Objective performance indicators differ in obese and nonobese patients during robotic proctectomy.

The protocol of this study was approved by the institutional review board (IRB) of Emory University (IRB #00111214). Across 39 RPs, thirteen patients were obese and twenty-six were non-obese. Patients with body mass index (BMI) of ≥ 30 kg/m 2 were classified as obese and those with a BMI of less than 30 kg/m 2 were considered non-obese. 9 39 RPs performed between January, 2020 and May, 2021 were included in this study. Endoscopic videos synchronized to robotic data streams were captured using Intuitive Data Recorders (Intuitive Surgical, Sunnyvale, CA). The data was then directly transmitted to research team members within the Intuitive Surgical Research Division via an internet connection. No identifiable personal health information was included in the data transfer. Trained video reviewers annotated each recorded video to delineate specific surgical steps utilizing a procedure-specific annotation card. OPIs were computed for each surgical step, and a step-specific OPI metrics report was sent to the research team at Emory University. Following this, the OPI data was paired with patient and surgeon information and underwent analysis by clinicians, clinical researchers, and biostatisticians at Emory. For each step of the procedure, the annotation card includes a description of the step and start/stop times, with 21 surgical steps possible during RP. Examples of these steps include medial-to-lateral left colon mobilization (LCM), inferior mesenteric artery dissection and posterior mesorectal dissection. Across procedures, surgeons may skip a step, complete them in one visit, or visit a step multiple times in order to complete, depending on the procedural workflow and case requirements. For instance, surgeons may skip splenic flexure mobilization if they can achieve adequate reach for anastomosis without this step. The following critical steps during RP were analyzed: splenic flexure mobilization (SFM); medial-to-lateral LCM; anterior, posterior, right and left pelvic dissection (PD) with total mesorectal excision; and rectal transection. OPIs can be classified into three descriptive categories: kinematic metrics (velocity, acceleration, and jerk), event metrics (task completion time, energy activation time, camera use, clutch use, and arm swaps) and wrist articulation metrics (roll, pitch, yaw). Alternatively, OPIs can be classified based on whether they relate to the patient cart or surgeon console. Patient cart metrics include OPIs for camera, dominant and non-dominant arms (including the 3 rd arm) and include moving times, path lengths, velocity, acceleration and jerk. Surgeon console metrics include surgeon hand economy of motion, workspace volume and wrist movements (roll, pitch, yaw). For each step, we captured approximately 150 OPIs. Wrist articulation metrics are reported in radians (rad), while other metrics are reported in meters (m), meters per second (m/sec), meters per second squared (m/sec 2 ), meters per second cubed (m/sec 3 ), and cubic meters (m 3 ), as applicable. Console surgeons were classified into two groups: expert surgeons (two senior attending surgeons) and intermediate surgeons (three junior attending surgeons). Expert surgeons completed over 500 robotic colorectal procedures, lead training courses in robotic colorectal surgery, are recognized by peers for their robotic surgery expertise, and are actively involved in training residents, fellows, and junior attendings in robotic colorectal surgery. Intermediate surgeons had performed 50–150 robotic colorectal operations. None of the surgeons in this study were classified as novice (i.e., < 50 robotic colorectal procedures) This classification method aligns with those used in previous research. 7 , 10 , 11 Statistical analyses were conducted using SAS Version 9.4 and SAS macros. 12 Descriptive statistics were utilized to summarize the covariates. Categorical variables were presented with counts and percentages and the numeric variables with median and interquartile range (IQR) or mean and standard deviation (SD), as appropriate. The comparison of numeric variables was performed using the ANOVA or the Kruskal Wallis test, as appropriate based on the Kolmogorov-Smirnov test for normality. Categorical variables were compared using the Chi-square test or Fisher’s exact test as appropriate. Statistical significance was assessed at the 0.05 alpha level. Data management and statistical analyses were performed by biostatisticians at Emory University.

Indications for the 39 RPs included cancer (29), diverticulitis (5), polyp (2), Crohn’s disease (1), endometriosis (1) and rectal prolapse (1). Median age of the cohort was 61 [IQR: 49 – 67] years, with the majority (23, 59%) being female. Median American Society of Anesthesiologist (ASA) score was 3 [IQR: 2 – 3]. Median BMI was significantly different across both groups, 36.6 [IQR: 34.1 – 40.3] kg/m 2 in obese vs 24.0 [IQR: 21.0 – 26.3] kg/m 2 in non-obese. All other demographic and surgical parameters were well matched between both groups ( Table 1 ). Expert surgeons performed the majority of surgery in both groups (92.3% in obese patients vs. 65.4% in non-obese patients, p = 0.120). Across 39 RRCs, a total of 1880 surgical steps were performed. When only the critical steps were included, 1166 steps were available for analysis. Of the 1166 steps analyzed, 146 (12.6%) were categorized as SFM, 111 (9.5%) as LCM, 132 (11.3%) as anterior PD, 306 (26.2%) as posterior PD, 258 (22.1%) as right PD, 163 (14.0%) as left PD and 50 (4.3%) as rectal transection. The 1166 steps included 174,900 corresponding OPI data points. The distribution of these steps across obese and non-obese patients is illustrated in Figure 1 . Time OPIs differed across obese and non-obese patients ( Table 2 ). Median time for step completion was significantly longer in obese patients during LCM (165.9 vs 96.4 sec; p = 0.008) and rectal transection (368.7 vs 135.1 sec; p = 0.014). Surgeons exhibited significantly longer moving times in obese patients for camera (11.8 vs 3.6 sec; p = 0.006) and dominant arm (158.8 vs 91.9 sec; p = 0.029) during LCM. Additionally, camera moving times were found to be higher in obese patients during anterior PD (3.1 vs 0.9 sec; p = 0.009) and right PD (1.3 vs 0.6 sec; p = 0.004). Energy activation duration was higher in obese patients during anterior PD (2.0 vs 1.2 sec; p = 0.015) and left PD (2.8 vs 1.9 sec; p = 0.022). Event OPIs also differed across the patient groups ( Table 3 ). In obese patients, surgeons utilized significantly more camera movements during LCM (9 vs 3; p = < 0.001), anterior PD (4 vs 2; p = 0.009), right PD (2 vs 1; p = 0.002) and rectal transection (15 vs 5; p = 0.016). Additionally, surgeons utilized more arm swaps in obese patients during SFM (1 vs 0; p = 0.024) and LCM (2 vs 0; p < 0.001). Higher utilization of both master clutch (8 vs 2.5; p = 0.023) and arm swaps (2 vs 0; p = 0.027) was noted in obese patients during rectal transection. Numerous kinematic OPIs differed in obese compared to non-obese patients ( Table 4 and 5 ). Longer camera path length was during the following steps: SFM (0.27 vs 0.16 m; p = 0.019), LCM (0.25 Vs 0.08 m; p < 0.001), anterior PD (0.08 vs 0.03 m; p = 0.018) and right PD (0.03 vs 0.01 m; p = 0.019). Surgeons also utilized a significantly longer path length for dominant arm (3.86 vs 1.73 m; p = 0.035) and 3 rd arm (0.23 vs 0 m; p = 0.034) during LCM. Significantly longer path length for dominant arm was seen during rectal transection (2.87 vs 1.23 m; p = 0.008) and for 3rd arm during SFM (0.03 vs 0 m; p = 0.019). Camera velocity was found to be significantly higher during SFM (0.09 vs 0.08 m/sec; p = 0.049), LCM (0.10 vs 0.06 m/sec; p < 0.001) and right PD (0.06 vs 0.03 m/sec; p = 0.033). Surgeons exhibited lower non-dominant arm velocity during posterior PD (0.06 vs 0.07 m/sec; p = 0.015), but higher 3rd arm velocity during LCM (0.06 vs 0 m/sec; p < 0.001). Surgeons used greater acceleration and jerk for both camera and 3 rd arm during LCM. In contrast, during posterior PD, surgeons exhibited lesser non-dominant arm acceleration (0.34 vs 0.41 m/sec 2 ; p = 0.022) and jerk (17.35 vs 20.06 m/sec 3 ; p = 0.020). However, non-dominant arm jerk was higher during rectal transection (28.56 vs 21.27 m/sec 3 ; p = 0.004). Dominant arm wrist articulation OPIs also differed between the two groups ( Table 6 ). During LCM, obese patients required greater yaw (33.01 vs 17.02; p = 0.029) and roll (47.86 vs 25.16 rad; p = 0.032). Whereas, during rectal transection in obese patients, surgeons utilized significantly increased yaw (48.81 vs 28.03 rad; p = 0.026), roll (72.27 vs 45.17 rad; p = 0.039) and pitch (49.83 vs 24.55 rad; p = 0.022).

Assessment of surgery complexity is critical for obtaining insights into variability in surgical workflow and correlating with patient outcomes. Traditionally, subjective methods have been employed to describe complex procedures. The operative report dictated by the surgeon is the most relied upon record of intraoperative events, despite its inherent subjectivity and bias. 13 Surgeons use phrases such as “extremely difficult” or “more complicated than usual” to describe surgery in obese patients. Operative reports have ramifications on post-operative expectations and care decisions, operating room logistics planning, resource allocation, and decision-making by surgeons involved in the same patient’s care at a future time. By utilizing OPIs, we hope to move beyond subjective descriptors and instead employ objective, quantifiable metrics to convey surgery complexity. One can imagine a future operative report which contains OPIs that fall 2 or 3 standard deviations away from known normal values, conveying a more complicated procedure. Recently, Kaoukabani, et al. evaluated step-specific OPIs in the context of case complexity and clinical outcomes across 87 robotic cholecystectomy cases. 4 Prior to our work, no studies exist correlating OPIs and surgery complexity within the field of colorectal surgery. In this feasibility study, we examined variations in OPIs between obese and non-obese patients undergoing RP. Our findings suggest statistical differences in OPIs, although the clinical significance of these variances requires careful interpretation. Future research will focus on elucidating the magnitude and directionality of OPI variances and determining which differences are clinically meaningful. The median time for step completion was significantly longer in obese patients during LCM and rectal transection. Prolonged times can be attributed to factors such as bulky mesentery, heavier organs, impaired visualization of vasculature, more difficulty with retraction, and more limited working space. 1 Zilberman, et al. found that higher BMI was associated with longer durations to complete certain surgical steps (preparation, nerve-sparing dissection, dorsal-vein complex control, anastomosis) during robotic prostatectomy. 14 Our findings revealed notably longer instrument moving times in obese patients across various steps. This may be due to more frequent adjustments and re-adjustments to effectively retract and maintain adequate exposure. Median energy activation duration was also significantly higher in obese patients during anterior PD and left PD. Surgeons employed significantly more camera movements and arm swaps in obese patients during mobilization and dissection steps of RP. Tousignant et al. found patients with elevated BMI necessitate more frequent camera adjustments during a surgical step to maintain a consistently safe view during bariatric surgery. 15 In other work, Kaoukabani et al. identified arm swap frequencies as an indicator of complexity during robotic cholecystectomy. 4 Prior studies have recognized camera control behaviors as crucial indicators of an operation’s success. 16 –18 Our findings revealed increased utilization of the master clutch and more arm swaps during rectal transection. This might be expected during RP in obese patients as the surgeon may have to use adjustments more often to maintain exposure when dealing with heavier, thicker tissues in a confined space. Numerous kinematic OPIs varied between obese and non-obese patients. Specifically, obese patients required greater camera path length, velocity, acceleration, and jerk across most steps. As stated above, higher camera kinematics would make sense in obese patients, as the surgeon might require more dynamic camera movements to achieve and maintain effective fields of view. Hung, et al. discovered a correlation between instrument kinematics and clinical outcomes during robotic radical prostatectomy. 19 However, a possible correlation with procedure complexity had not yet been explored prior to our work. Our findings indicated that obese patients required increased dominant wrist articulation during LCM and rectal transection. Greater wrist articulation may enable surgeons to achieve better leverage, safely, while making progress in narrowed workspaces. In urology, wrist articulation has been identified as a top-ranking OPI predictive of early urinary continence after robotic radical prostatectomy. 19 No studies prior to this have identified a possible relationship between wrist articulation and procedure complexity. There are limitations to this study. The small sample size of cases decreases its statistical power. Although we collected over 250,000 OPI data points for 39 RPs, our data is affected by inherent clustering, with OPIs nested within patients and patients nested within surgeons. We did not perform inferential modeling; instead, we present a pilot study exploring differences in OPIs between obese and non-obese patients. Similarly, we only examined surgery performed by five surgeons, two who are considered expert robotic colorectal surgeons and three who are classified as intermediate expertise. Although we saw no significant differences in level of expertise for procedures performed on obese and non-obese patients, the limited number of surgeons may confound our results. In future work, we will address layers of clustering by including a greater number of surgeons and their characteristics while also accounting for the nested structures within the data. Similarly, we could not analyze for presence of effect modification by surgical indication, but our descriptive comparisons revealed no significant differences in indications between obese and non-obese patients. Again, future studies with larger numbers of procedures are planned to address these limitations. Another limitation is that we only analyzed critical RP steps. These steps were chosen based on ongoing work identifying them as most significantly associated with procedure workflow, console time and total operative time. In parallel work, we are evaluating whole procedure OPI data for comparison. In our current analysis, we have employed only one cutoff point for defining obesity. In future studies, we plan to use a more nuanced assessment of central obesity (i.e., measurement of visceral adipose tissue). Other unmeasured non-automated metrics (i.e., tissue trauma, needle entering angle, and force sensitivity) may have influenced our findings. In the future, overcoming these limitations will play a pivotal role in unlocking the true potential of OPIs for colorectal surgery. Current literature on OPIs highlights a significant research gap in the field of colorectal surgery. Due to the specific steps involved in colorectal procedures, findings from studies on other surgical areas cannot simply be applied, underscoring the necessity for focused investigation in this specific domain. The prevalent use of subjective assessment tools to evaluate surgical complexity accentuates the need for objective tools to mitigate inherent biases. This shift toward objective methods may also facilitate progress towards synoptic reporting.

This work identifies step-specific OPIs that significantly differ between obese and non-obese during specific steps of RP. This is the first study to use OPIs to correlate a patient’s attribute with surgeon movements and robotic arm kinematics during robotic colorectal surgery. By leveraging the power of OPIs, we aim to lay the groundwork for understanding how varied patients and pathologies require different surgeon techniques in order to achieve excellent outcomes.

Pelvic surgery is perceived to be more complex in obese patients. Although the robotic platform offers several advantages (3D visualization, wristed instruments, tremor dampening), the heavier tissues, thicker mesentery, and larger organs in obese patients present unique technical challenges. 1 Despite surgeons’ subjective experiences, there remains a scarcity of objective intraoperative data establishing obesity as a predictor of higher surgical complexity. Currently, procedure details are classified according to the current procedural terminology (CPT) code and work relative value units (RVUs). 2 Although surgeons frequently employ qualifiers (i.e., modifier 22) for complex robotic colorectal operations, there is no specific qualifier for patients with elevated BMI. 3 These measures were originally created for reimbursement purposes and lack the descriptors necessary to accurately reflect higher surgical complexity due to obesity. 2 In order to support efforts of more accurate reporting of intraoperative workflow and events, there is a need for more objective intraoperative data relating to surgery complexity. Robotic surgery, in which computer hardware and software is situated between the patient and surgeon, allows a unique opportunity to capture machine-learning-enabled metrics, which are referred to as objective performance indicators (OPIs) . OPIs include surgeon console (hand, wrist and foot movements, energy and master clutch use, third arm swaps) and patient cart (camera movements, instrument path length, moving time, velocity, acceleration, jerk) metrics, as well as other calculated metrics (smoothness, economy of motion, workspace volume). OPIs have been utilized across various specialties to provide step-specific insights into surgical workflow, surgeon skill and surgical outcomes. 4 – 8 There is scarce evidence linking OPIs and surgical complexity during live surgery, and no studies on OPIs and surgical complexity for colorectal robotic surgeries. The aim of this feasibility study is to identify OPIs that differ between obese and non-obese patients during specific steps of robotic proctectomy (RP). In future work, we intend to determine which differences in OPIs are clinically significant, potentially offering automated, scalable data that can reflect operative complexity and correlate variance in surgeon techniques with individual patient attributes. This marks the first attempt to explore meaningful OPIs that correlate with case complexity at a step-specific level