TT interventions will be built based on teamwork training and nontechnical skills frameworks and will support the skills needed for teams to address RAS-specific communication challenges. The TT approach will consist of a TeamSTEPPS [24,25] driven training package (4- to 6-hour meeting for surgeons and anesthesiologists via small group teaching and successive 1-hour meetings for OR staff) complemented by on the spot coaching by human factors experts to offer reminders and encouragement.

TD interventions will focus on specifying, ordering, and allocating tasks to specific roles to improve efficiency, visibility, and reliability [26]. A previously performed failure modes and effects analysis [27,28] will be used to prioritize tasks for redesign. The Systems Engineering Initiative for Patient Safety model [21] will be used to determine a human-centered systems model of each task, and task analysis will be used to define roles, sequences, and allocation. Finally, we will practice and refine these redesigns using in situ simulation trials.

WD involves proposing and implementing new OR layout configurations to improve the use of space in RAS. OR layouts will be configured to ensure (1) the surgeon can see the patient and the team from the console, (2) the team can see the surgeon, (3) staff can move freely in the room, (4) robot docking can occur from multiple angles, (5) minimize cable tensions and trip hazards, and (6) optimization of OR equipment preparation and instrument storage. Key movement-oriented tasks will be used to plot ideal movement paths on existing room layouts, and new layouts will be proposed and tested to reduce unnecessary movement and disruption.

Given the close interactions between technology, tasks, teamwork, and process [29-32], we hypothesize that multiple interventions will function synergistically. Teamwork benefits from visual cues, sightlines, and face-to-face communication [33,34]; TD benefits from improved teamwork to allow better coordination of complex, interdependent tasks [29,31,35]; and better efficacy of teamwork-related checklists [16], improved equipment storage, and visibility through better WD allows for improved task performance [30,36,37]. Thus, this study is designed to specifically test each of these interactions.

This design allows for sufficient implementation and sampling of the interventions, introduces individual components of an overall improvement strategy, and evaluates how each change contributes to a larger whole (Table 1).

Using multiple regression with 10 predictor variables (4 sites, 5 data collection phases, and 3 interventions and 1 baseline period) and assuming a normal distribution, 40 observations per site per time per intervention will provide at least 80% power to find a statistically significant effect of the intervention on surgery duration. Achieving this level of statistical power remains possible with 23 observations per phase per site, making our planned sample of 40 robust, should data collection be more challenging than anticipated.

Contextual covariates of known influence include patient details (age, sex, BMI, American Society of Anesthesiologists [ASA] Physical Status classification), surgery details (procedure description, procedure category, date, robot model (S, Si, Xi), hospital, OR number, approximate room size), and personnel details (number of surgical trainees, OR staff trainees [circulating nurses and surgical technicians], and anesthesia trainees). These covariates will be collected during the observations and/or retrospectively from the patient’s electronic health record. We will also record operative and in-room time, intraoperative complications, blood loss, conversion to open surgery (which requires undocking the robot and making an abdominal incision), and returns to the OR via hospital electronic records for each intervention period.

Deviations from the natural progression of a task (ie, flow disruptions [FDs]) [39] were collected throughout all phases of each operation. Data collection includes a brief description of the event observed, time of occurrence, major category, and severity. FDs will be assigned a category and severity score during observation. With respect to classification, each FD will be assigned one of 8 possible categories: communication, coordination, equipment, training, external factors, environment, patient factors, and surgical task considerations (Table 3) based on an adapted taxonomy developed by Catchpole et al [40]. Minor categories may be developed for a more granular analysis of the data following data collection. FDs will also be assigned a severity score, ranging from 0 to 2: (0: potential disruption to the process, 1: disruption to the process, and 2: increased patient safety risk).

Each RAS procedure will be evaluated throughout 5 distinct surgical phases: (1) wheels in until incision, (2) incision to the surgeon on console (including the docking process), (3) surgeon on console to surgeon off console, (4) surgeon off console to patient closure, and (5) patient closure to wheels out. The duration of each phase will be recorded by the observers during data collection.

The Oxford NOTECHS 2 [41] rating system will be used to evaluate the nontechnical skills of the OR team. The scale includes 4 dimensions—leadership and management, teamwork and cooperation, problem-solving and decision-making, and situation awareness—rated on an 8-point scale. Observers will record NOTECHS ratings for each team member during the case.

The extent to which interventions are fully used following implementation will be assessed using the intervention adherence metric [42-44], a metric developed based on the developed interventions. It will consist of a series of observational scores (Likert and check boxes) that will be deployed during each surgical observation to evaluate the use of interventions, based on observable components of each intervention. This will be deployed uniformly at baseline and all intervention phases, allowing us to understand the use of intervention during each operation.

Subjective workload ratings will be obtained using the SURG-TLX (Task Load Index) [42]. This visual-analog workload measure asks each surgical team member to select a score from 1 to 20 on 6 parameters: mental demands, physical demands, temporal demands, task complexity, situational stress, and distractions, which are then aggregated and rescaled to generate a workload score between 0 and 100.

Safety culture will be assessed using the Safety Attitudes Questionnaire (SAQ) [45], which has been extensively used for nearly 2 decades. The teamwork subscale has been sensitive to teamwork interventions [46], whereas the perceptions of management subscale has identified barriers to such interventions [43]. This will be administered via REDCap (Research Electronic Data Capture, a web-based Health Insurance Portability and Accountability Act–compliant survey platform) [38] in the last 2 weeks of each data collection phase to all staff involved in robotic surgery during that trial period. Subanalysis via surgical specialty and specific operations performed will allow us to track subtle changes over time.

To gauge team members’ responses to the interventions, we will administer the concurrent acceptability [44] measure (7 items, 5-point Likert scale) to all involved staff after the first month and at the end of the last month of each intervention phase (estimate 30-50 staff per site per phase). The measure is based on the Theoretical Framework of Acceptability model (version 2), which reflects the extent to which people deliver or receive a health systems intervention consider it to be appropriate based on anticipated or experiential cognitive and emotional responses to the intervention.

Ensuring observers are effectively trained to perceive FDs and collect data on teamwork above the noise of otherwise normal system function is a critical requirement for this study [23]. During the prebaseline phase, observers will receive extensive training that includes initial classroom instruction (human factors and FD classification frameworks) and practice and familiarization (eg, identification of OR team members and the components of the operating room environment) in the OR with 2 human factors researchers with extensive experience with direct observation and FD measurement in surgery. Observers will be trained to understand the basic steps for each surgery type and are familiarized with the surgical subspecialties and components of the surgical robot. Trainees will also be provided with relevant reading material on FDs, NOTECHs, and RAS and given an example of a completed data collection tool.

Familiarization observations will take place across 3 stages: (1) orientation to the OR, (2) practice observations, and (3) simultaneous observation of interrater reliability (Table 4). Weekly meetings, including the observers and principal investigators, will be initiated to combat drift and allow observers to review their observations with the team.

Before conducting observations, 15-minute in-services will be conducted with the staff on each unit at each study site to explain the research, introduce them to the research team, and allow them to ask questions and express their concerns. In-services will be led by a human factors expert and surgeon team member(s). Furthermore, an information sheet will be provided to staff to educate them about the purpose of the study and provide contact information for members of the study team whether they have any questions.

A total of 4 trained human factors researchers will observe 680 RAS cases over the course of the study period. For MUSC, CSMC, and MDRH, observers will capture 40 cases during each of the 5 data collection phases (phases 2-6, each 8 months in duration). For VCMC, 40 cases will be captured in each of 2 phases: baseline (which spans across phases 2-4) and intervention (phases 5 and 6; Table 5).

Observers will collect FDs, NOTECH ratings, and all relevant case-related covariates, including patient details (age, sex, BMI, ASA classification), surgery details (procedure description, date, hospital, OR number, room size), personnel details (number of surgical trainees by type, OR staff trainees, and anesthesia trainees by type), and robot details (S, Si, or Xi model). During the intervention phases, the intervention adherence metric will also be collected during each surgical observation to evaluate the use of interventions.

Field notes will also be collected monthly by the observers at each of the 4 sites. Field notes generally consist of 2 parts: descriptive and reflective information. Descriptive information attempts to accurately document factual data (eg, date and time) and the settings, actions, behaviors, and conversations observed. Reflective information documents your thoughts, ideas, questions, and concerns as you are conducting the observation. These notes will provide additional context for the implementation of the intervention using the Consolidated Framework for Implementation Research (CFIR) [47].

Data will be collected in the OR using Microsoft Surface Pro 6 tablets. Urban Armor Gear Hand Strap & Shoulder Strap Military Drop Tested Cases are also used to provide ergonomic support and handling of tablets for observers standing or seated on stools for long period. XCOREsion 15-45 by J-Go Tech Microsoft Surface Portable Chargers were given to each observer to provide external battery life when collecting data over 2 or more consecutive cases with no opportunity to charge their tablets between cases.

The SURG-TLX will be collected in person during direct observation and will be administered on a Microsoft Excel form located on the observer’s Microsoft Surface Pro tablets. The SAQ and concurrent acceptability will each be collected via a REDCap survey emailed to surgeons and OR staff.

We will use multivariable regression models to explore the relationship between the covariates (ie, site, specialty, BMI, and teamwork) and process measures (ie, FDs and durations), examining how these relationships are modified by interventions. The following are the specific questions we seek to answer: (1) What interventions are used (intervention adherence metric)? (2) Did OR staff like the interventions (concurrent acceptability)? (3) Did the interventions change attitudes (SAQ)? (4) Did the interventions change individual workload (TLX) and/or improve teamwork (NOTECHS)? (5) Did the interventions result in a better process (FD)? and (6) Did the interventions reduce surgical durations and/or blood loss and/or OR returns? Statistical analysis will be conducted using the R programming language (R CORE TEAM, version 3.5.2) and assessed at the significance level of α .05.

Direct observation of surgical processes may be useful in modeling adverse event causation by looking at the concatenation of smaller, seemingly innocuous errors to larger, more clinically serious situations [48-50]. The primary purpose of the proposed analyses is to develop a quantitative framework that allows for the evaluation of the snowball hypothesis. The rationale behind this hypothesis is that accidents and injuries arise from the sequence of multiple, frequently occurring individual errors. Adverse outcomes can be seen both as the unlucky coincidence of multiple randomly occurring errors and/or as a causative chain of events where one error leads to the next, creating an error cascade (or snowball). A range of exploratory Markov chain, Poisson process, and changepoint modeling techniques will be applied with the R programming language to analyze data across more than 1000 procedures and identify error causation mechanisms as random coincidences or as a deterministic error cascade. This mode of analysis aims to be a profound advance, not solely in understanding and addressing surgical complications and adverse events but in the entire way in which accidents are viewed.

An inductive and deductive thematic content analysis approach will be used to analyze the qualitative data [51]. In total, 2 research team members will individually code all interviews and field notes, first in a deductive pass, using a codebook to tag segments of text specific to the CFIR constructs and systems model concepts. Next, an inductive pass through the data will identify new concepts to develop themes that have not been previously identified. Codes will identify causes, explanations, relationships, patterns, and themes related to the implementation of new RAS workflows. After iterative analyses, the 2 coders will immerse and crystallize [52] the final set of themes, confirm these findings with the research team, and develop a case study of comprehensive user experiences that promote successful implementation of RAS.