J. Kellogg Parsons, Division of Urologic Oncology, 200 West Arbor Drive #8897, University of California, San Diego, San Diego, CA 92103, USA. e-mail: firstname.lastname@example.org
Study Type – Cohort study
Level of Evidence 2b
What's known on the subject? and What does the study add?
Laparoscopic radical nephrectomy for renal cancer provides equivalent long-term cancer control with shorter hospital stays, less postoperative pain, and faster resumption of normal activities, but it has diffused slowly into clinical practice, perhaps as a result of perceptions about safety. Patient safety outcomes for laparoscopic and open radical nephrectomy using validated measures remain incompletely characterized.
This is the first study to investigate peri-operative outcomes of radical nephrectomy using validated patient safety measures. We found a 32% decreased probability of adverse patient safety events occurring in laparoscopic compared with open radical nephrectomy. The safety benefits of laparoscopy were attained only after 10% of cases were completed laparoscopically – a proportion some have proposed as the ‘tipping point’ for the adoption of surgical innovations. This observation could have implications for patient safety in the setting of diffusion of new surgical techniques.
• To compare peri-operative adverse patient safety events occurring in laparoscopic radical nephrectomy (LRN) with those occurring in open radical nephrectomy (ORN).
• We used the US Nationwide Inpatient Sample to identify patients undergoing kidney surgery for renal tumours from 1998 to 2008.
• We used patient safety indicators (PSIs), which are validated measures of preventable adverse outcomes, and multivariate regression to analyse associations of surgery type with patient safety.
• Open radical nephrectomy accounted for 235 098 (89%) cases while 28 609 (11%) cases were LRN.
• Compared with ORN, LRN patients were more likely to be male (P= 0.048), have lower Charlson comorbidity scores (P < 0.001), and to undergo surgery at urban (P < 0.001) and teaching (P < 0.001) hospitals.
• PSIs occurred in 18 714 (8%) of ORN and 1434 (5%) of LRN cases (P < 0.001).
• On multivariate analysis, LRN was associated with a 32% decreased probability of any PSI (adjusted odds ratio 0.68, 95% confidence interval: 0.6 to 0.77, P < 0.001). Stratification by year showed that this difference was initially manifested in 2003, when the proportion of LRN cases first exceeded 10%.
• We found that LRN was associated with substantially superior peri-operative patient safety outcomes compared with ORN, but only after the national prevalence of LRN exceeded 10%.
• Further study is needed to explain these patterns and promote the safe diffusion of novel surgical therapies into broad practice.
Laparoscopic radical nephrectomy (LRN) is regarded by many as the ‘gold standard’ for the surgical treatment of RCC requiring complete kidney removal . Compared with open radical nephrectomy (ORN), LRN provides equivalent long-term cancer control with shorter hospital stays, less postoperative pain, and faster resumption of normal activities [2–6].
Despite these advantages, LRN for renal cancer has diffused slowly into clinical practice in the USA and remains substantially underused [7,8]. Compared with other laparoscopic surgeries (including donor nephrectomy, cholecystectomy and fundoplication), adoption of LRN into clinical practice has been modest, and ORN remains the predominant surgical approach for kidney cancer . Potential barriers to LRN diffusion include perceived technical complexity, a limited pool of urologists trained in laparoscopy, costs associated with adoption of new technology and techniques, and individual surgeon practice style [9,10].
Another consideration possibly affecting the use of LRN is patient safety. Adoption of surgical innovations depends in large part upon provider acceptance of the new technology, and surgeon perception that ORN is safer than LRN may be a factor contributing to LRN's relatively sluggish diffusion [11,12]. The AUA's clinical guidelines for the management of stage T1 renal masses concluded from a systematic review that LRN was associated with a higher major urological complication rate than ORN ; however, other data suggest comparable, if not superior, peri-operative outcomes for LRN [14,15].
Comprehensive patient safety analyses comparing LRN with ORN, using process-driven measures, have yet to be performed. Safety analyses performed in a wider population of patients and incorporating a comprehensive array of validated patient safety outcomes would potentially inform the selection of operative approaches for nephrectomy, the process by which LRN has entered surgical practice, and the variables affecting the diffusion of innovative surgical techniques. The aim of the present study was to compare the prevalence of adverse patient safety events occurring in ORN and LRN in a national database over a 10-year period.
STUDY POPULATION: THE NATIONWIDE INPATIENT SAMPLE
We performed a cross-sectional analysis of inpatient hospital discharges after ORN or LRN from 1998 to 2008. Data on inpatient procedures and adverse events were obtained from the Healthcare Cost and Utilization Project-Nationwide Inpatient Sample (NIS), produced by the Agency for Healthcare Research and Quality (AHRQ). The NIS represents the largest hospital inpatient dataset in the USA and consists of a 20% sample of inpatient discharges from 1056 hospitals. Weighted sampling allows estimates for national trends. The NIS provides information on age, gender, race, income, insurance type, the Charlson comorbidity index , hospital location (rural or urban), and hospital type (teaching or non-teaching). We used only PSIs validated for surgery and patients with cancer.
NEPHRECTOMY CASE SELECTION
All discharges from 1998 to 2008 with a procedure code for total or radical kidney surgery were identified using principal or secondary procedure ICD-9CM codes for radical nephrectomy (55.5, 55.51, 55.52, 55.54) divided among ‘laparoscopic’ (concurrent 54.51, 54.21 or 17.42) and ‘open’ (without concurrent laparoscopy codes). Patients <18 years old, or with an admission type other than ‘elective’, were excluded, as were subjects with a diagnosis of transplant (v59.4), renal pelvis tumour (189.1), and pyelonephritis (590.00, 590.01, 590.10, 590.11), leaving only those with nephrectomy for benign and malignant tumours.
To reduce the potential for confounding by disease burden, we excluded from the analysis patients undergoing the following concomitant procedures: splenectomy (41.4, 41.42, 41.43), liver resection (50.2, 50.21–26, 50.29, 50.3), pancreas resection (52.5, 52.51–53, 52.59, 52.6, 52.7), bowel or colon resection (45.5, 45.50–52, 45.61–62, 45.7, 45.71, 45.73–76, 45.79, 45.8, 45.81–83), or thrombectomy with vascular reconstruction (37.10, 38.05, 38.07, 38.45, 38.47, 38.65, 38.67, 38.75, 38.77, 38.87, 39.6, 39.61, 39.63, 39.66).
MEASURES OF PATIENT SAFETY: PATIENT SAFETY INDICATORS
The AHRQ has developed patient safety indicators (PSIs) to capture clinical data regarding adverse inpatient outcomes [17–19]. These evidence-based measures are designed for use in administrative databases and are based on the Institute of Medicine's definition of patient safety as ‘freedom from accidental injury due to medical care, or medical errors.’ These validated instruments provide hospital- and national-level safety performance measures with greater specificity and less bias than other methods .
We identified PSIs according to the list published by AHRQ and they included anaesthetic complications, pneumothorax, haemorrhage/haematoma, decubitus ulcer, venous thrombosis/pulmonary embolism, hip fracture, postoperative physiological and metabolic derangements, postoperative respiratory failure, sepsis and accidental puncture or laceration of surrounding organs during a procedure. PSIs represent acute events that occur during the same hospitalization as the surgery.
The annual prevalence of ORN and LRN from 1998 to 2008 was examined to assess trends and determine inferential relationships between PSI and type of renal surgery. All numbers reported are estimates of the national number of hospital inpatients stays, generated by applying sample weighting to the 20% NIS sample. Univariate tests including chi-square (Rao & Scott second order correction), Student's t-test, and Fisher's exact test were used to compare groups and assess factors associated with increased frequency of PSIs. Variables significant at the 5% level were carried forward for assessment in binary logistic regression models to determine the associations of PSIs and time (year, categorical); variables which remained significant in multivariate analysis remained in the final models. The SVY coding in Stata v 11.1 (StataCorp, College Station, TX, USA) was used to account for NIS sampling methodology.
We identified 522 274 cases of radical nephrectomy performed from 1998 to 2008. Exclusion of those for age <18 (21 372), non-elective (136 365), transplant and infection (86 872), and concomitant surgeries suggesting locoregional or metastatic disease (13 958) left 263 707 cases for evaluation. Of these, there were 235 098 (89%) ORN cases and 28 609 (11%) LRN cases. The number of ORN and LRN cases increased steadily during the study period (Fig. 1). The prevalence of LRN increased at an annual rate of nearly 150%, with a change in proportion from 0.6% of all cases in 1998 to 18.9% of all cases in 2008 (Fig. 2).
Compared with those undergoing ORN, patients who underwent LRN were more likely to be male (P= 0.048), have lower Charlson comorbidity index scores (P < 0.001), and to undergo surgery at urban (P < 0.001) and teaching (P < 0.001) hospitals. Length of stay was shorter for LRN (P < 0.001). There were no significant differences in age, race, income and insurance status between groups (Table 1).
Table 1. Demographic characteristics of patients undergoing radical nephrectomy for renal cancer in the NIS, 1998 to 2008, stratified by open and laparoscopic surgery
The unadjusted prevalence of any PSI was higher in ORN compared with LRN (P < 0.001; Table 2). Specific PSIs associated with ORN included iatrogenic pneumothorax, postoperative respiratory failure, accidental injury or laceration to surrounding bowel or solid organs, and postoperative sepsis.
Table 2. Prevalence of PSIs in patients undergoing radical nephrectomy for renal cancer in the NIS, 1998 to 2008
Significant two-tailed. Superscript numbers correspond to numbered PSIs from the ARHQ.
Any PSI (%)
18 714/235 026 (8)
1.434/28 600 (5)
Complications of anaesthesia1 (%)
446/234 733 (0.2)
25/28 536 (0.1)
Death in low-mortality DRGs2 (%)
Decubitus ulcer3 (%)
495/117 136 (0.4)
50/6 696 (0.7)
Failure to rescue4 (%)
1 119/11 205 (10)
Iatrogenic pneumothorax6 (%)
2 510/225 665 (1.1)
50/28 054 (0.2)
Selected infections due to medical care7 (%)
76/24 569 (0.3)
0/3 218 (0)
Postoperative hip fracture8 (%)
18/210 588 (0)
5/26 686 (0)
Postoperative haemorrhage or haematoma9 (%)
955/234 544 (0.4)
51/28 532 (0.3)
Postoperative physiological and metabolic derangements10 (%)
724/233 872 (0.3)
61/28 464 (0.2)
Postoperative respiratory failure11 (%)
3 484/232 832 (1.5)
272/28 466 (1)
Postoperative PE/DVT12 (%)
2 030/234 258 (0.9)
235/28 512 (0.8)
Postoperative sepsis13 (%)
264/14 284 (1.8)
Postoperative wound dehiscence14 (%)
155/213 236 (0.1)
25/25 423 (0.1)
Accidental puncture or laceration15 (%)
9 021/234 735 (3.8)
706/28 581 (2.5)
Transfusion reaction16 (%)
There were no significant differences in the prevalence of decubitus ulcers, postoperative physiological or metabolic disorders, postoperative haemorrhage or haematoma, wound dehiscence, or postoperative deep venous thrombosis/pulmonary embolism.
On multivariate analysis, LRN was associated with a 32% reduction in the probability of any PSI for the entire study period (Table 3). The probability of any PSI was increased in older patients, men, Charlson comorbidty score ≥3, and teaching hospitals. When stratified by year, the odds of any PSI were lower for LRN starting in 2003, when the proportion of LRN cases increased for the first time above 10% of all nephrectomies performed. From 2003 to 2008, the association of LRN with decreased probability of PSI remained consistent (Fig. 3). Sensitivity analyses of PSIs stratified by demographic variables (sex, race, age, income, Charlson comorbidity score and hospital type) confirmed that LRN was consistently safer than ORN across all populations (odds ratio [OR] 0.62–0.9; data not shown).
Table 3. Multivariable-adjusted ORs for the occurrence of any PSI in patients undergoing radical nephrectomy for renal cancer in the NIS, 1998 to 2008, adjusted for number of eligible PSIs and year of surgery
A subgroup analysis of 143 802 patients ≥65 years of age showed a 38% reduction in the odds of any PSI when compared with ORN (adjusted OR [ORadj] 0.62, 95% CI 0.53 to 0.73, P < 0.001). A sensitivity analysis including all concomitant surgeries (liver, spleen, intestinal, vascular) but excluding thoracic and cardiac procedures showed similar results favouring LRN for any PSI (ORadj 0.69, 95% CI 0.61 to 0.73, P < 0.001).
The unadjusted prevalence of in-hospital mortality was higher in ORN (0.8%) than in LRN (0.5%, P= 0.03), but the adjusted prevalence was not (P= 0.90). On multivariate analysis, increased odds of in-hospital mortality was associated with older age (P < 0.001), male gender (P= 0.004), Charlson comorbidity score ≥3 (P < 0.001) and year of surgery (P= 0.007).
To our knowledge, this is the first study to investigate peri-operative outcomes of radical nephrectomy – or any urological surgery – using validated patient safety measures. While systematic analyses of patient safety and quality of care issues have gained broad acceptance within the fields of medicine and health policy [22,23], they have not disseminated as widely among the surgical disciplines . In this population-based analysis, we observed a 32% (ORadj 0.68) decreased probability of patient safety events occurring in LRN compared with ORN as measured by PSIs. Our results thus suggest that use of LRN results in substantially enhanced peri-operative patient safety.
A novel component of the present study is the stratification of the patient safety analyses by year in a period during which LRN was initially diffusing into broader clinical practice. The lower likelihood of adverse safety events with LRN emerged only after the national prevalence of LRN exceeded 10%, a proportion some have proposed as the ‘tipping point’ for the adoption of surgical innovations [12,25]. This phase of diffusion is a point during which early adopters (usually composed of expert opinion leaders) transition into the early majority (those who adopt a new technique just before the average surgeon) . Based on the present data, it is possible that the potential safety benefits of LRN were attained only after completion of a learning curve by the innovators and early adopters, an observation that could have implications for a broader understanding of patient safety as it relates to the diffusion of new surgical techniques.
The present study is also novel in that it is one of the first analyses of nephrectomy for renal tumours to use a large administrative dataset coupled with validated, process-focused, provider-side measures of patient safety. In previous studies in other settings, PSIs have proven to be robust metrics for quality improvement and are valuable in assessing safety because they correlate with adverse events while minimizing potential biases associated with variations in individual patients and care-givers [27–29]. Our results are consistent with those of Tan et al. , who observed among a national cohort of older US patients that LRN was associated with decreased probabilities of prolonged hospitalization, intensive care unit admission, and peri-operative transfusion compared with ORN; however, unlike Tan et al., we did not discern a higher mortality for LRN in the present cohort. Potential explanations include differences in our analytical cohort, which was younger and larger, and the fact that we assessed in-hospital, rather than 30-day, mortality.
Although we did not adjust for tumour size in the analytical models, there are several reasons why it is unlikely that differential bias in tumour size between groups accounted for the observed differences in PSIs. First, we excluded patients that underwent concomitant surgical procedures (thrombectomy and/or vascular reconstruction; pancreatic, splenic or hepatic procedures; bowel resection) indicative of large, complex renal tumours requiring extensive resection and potentially associated with poorer peri-operative outcomes related to aggressive disease. Second, during the study period, ORN remained the dominant operative approach to kidney cancer despite a period of significant stage migration towards smaller and lower stage tumours . An analysis of SEER-Medicare data from 2000 to 2007 noted that ORN continued to account for nearly 70% of renal cancer surgery in these patients at a time when localized or locoregional disease accounted for nearly 80% of cases . Similarly, in a recent case-control analysis conducted by the National Cancer Institute, 79% of renal tumours were ≤7 cm and 52% ≤4 cm; most of these cases were treated with ORN . In the present study, nearly 90% of all surgeries were ORNs, meaning that the vast majority of ORNs were performed for smaller tumours.
Furthermore, while many previous analyses of peri-operative outcomes for renal surgery have included variables for tumour stage and size, the question of whether these variables should be considered in all analytical comparisons of LRN and ORN has never been satisfactorily addressed. For the largest tumours, including those with venous thrombi, there is the potential for significant collinearity between variables for size and surgery type, which in turn can potentially introduce unpredictable biases into analyses comparing laparoscopic with open surgery. Within this context, it is worthwhile noting that, in the analysis by Tan et al. , exclusion of patients with larger renal tumours yielded results similar to those of the overall study.
Nevertheless, lack of patient-specific oncological data, including tumour size or tumour stage, is a potential limitation of the present study and possibly contributed to the observed PSI difference in hospital type, particularly since teaching hospitals would be expected to care for a greater proportion of complex tumours (Table 3). A second limitation is that variations in coding practices over the 10-year study period may have influenced the analyses. These changes included the addition, in more recent years, of laparoscopic-specific procedure codes, and may have been differential with respect to teaching and non-teaching hospitals – increased reporting of PSIs at teaching hospitals, for example, may have contributed to the observed 10% increased prevalence at teaching compared with non-teaching hospitals. Still, there is no evidence to suggest that such variations would have been differential with respect to the primary study outcomes. Finally, we could only assess adverse outcomes that occurred during the period of hospitalization after surgery. It is possible that differential outcomes of adverse events occurring after discharge, but still within the peri-operative period (i.e. 30 days), were not captured.
In conclusion, LRN was associated with substantially superior peri-operative patient safety outcomes than ORN, but only after the prevalence of LRN exceeded 10% of cases nationally. Further study is needed to explain these patterns and promote the safe diffusion of novel surgical therapies into broad practice.