- Email: [email protected]
Centralization of Radical Prostatectomy in the United States
Centralization of Radical Prostatectomy in the United States Christopher B. Anderson,* David F. Penson, Shenghua Ni, Danil V. Makarov and Daniel A. Barocas† From the Department of Urologic Surgery (CBA, DFP, SN, DAB), and Center for Surgical Quality and Outcomes Research (DFP, DAB), Vanderbilt University Medical Center, and United States Department of Veterans Affairs Tennessee Valley Geriatric Research, Education, and Clinical Center (DFP), Nashville, Tennessee, and Department of Urology, New York University School of Medicine and the United States Department of Veterans Affairs New York Harbor Healthcare System, New York, New York (DVM)
Abbreviations and Acronyms RALP ⫽ robotic assisted laparoscopic prostatectomy RP ⫽ radical prostatectomy Accepted for publication October 4, 2012. Supported by Grant 1 R01 HS019356-01 from the Agency for Healthcare Research and Quality (DFP). Supplementary material can be obtained at www.jurology.com. * Correspondence: Department of Urologic Surgery, Vanderbilt University Medical Center, A-1302 Medical Center North, Nashville, Tennessee 37232-2765 (telephone: 615-322-2101; FAX: 615-322-8990; e-mail: [email protected]). † Financial interest and/or other relationship with Allergan and Dendreon.
Purpose: Radical prostatectomy is a common treatment for organ confined prostate cancer and its use is increasing. We examined how the increased volume is being distributed and what hospital characteristics are associated with increasing volume. Materials and Methods: We identified all men age 40 to less than 80 years who underwent radical prostatectomy for prostate cancer from 2000 to 2008 in the NIS (Nationwide Inpatient Sample) (586,429). Ownership of a surgical robot was determined using the 2007 AHA (American Hospital Association) Annual Survey. The association between hospital radical prostatectomy volume and hospital characteristics, including ownership of a robot, was explored using multivariate linear regression. Results: From 2000 to 2008 there was a 74% increase in the number of radical prostatectomies performed (p ⫽ 0.05) along with a 19% decrease in the number of hospitals performing radical prostatectomy (p ⬍0.001), resulting in an increase in annual hospital radical prostatectomy volume (p ⫽ 0.009). Several hospital variables were associated with greater radical prostatectomy volume including teaching status, urban location, large bed size and ownership of a robot in 2007. On multivariate analysis the year, teaching status, large bed size, urban location and presence of a robot were associated with higher hospital radical prostatectomy volume. Conclusions: Use of radical prostatectomy increased significantly between 2000 and 2008, most notably after 2005. The increase in radical prostatectomy resulted in centralization to select hospitals, particularly those in the top radical prostatectomy volume quartile and those investing in robotic technology. Our findings support the hypothesis that hospitals with the greatest volume increases are specialty centers already performing a high volume of radical prostatectomy procedures. Key Words: prostatectomy, prostatic neoplasms, trends, robotics
500
www.jurology.com
PROSTATE cancer is the most frequently diagnosed malignancy among men and the second leading cause of male cancer mortality in the United States.1 Radical prostatectomy is the most commonly used treatment for clinically localized prostate cancer, especially among healthy men.2 Recent studies at the state3,4 and national5 level have demonstrated an in-
crease in RP use, raising questions about how the volume is being distributed. As opposed to a proportional volume increase among all hospitals, RP appears be increasing in a pattern consistent with centralization to certain institutions.3,4,6 While centralization is potentially desirable for certain operations given its association with improved surgical
0022-5347/13/1892-0500/0 THE JOURNAL OF UROLOGY® © 2013 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION
http://dx.doi.org/10.1016/j.juro.2012.10.012 Vol. 189, 500 –506, February 2013 RESEARCH, INC. Printed in U.S.A.
AND
CENTRALIZATION OF RADICAL PROSTATECTOMY
quality, some fear it may lead to decreased access to care, increased travel distances and diminished business at low volume hospitals.7–9 There are many proposed explanations for centralization, including referral patterns and hospital business strategy.10 –12 However, nationwide RP centralization in the absence of a policy mandate could have unintended consequences for quality or access to care if driven by market forces alone. Robotic assisted laparoscopic prostatectomy has rapidly grown to account for more than two-thirds of RP nationwide.13,14 Recent studies suggest that RP is becoming centralized to hospitals that own robots and some believe the popularity of RALP is related to increasing RP volume.3–5,13,15 Whether hospital volume and robot ownership are associated at the national level is not known. In this study we determined whether centralization of RP is occurring nationally and which hospital characteristics are associated with higher RP volume, with particular attention to robot ownership.
MATERIALS AND METHODS The NIS is a large administrative database with information on patient demographics, diagnosis and procedure codes, and several hospital characteristics for each inpatient discharge. The NIS represents a 20% stratified sample from more than 1,000 United States community hospitals in 42 states and is intended to reflect nationwide trends. We included all men in the NIS age 40 to less than 80 years who underwent RP between 2000 and 2008 for prostate cancer. Patients were identified by ICD procedure code (60.5 for RP) and diagnosis code (185 or 198.82 for prostate cancer). Patients were excluded from study if they underwent RP concurrently with radical cystectomy for bladder cancer. The NIS does not have data on hospital ownership of a robot and cannot be used to determine if RP was performed robotically until the final quarter of 2008. Therefore, we merged the NIS with the 2007 AHA Annual Survey based on AHA hospital identification number, which is present in approximately 60% of NIS discharges. The AHA survey records data on whether a hospital owned a robot in 2007, but not on the year the robot was purchased or how many robots a hospital owns. NIS hospitals without an AHA identifier were classified as having an unknown status for robotic ownership. To internally validate the AHA variable we identified all NIS hospitals that submitted a procedure code for robotic assistance (17.4x) in the final quarter of 2008, and reexamined the associations among robotic ownership, hospital characteristics and procedure volume. As the results were nearly identical to those using the AHA variable (data not shown), we were confident that the AHA variable was internally valid. The main outcome was annual number of RPs performed per hospital. Hospital volume was estimated for all hospitals performing 1 or more RP. Hospitals with 1 or
501
more NIS discharge but no RP discharge were labeled as having zero RP discharges that year. Hospitals with zero NIS discharges (indicating they were not sampled that year) were considered to have missing RP discharge data. Descriptive hospital data were weighted per NIS protocol and presented as averages during 3-year periods, generally corresponding to the phases of robot diffusion (2000 to 2002—introduction, 2003 to 2005— early adoption, 2006 to 2008 —widespread diffusion). Hospitals were stratified by volume quartile each year of analysis, and associations between RP volume and hospital characteristics were measured using ANOVA and chi-square tests for continuous and categorical variables, respectively. Linear regression models were used to examine bivariate associations of overall RP volume, number of hospitals performing RP, average annual hospital RP volume and proportion of cases performed at various types of hospitals over time. A mixed effects multivariate linear regression model was used to examine the relationship among hospital characteristics, time and hospital RP volume. We included a random effect for state to account for the possibility that the association between hospital characteristics and RP volume might vary by region. We performed a sensitivity analysis using 2 additional multivariate models to account for hospitals with unknown robotic ownership status. We performed a separate multivariate analysis stratified by time period to determine whether the relationship between hospital characteristics and outcome was consistent over time. Correlation coefficients were calculated between each hospital variable, and any variable that was highly correlated with another (r ⬎0.9) was excluded from the multivariate analysis to maintain the assumption of noncollinearity. However, no variables met this criterion (max r ⫽ 0.49) and all were included in the models. All statistical analysis was performed using SAS® statistical software (version 9.2) and all p values ⱕ0.05 were considered statistically significant.
RESULTS A total of 586,429 patients underwent RP from 2000 to 2008. Hospital volume thresholds ranged from 3 to 5 for the 25th percentile, 10 to 13 for the 50th percentile and 24 to 43 for the 75th percentile, depending on the year of analysis. The majority of patients were white, privately insured and healthy. There were significant differences in the types of patients treated across hospital volume quartiles (supplementary tables 1 and 2, www.jurology.com). The majority of hospitals performing RP were nonteaching, private not-for-profit, in an urban setting and with a medium or large bed size (see table). Compared to hospitals in the lowest volume quartile, those hospitals in the highest volume quartile tended to be teaching hospitals (67.9% vs 18.6%, p ⫽ 0.059), to be located in an urban setting (96.1% vs 73.7%, p ⬍0.001), to have a large bed size (72.4% vs 41.2%, p ⬍0.001), to be private not-for-profit
502
CENTRALIZATION OF RADICAL PROSTATECTOMY
Characteristics of hospitals that performed RP from 2000 to 2008 in the United States
Mean ⫾ SD No. hospitals Teaching (%): Yes No Location (%): Urban Rural Bed size (%): Small Medium Large Geographic location (%): Northeast Midwest South West Hospital ownership (%):* Government, nonfederal Private, not-for-profit Private, investor owned
2000–2002
2003–2005
2006–2008
2,549 ⫾ 14
2,349 ⫾ 73
2,191 ⫾ 97
29.4 70.6
27.3 72.7
35.8 64.2
78 22
82.3 17.7
82.4 17.6
17.8 31.1 51.1
17.2 29.9 52.9
15.5 29.3 55.2
20.3 22.4 36.9 20.4
18.9 23.6 37.5 20
17.9 23.5 37.6 21
11.2 72.8 14.6
11.3 69.1 18.6
10 55.6 18.2
* Numbers do not add up to 100% due to missing data.
(71.7% vs 57.4%, p ⫽ 0.04) and to have owned a robot in 2007 (46.1% vs 2.2%, p ⬍0.001). Approximately 15% of hospitals owned a robot in 2007, 46% did not and ownership status was unknown in 39%. Among AHA hospitals that owned a robot there were significant differences in teaching status, setting, robot ownership and bed size compared to hospitals that did not own a robot. Hospital teaching status, setting and bed size were similar between hospitals that did vs did not respond to the AHA survey. However, there were differences in robot ownership and geographic location as certain states do not report the AHA identifier (data not shown). From 2000 to 2008 nationwide annual RP volume increased by 74% (50,058 to 87,103; p ⫽ 0.05) and the proportion of cases performed at hospitals in the highest volume quartile jumped from 65% to 80% (p ⬍0.001, fig. 1). Concomitantly the number of hospitals performing RP decreased by 19% (2,558 to 2,080; p ⬍0.001), resulting in an increase in average annual hospital volume from 22.7 to 41.9 (p ⫽ 0.009). The majority of volume increase was seen at high volume centers (2000 to 2008, 61 to 135.7, volume quartile 4, p ⫽ 0.004), whereas decreases in average hospital volume were seen in 1st (p ⫽ 0.015) and 2nd (p ⫽ 0.03) volume quartile hospitals. Several hospital variables were associated with greater annual hospital volume from 2000 to 2008 including teaching status (41.2 to 77.8 teaching [p ⫽ 0.009] vs 15.5 to 21.9 nonteaching [p ⫽ 0.035]), urban setting (26.3 to 48.5 [p ⫽ 0.01] vs rural 10.3 to 10.8 [p ⫽ 0.199]) and large bed size (30.1 to 56.2 [p ⫽ 0.006] vs small bed size 11.2 to 19.7 [p ⫽ 0.639],
fig. 2). Average hospital RP volume increased among hospitals that owned a robot in 2007 (55.2 to 127.3, p ⫽ 0.003), while it remained unchanged in hospitals without a robot (16.0 to 16.7, p ⫽ 0.781). On multivariate analysis years 2006, 2007 and 2008; teaching hospitals; urban hospitals; hospitals with large bed size and hospitals that owned robots were associated with greater annual hospital RP volume (supplementary table 3, www.jurology.com). The results were not substantially different when we omitted robot ownership from the model or when we excluded hospitals with unknown robotic ownership status. In a separate multivariate analysis stratified by 3-year period, teaching hospitals, urban hospitals, hospitals with large bed size and hospitals that owned a robot in 2007 were associated with greater hospital RP volume during each period (data not shown). Since the AHA data were from 2007 alone, we repeated this analysis with the intervals 2000 to 2002, 2003 to 2006 and 2007 to 2008, and the results were unchanged.
DISCUSSION From 2000 to 2008 RP use increased by 74% and the number of hospitals performing RP decreased by 19%, resulting in an increase in average annual hospital volume. The majority of volume increase occurred at high volume hospitals while there were volume losses at low volume hospitals, a scenario consistent with centralization. This finding contrasts with a prior NIS analysis demonstrating no increase in RP volume in the pre-robotic era (1988 to 2002) and no evidence of centralization.16 Large bed size, urban setting, teaching status and robot ownership were independently associated with greater RP volume, suggesting centralization to these types of hospitals. Compared to volume creep, where increased volume is seen proportionally among all hospitals, centralization occurs due to a concentration of care to a smaller number of hospitals.6 While centralization may be desirable given the association between high hospital volume and improved outcomes for several operations including RP, it may also have unintended consequences such as decreased access to care, increased travel distances, diminished volume and preparedness at low volume hospitals, and overcrowding at high volume hospitals.6,8,9,17–20 Based on our analysis, patients treated at high volume hospitals differed from those treated at low volume hospitals, already demonstrating potential disparities in access to care. Thus, while centralization of RP to high volume hospitals may improve surgical quality for some, it may also detract from quality in areas without access to specialty centers, and the
CENTRALIZATION OF RADICAL PROSTATECTOMY
503
Figure 1. Trends in use of RP in United States stratified by hospital volume quartile
potential benefits have yet to be weighed against the real-world consequences. While some payer and employer groups such as Leapfrog have suggested using volume standards as a means to control quality for certain operations, no such guidelines exist for RP.7 Thus, the reason for centralization in the absence of a policy mandate is not known. However, several theories exist including increasing hospital specialization.8,10 In a competitive market, hospitals may attempt to improve efficiency and outcomes by strategic product line management, whereby they focus on specific services while other services are discontinued.21 Accordingly United States hospital specialization increased 31% from 1998 to 2008.12 The hospital characteristics associated with increased volume in recent years were also associated with increased volume from 2000 to 2002, when RALP accounted for less than 10% of all prostatectomies.14 This suggests that centralization was occurring toward hospitals that had already specialized in the RP product line before the widespread diffusion of the robot. Nevertheless, robotic ownership was strongly associated with hospital volume and likely led to fur-
ther RP centralization to specialty centers. In the last decade there has been a shift in approach from open to robotic, and today the majority of RPs are performed robotically.14,22 In fact, based on NIS data from the final quarter of 2008, we and others estimate that more than 50% of RPs were performed with robotic assistance.23 While RP use plateaued during the 1990s,24 volume has increased sharply since 2005, paralleling the diffusion of RALP and mirroring a tipping point pattern in technology adoption.14,25 Given the sizable investment, hospitals that purchase robots may feel pressured to increase robotic use to remain profitable, often through aggressive marketing.26,27 The diffusion of RALP contrasts sharply with that of laparoscopic nephrectomy.28 While there may be several explanations, the lack of aggressive marketing and patient demand may have been responsible. As such, robot ownership has been associated with increasing RP volume and centralization to high volume centers.3,4,13,15 Stitzenberg et al evaluated RP trends in 3 states from 2000 to 2009 and identified an increase in volume with centralization to high volume hospitals.4 Hospitals that owned a
504
CENTRALIZATION OF RADICAL PROSTATECTOMY
Figure 2. Trends in average annual RP hospital volume overall (A), and by hospital volume quartile (B), teaching status (C), geographic location (D), setting (E), bed size (F) and status of robot ownership (G). Asterisk indicates statistically significant increase in RP volume (p ⱕ0.05). All geographic locations (D) were associated with significant increase in RP volume except for Northeast (p ⫽ 0.075).
CENTRALIZATION OF RADICAL PROSTATECTOMY
robot performed nearly 85% of RPs in 2009 and had higher volume even before purchasing a robot. Neuner et al demonstrated similar results despite the decreasing incidence of prostate cancer in the last decade.3 Similar to our findings, these data suggest that certain high volume specialty hospitals accounted for the majority of the volume increase, especially after purchasing a robot. While robot dissemination probably had a role in the further concentration of RP to select centers, it was unlikely the initial catalyst for centralization. Some have raised the question of whether the availability and marketing of robotics have induced demand for RALP, and inappropriately drawn patients away from other suitable management strategies.5,15 While our study cannot address the issues of induced demand or appropriateness of intervention, it does support the notion of a market driven centralization of RP to select hospitals. Our study does have limitations that must be considered. Longitudinal analysis is not possible with the NIS, and we could not measure volume changes within individual hospitals before and after purchasing a robot. In addition, we were unable to correlate volume and disease trends given the absence of pathological data in the NIS, although it is unlikely that changing disease characteristics explain the increase in RP.3,29 Our conclusion regarding the impact of robot ownership is based on a single year of cross-sectional data from a merged data set that does not indicate the robot was being used for RP, nor does it include the year the robot was purchased. Despite robotic ownership data on only 60% of NIS hospitals, robot ownership continued to be associated with hospital volume when controlling for unknown ownership status in our multivariate analysis. Unknown robot ownership
505
status was associated with RP volume, which we suspect is because this heterogeneous group contains hospitals with and without robots, and the association with volume reflects the diluted effect of the robot owning hospitals. Lastly, our findings may be biased by endogeneity given the potential for self-selection of robots to higher volume institutions. Endogeneity implies a bidirectional causative loop between the dependent (hospital volume) and independent (ownership of a robot) variable. While our data suggest that owning a robot is associated with increased hospital volume, we were not able to determine the presence or directionality of a causative relationship. Despite these limitations we describe a dramatic increase in the use of RP from 2000 to 2008, particularly among hospitals that specialized in RP early in the decade. We also identified a centralization of RP to large, urban, academic centers. While the presence of a robot was not solely responsible for the greater volume, it seemed to have a role in increasing the market share of hospitals specializing in RP. As more hospitals purchase surgical robots and more surgeons are trained in robotic surgery, future trends in the distribution of hospital RP volume should be reexamined.
CONCLUSIONS The use of RP increased significantly from 2000 to 2008 with centralization to select hospitals, particularly those in the top RP volume quartile and those investing in robotic technology. Our findings support the hypothesis that hospitals with the greatest volume increases were specialty centers with an already high RP volume. Centralization has potential benefits but may also have unintended consequences that are not addressed by market driven forces.
REFERENCES 1. Jemal A, Siegel R, Xu J et al: Cancer statistics, 2010. CA Cancer J Clin 2010; 60: 277.
5. Barbash GI and Glied SA: New technology and health care costs–the case of robot-assisted surgery. N Engl J Med 2010; 363: 701.
2. Cooperberg MR, Broering JM and Carroll PR: Time trends and local variation in primary treatment of localized prostate cancer. J Clin Oncol 2010; 28: 1117.
6. Finks JF, Osborne NH and Birkmeyer JD: Trends in hospital volume and operative mortality for high-risk surgery. N Engl J Med 2011; 364: 2128.
3. Neuner JM, See WA, Pezzin LE et al: The association of robotic surgical technology and hospital prostatectomy volumes: increasing market share through the adoption of technology. Cancer 2012; 118: 371. 4. Stitzenberg KB, Wong YN, Nielsen ME et al: Trends in radical prostatectomy: centralization, robotics, and access to urologic cancer care. Cancer 2012; 118: 54.
7. Birkmeyer JD, Finlayson EV and Birkmeyer CM: Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative. Surgery 2001; 130: 415. 8. Stitzenberg KB, Sigurdson ER, Egleston BL et al: Centralization of cancer surgery: implications for patient access to optimal care. J Clin Oncol 2009; 27: 4671. 9. Birkmeyer JD: Should we regionalize major surgery? Potential benefits and policy considerations. J Am Coll Surg 2000; 190: 341.
10. Learn PA and Bach PB: A decade of mortality reductions in major oncologic surgery: the impact of centralization and quality improvement. Med Care 2010; 48: 1041. 11. Luft HS, Hunt SS and Maerki SC: The volumeoutcome relationship: practice-makes-perfect or selective-referral patterns? Health Serv Res 1987; 22: 157. 12. Eastaugh SR: Hospital costs and specialization: benefits of limiting the number of product lines. J Health Care Finance 2009; 36: 24. 13. Lowrance W, Eastham J, Savage C et al: Contemporary open and robotic radical prostatectomy practice patterns among United States urologists. J Urol, suppl., 2011; 185: e136, abstract 337.
506
CENTRALIZATION OF RADICAL PROSTATECTOMY
14. Intuitive Surgical media kit. Available at www. intuitivesurgical.com/company/media/. Accessed May 2011. 15. Makarov DV, Yu JB, Desai RA et al: The association between diffusion of the surgical robot and radical prostatectomy rates. Med Care 2011; 49: 333. 16. Cooperberg MR, Modak S and Konety BR: Trends in regionalization of inpatient care for urological malignancies, 1988 to 2002. J Urol 2007; 178: 2103. 17. Birkmeyer JD, Siewers AE, Finlayson EV et al: Hospital volume and surgical mortality in the United States. N Engl J Med 2002; 346: 1128.
tatectomy volume and patient outcomes: a systematic review. J Urol 2008; 180: 820.
prostate-specific antigen screening: 1986 –2005. J Natl Cancer Inst 2009; 101: 1325.
20. Barocas DA, Mitchell R, Chang SS et al: Impact of surgeon and hospital volume on outcomes of radical prostatectomy. Urol Oncol 2010; 28: 243.
25. Barkun JS, Aronson JK, Feldman LS et al: Evaluation and stages of surgical innovations. Lancet 2009; 374: 1089.
21. Zelman WN and Parham DL: Strategic, operational, and marketing concerns of product-line management in health care. Health Care Manage Rev 1990; 15: 29.
26. Bolenz C, Gupta A, Hotze T et al: Cost comparison of robotic, laparoscopic, and open radical prostatectomy for prostate cancer. Eur Urol 2010; 57: 453.
22. Williams SB, Prasad SM, Weinberg AC et al: Trends in the care of radical prostatectomy in the United States from 2003 to 2006. BJU Int 2011; 108: 49.
18. Begg CB, Cramer LD, Hoskins WJ et al: Impact of hospital volume on operative mortality for major cancer surgery. JAMA 1998; 280: 1747.
23. Yu HY, Hevelone ND, Lipsitz SR et al: Hospital volume, utilization, costs and outcomes of robotassisted laparoscopic radical prostatectomy. J Urol 2012; 187: 1632.
19. Wilt TJ, Shamliyan TA, Taylor BC et al: Association between hospital and surgeon radical pros-
24. Welch HG and Albertsen PC: Prostate cancer diagnosis and treatment after the introduction of
27. Scales CD Jr, Jones PJ, Eisenstein EL et al: Local cost structures and the economics of robot assisted radical prostatectomy. J Urol 2005; 174: 2323. 28. Miller DC, Wei JT, Dunn RL et al: Trends in the diffusion of laparoscopic nephrectomy. JAMA 2006; 295: 2480. 29. Jemal A, Center MM, DeSantis C et al: Global patterns of cancer incidence and mortality rates and trends. Cancer Epidemiol Biomarkers Prev 2010; 19: 1893.
Abbreviations and Acronyms RALP ⫽ robotic assisted laparoscopic prostatectomy RP ⫽ radical prostatectomy Accepted for publication October 4, 2012. Supported by Grant 1 R01 HS019356-01 from the Agency for Healthcare Research and Quality (DFP). Supplementary material can be obtained at www.jurology.com. * Correspondence: Department of Urologic Surgery, Vanderbilt University Medical Center, A-1302 Medical Center North, Nashville, Tennessee 37232-2765 (telephone: 615-322-2101; FAX: 615-322-8990; e-mail: [email protected]). † Financial interest and/or other relationship with Allergan and Dendreon.
Purpose: Radical prostatectomy is a common treatment for organ confined prostate cancer and its use is increasing. We examined how the increased volume is being distributed and what hospital characteristics are associated with increasing volume. Materials and Methods: We identified all men age 40 to less than 80 years who underwent radical prostatectomy for prostate cancer from 2000 to 2008 in the NIS (Nationwide Inpatient Sample) (586,429). Ownership of a surgical robot was determined using the 2007 AHA (American Hospital Association) Annual Survey. The association between hospital radical prostatectomy volume and hospital characteristics, including ownership of a robot, was explored using multivariate linear regression. Results: From 2000 to 2008 there was a 74% increase in the number of radical prostatectomies performed (p ⫽ 0.05) along with a 19% decrease in the number of hospitals performing radical prostatectomy (p ⬍0.001), resulting in an increase in annual hospital radical prostatectomy volume (p ⫽ 0.009). Several hospital variables were associated with greater radical prostatectomy volume including teaching status, urban location, large bed size and ownership of a robot in 2007. On multivariate analysis the year, teaching status, large bed size, urban location and presence of a robot were associated with higher hospital radical prostatectomy volume. Conclusions: Use of radical prostatectomy increased significantly between 2000 and 2008, most notably after 2005. The increase in radical prostatectomy resulted in centralization to select hospitals, particularly those in the top radical prostatectomy volume quartile and those investing in robotic technology. Our findings support the hypothesis that hospitals with the greatest volume increases are specialty centers already performing a high volume of radical prostatectomy procedures. Key Words: prostatectomy, prostatic neoplasms, trends, robotics
500
www.jurology.com
PROSTATE cancer is the most frequently diagnosed malignancy among men and the second leading cause of male cancer mortality in the United States.1 Radical prostatectomy is the most commonly used treatment for clinically localized prostate cancer, especially among healthy men.2 Recent studies at the state3,4 and national5 level have demonstrated an in-
crease in RP use, raising questions about how the volume is being distributed. As opposed to a proportional volume increase among all hospitals, RP appears be increasing in a pattern consistent with centralization to certain institutions.3,4,6 While centralization is potentially desirable for certain operations given its association with improved surgical
0022-5347/13/1892-0500/0 THE JOURNAL OF UROLOGY® © 2013 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION
http://dx.doi.org/10.1016/j.juro.2012.10.012 Vol. 189, 500 –506, February 2013 RESEARCH, INC. Printed in U.S.A.
AND
CENTRALIZATION OF RADICAL PROSTATECTOMY
quality, some fear it may lead to decreased access to care, increased travel distances and diminished business at low volume hospitals.7–9 There are many proposed explanations for centralization, including referral patterns and hospital business strategy.10 –12 However, nationwide RP centralization in the absence of a policy mandate could have unintended consequences for quality or access to care if driven by market forces alone. Robotic assisted laparoscopic prostatectomy has rapidly grown to account for more than two-thirds of RP nationwide.13,14 Recent studies suggest that RP is becoming centralized to hospitals that own robots and some believe the popularity of RALP is related to increasing RP volume.3–5,13,15 Whether hospital volume and robot ownership are associated at the national level is not known. In this study we determined whether centralization of RP is occurring nationally and which hospital characteristics are associated with higher RP volume, with particular attention to robot ownership.
MATERIALS AND METHODS The NIS is a large administrative database with information on patient demographics, diagnosis and procedure codes, and several hospital characteristics for each inpatient discharge. The NIS represents a 20% stratified sample from more than 1,000 United States community hospitals in 42 states and is intended to reflect nationwide trends. We included all men in the NIS age 40 to less than 80 years who underwent RP between 2000 and 2008 for prostate cancer. Patients were identified by ICD procedure code (60.5 for RP) and diagnosis code (185 or 198.82 for prostate cancer). Patients were excluded from study if they underwent RP concurrently with radical cystectomy for bladder cancer. The NIS does not have data on hospital ownership of a robot and cannot be used to determine if RP was performed robotically until the final quarter of 2008. Therefore, we merged the NIS with the 2007 AHA Annual Survey based on AHA hospital identification number, which is present in approximately 60% of NIS discharges. The AHA survey records data on whether a hospital owned a robot in 2007, but not on the year the robot was purchased or how many robots a hospital owns. NIS hospitals without an AHA identifier were classified as having an unknown status for robotic ownership. To internally validate the AHA variable we identified all NIS hospitals that submitted a procedure code for robotic assistance (17.4x) in the final quarter of 2008, and reexamined the associations among robotic ownership, hospital characteristics and procedure volume. As the results were nearly identical to those using the AHA variable (data not shown), we were confident that the AHA variable was internally valid. The main outcome was annual number of RPs performed per hospital. Hospital volume was estimated for all hospitals performing 1 or more RP. Hospitals with 1 or
501
more NIS discharge but no RP discharge were labeled as having zero RP discharges that year. Hospitals with zero NIS discharges (indicating they were not sampled that year) were considered to have missing RP discharge data. Descriptive hospital data were weighted per NIS protocol and presented as averages during 3-year periods, generally corresponding to the phases of robot diffusion (2000 to 2002—introduction, 2003 to 2005— early adoption, 2006 to 2008 —widespread diffusion). Hospitals were stratified by volume quartile each year of analysis, and associations between RP volume and hospital characteristics were measured using ANOVA and chi-square tests for continuous and categorical variables, respectively. Linear regression models were used to examine bivariate associations of overall RP volume, number of hospitals performing RP, average annual hospital RP volume and proportion of cases performed at various types of hospitals over time. A mixed effects multivariate linear regression model was used to examine the relationship among hospital characteristics, time and hospital RP volume. We included a random effect for state to account for the possibility that the association between hospital characteristics and RP volume might vary by region. We performed a sensitivity analysis using 2 additional multivariate models to account for hospitals with unknown robotic ownership status. We performed a separate multivariate analysis stratified by time period to determine whether the relationship between hospital characteristics and outcome was consistent over time. Correlation coefficients were calculated between each hospital variable, and any variable that was highly correlated with another (r ⬎0.9) was excluded from the multivariate analysis to maintain the assumption of noncollinearity. However, no variables met this criterion (max r ⫽ 0.49) and all were included in the models. All statistical analysis was performed using SAS® statistical software (version 9.2) and all p values ⱕ0.05 were considered statistically significant.
RESULTS A total of 586,429 patients underwent RP from 2000 to 2008. Hospital volume thresholds ranged from 3 to 5 for the 25th percentile, 10 to 13 for the 50th percentile and 24 to 43 for the 75th percentile, depending on the year of analysis. The majority of patients were white, privately insured and healthy. There were significant differences in the types of patients treated across hospital volume quartiles (supplementary tables 1 and 2, www.jurology.com). The majority of hospitals performing RP were nonteaching, private not-for-profit, in an urban setting and with a medium or large bed size (see table). Compared to hospitals in the lowest volume quartile, those hospitals in the highest volume quartile tended to be teaching hospitals (67.9% vs 18.6%, p ⫽ 0.059), to be located in an urban setting (96.1% vs 73.7%, p ⬍0.001), to have a large bed size (72.4% vs 41.2%, p ⬍0.001), to be private not-for-profit
502
CENTRALIZATION OF RADICAL PROSTATECTOMY
Characteristics of hospitals that performed RP from 2000 to 2008 in the United States
Mean ⫾ SD No. hospitals Teaching (%): Yes No Location (%): Urban Rural Bed size (%): Small Medium Large Geographic location (%): Northeast Midwest South West Hospital ownership (%):* Government, nonfederal Private, not-for-profit Private, investor owned
2000–2002
2003–2005
2006–2008
2,549 ⫾ 14
2,349 ⫾ 73
2,191 ⫾ 97
29.4 70.6
27.3 72.7
35.8 64.2
78 22
82.3 17.7
82.4 17.6
17.8 31.1 51.1
17.2 29.9 52.9
15.5 29.3 55.2
20.3 22.4 36.9 20.4
18.9 23.6 37.5 20
17.9 23.5 37.6 21
11.2 72.8 14.6
11.3 69.1 18.6
10 55.6 18.2
* Numbers do not add up to 100% due to missing data.
(71.7% vs 57.4%, p ⫽ 0.04) and to have owned a robot in 2007 (46.1% vs 2.2%, p ⬍0.001). Approximately 15% of hospitals owned a robot in 2007, 46% did not and ownership status was unknown in 39%. Among AHA hospitals that owned a robot there were significant differences in teaching status, setting, robot ownership and bed size compared to hospitals that did not own a robot. Hospital teaching status, setting and bed size were similar between hospitals that did vs did not respond to the AHA survey. However, there were differences in robot ownership and geographic location as certain states do not report the AHA identifier (data not shown). From 2000 to 2008 nationwide annual RP volume increased by 74% (50,058 to 87,103; p ⫽ 0.05) and the proportion of cases performed at hospitals in the highest volume quartile jumped from 65% to 80% (p ⬍0.001, fig. 1). Concomitantly the number of hospitals performing RP decreased by 19% (2,558 to 2,080; p ⬍0.001), resulting in an increase in average annual hospital volume from 22.7 to 41.9 (p ⫽ 0.009). The majority of volume increase was seen at high volume centers (2000 to 2008, 61 to 135.7, volume quartile 4, p ⫽ 0.004), whereas decreases in average hospital volume were seen in 1st (p ⫽ 0.015) and 2nd (p ⫽ 0.03) volume quartile hospitals. Several hospital variables were associated with greater annual hospital volume from 2000 to 2008 including teaching status (41.2 to 77.8 teaching [p ⫽ 0.009] vs 15.5 to 21.9 nonteaching [p ⫽ 0.035]), urban setting (26.3 to 48.5 [p ⫽ 0.01] vs rural 10.3 to 10.8 [p ⫽ 0.199]) and large bed size (30.1 to 56.2 [p ⫽ 0.006] vs small bed size 11.2 to 19.7 [p ⫽ 0.639],
fig. 2). Average hospital RP volume increased among hospitals that owned a robot in 2007 (55.2 to 127.3, p ⫽ 0.003), while it remained unchanged in hospitals without a robot (16.0 to 16.7, p ⫽ 0.781). On multivariate analysis years 2006, 2007 and 2008; teaching hospitals; urban hospitals; hospitals with large bed size and hospitals that owned robots were associated with greater annual hospital RP volume (supplementary table 3, www.jurology.com). The results were not substantially different when we omitted robot ownership from the model or when we excluded hospitals with unknown robotic ownership status. In a separate multivariate analysis stratified by 3-year period, teaching hospitals, urban hospitals, hospitals with large bed size and hospitals that owned a robot in 2007 were associated with greater hospital RP volume during each period (data not shown). Since the AHA data were from 2007 alone, we repeated this analysis with the intervals 2000 to 2002, 2003 to 2006 and 2007 to 2008, and the results were unchanged.
DISCUSSION From 2000 to 2008 RP use increased by 74% and the number of hospitals performing RP decreased by 19%, resulting in an increase in average annual hospital volume. The majority of volume increase occurred at high volume hospitals while there were volume losses at low volume hospitals, a scenario consistent with centralization. This finding contrasts with a prior NIS analysis demonstrating no increase in RP volume in the pre-robotic era (1988 to 2002) and no evidence of centralization.16 Large bed size, urban setting, teaching status and robot ownership were independently associated with greater RP volume, suggesting centralization to these types of hospitals. Compared to volume creep, where increased volume is seen proportionally among all hospitals, centralization occurs due to a concentration of care to a smaller number of hospitals.6 While centralization may be desirable given the association between high hospital volume and improved outcomes for several operations including RP, it may also have unintended consequences such as decreased access to care, increased travel distances, diminished volume and preparedness at low volume hospitals, and overcrowding at high volume hospitals.6,8,9,17–20 Based on our analysis, patients treated at high volume hospitals differed from those treated at low volume hospitals, already demonstrating potential disparities in access to care. Thus, while centralization of RP to high volume hospitals may improve surgical quality for some, it may also detract from quality in areas without access to specialty centers, and the
CENTRALIZATION OF RADICAL PROSTATECTOMY
503
Figure 1. Trends in use of RP in United States stratified by hospital volume quartile
potential benefits have yet to be weighed against the real-world consequences. While some payer and employer groups such as Leapfrog have suggested using volume standards as a means to control quality for certain operations, no such guidelines exist for RP.7 Thus, the reason for centralization in the absence of a policy mandate is not known. However, several theories exist including increasing hospital specialization.8,10 In a competitive market, hospitals may attempt to improve efficiency and outcomes by strategic product line management, whereby they focus on specific services while other services are discontinued.21 Accordingly United States hospital specialization increased 31% from 1998 to 2008.12 The hospital characteristics associated with increased volume in recent years were also associated with increased volume from 2000 to 2002, when RALP accounted for less than 10% of all prostatectomies.14 This suggests that centralization was occurring toward hospitals that had already specialized in the RP product line before the widespread diffusion of the robot. Nevertheless, robotic ownership was strongly associated with hospital volume and likely led to fur-
ther RP centralization to specialty centers. In the last decade there has been a shift in approach from open to robotic, and today the majority of RPs are performed robotically.14,22 In fact, based on NIS data from the final quarter of 2008, we and others estimate that more than 50% of RPs were performed with robotic assistance.23 While RP use plateaued during the 1990s,24 volume has increased sharply since 2005, paralleling the diffusion of RALP and mirroring a tipping point pattern in technology adoption.14,25 Given the sizable investment, hospitals that purchase robots may feel pressured to increase robotic use to remain profitable, often through aggressive marketing.26,27 The diffusion of RALP contrasts sharply with that of laparoscopic nephrectomy.28 While there may be several explanations, the lack of aggressive marketing and patient demand may have been responsible. As such, robot ownership has been associated with increasing RP volume and centralization to high volume centers.3,4,13,15 Stitzenberg et al evaluated RP trends in 3 states from 2000 to 2009 and identified an increase in volume with centralization to high volume hospitals.4 Hospitals that owned a
504
CENTRALIZATION OF RADICAL PROSTATECTOMY
Figure 2. Trends in average annual RP hospital volume overall (A), and by hospital volume quartile (B), teaching status (C), geographic location (D), setting (E), bed size (F) and status of robot ownership (G). Asterisk indicates statistically significant increase in RP volume (p ⱕ0.05). All geographic locations (D) were associated with significant increase in RP volume except for Northeast (p ⫽ 0.075).
CENTRALIZATION OF RADICAL PROSTATECTOMY
robot performed nearly 85% of RPs in 2009 and had higher volume even before purchasing a robot. Neuner et al demonstrated similar results despite the decreasing incidence of prostate cancer in the last decade.3 Similar to our findings, these data suggest that certain high volume specialty hospitals accounted for the majority of the volume increase, especially after purchasing a robot. While robot dissemination probably had a role in the further concentration of RP to select centers, it was unlikely the initial catalyst for centralization. Some have raised the question of whether the availability and marketing of robotics have induced demand for RALP, and inappropriately drawn patients away from other suitable management strategies.5,15 While our study cannot address the issues of induced demand or appropriateness of intervention, it does support the notion of a market driven centralization of RP to select hospitals. Our study does have limitations that must be considered. Longitudinal analysis is not possible with the NIS, and we could not measure volume changes within individual hospitals before and after purchasing a robot. In addition, we were unable to correlate volume and disease trends given the absence of pathological data in the NIS, although it is unlikely that changing disease characteristics explain the increase in RP.3,29 Our conclusion regarding the impact of robot ownership is based on a single year of cross-sectional data from a merged data set that does not indicate the robot was being used for RP, nor does it include the year the robot was purchased. Despite robotic ownership data on only 60% of NIS hospitals, robot ownership continued to be associated with hospital volume when controlling for unknown ownership status in our multivariate analysis. Unknown robot ownership
505
status was associated with RP volume, which we suspect is because this heterogeneous group contains hospitals with and without robots, and the association with volume reflects the diluted effect of the robot owning hospitals. Lastly, our findings may be biased by endogeneity given the potential for self-selection of robots to higher volume institutions. Endogeneity implies a bidirectional causative loop between the dependent (hospital volume) and independent (ownership of a robot) variable. While our data suggest that owning a robot is associated with increased hospital volume, we were not able to determine the presence or directionality of a causative relationship. Despite these limitations we describe a dramatic increase in the use of RP from 2000 to 2008, particularly among hospitals that specialized in RP early in the decade. We also identified a centralization of RP to large, urban, academic centers. While the presence of a robot was not solely responsible for the greater volume, it seemed to have a role in increasing the market share of hospitals specializing in RP. As more hospitals purchase surgical robots and more surgeons are trained in robotic surgery, future trends in the distribution of hospital RP volume should be reexamined.
CONCLUSIONS The use of RP increased significantly from 2000 to 2008 with centralization to select hospitals, particularly those in the top RP volume quartile and those investing in robotic technology. Our findings support the hypothesis that hospitals with the greatest volume increases were specialty centers with an already high RP volume. Centralization has potential benefits but may also have unintended consequences that are not addressed by market driven forces.
REFERENCES 1. Jemal A, Siegel R, Xu J et al: Cancer statistics, 2010. CA Cancer J Clin 2010; 60: 277.
5. Barbash GI and Glied SA: New technology and health care costs–the case of robot-assisted surgery. N Engl J Med 2010; 363: 701.
2. Cooperberg MR, Broering JM and Carroll PR: Time trends and local variation in primary treatment of localized prostate cancer. J Clin Oncol 2010; 28: 1117.
6. Finks JF, Osborne NH and Birkmeyer JD: Trends in hospital volume and operative mortality for high-risk surgery. N Engl J Med 2011; 364: 2128.
3. Neuner JM, See WA, Pezzin LE et al: The association of robotic surgical technology and hospital prostatectomy volumes: increasing market share through the adoption of technology. Cancer 2012; 118: 371. 4. Stitzenberg KB, Wong YN, Nielsen ME et al: Trends in radical prostatectomy: centralization, robotics, and access to urologic cancer care. Cancer 2012; 118: 54.
7. Birkmeyer JD, Finlayson EV and Birkmeyer CM: Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative. Surgery 2001; 130: 415. 8. Stitzenberg KB, Sigurdson ER, Egleston BL et al: Centralization of cancer surgery: implications for patient access to optimal care. J Clin Oncol 2009; 27: 4671. 9. Birkmeyer JD: Should we regionalize major surgery? Potential benefits and policy considerations. J Am Coll Surg 2000; 190: 341.
10. Learn PA and Bach PB: A decade of mortality reductions in major oncologic surgery: the impact of centralization and quality improvement. Med Care 2010; 48: 1041. 11. Luft HS, Hunt SS and Maerki SC: The volumeoutcome relationship: practice-makes-perfect or selective-referral patterns? Health Serv Res 1987; 22: 157. 12. Eastaugh SR: Hospital costs and specialization: benefits of limiting the number of product lines. J Health Care Finance 2009; 36: 24. 13. Lowrance W, Eastham J, Savage C et al: Contemporary open and robotic radical prostatectomy practice patterns among United States urologists. J Urol, suppl., 2011; 185: e136, abstract 337.
506
CENTRALIZATION OF RADICAL PROSTATECTOMY
14. Intuitive Surgical media kit. Available at www. intuitivesurgical.com/company/media/. Accessed May 2011. 15. Makarov DV, Yu JB, Desai RA et al: The association between diffusion of the surgical robot and radical prostatectomy rates. Med Care 2011; 49: 333. 16. Cooperberg MR, Modak S and Konety BR: Trends in regionalization of inpatient care for urological malignancies, 1988 to 2002. J Urol 2007; 178: 2103. 17. Birkmeyer JD, Siewers AE, Finlayson EV et al: Hospital volume and surgical mortality in the United States. N Engl J Med 2002; 346: 1128.
tatectomy volume and patient outcomes: a systematic review. J Urol 2008; 180: 820.
prostate-specific antigen screening: 1986 –2005. J Natl Cancer Inst 2009; 101: 1325.
20. Barocas DA, Mitchell R, Chang SS et al: Impact of surgeon and hospital volume on outcomes of radical prostatectomy. Urol Oncol 2010; 28: 243.
25. Barkun JS, Aronson JK, Feldman LS et al: Evaluation and stages of surgical innovations. Lancet 2009; 374: 1089.
21. Zelman WN and Parham DL: Strategic, operational, and marketing concerns of product-line management in health care. Health Care Manage Rev 1990; 15: 29.
26. Bolenz C, Gupta A, Hotze T et al: Cost comparison of robotic, laparoscopic, and open radical prostatectomy for prostate cancer. Eur Urol 2010; 57: 453.
22. Williams SB, Prasad SM, Weinberg AC et al: Trends in the care of radical prostatectomy in the United States from 2003 to 2006. BJU Int 2011; 108: 49.
18. Begg CB, Cramer LD, Hoskins WJ et al: Impact of hospital volume on operative mortality for major cancer surgery. JAMA 1998; 280: 1747.
23. Yu HY, Hevelone ND, Lipsitz SR et al: Hospital volume, utilization, costs and outcomes of robotassisted laparoscopic radical prostatectomy. J Urol 2012; 187: 1632.
19. Wilt TJ, Shamliyan TA, Taylor BC et al: Association between hospital and surgeon radical pros-
24. Welch HG and Albertsen PC: Prostate cancer diagnosis and treatment after the introduction of
27. Scales CD Jr, Jones PJ, Eisenstein EL et al: Local cost structures and the economics of robot assisted radical prostatectomy. J Urol 2005; 174: 2323. 28. Miller DC, Wei JT, Dunn RL et al: Trends in the diffusion of laparoscopic nephrectomy. JAMA 2006; 295: 2480. 29. Jemal A, Center MM, DeSantis C et al: Global patterns of cancer incidence and mortality rates and trends. Cancer Epidemiol Biomarkers Prev 2010; 19: 1893.