SEARCH

SEARCH BY CITATION

Keywords:

  • radical prostatectomy;
  • outcomes;
  • surgical volume

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. REFERENCES

OBJECTIVE

To determine the impact of hospital variables on immediate surgical outcomes for patients treated with radical prostatectomy (RP) in academic centres.

PATIENTS AND METHODS

The University HealthSystem Consortium (UHC) Clinical Data Base was queried for data corresponding to patients who had RP at one of 130 academic medical centres nationwide between 2003 and the second quarter of 2007 (48 086). RP case volume (1–99, 100–499 and >500), total discharges (1–49 999, 50 000–99 999, >100 000), and geographical region (five categories) were determined and categorized for each academic centre. Analysis of variance and the Tukey statistic were used to assess the results. Length of stay (LOS), intensive care unit (ICU) rate, complication rate (CR) and in-hospital mortality (IHM) were analysed.

RESULTS

Case volume was a significant predictor of LOS, ICU and CR. The mean LOS was 3.77, 2.65 and 2.09 days, respectively, for centres from three tiers of lowest to highest case volumes (P < 0.001). ICU rates for the three tiers were 18.57, 3.61, and 1.30 (P < 0.001); CRs were 15.93, 8.79 and 5.76 (P < 0.001). Tukey analysis showed a ‘ceiling’ effect for ICU and CRs; there were no differences between the two higher case-volume groups. IHM was not significantly different between groups stratified by case volume. Stratification by total discharges showed differences in ICU rates only (P = 0.003). Stratification by geographical region showed no differences in outcome.

CONCLUSIONS

RP case volume was an important variable in predicting three of the four outcome variables. CRs and ICU rates showed a ‘ceiling effect’ suggesting that an unknown ‘critical volume’ of cases portends improved surgical outcomes.


Abbreviations
RP

radical prostatectomy

LOS

length of hospital stay

CR

complication rate

IHM

in-hospital mortality

UHC

University HealthSystem Consortium

ICD

International Classification of Diseases.

INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. REFERENCES

Previous studies in several different disease systems have indicated that for major surgical procedures, outcomes might be better for patients who have a particular surgical procedure in a hospital in which very many of those procedures are performed [1,2]. Other investigators have posited that specialized hospitals, or hospitals in which large numbers of any type of complicated surgery are conducted, are associated with favourable surgical outcomes. Most of these studies use Medicare billing databases or the Nationwide Inpatient Sample as data sources [3–5]. Radical prostatectomy (RP) is unique surgery, both because of its technical complexity and because of its relative ubiquity across the spectrum of hospital centres. In the present study we drew data from a national database of academic medical centres, and the study was designed to determine whether outcomes after RP depend on hospital case volume, hospital size, and geographical region of the United States.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. REFERENCES

The University HealthSystem Consortium (UHC) is a non-profit, member-driven alliance of >90% of the academic medical centres in the USA. The UHC offers tools and services that support clinical resource management for members, and quality-improvement initiatives. For this study we used the Clinical Data Base of the UHC; this is an electronic repository that combines administrative, clinical and financial data from each of the member institutions of the UHC. Data are gathered from participant hospital discharge summaries and Uniform Billing-04 data.

Data from 130 hospital centres at which RP was performed were used for this analysis (there are 184 hospital centres in the UHC network). To capture patient data, the International Classification of Diseases, 9th revision (ICD-9) procedure code for RP (605) was used as a search criterion. This code captures all RP procedures, including open, laparoscopic and robotic-assisted RP. Data were available from 2003 to the second quarter of 2007. Hospital centres were stratified categorically by the total number of RPs performed over the period of the study, total hospital discharges (as a surrogate for hospital size) and by geographical region of the country. For case volume, three categories were defined before analysis to divide the participant hospitals into approximate tertiles: 0–99, 100–499, and >500. Similarly, for hospital discharges, limits were defined as 0–49 999, 50 000–99 999 and >100 000, to group hospitals into approximate tertiles. The country was divided into five geographical regions as defined by the USA Census Bureau (West, Midwest, Central, Southeast, and Northeast).

Outcome variables of interest were the hospital length of stay (LOS), intensive care unit (ICU) admission rate (defined as the percentage of patients admitted to an ICU after RP), in-hospital mortality (IHM) and complication rate (CR). Complications were defined as those which triggered a separate ICD-9 designation. These were distinguished from comorbid conditions because they were cited immediately after RP. Outcome variables were available as mean values for each medical centre; these means were used in the subsequent analyses. anova was used to determine if there was an overall difference between volume categories for a given outcome variable. Where differences existed, the Tukey analysis was used to compare each volume category with every other category. All P values were two-tailed and a P ≤ 0.05 was taken to indicate statistical significance for each analysis.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. REFERENCES

In all, 48 086 patients had RP during the period covered by the study, with the surgery conducted at one of 130 hospital centres; 72% of patients were aged <65 years and only 1.1% were aged >75 years; 74.7% were Caucasian, while African-Americans and Hispanics comprised 9.4% and 2.4%, respectively. The hospital centre characteristics are listed in Table 1; importantly, RP case volume and total hospital discharges are cumulative over the study period. Therefore, a case volume of 100 in this study equates to ≈20 cases per year, and a case volume of 500 to ≈110 cases per year, etc.

Table 1.  Hospital centre characteristics
CharacteristicN (%) of hospitalsN (%) of cases
Geographical region  
 Central19 (14.6) 4 279 (8.8)
 Midwest30 (23.1)13 543 (28.1)
 North-east45 (34.6)18 606 (38.7)
 South-east16 (12.3) 4 697 (9.8)
 West20 (15.4) 6 961 (14.5)
Total hospital discharges  
 0–49 99933 (25.4) 3 165 (6.6)
 50 000–99 99939 (30.0) 8 739 (18.2)
 >100 00058 (44.6)36 182 (75.2)
RP case volume  
 0–9948 (36.9) 1 927 (4.0)
 100–49957 (43.8)14 540 (30.2)
 >50025 (19.2)31 619 (65.6)

The results of the anova are shown in Table 2; RP case volume was significantly related to LOS, ICU admission rate and CR, but not to IHM (P = 0.224). Specifically, the LOS ranged from 2.09 days at the highest case volume centres, to 3.77 days at the lowest (P < 0.001). ICU admission rates were lowest for the highest volume centres (1.3%), and highest for the lowest case volume centres (18.6%) (P < 0.001). CRs were also disparate when stratified by case volume, at 5.76% to 15.93% from the highest to lowest case volume, respectively (P < 0.001). Hospital size (as measured by total hospital discharges) was related only to ICU admission rate (P = 0.003). Geographical region of the country was not significantly related to any of the outcome variables studied.

Table 2. anova for each outcome variable (categorized by case volume, hospital size, and geographical region)
Mean (sd) values, byLOSICU rate, (%)CR, (%)IHM, (%)
Case volume (hospitals/cases)    
 0–99 (48/1927)3.77 (1.12)18.57 (31.22)15.93 (16.05)0.15 (0.52)
 100–499 (57/14540)2.65 (0.63)3.61 (4.35)8.79 (4.01)0.05 (0.17)
 >500 (25/31619)2.09 (0.52)1.30 (0.68)5.76 (2.08)0.02 (0.04)
 P<0.001<0.001<0.0010.224
Hospital size    
 0–49 999 (33/3165)3.26 (1.17)17.82 (32.09)14.21 (19.16)0.01 (0.04)
 50 000–99 999 (39/8739)3.02 (1.16)9.88 (20.63)10.50 (6.19)0.11 (0.37)
 >100 000 (58/36182)2.74 (0.86)2.73 (3.26)9.16 (5.13)0.11 (0.40)
 P0.0700.0030.1000.322
Region    
 West (20/6961)2.97 (0.91)2.77 (3.33)9.92 (5.49)0.21 (0.59)
 Central (19/4279)3.35 (1.38)13.32 (27.25)9.80 (7.62)0.16 (0.47)
 Midwest (30/13543)29.7 (1.17)12.85 (24.92)12.79 (8.61)0.03 (0.08)
 Southeast (16/4697)2.85 (1.16)12.58 (27.27)13.70 (23.62)0.14 (0.45)
 Northeast (45/18606)2.82 (0.82)4.60 (12.97)9.39 (7.88)0.02 (0.11)
 P0.4670.1940.5310.172

After significant differences were detected by anova, the Tukey statistic was used to probe differences between tiers; the results are shown qualitatively in Fig. 1, and CIs are given in Table 3. LOS was significantly different between each tier of hospitals when stratified by case volume. ICU admissions and CRs were significantly different only between the first and second tiers. Differences between the higher case volume tiers were not significant. When stratified by hospital size, ICU admissions were only significantly different between tiers 1 and 3.

image

Figure 1. Outcome variables stratified by LOS.

Download figure to PowerPoint

Table 3.  Tukey statistic results and 95% CI for outcomes based on case volume
Variable/tier (cases)MeanDifference in mean (95% CI)
  • *

    Significant.

LOS, days  
 Tier 1 (0–99)3.77(1–2) −1.11 (−1.50 to −0.73)*
 Tier 2 (100–499)2.65(1–3) −1.67 (−2.16 to −1.18)*
 Tier 3 (>500)2.09(2–3) −0.55 (−1.03 to −0.81)*
ICU rates  
 Tier 118.57(1–2) −14.95 (24.00 to −5.91)
 Tier 2 (1–3) −17.26 (−28.49 to −6.04)*
 Tier 3 (2–3) −2.31 (−13.17 to 8.55)
CRs, %  
 Tier 115.93(1–2) −7.14 (−11.86 to −2.42)*
 Tier 28.79(1–3) −10.16 (−16.11 to −4.22)*
 Tier 35.76(2–3) −3.02 (−8.80 to 2.75)
IHM, %  
 Tier 10.15(1–2) −0.09 (−0.25 to 0.06)
 Tier 20.055(1–3) −0.12 (−0.32 to 0.73)
 Tier 30.026(2–3) 0.02 (−0.22 to 0.16)

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. REFERENCES

The present findings affirm the positive relationship between case-specific surgical volume and short-term outcomes after RP. Specifically, these findings show a relationship between RP case-specific volume and hospital LOS, CRs and ICU admission rates. The association between surgical outcomes and hospital centre characteristics has been scrutinized by many investigators over the last decade. Hospital size, case-specific volume, hospital speciality status, and individual surgeon volume have all been cited as important factors affecting outcomes after surgery. Such studies raise many complicated questions about exactly what conclusions can be drawn from retrospective database studies, and how to apply those conclusions to the current system of hospital care.

Several authors have examined the outcomes of RP with regard to hospital centre characteristics. Using the Surveillance, Epidemiology, and End Results Medicare database, Begg et al.[6] studied rates of postoperative complications and late urinary complications for almost 12 000 men for hospital case-specific volume. They found no relationship between hospital volume and mortality, but found a slight reduction in postoperative complications when surgery was performed in very high volume centres compared to low volume centres (27% vs 32%). They described a similar trend when CRs were stratified by surgeon volume. The number of samples was large, and differences small, but still significant. Yao et al.[4], almost 10 years ago, used Medicare records to examine the association between LOS, surgical CRs, readmission rate and mortality for >100 000 men treated with RP. Their data are somewhat historical in value, as the LOS was 7–9 days, but they did report significant differences in all of their outcome variables when stratified by case-specific hospital volume. A more modern study by Ku et al.[7] examined outcomes after RP in the Veterans Affairs system, stratified by case-specific hospital volume. They found that among academic institutions, high-volume centres were associated with a shorter LOS and lower risk of transfusion. Ellison et al.[3] studied the effect of hospital volume on mortality; they confirmed a low overall IHM rate of 0.25%, but found a significant increase in mortality for low-volume centres relative to high-volume centres (<25 vs >54 cases annually).

Nuttall et al.[8] reviewed the reports relating hospital volume to surgical outcomes, and 12 studies were found which focused on urological cancer procedures, eight of which studied RP. For each study, hospital case volume had either a positive or neutral effect or at least one outcome variable. Outcomes included mortality, LOS, CRs, and late urinary CRs. It would be impossible to combine data, even in a Forrest plot, because the studies were heterogeneous in terms of definitions of outcome variables and patient populations.

The current study is unique in that the data used were compiled exclusively from academic medical centres, and it is a more recent series than virtually any other study relating outcomes to hospital characteristics. It is, to our knowledge, the first such study to be drawn from the UHC Clinical Data Base. Along with the cited studies, the present study suggests a small but definite association between case-specific volume and short-term outcomes after RP. This study does not characterize outcomes by individual surgeon or patient health status. In addition, a patient’s long-term outcome is not captured. Moreover, CRs are heavily dependent on coding and documentation rates at each hospital, making cross-institutional comparisons more difficult to interpret. However, the study raises important questions about optimizing patient outcomes and medical resources. How to apply these findings to patient care is unclear. It should be understood that none of the outcome variables in the current or any of the other studies is necessarily a direct surrogate for quality. However, the association between hospital case volume and outcome would seem to support the practice of volume-based referrals. This concept was recently addressed by Hollenbeck et al.[9] in a study examining mortality and LOS for six cancer surgeries, including RP and cystectomy. The percentage of attributable risk was calculated to determine the potential impact of ‘regionalization’ of care. They calculated that 0.11 lives per 100 000 population in the USA would be saved if cases were regionalized from the lowest volume centres to the highest volume centres. Obviously, the absolute mortality after RP is low. They also estimated that >10 000 patients annually would avoid a prolonged hospitalization if they were regionalized to high-volume centres (determined by a LOS of >90% percentile for the year in which surgery was performed). Their conclusion was, in part, that regionalization of care would result in different outcomes depending on the measure being studied. Not many lives would be saved by regionalization of RP, but the LOS would be shorter for many patients. Whether a shorter LOS warrants the increased time, travel, inconvenience and expense of travelling to a regional centre is impossible to determine based on data alone. Importantly, from a patient’s perspective, decreasing LOS is not necessarily the highest priority.

As a more reasonable and feasible alternative to the regionalization of RP care, it seems prudent to determine which attributes of higher-volume centres can be adopted by lower-volume centres. We previously described the initiation of a clinical-care pathway for patients undergoing RP; that study suggested that patients can be safely changed to a 2-day LOS from a 3-day pathway. The use of standardized diet and pain-control orders, with education by nurse specialists, allowed for a safe transition [10]. Since then, the pathway has been shortened to a 1-day LOS for most patients. Pathway-driven care is becoming more commonplace, driven by quality-improvement initiatives. Whether or not such pathways would reduce LOS when instituted at lower-volume centres is unknown, but should be explored, and represents only one avenue by which to apply the results of studies such as the present.

The present study has several limitations. It does not characterize outcomes by individual surgeon or patient health status. In addition, a patient’s long-term outcome is not captured. Moreover, CRs are heavily dependent on coding and documentation rates at each hospital, making cross-institutional comparisons more difficult to interpret. The UHC Clinical Data Base generates only summary data (mean values) for each hospital centre. Each of the outcome variables included in the study was analysed as a raw, mean value for each medical centre, and values were not weighted by case volume, hospital volume, or standard deviation. It was not possible therefore to characterize the relationships between variables by controlling for the impact of covariables.

In conclusion, this study strengthens a growing body of evidence to suggest that hospital case volume for radical prostatectomy has a significant impact on postoperative outcomes. Length of stay, complication rates, and ICU admission rates are more favourable for patients who have surgery performed at a relatively high-volume center.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. REFERENCES