Objective To prospectively assess the construct and criterion validity of ClassIntra version 1.0, a newly developed classification for assessing intraoperative adverse events. Design International, multicentre cohort study. Setting 18 secondary and tertiary centres from 12 countries in Europe, Oceania, and North America. Participants The cohort study included a representative sample of 2520 patients in hospital having any type of surgery, followed up until discharge. A follow-up to assess mortality at 30 days was performed in 2372 patients (94%). A survey was sent to a representative sample of 163 surgeons and anaesthetists from participating centres. Main outcome measures Intraoperative complications were assessed according to ClassIntra. Postoperative complications were assessed daily until discharge from hospital with the Clavien-Dindo classification. The primary endpoint was construct validity by investigating the risk adjusted association between the most severe intraoperative and postoperative complications, measured in a multivariable hierarchical proportional odds model. For criterion validity, inter-rater reliability was evaluated in a survey of 10 fictitious case scenarios describing intraoperative complications. Results Of 2520 patients enrolled, 610 (24%) experienced at least one intraoperative adverse event and 838 (33%) at least one postoperative complication. Multivariable analysis showed a gradual increase in risk for a more severe postoperative complication with increasing grade of ClassIntra: ClassIntra grade I versus grade 0, odds ratio 0.99 (95% confidence interval 0.69 to 1.42); grade II versus grade 0, 1.39 (0.97 to 2.00); grade III versus grade 0, 2.62 (1.31 to 5.26); and grade IV versus grade 0, 3.81 (1.19 to 12.2). ClassIntra showed high criterion validity with an intraclass correlation coefficient of 0.76 (95% confidence interval 0.59 to 0.91) in the survey (response rate 83%). Conclusions ClassIntra is the first prospectively validated classification for assessing intraoperative adverse events in a standardised way, linking them to postoperative complications with the well established Clavien-Dindo classification. ClassIntra can be incorporated into routine practice in perioperative surgical safety checklists, or used as a monitoring and outcome reporting tool for different surgical disciplines. Future studies should investigate whether the tool is useful to stratify patients to the appropriate postoperative care, to enhance the quality of surgical interventions, and to improve long term outcomes of surgical patients. Trial registration ClinicalTrials.gov NCT03009929 .

Objective: To assess the discrepancy between perioperative complications, prospectively recorded during a cohort study versus retrospectively from health records. Background: Perioperative adverse events are relevant for patient outcome, but incomplete reporting is common. Methods: Two physicians independently recorded all intraoperative adverse events according to ClassIntra and all postoperative complications according to the Clavien-Dindo classification based on all available health records. These retrospective assessments were compared with the number and severity of those prospectively assessed in the same patients during their inclusion in 1 center of a prospective multicenter cohort study. Results: Interrater agreement between both physicians for retrospective recording was high [intraclass correlation coefficient: 0.89 (95% CI, 0.86, 0.91) for intraoperative and 0.88 (95% CI, 0.85, 0.90) for postoperative complications]. In 320 patients, the incidence rate was higher retrospectively than prospectively for any intraoperative complication (incidence rate ratio: 1.79; 95% CI, 1.50, 2.13) and for any postoperative complication (incidence rate ratio: 2.21; 95% CI, 1.90, 2.56). In 71 patients, the severity of the most severe intraoperative complication was higher in the retrospective than in the prospective data collection, whereas in 69 the grading was lower. In 106 patients, the severity of the most severe postoperative complication was higher in the retrospective than in the prospective data collection, whereas in 19 the grading was lower. Conclusions: There is a noticeable discrepancy in the number and severity of reported perioperative complications between these 2 data collection methods. On the basis of the double-blinded assessment of 2 independent raters, our study renders prospective underreporting more likely than retrospective overreporting.