic strictures. Physician or hospital affiliation with an academic center was not independently associated with outcomes.
The mortality from radical prostatectomy at major academic centers ranges from 0% to 0.42%.24–,26 Previous studies of hospital volume and radical prostatectomy mortality found that patients in low-volume hospitals were 1.78 times more likely to suffer in-hospital mortality and had 1.51 times increased risk of 30-day mortality compared with high-volume hospitals.15,17 The low mortality (three deaths, 0.13%) in our study precluded mortality comparisons between high- and low-volume providers.
Single-surgeon, academic series report radical prostatectomy complication rates ranging from 2.2% to 27.8%24–,26; however, differences in defining complications make meaningful comparisons difficult. Using the Surveillance, Epidemiology, and End Results Medicare-linked data from 1992 to 1996, Begg et al16 noted a 30-day postoperative complication rate of 32% at low-volume hospitals compared with 27% at high-volume hospitals. In addition, high-volume surgeons, compared with low-volume surgeons, experienced a lower postoperative complication rate of 26% v 32%.16 Similarly, Yao17 found an overall 30-day complication rate that ranged from 31% in low-volume hospitals to 26% in high-volume hospitals when using Medicare claims data from 1991 to 1994. In our study, although both high-volume hospitals and surgeons had lower in-hospital complications, hospital volume was not significantly associated with complications in multivariate analyses.
Anastomotic strictures are reported to be the most common complication following radical prostatectomy, with rates ranging from 4% to 17.5% at centers of excellence,24,27,28 to 20% from patient self-reported questionnaires.29,30 In our study, the overall anastomotic stricture rate was 25.7%. Risk factors for developing anastomotic strictures relate to surgical factors (eg, excessive intraoperative bleeding and urinary extravasation) and patient comorbidities.27,31 Advances in surgical technique (eg, bladder mucosal eversion, mucosa-to-mucosa apposition, and tension-free anastomosis) have decreased the incidence of anastomotic strictures.24,27 Patients with coronary artery disease, hypertension, and diabetes mellitus also have an increased rate of anastomotic strictures attributed to microvascular disease and resultant tissue ischemia during wound healing.32 Despite these reports, patient comorbidity was not significantly associated with anastomotic strictures in our analysis.
High-volume hospitals tended to have fewer anastomotic strictures, whereas surgeon volume had no impact on their incidence. There are two possible explanations for this counterintuitive finding. First, radical prostatectomy patients in high-volume hospitals were younger in our study, and age was significantly associated with anastomotic stricture formation, a finding that has not been previously shown. Second, high-volume hospitals may disseminate advances in surgical technique more readily. Teaching conferences may increase physician awareness of current medical approaches. For example, high-volume hospitals are more compliant with widely accepted clinical guidelines, such as the use of aspirin after myocardial infarction.33
Length of stay, a significant determinant of radical prostatectomy hospital charges,34,35 has been a target for managed care; using clinical care pathways, hospital stays after radical prostatectomy have been shortened from greater than 2 weeks in 198536 to 2 days at present.25 Previous studies have demonstrated an inverse relationship between hospital volume, length of stay, and costs for men undergoing radical prostatectomy.15,17 We found that greater physician volume is indeed associated with shorter length of stay. In addition, old age and northeastern location were associated with longer stays. This finding mirrors the work of Lu-Yao and others, who demonstrated substantial geographic variation in the utilization of radical prostatectomy and resultant hospital experience in postoperative care.13,37
Our use of Medicare claims data has certain limitations. First, claims files are designed primarily to provide billing information, not detailed clinical information. More comprehensive clinical data on severity of illness and comorbidity might have affected the associations we identified. Furthermore, we could not control for patient selection bias and referral patterns. Younger, potent patients desiring nerve-sparing radical prostatectomy may seek high-volume, academic centers of excellence. This is important, given our finding that younger age is strongly associated with better outcomes.
Our definition of provider volume encompasses primarily patients aged 65 years and older; hence, high-quality surgeons who operate on younger patients might be underrepresented in our sample, potentially biasing our results. With 58% of patients undergoing radical prostatectomy less than 65 years of age,38 our findings in the Medicare population may not be generalizable to younger men undergoing radical prostatectomy. Finally, our study focused on patient-level analyses. Aggregating to the provider level does not ensure a random sample of surgeons and hospitals. Low-volume providers whose patients were not sampled were therefore not characterized, potentially resulting in an underrepresentation of outcomes for this volume category. Despite these limitations, Medicare claims data allow a population-based, heterogeneous, nationwide evaluation of radical prostatectomy results in various clinical settings.
The policy impact of these results is noteworthy. Although the Leapfrog initiative9,10 favors regionalization of care at a hospital level, the American Urological Association is currently exploring the merits of a two-tier system whereby urologic surgeons perform major operative procedures and office-based urologists handle diagnosis and follow-up.39,40 Although our results seem to support this paradigm, it is difficult to establish a causal relationship between volume and outcome. Two hypotheses have been proposed to explain this effect. The practice-makes-perfect theory dictates that high-volume surgeons have better outcomes because of greater experience. Conversely, the selective referral theory holds that better surgeons attract more patients.41 Before policy recommendations are made or actual changes in medical care delivery are suggested, further study is needed to determine the direction of this relationship with volume as a surrogate for physician skill and expertise.
Surgeon rather than hospital volume determines short-term radical prostatectomy outcomes. Additional research is needed to establish thresholds that answer the question, “How many is enough?” In contrast to high-risk procedures or cancer surgery for aggressive tumors, radical prostatectomy operative mortality is low and 5-year survival is high. Consequently, complications are an important consideration for patients choosing a provider for prostate cancer surgery. Additional study is warranted to determine the effect of surgeon volume on potency, continence, and cancer control.
- Received May 24, 2002.
- Accepted October 4, 2002.