J.M.N., W.A.S., L.E.P., and A.B.N. conceived and designed the study; S.T., J.M.N., W.A.S., L.E.P., and A.B.N. analyzed and interpreted the data; J.M.N. and A.B.M. drafted the article; W.A.S., L.E.P., and S.T. critically revised the article for important intellectual content; and J.M.N., W.A.S., L.E.P., S.T., and A.B.N. provided final approval of the version to be published. J.M.N. had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Despite limited and conflicting evidence for the efficacy of newly developed robotic technology for laparoscopic prostatectomy, this technology is spreading rapidly. Because the newer technology is more costly, reasons for this rapid adoption are unclear. The authors of this report sought to determine whether hospital acquisition of robotic technology was associated with volume of prostate cancer surgery.
The inpatient dataset of claims records from 2002 to 2008 and the acquisition dates of robotic technology were used to examine the rates of prostatectomy in Wisconsin hospitals. In analyses that accounted for hospital and referral region characteristics, changes in hospital prostatectomy volume were examined for their association with technology acquisition. Overall trends in the rate of prostatectomy also were examined over the study period.
In total, 10,021 prostatectomies were performed in 52 hospitals in Wisconsin's 8 health referral regions during the study period. The mean quarterly prostatectomy volume in hospitals that did not acquire the technology was 4.5 in 2002 and 3.1 in 2007/2008. In contrast, the mean quarterly prostatectomy volume in hospitals that went on to acquire robotic technology was 16.5 in 2002 and 24.8 in 2007/2008. In adjusted models, the acquisition of a robot was associated with a 114% annual increase (95% confidence interval, 62%-177% annual increase) in hospital prostatectomy volume. The average Wisconsin hospital prostatectomy volume was unchanged during 2002 through 2006 but increased by 25.6% in 2007.
Technology is the largest single factor driving increases in health care costs.1 Some of the concerns about technology costs relate to overuse; that is, patients undergoing unnecessary procedures.2, 3 However, another concern is that some new technologies provide little or no greater health benefit than older technologies. Both physicians and hospitals have been criticized for adopting new, expensive technologies in advance of evidence that they improve health outcomes.1, 4, 5 In addition, the widespread adoption of technologic innovations before the completion of trials demonstrating their safety and effectiveness may lead to a situation in which it is difficult to conduct these trials.
We were interested in the possible incentives for “premature” technology adoption, that is, the adoption of incompletely tested technology. Specifically, we were interested in the extent to which market competition considerations may promote the adoption of unproven technologic innovations. We examined this by studying the surgical treatment of prostate cancer. Prostate cancer is the most common cancer among men in the United States, and the traditional surgical treatment for prostate cancer is open total prostatectomy. A robot that facilitates a less invasive laparoscopic option for prostatectomy was approved by the US Food and Drug Administration (FDA) and introduced to the US market in 2000.
Evidence for the effectiveness of robotic technology appears to be in its early stages. Early adopters of this robot-assisted laparoscopic technology reported some benefits in terms of reduced blood loss and shorter recovery time,6, 7 but the studies generally were based on small, selected samples. Several follow-up studies have demonstrated higher rates of adverse outcomes, including incontinence and impotence,8, 9 and the estimated costs of the surgery are 13% higher than the costs with nonrobotic techniques.10 The effect of the new technology on long-term cancer outcomes, such as mortality and recurrence, remains unknown. A recent study indicated greater use of androgen blockade treatment in the months after laparoscopic prostatectomy compared with the use of such treatment after open prostatectomy, suggesting that some physicians may be concerned with residual cancer.9
Thus, a full decade after the introduction of robotically assisted total prostatectomy, evidence for its risks and benefits is conflicting, and no randomized controlled trial has been undertaken. Despite the lack of evidence supporting its adoption, the use of robot-assisted laparoscopy for the treatment of prostate cancer has increased rapidly in the past few years.4 One possible explanation for the rapid uptake of robot-assisted technology despite inconsistent supporting evidence is that hospitals acquire and market the technology to increase their prostate cancer surgical volumes. For the current study, we tested this hypothesis using statewide Wisconsin data. We also examined whether recently reported increases in prostatectomy10, 11 also occurred in Wisconsin, hypothesizing that robotic technology is associated not only with changes in the location of care but also with an increase in the overall number of men undergoing surgical therapy for prostate cancer.
MATERIALS AND METHODS
Our primary source of information was the Wisconsin Inpatient Discharge database. The Wisconsin Inpatient Database consists of all inpatient discharges from acute care, nonfederal hospitals in the state of Wisconsin. These data are collected under a state legislative mandate and are made available to researchers by the Wisconsin Hospital Association. The database includes patient demographic data, admission and discharge data, and diagnostic and procedure data, including International Classification of Diseases, Ninth Revision, Clinical Modification (ICD9-CM) codes. The Wisconsin Hospital Association Survey also was used for information regarding the for-profit status of hospitals.
Study Population and Definitions
We focused on patients who underwent total retropubic prostatectomy between January 2002 and June 2008, years when the adoption progressed rapidly in other areas of the United States. This procedure, which is very extensive compared with partial prostatectomy (eg, transurethral resection of the prostate), is used solely for cancer surgery and is identified by ICD9-CM code 605.
We identified all general hospitals in which these patients underwent surgery as well as each hospital's health referral region (HRR), which is defined as the geographic unit that represents each hospital's market for tertiary care. HRRs were developed first to capture patterns of hospital use for major cardiac surgery,12 but they have been used subsequently to study a variety of treatments, including medical treatments of prostate cancer.13-15 Wisconsin hospitals that were located in neighboring states' HRRs were excluded. One Wisconsin hospital that closed during the study and performed only 33 prostatectomies in the 4 years before closure also was excluded from the study. Pairs of hospitals that merged during the study period were considered to be individual hospitals before the merger and a single new hospital after the consolidation.
The primary outcome of interest was quarterly surgical volume measured as the number of prostatectomies performed by each hospital in each of the 26 calendar quarters in the 6.5-year study period. Information regarding whether each hospital had purchased the robotic technology and the date of acquisition of the system was obtained from the system's (sole) manufacturer and confirmed with each hospital. To accommodate training or other procedures associated with adoption of the new technology, the correlation between robot acquisition and the annual prostatectomy rate was assessed assuming that actual use of the new technology lagged by 3 months from the manufacturer's reported date of purchase. Additional sensitivity analyses assumed that actual robotic technology use started at 6 months after robot acquisition.
For data regarding prostate cancer incidence, Wisconsin state tumor registry estimates were used.16 The Institutional Review Board of the Medical College of Wisconsin approved the study design.
The association between hospitals' robotic acquisition and prostatectomy surgical volume was examined using generalized estimating equation (GEE) Poisson models. The Poisson count data specification was chosen in view of the relatively low quarterly surgical volume of certain hospitals during the study period. The GEE method enabled us to account for within-hospital correlation in prostatectomy surgical volumes over time. In addition to controlling for hospital-invariant time trends on prostatectomy rates, which were measured by a set of binary indicators of calendar quarter, our multivariate models also examined variables that may confound the correlation between adoption of robotic technology and prostatectomy volume. These included mean age of prostatectomy patients, proportion of the hospital's prostatectomy surgical patients covered by public programs (Medicare and Medicaid, separately), and indicators for the hospitals' referral region (to account for competition or other differences between hospitals' prostatectomy markets). Because only 1 Wisconsin hospital that performed prostatectomy was a for-profit institution during the study years, this variable was not included in the models. One of the authors (S.T.) performed all analyses.
In total, 52 hospitals performed 10,021 prostatectomies in the state of Wisconsin and its HRRs between January 1, 2002 and June 30, 2008. Of those, 12 hospitals (23%) acquired the robot technology by December 31, 2007. Hospitals that acquired robot technology were distributed in 6 of the 8 Wisconsin HRRs.
Table 1 lists the 2002 “baseline” values of surgical volume and other hospital characteristics for the 12 hospitals that were classified as “adopters” compared with “nonadopters” (ie, hospitals that did not acquire the robot system over the study period). Although there were no statistically significant differences between adopter and nonadopter hospitals with respect to the age and payer composition of their prostatectomy patients, adopter hospitals had slightly younger patients (P = .005) and a higher volume of prostatectomies (P < .001) before robot acquisition.
Table 1. Hospital Characteristics by Adopter Status of Hospitals
Abbreviations: HRR, health referral region; SD, standard deviation.
Adopter hospitals are those that acquired a robotic laparoscopic system between January 2001 and December 2007.
Quarterly prostatectomy volume in 2002: Mean ± SD
16.52 ± 8.98
4.54 ± 4.34
Age of patients undergoing prostatectomy: Mean ± SD, y
60.26 ± 6.80
61.30 ± 7.10
Median no. of hospitals in HRR
Association of Robotic Technology and Prostatectomy Surgical Volume
The trajectory of prostatectomy volume in robotic adopter and nonadopter hospitals is illustrated in Figures 1 and 2. In 2002, a total of 1404 prostatectomies were performed in Wisconsin hospitals (Fig. 1). Between the first and final 12-month periods of the study (January 1, 2002 to December 30, 2002 and July 1, 2007 to June 30, 2008, respectively), the mean quarterly (3-month) prostatectomy volume among adopter hospitals rose from 16.5 to 24.8. The mean prostatectomy surgical volume among nonadopter hospitals decreased during the same period from 4.5 to 3.1. Consistent with these findings, centralization of prostatectomies by adopter hospitals, as measured by the proportion of all prostatectomies performed by hospitals that eventually acquired the robot system, rose from 48.1% in the first quarter of 2002 to 72% in the second quarter of 2008.
There was substantial variability in volume between hospitals. For example, in 2007, the adopter hospitals' median quarterly prostatectomy volume was 22 with an interquartile range (IQR) of 14 to 34. Nonadopter hospitals had a 2007 median quarterly prostatectomy volume of 3 (IQR, 0-4; minimum 0, maximum 18).
In multivariate models that were adjusted for characteristics of the hospital and HRRs, acquisition of a robot was associated with more than a doubling of a hospital's prostatectomy volume. Specifically, a 114% increase (95% confidence interval [CI], 62%-177% increase) in the quarterly rate of prostatectomy occurred after 3 months (Fig. 2). Results of sensitivity analyses in which the lag between robot purchase date and its use was changed to 6 months revealed a similar increase in prostatectomy rates (112% increase; 95% DI, 60%-179% increase).
Overall Prostatectomy Use
To investigate the hypothesis that the acquisition of surgical robots increased the overall use of prostatectomy, we examined overall use of prostatectomy in the context of Wisconsin prostate cancer incidence. There was no change in the total number of prostatectomies performed in our cohort hospitals between January 2002 and 2006 (Fig. 1). However, in 2007, the number of prostatectomies at cohort hospitals increased by 25.6% compared with 2002 (adjusted P = .07). This increase was sustained through June 2008. The incidence of prostate cancer decreased in Wisconsin during the study period from an estimated 4238 men in 2002 (age-adjusted incidence of 169 in 100,000 men) to 3900 men in 2006 (142 in 100,000 men).16 Therefore, we estimate that 33.1% of Wisconsin men with prostate cancer underwent surgery (prostatectomy) in 2002, and 38.1% underwent surgery in 2006.
During 2002 through 2008, 23% of Wisconsin hospitals acquired robotic technology that was used for laparoscopic prostatectomies; and, by 2007, more men underwent prostatectomies at hospitals with robotic technology than at hospitals without it. The acquisition of robotic technology was associated with a 114% increase in hospital prostatectomy volume within 3 months. In contrast, the average prostatectomy volume of hospitals that did not acquire the technology fell. Hospitals' insurance patterns and the regions in which they were located did not explain these results. The average Wisconsin total prostatectomy volume rose >20% between 2002 and 2008 despite decreasing prostate cancer incidence, which raises the question of whether the care of men with prostate cancer also was shifted toward surgery and away from alternative treatments, such as radiotherapy or watchful waiting.
The hospitals that acquired robotic technology during the study time frame arguably adopted the technology prematurely, because the evidence regarding its effectiveness still is incomplete. The first studies of robotic prostatectomy, mostly from early centers for the technology, reported reductions in blood loss and length of hospital stay compared with open prostatectomy.6 A recent study from 1 large center demonstrated reassuringly similar 3-year rates of biochemical recurrence-free survival after robotic prostatectomy compared with open procedures.17 Most population-based studies, however, have been less encouraging: One study noted more postoperative urethral strictures with laparoscopic techniques (either robot-supported or not) than with open techniques,9 and another noted more incontinence and impotence.8 Studies of long-term outcomes, including cancer recurrence and mortality, can take a decade or more and are not yet available.
There is also some evidence that the favorable effect of experience reported for open prostatectomy18, 19 also may occur with robotic prostatectomy20; if so, then the centralization of prostatectomy that occurred in our study may be beneficial and even cost-effective.21 However, of concern is the large number of surgeries that would be required for optimal outcomes, a number probably higher than that achieved by most hospitals19 and surgeons10, 20 in our study. Our study was limited by our inability to examine individual physician surgical volumes. However, physicians themselves appear to have concerns about their outcomes with the newer procedures. One population-based study8 indicated increased postoperative use of hormone therapies for patients who underwent laparoscopic prostatectomies.
We hypothesize that the rapid adoption of robotic technology by multiple Wisconsin hospitals is caused at least in part by awareness that technology increases hospitals' appeal to patients. Manufacturers increasingly market devices directly to patients.22 Patients request pharmaceuticals from health care providers, and it is possible that they also have begun to respond to this marketing by directly requesting specific devices. Physicians do advocate for technology to both hospital administrators and local insurers,23 reporting that they associate technology with quality23, 24 and that technology increases their sense of community leadership.23 Our study adds to this literature through its finding that the effects of a specific technology on hospital surgical volume are rapid and large enough to be perceived by hospital decision-makers. Even the first adopter hospitals in a community had doublings in volume after robotic acquisition that other hospitals' administrators are likely to have noticed.
The magnitude of the effects of new technology acquisition observed in the current study suggests that policy interventions to support proven treatments would have to be strong to be effective. One possible strategy to promote the use of proven technologies is stronger ties between proof of effectiveness and reimbursement. Although effective in countries with a centralized health care system, this approach may be more challenging in the United States. For example, 2 recent Medicare panels, the first for computed tomographic (CT) coronary angiography and the second for CT colonography, reviewed comparative effectiveness data and initially advised that Medicare decline coverage for the procedures. There was much controversy surrounding both decisions, however, and the CT angiography decision eventually was reversed.25 The factors behind such controversies have not been well described, but physician interests and perhaps patient interests in technology, the many stakeholders involved once a product has been approved, or the great psychological value placed on “losing” something once gained26 may hinder insurers from restraining technology diffusion.
Another potential and perhaps stronger intervention would be strengthened FDA regulation of devices. In contrast to pharmaceuticals, for which initial randomized trials are required, Congress currently mandates that the FDA use the “least burdensome” of 3 possible evaluation categories for devices.27 Robotic technology for prostatectomies was approved under the least restrictive of these categories, 1 that allows manufacturers to state that a product is nearly identical to previously approved devices. It is possible that a requirement for the use of more stringent categories may improve care. However, even for the most stringent FDA category (premarket reviews), evidence of efficacy may be minimal. A recent systematic review of cardiovascular device approvals indicated that only 27% referenced a randomized study, and only 52% referenced a study with any controls.28, 29 Thus, further FDA changes also may be required.
In 2010, there was still no evidence from observational studies that robotic technology is clearly beneficial for prostatectomy; and, to date, no randomized trial is in progress. Given recent studies suggesting that survival in most (eg, Gleason ≤7 disease), but not all (eg, Gleason 8-10 disease) patients with localized disease managed conservatively is now similar to that of age-matched control individuals,30, 31 trials of any prostate cancer treatments are particularly important. Recruitment for a randomized trial of robotic technology may be difficult, however, given the rapid spread of the technology. Our study indicated that the majority of all prostate cancer surgeries in Wisconsin were performed in hospitals with robotic technology in 2008. Consistent with our findings, 43% of prostatectomies in Surveillance, Epidemiology, and End Results cancer registries were performed laparoscopically in 2006 and 2007, a large increase from the 9.2% of surgeries in 2003.8 The manufacturer reported in 2009 that 86% of all prostatectomies were performed using their technology.4
Our study's findings also are notable for the increase in Wisconsin hospitals' prostatectomy use in 2007 and 2008, suggesting that new technology may create greater demand for procedures. The short time frame of our study is a limitation; however, the finding that robot-assisted prostatectomies are a recent development in the state only adds to the timeliness of our analysis. Our findings regarding possible increases in prostatectomy use are consistent with a study that noted a substantial increase among a national sample of inpatients,10 and we add to that study's findings by demonstrating that the increase is limited to hospitals acquiring robotic technology. Furthermore, it has been demonstrated that other technology also changes surgical rates. For example, after the introduction of laparoscopic cholecystectomy in the 1990s, the overall cholecystectomy rate increased by 28%.2 Wisconsin cancer statistics indicate that prostate cancer rates had been decreasing in the state—as in much of the United States—during the 5 years leading up to 2007,16 suggesting that increasing cancer incidence is unlikely to explain the findings.
Because there was no specific hospital procedure (ICD9) code for robotic technology use at the time of our study, we were unable to differentiate between surgeries performed with versus without robotic technology at hospitals that had acquired the technology. It is possible that some hospitals that acquired the technology were not using it for all patients. However, this would not change our finding that hospitals that did invest in the technology had increased surgical volumes and the policy implications of that finding. In addition, like another recent analysis,32 we were not able to fully account for the possibility that hospitals that adopted the robotic technology may have differed by some characteristics in our analysis. However, our analysis did account for several characteristics that plausibly could affect robotic acquisition decisions (competition represented by HRR, case mix of insurance type, and patient age). We also adjusted for within-hospital correlation, in which we allowed each hospital to act as its own “control.” Future studies focused on clinical outcomes of prostate cancer at the patient level could use techniques like instrumental variable analyses, and also should examine other technologies, such as intensity-modulated radiation therapy, which is estimated to increase costs yearly by more than $282 million compared with older radiation techniques.33
In conclusion, we observed that robotic technology acquisition occurred rapidly in Wisconsin hospitals during a 6-year period and was associated with a substantial increase in hospitals' prostatectomy volumes. The rapid acquisition and deployment of expensive technology of unproven value has the potential for far-reaching implications for patients and health care delivery systems. Increases in cost per episode of care, together with overuse (redistribution of patients previously treated with other modalities) may be associated with significant cost increases.10 Although the centralization of care delivery may benefit patients with access to early adopter hospitals, it risks of harm to patients who are treated at nonadopter sites whose surgical volumes fall below those required for optimal care delivery. In circumstances like robotic prostatectomy in which it may no longer be possible to demonstrate benefit through prospective, randomized, controlled trials, rigorous comparative-effectiveness studies may be required.34 A recently funded study may be able to compare different surgical and radiotherapy options for prostate cancer.35 Given our findings of the widespread use of new technologies, such studies should be done early, should be drawn from wide populations, and should include the measurement of centralization of care and other mutable factors that affect outcomes.
This research was conducted using internal funding from the Medical College of Wisconsin.