Journal Article Summary and Critical Evaluation

0.JornalArticle.pdf

Home >Information Systems homework help >Journal Article Summary and Critical Evaluation

Hospital volume and outcomes of robot-assisted partial nephrectomy Leilei Xia* , Jose E. Pulido*, Raju R. Chelluri*, Marshall C. Strother*, Benjamin L. Taylor† , Jay D. Raman‡ and Thomas J. Guzzo* *Division of Urology, Department of Surgery, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA, †Department of Urology, Weill Cornell Medical College, New York, NY, USA, and ‡Division of Urology, Department of Surgery, College of Medicine, Pennsylvania State University, Hershey, PA, USA

Objective To evaluate the impact of hospital volume on outcomes of robot-assisted partial nephrectomy (RAPN).

Materials and Methods Patients with renal cell carcinoma who underwent RAPN between 2010 and 2013 were identified in the National Cancer Database. Hospital yearly RAPN volume was categorized into groups by sorting patients as closely as possible into five groups of equal size (quintiles): very low; low; medium; high; and very high volume. Outcomes included 30-day mortality, 90-day mortality, open conversion, prolonged length of hospital stay (PLOS; defined as >3 days), 30-day readmission rate, and positive surgical margin (PSM) rate. Unadjusted analyses and multivariable logistic regressions were used to compare outcomes. Sensitivity analyses with hospital volume considered as a continuous variable were also performed.

Results A total of 18 724 RAPN cases were included. Hospital volume quintiles were: very low volume, 1–7 cases (n = 3 693); low volume, 8–14 cases (n = 3 719); medium volume, 15–23 cases (n = 3 833); high volume, 24–43 cases (n = 3 649); and very high volume, ≥44 cases (n = 3 830). There was no significant difference in 30-day or 90-day mortality among the five groups. Multivariable logistic regression

analysis (reference: very low volume) showed that higher hospital volume was associated with lower odds of conversion (low [odds ratio {OR}: 0.88; P = 0.377]; medium [OR: 0.60; P = 0.001]; high [OR: 0.57; P < 0.001]; very high [OR: 0.47; P < 0.001]), lower odds of PLOS (low [OR: 0.93; P = 0.197], medium [OR: 0.75; P < 0.001]; high [OR: 0.62; P < 0.001]; very high [OR: 0.45; P < 0.001]), and lower odds of PSMs (low [OR: 0.76; P < 0.001]; medium [OR: 0.76, P < 0.001]; high [OR: 0.59; P < 0.001]; very high [OR: 0.34; P < 0.001]). Sensitivity analyses confirmed increasing hospital volume (per 1-case increase) was associated with lower odds of conversion (OR: 0.986; P < 0.001), PLOS (OR: 0.989; P < 0.001) and PSMs (OR: 0.984; P < 0.001). A difference in 30-day readmission rate was found in unadjusted analysis but not in adjusted analyses.

Conclusion Undergoing RAPN at higher-volume hospitals may have better peri-operative outcomes (conversion to open and LOS) and lower PSM rates. Future studies are needed to explore the detailed components that lead to the superior outcomes in higher-volume hospitals.

Keywords partial nephrectomy, robotics, hospital volume, outcomes, positive surgical margin, length of stay

Introduction Urology has always been at the forefront of embracing new technologies, and one clear example is robotic surgery. Robot- assisted radical prostatectomy (RARP) has already become the predominant approach in the surgical management of prostate cancer in the USA [1]. Continuing this trend, the utilization of robot-assisted partial nephrectomy (RAPN) for renal mass and RCC has also grown rapidly in recent years. One study showed that ~64% of partial nephrectomies (PNs) in the USA were performed robotically in 2013 [2].

It is well established that PN is the recommended surgical management for localized T1a renal masses and RCCs [3,4]. When technically feasible, PN is also favoured over radical nephrectomy (RN) in patients with T1b tumours [3]. The robotic platform may directly contribute to the increasing usage of PN for the management of renal masses [5,6]. Recent meta-analyses have shown that RAPN provides better peri-operative outcomes than laparoscopic and open PN [7,8]. Aside from the medical factors, marketing has also contributed to the increased utilization of robotic surgery [9]. Taking these factors together, it is foreseeable that RAPN

Robotics and Laproscopy

http://orcid.org/0000-0003-2954-7256

http://orcid.org/0000-0003-3635-003X

http://orcid.org/0000-0001-8480-8397

usage will continue to increase and may become the standard option for the management of small renal masses.

With more surgeons being trained to perform robotic surgeries and more hospitals being equipped with robotic platforms, it is possible that robotic surgeries will disseminate more into the community setting. A recent study has already shown this decentralization trend of RARP in the USA [10]. Although it is unclear whether this decentralization could also happen in RAPN, there is a need to evaluate the volume– outcome associations in RAPN, considering its rapidly increasing usage. In addition, the minimally invasive nature of the RAPN procedure and significant advantages of the robotic system may skew previously reported volume– outcome associations in nephrectomy [11,12]. To address this knowledge gap, we analysed a contemporary cohort from a national database to determine the impact of hospital volume on RAPN outcomes.

Materials and Methods Data Source

This study was a retrospective cohort analysis of the National Cancer Database (NCDB), which is a joint project of the Commission on Cancer of the American College of Surgeons and American Cancer Society. The NCDB is a prospectively maintained nationwide and hospital-based database that captures 70% of all newly diagnosed malignancies in the USA each year. The NCDB contains comprehensive data on reporting facilities, patient demographics, cancer identification, staging, treatments and outcomes [13]. Data in the NCDB were de-identified and institutional review board approval was not required.

Hospital Volume and Cohort Selection

All patients with primary renal parenchymal malignancy who were treated with RAPN at the reporting facility between 2010 and 2013 were firstly identified (n = 20 975). Renal parenchymal malignancy was identified by International Classification of Diseases for Oncology, third edition (ICD�O�3) site code C649. PN was determined by surgery of primary site (at this facility) code 30 (partial or subtotal nephrectomy) and robotic approach were determined by the surgical approach codes 1 (robot-assisted) and 2 (robot- assissted converted to open). The surgical approach was firstly available in the NCDB in the year 2010. This complete RAPN cohort was used to calculate hospital volume in each respective year based on the unique facility identifier [13]. We determined hospital volume on an annual basis because the hospitals included in the NCDB vary every year. Hospital volume (cases/year) was assigned to each patient/case as a continuous variable. For analysis and presentation purposes, we created and defined five categories of hospital volume by

most closely sorting patients into five groups of equal size (quintiles): very low; low; medium; high; and very high [11].

We then applied our inclusion and exclusion criteria to create a final study cohort for the outcomes analysis. We included only patients with tumour size ≤7 cm (T1) and patients with a final diagnosis of RCC using the ICD�O�3 histology codes 8000–8980 [14]. Patients with metastatic involvement of distant site(s) at the time of diagnosis were excluded. Patients with missing data on any of the outcomes of interest were also excluded.

Outcomes and Co-variables

The outcomes of interest were determined based on clinical relevance and data availability in the NCDB. Outcomes of this study included 30-day mortality, 90-day mortality, open conversion, prolonged length of hospital stay (PLOS), 30-day (unplanned) readmission, and positive surgical margin (PSM) rates. PLOS was defined as length of stay greater than the 75th percentile (3 days) of length of stay in the final included cohort [15].

Patient covariables included age, sex, race, Charlson–Deyo (comorbidity) score, insurance status, income level (median household income in each patient’s area of residence), educational level (number of adults who did not graduate from high school in the patient’s area of residence), and residence location (metropolitan vs urban vs rural). Tumour covariables included Fuhrman grade, histology (clear-cell vs other) and tumour size. Fuhrman grade was determined using the Collaborative Stage Data Collection System Site-Specific Factor 6. Clear-cell RCC was determined by ICD�O�3 histology codes 8000, 8005, 8310, 8312–8316, and 8359 [14]. Year of (RCC) diagnosis was also considered as a covariable. Because year of surgery (RAPN) is not included in the NCDB, we used the year of diagnosis as an alternative for reporting and analysis in our study.

Statistical Analysis

Patient characteristics, tumour characteristics and unadjusted outcomes were reported and compared using chi-squared and Kruskal–Wallis tests for categorical variables and continuous variables, respectively. Multivariable logistic regressions were used to assess the associations between hospital volume and outcomes. We then performed sensitivity analyses, with hospital volume (per 1-case increase) considered as a continuous variable for the logistic regression models. To account for the potential confounding from the increased case volume by the year of diagnosis, we repeated the sensitivity analyses in a subcohort of patients from the most recent diagnosis year (2013). P values <0.05 were taken to indicate statistical significance. All analyses were performed using STATA 14 (StataCorp LP, College Station, TX, USA).

Hospital volume and robotic partial nephrectomy

Results A total of 18 724 patients/RAPNs were included in the final cohort. From 2010 to 2013 there was a yearly increase in the number of RAPNs performed (Fig. 1A). Hospital volume categories were determined according to the following criteria: very low volume, 1–7 cases (n = 3 693); low volume, 8–14 cases (n = 3 719); medium volume, 15–23 cases (n = 3 833); high volume, 24–43 cases (n = 3 649); and very high volume, ≥44 cases (3 830). The median (interquartile range) hospital volume increased from 15 (7–26) cases in 2010 to 21 (11–42) cases in 2013. Hospital volume distribution by year of diagnosis is shown in Fig. 1B. The percentage of patients who underwent RAPN in the very-low-volume (1–7 cases) hospitals decreased from 27.2% in 2010 to 15.5% in 2013. The percentage of patients who underwent RAPN in the very-high-volume (≥44 cases) hospitals increased from 14.9% in 2010 to 24.6% in 2013. A total of 699 hospitals were included in the final cohort: 418 in 2010, 492 in 2011, 558 in 2012 and 610 in 2013.

Baseline characteristics of the cohort are summarized in Table 1. The unadjusted outcome comparison across the strata of hospital volume is shown in Fig. 2. Overall 30-day mortality (n = 48, 0.26%) and 90-day mortality (n = 76, 0.41%) were low in the final cohort. There were no significant differences in 30- day mortality (P = 0.151) or 90-day mortality (P = 0.347) among the five hospital volume groups. Significant differences were found in conversion (P < 0.001), PLOS (P < 0.001), 30- day readmission (P = 0.024) and PSM rates (P < 0.001). The trend of association between increasing hospital volume categories and improved outcomes were seen in conversion, PLOS and PSM but not 30-day readmission rates.

The multivariable logistic regression results for conversion, PLOS, 30-day readmission and PSM rates are shown Tables

S1–S4. Adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) of hospital volume strata are shown in Fig. 3. Compared with the very-low-volume group, patients in the higher-volume groups (medium to very high for conversion and PLOS; low to very high for PSM) had lower odds of conversion, PLOS and PSM. No association between hospital volume (stratum) and 30-day readmission was found in the logistic regression model (P > 0.05 in all hospital volume groups).

Sensitivity analyses (multivariable logistic regressions) showed that increasing hospital volume (per 1-case increase) was associated with lower odds of conversion (OR 0.986, 95% CI 0.981–0.991; P < 0.001), PLOS (OR 0.989, 95% CI 0.987– 0.990; P < 0.001) and PSM (OR 0.984, 95% CI 0.981–0.987; P < 0.001). No association between hospital volume (per 1- case increase) and 30-day readmission was found in the sensitivity analysis (OR 0.997, 95% CI 0.994–1.001; P = 0.119). We then repeated the sensitivity analyses in the subcohort of patients who were diagnosed in the year of 2013 (n = 6 139). The results still showed that increasing hospital volume (per 1-case increase) was associated with lower odds of conversion (OR 0.984, 95% CI 0.974–0.994; P = 0.002), PLOS (OR 0.989, 95% CI 0.986–0.992; P < 0.001) and PSM (OR 0.983, 95% CI 0.979–0.988; P < 0.001). No association between hospital volume (per 1-case increase) and 30-day readmission was found in the subcohort sensitivity analysis (OR 1.001, 95% CI 0.995–1.007; P = 0.651).

Discussion In the present NCDB-based study, we examined the associations between hospital volume and several outcomes of RAPN for RCC. Hospital volume was modelled both as a categorical variable and a continuous variable. We showed

2,981

4,173

5,431

6,139

1,000

2,000

3,000

4,000

5,000

6,000

7,000

2010 2011 2012 2013

N um

be r o

f R A

PN s

Year of diagnosis

27.2 22.7 18.1 15.5

21.8 20.3

19.0 19.4

23.5 20.5

18.0 21.2

12.6 17.6

25.0 19.3

14.9 19.0 20.0 24.6

100

2010 2011 2012 2013

Pe rc

en ta

ge o

f R A

PN s,

Year of diagnosis

Very high (≥ 44 cases) High (24-43 cases) Medium (15-23 cases) Low (8-14 cases) Very low (1-7 cases)

A B

Fig. 1 (A) Robot-assisted partial nephectomies (RAPNs) included in the study from 2010 to 2013 and (B) hospital volume distribution by year of diagnosis.

Xia et al.

that higher hospital RAPN volume was associated with lower risks of open conversion, PLOS and PSM. No difference was found in 30-day mortality, 90-day mortality or 30-day readmission rates. To the best of our knowledge, this is the first study evaluating the impact of RAPN volume on open conversion and PSM rates. The significant differences in PSM rates across hospital volume groups (11.8% in very-low- volume hospitals vs 4.6% in very-high-volume hospitals) is a key finding.

Volume–outcome association has been studied for many surgical procedures, including major cancer surgeries [11,16]. In the field of urological oncology, most of the studies have focused on radical cystectomy and radical prostatectomy [17– 19]. There are still limited data on the association between surgical volume and PN outcomes. Sun et al. [20] analysed a PN sub-cohort from the 1998–2007 Nationwide Inpatient Sample (NIS) and found that hospitals with higher nephrectomy (PN + RN) volume had shorter LOS and lower

Table 1 Baseline characteristics of the overall included cohort, stratified by hospital volume.

Characteristics Overall Very low Low Medium High Very high P

Age, years 60 (51–68) 60 (51–67) 60 (51–68) 60 (51–68) 60 (51–68) 59 (50–67) 0.168 Sex Male 11 480 (61.3) 2 218 (60.1) 2 230 (60.0) 2 310 (60.3) 2 273 (62.3) 2 449 (63.9) 0.001 Female 7 244 (38.7) 1 475 (39.9) 1 489 (40.0) 1 523 (39.7) 1 376 (37.7) 1 381 (36.1)

Race White 15 825 (84.5) 3 202 (86.7) 3 098 (83.3) 3 273 (85.4) 3 016 (82.7) 3 236 (84.5) <0.001 Black 2 093 (11.2) 341 (9.2) 462 (12.4) 421 (11.0) 427 (11.7) 442 (11.5) Other 596 (3.2) 118 (3.2) 139 (3.7) 97 (2.5) 121 (3.3) 121 (3.2) Unknown 210 (1.1) 32 (0.9) 20 (0.5) 42 (1.1) 85 (2.3) 31 (0.8)

Charlson–Deyo score 0 13 232 (70.7) 2 623 (71.0) 2 617 (70.4) 2 747 (71.7) 2 585 (70.8) 2 660 (69.5) <0.001 1 4 334 (23.2) 841 (22.8) 864 (23.2) 873 (22.8) 827 (22.7) 929 (24.3) ≥2 1 158 (6.2) 229 (6.2) 238 (6.4) 213 (5.6) 237 (6.5) 241 (6.3)

Insurance Private 10 337 (55.2) 2 043 (55.3) 2 053 (55.2) 2 046 (53.4) 2 033 (55.7) 2 162 (56.5) 0.179 Medicare 6 600 (35.3) 1 281 (34.7) 1 296 (34.9) 1 413 (36.9) 1 270 (34.8) 1 340 (35.0) Medicaid and other government 1 156 (6.2) 228 (6.2) 243 (6.5) 246 (6.4) 225 (6.2) 214 (5.6) Uninsured 437 (2.3) 93 (2.6) 83 (2.2) 86 (2.2) 85 (2.3) 90 (2.4) Unknown 194 (1.0) 48 (1.3) 44 (1.2) 42 (1.1) 36 (1.0) 24 (0.6)

Education level* ≥21% 2 866 (15.3) 522 (14.1) 645 (17.3) 572 (14.9) 629 (17.2) 498 (13.0) <0.001 13–20.9% 4 748 (25.4) 942 (25.5) 895 (24.1) 1 015 (26.5) 966 (26.5) 930 (24.3) 7–12.9% 6 296 (33.6) 1 319 (35.7) 1 194 (32.1) 1 214 (31.7) 1 169 (32.0) 1 400 (36.6) <7% 4 789 (25.6) 901 (24.4) 981 (26.4) 1 026 (26.8) 882 (24.2) 999 (26.1) Unknown 25 (0.1) 9 (0.2) 4 (0.1) 6 (0.2) 3 (0.1) 3 (0.1)

Income level <$38 000 2 932 (15.7) 518 (14.0) 609 (16.4) 588 (15.3) 591 (16.2) 626 (16.3) 0.021 $38 000–$47 999 4 181 (22.3) 839 (22.7) 774 (20.8) 860 (22.4) 811 (22.2) 897 (23.4) $48 000–$62 999 5 054 (27.0) 1 040 (28.2) 974 (26.2) 1 055 (27.5) 957 (26.2) 1 028 (26.8) ≥$63 000 6 527 (34.9) 1 286 (34.8) 1 357 (36.5) 1 324 (34.5) 1 286 (35.2) 1 274 (33.3) Unknown 30 (0.2) 10 (0.3) 5 (0.1) 6 (0.2) 4 (0.1) 5 (0.1)

Residence location Metropolitan 15 247 (81.4) 3 000 (81.2) 3 095 (83.2) 3 076 (80.3) 3 000 (82.2) 3 076 (80.3) <0.001 Urban 2 717 (14.5) 536 (14.5) 497 (13.4) 606 (15.8) 464 (12.7) 614 (16.0) Rural 290 (1.6) 69 (1.9) 67 (1.8) 62 (1.6) 38 (1.0) 54 (1.4) Unknown 470 (2.5) 88 (2.4) 60 (1.6) 89 (2.3) 147 (4.0) 86 (2.3)

Fuhrman grade 1–2 13 225 (70.6) 2 718 (73.6) 2 711 (72.9) 2 827 (73.8) 2 589 (71.0) 2 380 (62.1) <0.001 3–4 4 059 (21.7) 706 (19.1) 744 (20.0) 770 (20.1) 815 (22.3) 1 024 (26.7) Unknown 1 440 (7.7) 269 (7.3) 264 (7.1) 236 (6.2) 245 (6.7) 426 (11.1)

Tumour size, mm 26 (20–35) 25 (19–35) 27 (20–36) 25 (20–35) 26 (20–36) 28 (20–38) Tumour histology Clear-cell 13 481 (72.0) 2 740 (74.2) 2 702 (72.7) 2 770 (72.3) 2 577 (70.6) 2 692 (70.3) 0.001 Other 5 243 (28.0) 953 (25.8) 1 017 (27.4) 1 063 (27.7) 1 072 (29.4) 1 138 (29.7)

Year of diagnosis 2010 2 981 (15.9) 812 (22.0) 650 (17.5) 701 (18.3) 375 (10.3) 443 (11.6) <0.001 2011 4 173 (22.3) 948 (25.7) 846 (22.8) 855 (22.3) 733 (20.1) 791 (20.7) 2012 5 431 (29.0) 981 (26.6) 1 031 (27.7) 975 (25.4) 1 356 (37.2) 1 088 (28.4) 2013 6 139 (32.8) 952 (25.8) 1 192 (32.1) 1 302 (34.0) 1 185 (32.5) 1 508 (39.4)

Length of stay, days 2 (2–3) 3 (2–4) 3 (2–4) 2 (2–3) 2 (2–3) 2 (1–3) <0.001

Age, tumour size, and length of stay are median (interquartile range), all other variables are presented as n (%). Due to rounding, percentages may not always add up to 100%. *Percentage of adults who did not graduate from high school in the patient’s area of residence.

Hospital volume and robotic partial nephrectomy

rates of blood transfusion. A French multicentre study showed that higher hospital PN volume was associated with better peri-operative outcomes, including favourable LOS,

warm ischaemia time, medical complications and haemoglobin decrease [21]. One limitation of both studies was the compilation of PN approaches (open, laparoscopic

0.2 0.3

3.1

26.2

2.8

11.8

0.4 0.5 2.7

25.7

3.5

9.5

0.4 0.5 1.9

21.7

3.2

9.4

0.3 0.4 1.7

18.7

2.3

7.5

0.1 0.3 1.4

14.9

3.0 4.6

Pe rc

en ta

ge o

f p at

ie nt

s, %

Very low (1-7 cases) Low (8-14 cases) Medium (15-23 cases) High (24-43 cases) Very high (≥ 44 cases)

30-day mortality 90-day mortality Conversion PLOS 30-day readmission PSM

Fig. 2 Unadjusted outcomes comparison across the strata of hospital volume. PLOS, prolonged length of hospital stay; PSM, positive surgical margin.

0.0 0.5 1.0 1.5 OR (95%CI)

0.0 0.5 1.0 1.5

Very high (≥44 cases)

High (24-43 cases)

Medium (15-23 cases)

Low (8-14 cases)

Very low (1-7 cases)

0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5

Favors higher volume Favors lower volume Favors higher volume Favors lower volume

Favors higher volume Favors lower volumeFavors higher volume Favors lower volume

Conversion PLOS

30-day readmission PSM

Very high (≥44 cases)

High (24-43 cases)

Medium (15-23 cases)

Low (8-14 cases)

Very low (1-7 cases)

Very high (≥44 cases)

High (24-43 cases)

Medium (15-23 cases)

Low (8-14 cases)

Very low (1-7 cases)

Very high (≥44 cases)

High (24-43 cases)

Medium (15-23 cases)

Low (8-14 cases)

Very low (1-7 cases)

OR (95%CI)

Fig. 3 Adjusted odds ratios (ORs) and 95% confidence intervals (95%CIs) of hospital volume stratum from the multivariable logistic regressions for outcomes. PLOS, prolonged length of hospital stay; PSM, positive surgical margin.

Xia et al.

and robot-assisted), with most patients in their cohorts undergoing open surgery.

To our knowledge, there have been only two studies that evaluated the volume–outcome associations in RAPN; one on hospital volume and one on surgeon volume. Monn et al. [22] used the NIS data (2009–2011) and categorized hospital volume into high, medium and low. Tumour characteristics (size, grade, pathology) are not available in the NIS database and their study mainly focused on in-hospital complications. They showed that patients who underwent RAPN at high- volume hospitals had lower rates of overall in-hospital complications and transfusion. Khandwala et al. [23] showed that higher surgeon volume was associated with lower odds of prolonged operating time and PLOS. Compared with these two studies, the present study has several notable differences. Because the NCDB is a comprehensive cancer registry database, our study cohort only included patients with a final diagnosis of RCC and we adjusted for some of the tumour characteristics in our analyses. In addition, our outcomes were different (except PLOS) and we included one important oncological quality indicator, PSM. In terms of statistical methods, we performed sensitivity analyses, with hospital volume considered as a continuous variable to test the robustness of our results.

The most important and worrisome finding of the present study is the higher PSM rate in the lower-volume hospitals. According to the AUA guideline, negative surgical margins should be a priority for patients undergoing PN [4]. In other words, complete tumour excision is the most critical requirement of PN, and PSM may still the best immediate surrogate for oncological control in localized RCC. Although historical studies reported conflicting results regarding the long-term clinical significance of PSMs, more recent studies have shown evidence that PSMs after PN increase the risk of disease recurrence and are associated with inferior outcomes [24–27]. As for RAPN, Khalifeh et al. [26] included 943 cases from five major tertiary care institutions and showed that PSMs after RAPN were associated with higher risks of cancer recurrence and metastasis. Our finding in the present study of volume–PSM association, therefore, should not be taken lightly, especially when considering that more complex RCCs are now being managed with RAPN [2,28,29].

The present study also showed the association between hospital volume and open conversion rate. Although the absolute differences in conversion rate among the hospital volume groups may not be clinically significant, it still somewhat implies that higher-volume hospitals may have better intra-operative safety profiles. The lower odds of PLOS in the higher-volume hospitals is consistent with a previous study [22]. PLOS can be considered as a proxy for inpatient complications, which were also more likely to be seen in lower-volume hospitals [22]. Although discharge practice may

be different across various hospitals and surgeons, higher odds of PLOS would be very likely to increase the final overall cost [23].

Considering hospital volume and surgeon volume are generally correlated, the volume–outcome associations we showed in the present study may still hold true at the individual surgeon level [12,23]. More specifically, the conversion and PSM rates are more likely related to surgeon experience than hospital setting. It is possible that a significant number of RAPNs were performed at low-volume hospitals by surgeons without adequate case volume. Based on our results, centralization of RAPN to high-volume hospitals may provide superior peri-operative outcomes and potentially better oncological control. When taken together with previous studies, centralization may also decrease the procedure-related cost in the overall population [22,23]; however, robot-assisted surgery has expanded rapidly, and some degree of decentralization has already occurred in RARP [10]. It is unclear, therefore, whether centralization of RAPN would happen naturally as with radical cystectomy, or whether RAPN would eventually decentralize as has RARP [10,30]. Although the present study to some extent justifies the regionalization of RAPN, the main implication is that concern should be raised regarding the inferior outcomes in the lower-volume hospitals. Our findings also have policy implications with regard to the further diffusion of RAPN. One additional area of research could focus on uncovering the detailed components that lead to the superior outcomes of the higher-volume hospitals/surgeons and, accordingly, on developing strategies to improve surgical quality of the lower- volume hospitals/surgeons. Lessons should be learned from the present study, thereby helping to improve the quality of care provided to patients.

The present study has several limitations. First, detailed information on tumour complexity is not available in the NCDB. We could only control for tumour size instead of R.E.N.A.L nephrometry score; however, it is possible that higher-volume hospitals may receive referrals for more complex tumours and this ‘bias’ may bolster our results. Second, other selection biases may still exist considering the observational nature of this study. Third, some important outcomes were not available, such as intra-operative/ postoperative complications, warm ischaemia time, change in estimated GFR, and long-term cancer-specific survival. Volume-outcome studies, such as the present study, should always be interpreted together with other studies on this topic (e.g. using another database and having other outcomes). Fourth, information regarding the learning curve is not available in the NCDB. It is possible that lower-volume hospitals may have surgeons in their early learning curve, which may introduce potential bias with regard to the volume–outcome association. Lastly, the NCDB is a cancer registry and no benign tumours were included in the present

Hospital volume and robotic partial nephrectomy

study. The surgical outcomes in patients with benign tumours may be different.

In conclusion, undergoing RAPN at higher-volume hospitals may lead to better peri-operative outcomes (conversion to open and LOS) and lower PSM rates. Future studies are needed to explore the detailed components that lead to the superior outcomes in higher-volume hospitals.

Conflicts of Interest None declared.

Disclaimer The data used in this study were derived from a de-identified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytical or statistical methodology employed, or the conclusions drawn from these data by the investigators.

References 1 Leow JJ, Chang SL, Meyer CP et al. Robot-assisted versus open radical

prostatectomy: a contemporary analysis of an all-payer discharge database. Eur Urol 2016; 70: 837–45

2 Cheung H, Wang Y, Chang S, Khandwala YS, Del GF, Chung BI. Adoption of robotic-assisted partial nephrectomies: a population-based analysis of U.S. surgeons from 2004–2013. J Endourol 2017; 31: 886–92

3 Ljungberg B, Bensalah K, Canfield S et al. EAU guidelines on renal cell carcinoma: 2014 update. Eur Urol 2015; 67: 913–24

4 Campbell S, Uzzo RG, Allaf ME et al. Renal mass and localized renal cancer: AUA guideline. J Urol 2017; 198: 520–9

5 Patel HD, Mullins JK, Pierorazio PM et al. Trends in renal surgery: robotic technology is associated with increased use of partial nephrectomy. J Urol 2013; 189: 1229–35

6 Kardos SV, Gross CP, Shah ND et al. Association of type of renal surgery and access to robotic technology for kidney cancer: results from a population-based cohort. BJU Int 2014; 114: 549–54

7 Choi JE, You JH, Kim DK, Rha KH, Lee SH. Comparison of perioperative outcomes between robotic and laparoscopic partial nephrectomy: a systematic review and meta-analysis. Eur Urol 2015; 67: 891–901

8 Xia L, Wang X, Xu T, Guzzo TJ. Systematic review and meta-analysis of comparative studies reporting perioperative outcomes of robot-assisted partial nephrectomy versus open partial nephrectomy. J Endourol 2017; 31: 893–909

9 Wright JD, Tergas AI, Hou JY et al. Effect of regional hospital competition and hospital financial status on the use of robotic-assisted surgery. JAMA Surg 2016; 151: 612–20

10 Gershman B, Meier SK, Jeffery MM et al. Redefining and contextualizing the hospital volume-outcome relationship for robot-assisted radical prostatectomy: implications for centralization of care. J Urol 2017; 198: 92–9

11 Birkmeyer JD, Siewers AE, Finlayson EV et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002; 346: 1128–37

12 Abouassaly R, Finelli A, Tomlinson GA, Urbach DR, Alibhai SM. Volume-outcome relationships in the treatment of renal tumors. J Urol 2012; 187: 1984–8

13 Boffa DJ, Rosen JE, Mallin K et al. Using the national cancer database for outcomes research: a review. JAMA Oncol 2017; 3: 1722–8

14 Hanna N, Sun M, Meyer CP et al. Survival analyses of patients with metastatic renal cancer treated with targeted therapy with or without cytoreductive nephrectomy: a national cancer data base study. J Clin Oncol 2016; 34: 3267–75

15 Collins TC, Daley J, Henderson WH, Khuri SF. Risk factors for prolonged length of stay after major elective surgery. Ann Surg 1999; 230: 251–9

16 Begg CB, Cramer LD, Hoskins WJ, Brennan MF. Impact of hospital volume on operative mortality for major cancer surgery. JAMA 1998; 280: 1747–51

17 Leow JJ, Leong EK, Serrell EC et al. Systematic review of the volume- outcome relationship for radical prostatectomy: a 2017 update. Eur Urol Focus 2017; pii: S2405-4569(17)30076-7; [Epub ahead of print]. https:// doi.org/10.1016/j.euf.2017.03.008

18 Goossens-Laan CA, Gooiker GA, van Gijn W et al. A systematic review and meta-analysis of the relationship between hospital/surgeon volume and outcome for radical cystectomy: an update for the ongoing debate. Eur Urol 2011; 59: 775–83

19 Peyronnet B, Couapel JP, Patard JJ, Bensalah K. Relationship between surgical volume and outcomes in nephron-sparing surgery. Curr Opin Urol 2014; 24: 453–8

20 Sun M, Bianchi M, Trinh QD et al. Hospital volume is a determinant of postoperative complications, blood transfusion and length of stay after radical or partial nephrectomy. J Urol 2012; 187: 405–10

21 Couapel JP, Bensalah K, Bernhard JC et al. Is there a volume-outcome relationship for partial nephrectomy? World J Urol 2014; 32: 1323–9

22 Monn MF, Bahler CD, Flack CK, Dube HT, Sundaram CP. The impact of hospital volume on postoperative complications following robot- assisted partial nephrectomy. J Endourol 2014; 28: 1231–6

23 Khandwala YS, Jeong IG, Kim JH et al. The impact of surgeon volume on perioperative outcomes and cost for patients receiving robotic partial nephrectomy. J Endourol 2017; 31: 851–7

24 Laganosky DD, Filson CP, Master VA. Surgical margins in nephron- sparing surgery for renal cell carcinoma. Curr Urol Rep 2017; 18: 8

25 Maurice MJ, Zhu H, Kim SP, Abouassaly R. Reexamining the association between positive surgical margins and survival after partial nephrectomy in a large American cohort. J Endourol 2016; 30: 698–703

26 Khalifeh A, Kaouk JH, Bhayani S et al. Positive surgical margins in robot-assisted partial nephrectomy: a multi-institutional analysis of oncologic outcomes (leave no tumor behind). J Urol 2013; 190: 1674–9

27 Shah PH, Moreira DM, Okhunov Z et al. Positive surgical margins increase risk of recurrence after partial nephrectomy for high risk renal tumors. J Urol 2016; 196: 327–34

28 Paulucci DJ, Abaza R, Eun DD, Hemal AK, Badani KK. Robot-assisted partial nephrectomy: continued refinement of outcomes beyond the initial learning curve. BJU Int 2017; 119: 748–54

29 Wang Y, Ma X, Huang Q et al. Comparison of robot-assisted and laparoscopic partial nephrectomy for complex renal tumours with a RENAL nephrometry score ≥7: peri-operative and oncological outcomes. BJU Int 2016; 117: 126–30

30 Anderson CB, Gennarelli R, Herr HW, Elkin EB. Regionalization of radical cystectomy in the United States. Urol Oncol 2017; 35: 528. e7–13

Correspondence: Leilei Xia, Abramson Research Center, Room 915-H, 3615 Civic Center Boulevard, Philadelphia, Pennsylvania 19104, USA.

e-mail: [email protected]

Abbreviations: ICD-O-3, International Classification of Diseases for Oncology, third edition; NCDB, National Cancer Database; NIS, Nationwide Inpatient Sample; PLOS,

Xia et al.

https://doi.org/10.1016/j.euf.2017.03.008

prolonged length of hospital stay; PN, partial nephrectomy; PSM, positive surgical margin; RAPN, robot-assisted partial nephrectomy; RARP, robot-assisted radical prostatectomy; RCC, renal cell carcinoma; RN, radical nephrectomy.

Supporting Information Additional Supporting Information may be found in the online version of this article:

Table S1. Multivariable logistic regression results for conversion. Table S2. Multivariable logistic regression results for prolonged length of stay. Table S3. Multivariable logistic regression results for 30-day readmission. Table S4. Multivariable logistic regression results for positive surgical margin.

Hospital volume and robotic partial nephrectomy