An economic analysis of robotically assisted hysterectomy.

A total of 180,230 women were identified. The cohort included 169,324 women who underwent minimally invasive hysterectomy for benign indications and 10,906 patients whose hysterectomy was performed for endometrial cancer. Robotically assisted hysterectomy accounted for 30.4% of the hysterectomies for benign disorders and 59.6% of hysterectomies for endometrial cancer. The clinical and demographic characteristics of the cohort are displayed in Table 1 . The unadjusted median cost of robotically assisted hysterectomy for benign indications was $8,152 (interquartile range [IQR] $6,011–10,932) compared with $6,535 (IQR $5,127–8,357) for laparoscopic hysterectomy ( Table 2 ). Median fixed costs were $3,591 (IQR $2,347–5,345) for robotic-assisted hysterectomy compared with $2,965 (IQR $2,137–4,051) for laparoscopic hysterectomy, whereas variable costs were $4,384 (IQR $3,170–5,883) and $3,440 (IQR $2,633–4,513) for the two procedures, respectively. The cost of both robotic-assisted and laparoscopic hysterectomy for benign indications decreased with increasing procedural volume. Figure 1 displays the median total cost of robotically assisted and laparoscopic hysterectomy for benign indications based on previous surgeon ( Fig. 1A ) and hospital ( Fig. 1B ) volume. Table 2 displays the unadjusted median costs for hysterectomy for benign by procedural volume. At hospitals that had performed fewer than five previous cases, robotically assisted hysterectomy was $2,053 more expensive. The cost differential decreased with increasing hospital volume as follows: 5 to 15 cases (+$2,007), 16 to 30 cases (+$1,729), 31 to 50 cases (+$1,657), and more than 50 cases (+$1,560). Among surgeons, robotically assisted hysterectomy was $1,705 more expensive than laparoscopic hysterectomy for those who had performed fewer than five cases, whereas costs declined to +$1,467 for 5 to 15 cases, +$1,559 for 16 to 30 cases, +$1,390 for 31 to 50 cases, and +$1,619 for more than 50 procedures. The unadjusted median cost of robotic-assisted hysterectomy for endometrial cancer was $9,691 (IQR $7,591–12,428) compared with $8,237 (IQR $6,400–10,807) for laparoscopic hysterectomy ( Table 3 ). Median fixed costs were $4,543 (IQR $3,201–6,164) for robotic-assisted hysterectomy for endometrial cancer compared with $3,790 (IQR $2,832–5,406) for laparoscopic hysterectomy, whereas variable costs were $5,065 (IQR $3,994–6,536) and $4,215 ($3,098–5,644) for the two procedures, respectively. The cost of both robotically assisted and laparoscopic hysterectomy for endometrial cancer decreased with increasing procedural volume. Figure 2 displays the median total cost of robotically assisted and laparoscopic hysterectomy for endometrial cancer based on previous surgeon ( Figure 2A ) and hospital ( Figure 2B ) volume. At hospitals that had performed fewer than five previous cases for endometrial cancer, robotically assisted hysterectomy was $2,471 more expensive than laparoscopic hysterectomy. The cost differential decreased with increasing hospital volume: 5 to 15 cases (+1,984), 16 to 30 cases (+$1,753), 31 to 50 cases (+$1,403), and more than 50 cases (+$924). Among surgeons, robotically assisted hysterectomy for endometrial cancer was $1,761 more expensive for those who had performed fewer than five cases, whereas costs declined to +$1,266 for 5 to 15 cases, +$1,565 for 16 to 30 cases, +$1,451 for 31 to 50 cases, and +$688 for more than 50 procedures. In a series of adjusted models, these results were largely unchanged ( Table 4 ). In a fully adjusted model, compared with laparoscopic hysterectomy, robotically assisted hysterectomy for benign indications was $1,225 (95% CI $1,177–1,272) more expensive and $1,328 (95% CI $1,164–1,491) more costly for endometrial cancer. When the analysis was limited to hospitals that used direct costing methods, the cost differential associated with robotic procedures was greater; robotically assisted benign hysterectomy was $1,995 (95% CI $1,948–2,041) more than laparoscopic hysterectomy. Removing patients who experienced complications had minimal effects on the estimates. Like the unadjusted models, cost declined with volume. Results were similar for fixed and variable costs and after log transformation of the data (not shown).

Our data suggests that the cost of robotic gynecologic surgery decreases with increasing procedural volume. The reduction in costs was most pronounced for women undergoing robotically assisted hysterectomy for endometrial cancer and was more modest when the surgery was performed for benign indications. Despite the reductions in cost associated with increased procedural volume, in all of the scenarios modeled, robotic-assisted hysterectomy remained substantially more costly than laparoscopic hysterectomy. The introduction of new surgical technologies is often associated with a learning curve. 13 , 15 – 17 An institutional analysis of 325 patients who underwent robotic-assisted hysterectomy noted that although increasing surgical experience was associated with shorter operative times and decreased length of stay, surgical experience had no effect on complications. The investigators noted that operative time decreased from 3.5 hours during the first 6 months of robot use to 2.7 hours per procedure. 13 We noted that cost decreased with the number of previous procedures performed for both hospitals and surgeons. The association between surgical volume and outcomes has been well-described over the course of the past two decades. 28 , 29 For high-risk oncologic and cardiovascular procedures, increased surgeon and hospital procedural volume are associated with decreased morbidity and mortality. 28 , 29 For lower-morbidity procedures, including most gynecologic operations, the association between increased volume and decreased complications is more modest. 30 – 33 However, for many gynecologic procedures, increased procedural volume is associated with lower resource use and decreased costs. 30 – 33 The current study suggests similar trends for robotically assisted hysterectomy. The relationship between increased volume and decreased cost appeared to be greater when robotically assisted hysterectomy was performed for endometrial cancer than when the operation was used for benign indications. Further, for hysterectomy for benign indications, the relative magnitudes of cost reductions for robotically assisted and laparoscopic hysterectomy associated with increasing surgeon volume were similar. We noted that surgeon volume had a more meaningful influence on cost than hospital volume. Given that hysterectomy is performed much more frequently for benign indications than for cancer, the limited reduction in the cost differential for benign indications with higher volume has important public health implications. 34 Our study addresses a number of methodologic concerns in the analysis of the cost. First, both hospital volume and surgeon procedural volume were measured as the number of operations performed before the index case and as continuous variables. 24 This approach was meant to account for the learning curve for the operation. Second, separate analyses were performed for hospitals that used direct internal accounting systems compared with those that estimate cost based on cost-to-charge ratios. When analyzing only those hospitals that directly report actual costs, we noted that the cost differential of robotic-assisted hysterectomy was greater than that in the analysis of all hospitals. We used a number of statistical methodologies to account for right skewed data such as cost. 19 , 20 We recognize a number of important limitations. We cannot exclude the possibility that some procedures were misclassified. However, the classification system we used to identify robotic procedures has been used in previous work and was previously validated. 2 Our classification of physician volume only captured a surgeon’s patients within a given hospital, and we were unable to link physicians across hospitals. There is likely variation across hospitals in fixed costs for the robotic platform and instrumentation based on different negotiated prices. We were unable to capture some factors, including tumor characteristics, weight, and surgical history, that undoubtedly influenced outcomes. Finally, our data present cost from a hospital perspective. Previous work has suggested that increased use of robotic-assisted hysterectomy is associated with a decreased rate of abdominal hysterectomy that likely provides important cost reductions from a societal standpoint. Our study suggests that costs are reduced with both increased surgeon experience and hospital experience, although the reduction in cost is affected to a greater degree by surgeon rather than hospital volume. Although the cost reduction is multifactorial, the cost savings are likely from a combination of shorter operative times and reduced length of stay. Our findings are also notable in that no matter how the two procedures were modeled, in similar circumstances laparoscopic hysterectomy always remained less costly than robotically assisted hysterectomy. Even for very high-volume surgeons and centers, robotically assisted hysterectomy remained more costly. Based on these data, it appears unlikely that robotic-assisted hysterectomy can achieve cost parity with laparoscopic hysterectomy based on surgical experience alone and that reductions in the cost of robotic instrumentation will be required for the procedure to become cost-effective. 2 , 35 Strategies to reduce the cost of robotic instrumentation as well as initiatives to promote access to high-quality laparoscopic surgery are warranted.

The Perspective database was used for analysis. Perspective captures comprehensive billing data of all hospital admissions from more than 500 acute care facilities from throughout the United States. The database collected data for nearly 5.5 million discharges in 2006, which represents approximately 15% of hospitalizations in the United States. 21 The study was deemed exempt by the Columbia University Institutional Review Board. Women 18 to 90 years of age who underwent a minimally invasive hysterectomy from 2006 to 2012 were analyzed. We initially selected patients who had a code for a laparoscopic hysterectomy (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] codes 68.31, 68.41, 68.51). Those women who had either an ICD-9-CM procedure code for a robotic-assisted procedure (ICD-9-CM 17.42 or 17.44) or a recorded charge code for robotic instrumentation were classified as having undergone a robotically assisted hysterectomy as previously described. 2 , 22 Women with a gynecologic malignancy other than endometrial cancer were excluded. The cohort was then stratified into the following two groups: those with endometrial cancer (ICD-9-CM 182.x) and those without gynecologic cancer who underwent hysterectomy for benign indications. Clinical and demographic characteristics including age at the time of the procedure (younger than 50, 50–59, 60–69, and older than 70 years), race (white, black, other), marital status (married, single, unknown), year of diagnosis (2006 to 2012), and insurance status (commercial, Medicare, Medicaid, uninsured, and unknown) were recorded. For women who underwent hysterectomy for benign indications, we noted the following gynecologic conditions: leiomyomas; endometriosis; abnormal bleeding; benign ovarian neoplasms; and pelvic organ prolapse. The performance of concomitant gynecologic procedures, including anterior colporrhaphy, posterior colporrhaphy, salpingo-oophorectomy, incontinence surgery, and lymphadenectomy, were also noted. Hospital characteristics including location (metropolitan and nonmetropolitan), region of the country (northeast, midwest, west, and south), size (fewer than 400 beds, 400–600 beds, and more than 600 beds), and teaching status (teaching and nonteaching) were recorded for each patient. Risk adjustment for comorbid medical conditions was performed using the Elixhauser comorbidity index. Women were classified based on the number of medical comorbidities as 0, 1, or 2 or more, as previously reported. 23 Physician and hospital volume were determined for each patient. Both hospital and surgeon volume were calculated individually for each patient and estimated as the number of procedures performed at a given patient’s hospital or by a given patient’s surgeon before the index procedure. Separate volume-based calculations were performed for robotic-assisted and laparoscopic hysterectomy. Volume was calculated separately for procedures for endometrial cancer and benign indications. Volume was included as a continuous variable in all of the multivariable models. 24 To determine the effect of complications on cost, we examined perioperative morbidity. The following perioperative complications were analyzed: intraoperative complications (bladder injury, ureteral injury, intestinal injury, vascular injury, and other operative injury); surgical site complications (wound complications, abscess, hemorrhage, bowel obstruction, ileus); and medical complications (venous thromboembolism, myocardial infarction, cardiopulmonary arrest, acute renal failure, respiratory failure, stroke, bacteremia or sepsis, shock, and pneumonia). 2 , 9 Any morbidity, a composite score of any of these complications, was analyzed. The primary outcome of the analysis was cost. Cost represents the monetary value to perform a service, whereas charges are based on what a hospital bills for the service. We directly analyzed cost. Perspective captures cost data through an itemized log of all items and services billed to a patient during the acute hospitalization. Cost data represent the cost of the entire index hospitalization. Within the database, hospitals report cost either through direct internal accounting systems or through Medicare cost-to-charge ratios. 21 , 25 The type of accounting performed by each hospital was recorded and separate sensitivity analyses based on the type of accounting system were performed as described. All costs were adjusted for inflation using the Consumer Price Index and reported in 2012 U.S. dollars. 26 Converted cost data were inspected and patients with spurious costs (less than $500) were removed from the cost analyses. 2 We examined total costs and performed separate analyses for fixed and variable costs. Fixed costs are those costs attributable to capital equipment and maintenance, whereas variable costs are attributable to the operation of the hospital irrespective of fixed costs. 27 Women who underwent hysterectomy for benign disease and those who underwent hysterectomy for endometrial cancer were analyzed separately. Frequency distributions between categorical variables for those who underwent laparoscopic hysterectomy and those who underwent robotically assisted hysterectomy were compared using χ 2 tests, and median values of continuous variables were compared using Wilcoxon rank-sum tests. Cost data for each of the groups are reported as medians with interquartile ranges. Distributions of cost based on previous procedural volume are displayed graphically with previous surgical volume broken down into deciles. Multivariable adjustments of cost were performed using quantile (median) regression methodology. 19 Quantile regression directly estimates the adjusted median costs and 95% confidence intervals (CIs) were derived based on bootstrap resampling methods. A series of cost models were developed. We first developed an unadjusted model based only on the route of hysterectomy. A similar model was constructed that only included hospitals that reported cost based on direct internal accounting systems. A fully adjusted model that reports cost after adjustment for all of the clinical, demographic, physician, and hospital characteristics was then described. Similarly, a fully adjusted model excluding patients with any perioperative complication was shown. Finally, a series of stratified models were developed that included only patients based on the volume of the attending surgeon (fewer than 5 procedures, 5–15 procedures, 16–30 procedures, and more than 50 procedures). A series of sensitivity analyses of the adjusted and unadjusted cost models were performed after log-transformation of the data. All analyses were performed with SAS 9.2. All statistical tests were two-sided. P <.05 was considered statistically significant.