Video-Based Grading of Intraoperative Adverse Events in Pediatric Laparoscopic Appendectomy

Background and RationaleSystematic classification of intraoperative adverse events (iAEs) is a prerequisite formeasuring surgical safety. In adults, ClassIntra (Dell‐Kuster et al., 2020) gradesintraoperative events on five severity levels (Grade I‐V) and has been internationallyvalidated. In children, no population‐specific instrument exists and pediatric experiencewith ClassIntra is limited to a small number of reports: a neurosurgical pilot congressabstract (Drexler et al., 2022; n=21), a full‐text prospective neurosurgical cohort fromthe same group (Middelkamp, Drexler et al., World Neurosurgery 2025; n=47), and aprospective multidisciplinary pediatric robotic surgery programme in which ClassIntra wasrecorded alongside Clavien‐Dindo grading (Vinit et al., Annals of Surgery 2023; n=300, ofwhich 83 digestive and 105 urological or gynaecological procedures). In none of these wasClassIntra applied to video recordings, in none was its inter‐rater reliability assessed,and none focused on appendectomy. The iAE profile of pediatric laparoscopic appendectomy‐ the most common pediatric surgical procedure ‐ has not been systematically documented.For postoperative complications, a pediatric‐validated instrument does exist: theClavien‐Madadi classification (Madadi‐Sanjani et al., 2023; ERNICA validation 2024). Therelationship between intraoperative event severity and postoperative complicationseverity has not been studied in children.DesignProspective, single‐center, single‐surgeon, video‐based pilot study, positioned at Stage2a (Development) of the IDEAL Framework for surgical innovation. Single‐arm observationalcohort; no control group and no comparison arm. The principal investigator performs alloperations and does not serve as a rater; the investigator acts only as arbiter in caseof unresolved disagreement between raters.Video Assessment ProtocolOperations are recorded in 1920x1080 H.264 MP4 format. Recordings are de‐identified,assigned a case code (PA‐XXX), and distributed to raters in randomized order. Twoindependent pediatric surgery specialists from outside the operating institution serve asraters; a third specialist is pre‐designated as reserve. Raters have no access topostoperative clinical data, which prevents halo effects. Raters undergo a three‐stagetraining program before assessment begins: theoretical training (2 weeks), calibrationsessions (2 weeks), and certification; an inter‐rater reliability threshold of 0.60 istargeted at first calibration, with an additional calibration round if not reached.The assessment window is deliberately restricted to the laparoscopic phase, frominsertion of the laparoscope into the abdomen until its withdrawal. Skin incision, portplacement, fascial closure and anesthesia‐related events are outside the assessmentwindow. This partial application of ClassIntra is a pre‐specified scope decision made forreasons of pilot scope management and rater workload, and is reported as a limitation.The study is not a retrospective analysis of an existing video archive: all videos areobtained prospectively after ethics approval, in accordance with the study protocol.Three‐Level Event CharacterizationEach identified intraoperative event is characterized on three levels: ‐ Level 1, event type: E1 bleeding, E2 thermal injury, E3 avulsion, E4 serosal tear,E5 injury to adjacent organ, E6 detached clip, E7 spillage (pediatric adaptationafter Sanmoto 2025). ‐ Level 2, mechanism: GERT categories M1 excessive force/distance, M2 insufficientforce/distance, M3 wrong orientation, M4 insufficient visualization (Bonrath 2013),a procedure‐independent internationally validated framework. ‐ Level 3, severity: ClassIntra Grade I‐V (Dell‐Kuster 2020), applied to thelaparoscopic phase.In parallel and independently of the video assessment, postoperative complicationsoccurring within 30 days are graded by the principal investigator using theClavien‐Madadi classification from routine clinical follow‐up data. No automaticconversion between complication classification systems is performed.Relationship to Institutional ERAS CareThe department applies a 20‐item Enhanced Recovery After Surgery (ERAS) protocol asroutine care for all children undergoing appendectomy, independently of this study.Protocol adherence and recovery data (including time to medical readiness for discharge)are therefore documented for all participants as part of routine care, and are used inthis study only as exploratory secondary variables. Enrollment in any other study isneither an inclusion criterion nor a requirement for participation in this study.Statistical ApproachInter‐rater reliability is assessed with Gwet's AC1 as the primary measure, with weightedKappa and the intraclass correlation coefficient reported as supporting measures. As apre‐specified sensitivity analysis, AC1 is additionally computed in the subset of videoswith at least one identified intraoperative adverse event; if this subset comprises fewerthan 15 videos, the subset analysis is reported descriptively only. Event incidence isreported with Wilson score confidence intervals. Exploratory hypotheses are analyzed withlogistic regression (ERAS protocol adherence) and Spearman rank correlation (time tomedical readiness for discharge; Clavien‐Madadi severity). No adjustment for multipletesting is applied, as these analyses are explicitly exploratory; effect sizes with 95%confidence intervals are emphasized over p‐values. Missing data are managed with apre‐specified, proportion‐dependent three‐scenario strategy. Statistical analysis isconducted under the responsibility of the principal investigator.Sample SizeA target of 80 patients is based on published precedent for reliability pilot studiesrather than on a confirmatory power calculation. This sample supports estimation of AC1with acceptable precision (approximately +/‐0.10 around an expected AC1 of 0.65) but doesnot provide statistical power for confirmatory hypothesis testing; secondary results aretherefore reported as effect‐size estimates.Known LimitationsThe single‐surgeon, single‐center design limits generalizability and is a deliberatepilot choice. Findings are restricted to non‐complicated appendicitis and cannot beextrapolated to other pediatric procedures. Video‐based assessment may under‐detectanesthesia‐related and organizational events. Post‐hoc exclusion of conversion caseslowers the estimated incidence of high‐grade events. The hybrid ClassIntra/Clavien‐Madadistructure has not previously been tested in a pediatric population, so no accuracybenchmark exists. Accordingly, this study provides feasibility evidence, not validityevidence; multicenter studies are required for the latter.