Anne van Dalen

120 I Chapter 4 METHODS Subjects and Setting of the Intervention In this cross-sectional study, a consecutive sample of 35 consented adults (>18 years) patients who underwent general laparoscopic surgery between February 2017 and January 2018 was used. 28 Their surgeries were recorded using an MDR for the purpose of generating and researching the MDR outcome report to be used in team debriefing. Seven standard abdominal laparoscopic procedures were selected and performed by 4 staff surgeons and 1 surgical fellow, working at one tertiary academic medical center. The OR team was completed with anesthesiologists, surgical- and anesthesiology residents, medical students and OR nurses, in various constellations. Medical students assisted during the procedures by, for example, holding instruments. Cases were only recorded and included if every team member had given their formal written informed consent to use an MDR for the purpose of the study, prior to start of the procedure. 28 29 The project was approved by theWorks Council andHospital Board of Directors. To address all ethical considerations, a Privacy Impact Assessment (PIA) was conducted. This made sure the study was compliant with all applicable privacy, legal and regulatory requirements. 29 The MDR obtained diverse intra-operative data feeds, including audio-visual, environmental and patient physiologic data. 30 Visual data feeds originated from capturing the surgical field, nursing station, laparoscopic camera, and anesthesia station, including the anesthesia patient-monitoring device. Recording began just after anesthesia induction, when the patient was put to sleep, and ended after skin closure, just before the drapes were removed. This, in order to make sure that the patient’s face was fully covered and thus not identifiable during the recordings of the MDR at our tertiary referral center. Patient-related data were stripped from personal identifiers immediately upon capture. Then following, all data were synchronized, encrypted, and sent via a secure digital channel to the MDR analysis centre. 31 There, the dataset was used for post-processing, generating the MDR performance report. Post-processing was partly automatic, using AI software and rating scales (i.e non- and technical skills, distractions) 30 32-34 , identifying the events. The AI techniques that are used include machine learning and computer vision, which enable computers to learn from images and videos that are fed into them. 35