Perioperative Risk Assessment in Robotic General Surgery: Lessons Learned From 884 Cases at a Single Institution

Objective To assess factors associated with morbidity and mortality following the use of robotics in general surgery.

Design Case series.

Setting University of Illinois at Chicago.

Patients and Intervention Eight hundred eighty-four consecutive patients who underwent a robotic procedure in our institution between April 2007 and July 2010.

Main Outcomes Measures Perioperative morbidity and mortality.

Results During the study period, 884 patients underwent a robotic procedure. The conversion rate was 2%, the mortality rate was 0.5%, and the overall postoperative morbidity rate was 16.7%. The reoperation rate was 2.4%. Mean length of stay was 4.5 days (range, 0.2-113 days). In univariate analysis, several factors were associated with increased morbidity and included either patient-related (cardiovascular and renal comorbidities, American Society of Anesthesiologists score ≥3, body mass index [calculated as weight in kilograms divided by height in meters squared] <30, age ≥70 years, and malignant disease) or procedure-related (blood loss ≥500 mL, transfusion, multiquadrant operation, and advanced procedure) factors. In multivariate analysis, advanced procedure, multiquadrant surgery, malignant disease, body mass index of less than 30, hypertension, and transfusion were factors significantly associated with a higher risk for complications. American Society of Anesthesiologists score of 3 or greater, age 70 years or older, cardiovascular comorbidity, and blood loss of 500 mL or more were also associated with increased risk for mortality.

Conclusions Use of the robotic approach for general surgery can be achieved safely with low morbidity and mortality. Several risk factors have been identified as independent causes for higher morbidity and mortality. These can be used to identify patients at risk before and during the surgery and, in the future, to develop a scoring system for the use of robotic general surgery.

While complications can arise from any surgery, a lack of consensus exists within the surgical community as to the best way to report surgical complications, and this has hampered proper evaluation of the surgeon's work.^1,2 Since the adoption of the Clavien-Dindo Classification,¹ much progress has been made toward categorizing the various surgical complications that can occur. At the same time, several scoring systems have been used to predict the postoperative complications, including the American Society of Anesthesiologists (ASA) score,³ Acute Physiology and Chronic Health Evaluation II,⁴ Physiological and Operative Severity Score for the Enumeration of Mortality and Morbidity, Simplified Acute Physiology Score,⁴ and the Estimation of Physiologic Ability and Surgical Stress,^4,5 among others.^6,7 While the ASA score is easy to use, others can require the use of a computer and are not always user friendly. Still, a preoperative assessment is crucial in planning the operative approach and anticipating possible postoperative complications.⁴

With the era of minimally invasive surgery came the hope that its use would result in reduced complications. Since their introduction in 2001, robotics have not only gained increasing acceptance for general surgery, but their feasibility and safety have also been reported for various and advanced procedures.^8-10 Thus far, however, few data are available to evaluate the risk factors associated with the robotic approach.

This study was designed to assess those factors associated with morbidity and mortality following a robotic general procedure.

Between April 2007 and July 2010, all robotic procedures performed at the Department of Surgery, University of Illinois Medical Center at Chicago, were analyzed. A retrospective review of prospectively collected data from patient medical records was performed as part of an approved institutional review board protocol.

During the study period, a total of 884 consecutive elective robotic procedures were performed (Table 1).

For the purposes of our study, comorbidities were defined as follows. Hypertension included those with a persistent elevated blood pressure (>140/90 mm Hg) or those who were taking medication to control their blood pressure. Respiratory diseases included but were not limited to chronic obstructive pulmonary disease, asthma, respiratory insufficiency, or obstructive sleep apnea. Diabetes mellitus was defined as elevated glycemia or the need to take medication to control it. Cardiovascular comorbidity included but was not limited to valvulopathy, myocardial ischemia or infarct, cardiac failure, or arrhythmia. Renal disease included but was not limited to renal insufficiency or glomerulonephritis. Neurologic diseases included stroke, epilepsy, brain tumor, migraine, Parkinson disease, or Alzheimer disease.

Other comorbidities included various diseases not commonly reported. These were not included in the statistical model owing to their rarity and heterogeneity.

The 884 robotic procedures were performed by various experienced surgeons from the Department of Surgery.

The procedures were divided into 3 categories according to surgical complexity (Table 2), using a modified version of the department's own classification:¹¹

1. A basic procedure is an operation requiring little dissection and/or reconstruction.

2. An intermediate procedure is an intervention requiring substantial dissection and/or reconstruction, usually requiring a mechanical device for the anastomosis.

3. An advanced procedure is an operation requiring advanced dissection in difficult to reach areas as well as complex reconstruction, usually requiring hand-sewn anastomosis.

A multiquadrant surgery was defined as a procedure performed in more than 1 anatomical quadrant. This typically required moving the robotic cart and/or the trocars (colorectal resection, gastric bypass, bilateral procedure, total pancreatectomy, and gastrectomy).

The operative time was calculated as the time between skin incision and the last port skin closure, including the docking time and any associated procedures. Conversion was defined as the need to terminate the operation with an open approach.

Morbidity and mortality were defined as any complication or death that occurred during hospitalization or within 30 days of discharge following the robotic procedure.

A complication was defined as any deviation from the normal postoperative course. To classify the severity of the complications, we used the Clavien-Dindo Classification¹ system.

The results of parametric and nonparametric data were expressed as mean (standard deviation) and median (range), respectively. We used GraphPad Software (GraphPad) for the first part of the statistical analysis. Confidence intervals were set at 95%. A 2-sided P ≤ .05 was considered statistically significant.

Data were summarized using frequencies and percentages for all factors. With the binary morbidity/mortality end point, various factors were examined for their relationship to the end points. The factors considered in the modeling process included classification of surgical difficulty (advanced, intermediate, or basic procedure), malignant or benign disease, sex, ASA score (<3 and ≥3), body mass index (BMI, calculated as weight in kilograms divided by height in meters squared) (<30 or ≥30), diabetes mellitus, neurologic comorbidity, cardiovascular comorbidity, hypertension, renal disease, respiratory disease, age (<70 or ≥70 years), conversion, estimated blood loss (<500 or ≥500 mL), transfusion, previous chemotherapy, and multiquadrant surgery.

Logistic regression modeling techniques were used to explore the relationships. First, univariable logistic regression was used to identify the potential factors to include in the model-building process. Factors with univariable P ≤ .15 were considered for inclusion in multivariable modeling. Multicollinearity as a potential factor and a possible interaction was explored. All factors with univariable P ≤ .15 were included in a model and backward elimination was then completed in a stepwise fashion. The least significant factor was removed from the model and the model was rerun. This process was repeated until only factors with P ≤ .05 remained in the model. Parameter estimates, standard errors, odds ratios, and 95% confidence intervals for the odds ratios were reported along with the P values from the logistic regression modeling.

Among the 884 robotic procedures, 158 were classified as basic (17.9%), 516 as intermediate (58.4%), and 210 as advanced (23.7%) (Table 2). The differences among the patients in the 3 groups are summarized in Table 3.

The perioperative data are reported in Table 4. The mean estimated blood loss was 121.9 mL (range, 0-10 000 mL). The patient who bled 10 L underwent a right hepatectomy with portal vein resection for hilar cholangiocarcinoma and underwent conversion owing to caval bleeding. In the postoperative course, he developed a biliary leak and was treated conservatively with good results.

A total of 18 conversions were performed (2%), including 8 for oncologic reasons (advanced pancreatic lesions requiring difficult vascular resections and posterior hepatic lesions), 5 for exposure or technical problems, 2 for bleeding, and 3 for other causes (iatrogenic enterotomy or lysis of adhesions).

Intraoperative complications included 4 iatrogenic injuries (2 enterotomies, 1 hepatic injury, and 1 bladder lesion), 2 bleedings requiring a conversion, and 1 intraoperative hypoglycemia necessitating close monitoring in the intensive care unit.

A total of 21 reoperations were performed including 4 postoperative hernias, 3 pancreatic/biliary fistulas, 3 gastrointestinal leaks, 2 postoperative ileus, 2 wound infections, 1 hematoma, and 6 other causes (pleural effusion, sepsis, cholecystitis, duodenal ulcer, piriform sinus laceration, and collection).

The differences in the perioperative data among the 3 groups of procedures are reported in Table 5.

A total of 148 postoperative complications (16.7%) occurred during the study period. Using the Clavien-Dindo Classification, the severities of the complications are summarized in Table 6, with a specific description of complications for basic procedures (Table 7).

The univariate analysis for morbidity (Table 8) showed that there were both patient-related and procedure-related risk factors. The patient-related factors included age of 70 years or older, ASA score of 3 or greater, BMI of less than 30, hypertension, cardiovascular disease, renal comorbidity, and malignant disease. The procedure-related factors included the type of procedure (advanced > intermediate > basic), a multiquadrant surgery, blood loss of 500 mL or greater, and the need for transfusion.

The multivariate analysis (Table 9) showed several significant independent risk factors. These included a BMI of less than 30, hypertension, malignant disease, type of procedure, multiquadrant surgery, and transfusion.

In terms of mortality, 4 patients died during the study period (0.5%). Of those, 2 patients died following cardiac arrhythmia. Both were older than 75 years with an ASA score of 3. They underwent a pancreaticoduodenectomy and a complex paraesophageal hernia repair, respectively. The third patient underwent a hepaticojejunostomy for common bile duct injury and died from septic condition after a prolonged hospital stay. The last patient underwent a pancreaticoduodenectomy and died as a result of upper gastrointestinal bleeding.

The univariate analysis (Table 10) showed that age of 70 years or older, an ASA score of 3 or greater, cardiovascular disease, advanced procedure, and blood loss of 500 mL or greater were associated with an increased risk for mortality.

It is important to note that the multivariate analysis was not performed because there was not enough statistical power to run hypothesis tests or statistical modeling.

We present one of the largest series of reported morbidity and mortality following robotic general surgery. The overall results are more than encouraging, with an acceptable morbidity as well as low mortality, conversion, and reoperation rates, even when there was very little patient selection.

The goal of our study was to evaluate the risk factors associated with mortality and morbidity in robotic general surgery. Similar studies have been published for open and laparoscopic surgery.^5,7,12-14 Two series^15,16 have reported the risk factors associated with robotic urology, however, neither was designed to assess the complications following robotic general surgery.

We agree with Haga et al⁵ in believing that postoperative complications can result from 3 major factors: the patient's physiological status, the quality of surgical performance, and the degree of surgical stress applied.

Among the tools available to evaluate patients' physiological status, the ASA score is probably the most widely used and easy to use, despite it being subjective with a large interobserver variability.¹⁷ American Society of Anesthesiologists scores have been shown to be correlated with postoperative complications and mortality rates,^{3,12,13,18-22} and we confirmed those results, even if the multivariate analysis did not corroborate this.

The elderly population has been reported by many^{6,13,14,20,23,24} to be a population at risk for complications, although advances in anesthesiology and surgical techniques have led to improved outcomes in this group following major surgery. Moreover, the use of a minimally invasive approach is particularly interesting as the reduced surgical trauma can potentially lead to a reduction in postoperative complications²⁵ especially among elderly individuals.²⁶ Likewise, we have recently shown that robotic general surgery and more specifically robotic pancreaticoduodenectomy can be performed safely in patients older than 70 years.^8,11

Yet, in the present study, using the univariate analysis, age greater than 70 years was found to be a risk factor for increased morbidity and mortality. As a result, it is believed that elderly patients should be considered as a population at risk for complications, even if the multivariate analysis failed to demonstrate an independent relationship.

Overweight patients have typically been associated with an augmented risk for surgical complications.^27,28 Interestingly, our series showed a BMI of less than 30 to be an independent factor for morbidity, even after multivariate analysis. This can be explained by at least 2 reasons. First, the malignant patients (who were found to be at risk for complications) often have a poor nutritional status, thus a lower BMI.²⁹ In our series, the mean BMI for malignant patients was 28.8 (vs 33.5 for benign cases; P < .001). In fact, poor nutritional status³⁰ and rapid weight loss^6,13 were both associated with an increased risk for complications. Second, of the 451 robotic procedures for patients with a BMI of 30 or greater, 151 were bariatric (33.5%) with only 4 complications. Thus, there could be a bias since this subgroup had relatively few complications.

The cardiovascular and hypertensive comorbidities have previously been reported as risk factors associated with postoperative complications.^24,31 The cardiac risk event has been stratified using a different risk index.^4,32 The Goldman Cardiac Risk Index includes several factors and most of them are similar to those reported in our study. Thus, as we have seen in the present series, the cardiac risk stratification is important even for a minimally invasive procedure.

Lastly, it is worthwhile to mention that a malignant pathology was strongly associated with an increased risk for morbidity, as confirmed by the multivariate analysis. The general condition of these patients is typically poor and associated with a bad nutritional and immune status.³³ The relationship between malignant pathology and complications has previously been reported by several groups.^12,34

Several risk factors can be associated with the quality of surgical performance. First, blood loss is recognized as a good indicator of the surgical technique as well as the experience of the surgeon.³⁵ Several studies have shown that increased blood loss was associated with an augmented morbidity or mortality^16,18 and the same connection can be made with the need for transfusions.^19,23 Perioperative bleeding is often an indicator of the surgical difficulty and complexity of the procedure. In our review, more blood loss was seen in the advanced group. It can also be an indicator of the surgical technique used and, in fact, has been confirmed as an advantage of not only the minimally invasive approach,³⁶ but the robotic approach as well.³⁷

The experience of the surgical team is another factor to consider. A low volume hospital⁵ or a low surgeon case volume²² is related to a higher mortality rate.^14,15,38 At our center, we perform more than 270 robotic procedures per year, thus our caseload corresponds to a high-volume center. Our mortality rate (0.5%) is also similar to those reported previously for an open high-volume hospital.⁵

To categorize the surgical complexity, we decided to modify our own classification system¹¹ because other classification systems were not satisfactory for our use. For example, using the Hoehn classification,³ most robotic procedures would have been categorized as advanced interventions. Conversely, Klotz et al¹² proposed dividing the surgical complexity into different groups. While practical, this classification was proposed in 1996 and did not correspond anymore to our present activity. Currently, almost all of the abdominal procedures, even the most complex, can be performed using a minimally invasive robotic approach.^8-10,39 Lastly, Kaafarani et al⁴⁰ adapted the classification from the American College of Cardiology and the American Heart Association,⁷ which divided the procedures into expected cardiac risk. The high-risk group included major and emergent procedures, particularly in elderly individuals, as well as prolonged operations associated with large fluid shift and/or blood loss. While we rejoined this classification particularly for the advanced procedures, we did not focus solely on postoperative cardiac complications.

It is clear that patients who underwent an advanced procedure were more fragile than the other groups for reasons of age, higher ASA score, more comorbidity, more malignant pathology, etc. This can in part explain the higher complication rates as well. Still, the multivariate analysis found advanced procedures as a clear and independent risk. This should be taken into consideration when planning a complex robotic case.

Interestingly, the Estimation of Physiologic Ability and Surgical Stress system incorporates the presence of surgical stress and includes the blood loss (divided by the body weight), the operative time, and the extent of skin incision as risk factors.⁵ In fact, the surgical stress can be largely reduced by the minimally invasive approach.⁵ Still others found that the minimally invasive approach resulted in less inflammation response⁴¹ and decreased blood loss^36,37 when compared with the open procedure.

One challenge that raised concern was the robotic multiquadrant procedure.^42,43 During these procedures, the need to move the robotic cart and/or the trocars can lead to an increased operative time. In our study, we also showed that a multiquadrant intervention was associated with an augmented risk for complications. This scenario is unique to robotic surgery and can often be seen during colorectal resections or when multiple or bilateral procedures are performed during the same operation. To date, this risk factor has been rarely reported.⁴²

It is interesting to note that when considering surgical stress, the conversion rate was not associated with an increased risk for postoperative complications. Still, and not surprisingly, the conversion rate was associated with the intraoperative complications, with 3 procedures leading to an open conversion. The 4 remaining intraoperative complications were managed robotically.

We reported few specific complications for robotics. The multiquadrant approach and the advanced procedure were both found to be significantly associated with an increased risk for complications. Otherwise, the remaining risk factors were similar to those reported for open and laparoscopic surgery.

It is hoped that during the next few years, we will be able to develop a prospective scoring system to include the different factors that are significantly associated with morbidity. The existing scores are too complex and not user friendly for this purpose. The Acute Physiology and Chronic Health Evaluation II, the Physiological and Operative Severity Score for the Enumeration of Mortality and Morbidity, and the Estimation of Physiologic Ability and Surgical Stress do not fit the specific needs of the surgeon and were shown to be inadequate in predicting the postoperative complications and outcomes.⁴

While bringing new and encouraging results, our study has some limitations that deserve comment. First, it is not a prospective study. It is a large retrospective series of procedures performed by various experienced surgeons at a single institution. Yet, the data are unique and would allow for a prospective scoring system specific to robotic general surgery. Another limitation was that because the mortality rate was very low, there was not enough power to run multivariate analysis for this outcome. This is a limitation for the assessment of the risk factors associated with a fatal issue. However, the univariate analysis did show several potential risk factors, most of which have been reported before.

Finally, one can argue that the classification or subdivision of the type of procedure is quite subjective. We agree that a classification should take into consideration only objective parameters. As a result, and to avoid the risk of interobserver differences, we decided by consensus agreement to use the complexity of the procedure (dissection and anastomosis) as a classification.

To our knowledge, this is one of the largest series reporting morbidity and mortality following robotic general surgery. The robotic approach for general surgery can be achieved safely with low morbidity and mortality and several risk factors have been identified as independent factors for morbidity and mortality. These can be used to identify patients at risk before and during the surgery and, in the future, to develop a scoring system for the use of robotic general surgery.

Correspondence: Pier C. Giulianotti, MD, Division of General, Minimally Invasive, and Robotic Surgery, University of Illinois at Chicago, 840 S Wood St, Ste 435 E, Chicago, IL 60612 (piercg@uic.edu).

Accepted for Publication: January 18, 2012.

Published Online: April 16, 2012. doi:10.1001/archsurg.2012.496

Author Contributions:Study concept and design: Buchs, Bianco, Ayloo, Elli, and Giulianotti. Acquisition of data: Buchs, Addeo, and Gorodner. Analysis and interpretation of data: Buchs, Addeo, Oberholzer, and Benedetti. Drafting of the manuscript: Buchs, Addeo, and Ayloo. Critical revision of the manuscript for important intellectual content: Addeo, Bianco, Gorodner, Ayloo, Elli, Oberholzer, Benedetti, and Giulianotti. Statistical analysis: Buchs and Addeo. Administrative, technical, and material support: Bianco, Gorodner, and Giulianotti. Study supervision: Addeo, Ayloo, Elli, Oberholzer, Benedetti, and Giulianotti.

Previous Presentation: This study was presented at the 2nd Worldwide Meeting of Clinical Robotic Surgery Association; October 2, 2010; Chicago, Illinois.

Additional Contribution: We thank Michelle Secic from Secic Statistical Consulting Inc for the statistical work.