PREDICTING SURGEON’S NON-TECHNICAL SKILLS THROUGH COMMUNICATION ANALYSIS IN ROBOTIC-ASSISTED SURGERY

Surgical operations assisted by robots differ from traditional human surgical procedures by combining robotic elements with human performance that introduce greater complexity. The rapid evolution of these surgical procedures demands surgeons to apply both clinical expertise together with non-technical skills (NTS) which scientists need to study because these skills remain underappreciated. Preliminary studies have limitations because generalizability is restricted, procedures are simulated, and the analysis framework needs improvement to include a broader range of communication characteristics.

The primary goal of this research work is to analyze communication features (frequency and occurrence) such as closed-loop communication, questioning, task-focused dialogues, speech speed, and pauses during communication among healthcare professionals in the operating room. The research also aims at examining the connection between communication features and healthcare providers’ (mostly surgeons’) NOTSS (Non-Technical Skills for Surgeons) performance in robotic settings and the goal is to identify objective predictors of NOTSS.

A total of 13 surgeons from general surgery, bariatric surgery, obstetrics and gynecology, urology and colorectal surgery who perform robotic assisted surgery were recruited from the Indiana University Health. A total of 36 robotic-assisted surgeries were collected. The attending surgeons were fitted with microphones and eye-tracking devices to record audio from their surroundings and video from their field of view (note that eye-tracking was only used for selected portion of the procedure). A team of 4 to 5 experts who were trained in evaluating surgeons' non-technical skills using the NOTSS training materials and experienced experts, provided consensus-based NOTSS ratings for each of the 36 surgeries, which served as outcome measures for NTS prediction.

The ability of non-technical skills (NTS) prediction was investigated from different angles. The first and second hypotheses supported that the prevalence and frequency of the CLC components except call-outs were positively related to overall NOTSS scores and decision-making performance. This underlines the important role of CLC in healthcare communication. In contrast, excessive call-outs were negatively correlated with decision-making, which suggests that frequent call-outs might be associated with poor surgical decision-making. Additionally, increased questioning tones during surgery were negatively correlated with situation-awareness.

The third and fourth hypotheses were tested with five composite communication features which were found to be positively correlated with NOTSS scores across overall, decision-making, and leadership dimensions. Surgeons who clearly responded to team members and assistants who actively engaged in CLC, and task-focused communication were associated with better non-technical skills. Moreover, a faster speech rate by surgeons was positively correlated with higher decision-making performance.

For the fifth hypothesis, logistic regression identified key predictors of NTS performance: A higher rate of surgeon check-backs per minute significantly increased the likelihood of having exemplary overall NOTSS scores, whereas a greater number of case-related non-questions by surgeons and more pauses by non-surgeons were associated with reduced decision-making performance. Additionally, a higher rate of assistant check-backs per minute was associated with improved leadership scores.

This thesis established that there are both positive and negative associations between the content-based and periodicity-based communication features and the non-technical skill performance of attending surgeons in robotic-assisted surgeries. The communication features derived from surgical conversations were both manually coded and automatically extracted to capture the dynamics of interactions among surgical team members. Exemplary NOTSS scores, defined by a study-specific threshold, were successfully classified most of the time, thus enabling objective assessment of surgeons’ non-technical performance. In addition, the study was able to classify some of the NOTSS subscales, with different communication features emerging as the most representative predictors for each subscale, compared to those for the overall NOTSS score.