engleski

Predictive model for postoperative vomiting based on CoPlot visualization technique

Background and Goal of Study: Predictive accuracy of two known risk scores for postoperative vomiting (POV), Koivuranta and Apfel's, are limited.Improvement of POV predictive model requires exploration of interactions and relationships among predictive variables. Materials and methods: After obtaining IRB approval and informed consent, 421 women (ASA I-II) undergoing laparoscopic gynecological surgery were enrolled in a prospective study.Of these women, 47 were excluded and 374 completed the study.No POV prophylaxis was given.Thiopental was used for induction and isoflurane or sevoflurane with or without N2O for maintenance of anesthesia.POV and pain scores were measured at 2 and 24 hours postoperatively.16 predictive factors were selected based on the result of univariate logistic regression (P< 0.05) and clinical relevance.For CoPlot visualization, distance metrics between observations were defined as "city-block metrics".For goodness-of-fit diagnostics coefficient of alienation, and correlation between the original data for each variable and projection of each observation on CoPlot vector were used. Results and discussion: CoPlot analysis of 16 possible predictors describing orientation and corresponding correlation of each predictor are shown in tables. Predictor Orientation Correlation Age 8 0.43 BMI 32 0.38 Medication -23 0.35 POV history 95 0.16 Kinetosis 133 0.06 Smoking -129 0.13 Alergies 74 0.10 Blood pressure -17 0.23 [Table 1] Predictor Orientation Correlation Menstrual status 151 0.24 Type of surgery 57 0.23 Duration of surgery -7 0.34 Duration of anaesthesia -1 0.34 Anaesthesia technique 16 0.46 Intraoperative opioids 3 0.27 Postoperative pain (0-2h) 35 0.19 Postoperative opioids 31 0.34 [Table 2.] Final CoPlot model includes 4 predictors:age>40 years, BMI≥30 kg/m2, medication, anaesthesia technique with N2O.Coefficient of alienation was 0.17 and mean correlation was 0.819. Conclusion: Further validation of this original predictive model on a new data set as well as comparison with two known predictive models for POV is needed.

Postoperative vomiting ; Predictive model

nije evidentirano