Assessing risk factors of delirium and its effects on adverse outcomes in patients admitted to the ICU after craniotomy

Editor, In a prospective cohort study, Wang et al.1 attempted to determine incidence and risk factors of postoperative delirium in patients admitted to the ICU after intracranial surgery and assessed its effects on postoperative adverse outcomes. This study has potential clinical implications, but we note several issues that need attention. First, postoperative delirium was assessed using the Confusion Assessment Method for the ICU (CAM-ICU) on postoperative day 1 or 3. It was unclear whether or not patients were mechanically ventilated when assessing postoperative delirium. The CAM-ICU is rational for ventilated patients, but the Confusion Assessment Method is better in nonventilated patients.2 Most importantly, new recommendations for the nomenclature of cognitive change associated with anaesthesia and surgery, suggest reporting on postoperative delirium during hospital stay up to 1 week after surgery or until discharge. Furthermore, definitions of postoperative delirium must meet the diagnostic criteria of delirium in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition definition (DSM-5).3 In addition, in the Wang study, part of postoperative delirium assessments were completed on the ward. Finally, recent evidence indicates that compared with the diagnostic criteria of delirium in the DSM-5, the CAM-ICU performs poorly for diagnosis of delirium in non-ICU patients.4 We argue that these factors may have confounded the incidence and type of postoperative delirium after intracranial surgery in this study. Second, to determine independent risk factors of postoperative delirium, the authors entered all factors that had a P value less than 0.05 in the univariate and categorical analyses, into a multivariate analysis with a stepwise backward logistic regression. This method of building a multivariate model may be questioned. Furthermore, the data provided by Wang in their Tables 1 to 4 should be initial comparisons of demographics and peri-operative variables between patients with and without postoperative delirium, rather than the final results of the univariate analyses for demographics and peri-operative data, which include P values and unadjusted odds ratios (ORs) with their 95% confidence intervals (CIs). As a general principle, a normal process of performing multivariate analysis includes: first, initial comparison of cohorts experiencing and cohort not experiencing a postoperative delirium, performed by the Student's t test for continuous variables and χ2 test for categorical variables, as described in Wang's Tables 1 to 4; second, incorporation of variables with statistical significance in initial comparisons, defined as P less than 0.05, into the univariate analysis to examine multicollinearity among candidate independent variables; third, finally, inclusion of the variables with large P values (P < 0.2) in the univariate analyses into the multivariate model using postoperative delirium as the dependent outcome variable to identify independent risk factors of postoperative delirium, with their P values and adjusted ORs with their 95% CIs.5 We consider that clarifying these statistical issues would have improved the transparency of this study and the interpretation of the findings. Third, by direct comparisons of postoperative adverse outcomes between patients with and without postoperative delirium, the authors concluded that postoperative delirium was associated with postoperative adverse outcomes. Evidently, this conclusion is not robust, as many demographics and peri-operative variables are significantly different between patients with and without postoperative delirium. Furthermore, these factors are also attributable to postoperative adverse outcomes. To obtain the true contribution of postoperative delirium to postoperative adverse outcomes, multivariate analysis is needed for adjusting patients' baseline characteristic and controlling selection biases. Finally, this study observed the development of incidences of different types of postoperative delirium, but not severity and duration of postoperative delirium. Other than the development of delirium, both duration and severity of delirium have also been associated with significantly increased adverse outcomes of ICU patients.6 Furthermore, identification and study of safe and efficacious interventions to reduce the incidence, duration and severity of ICU delirium are hot topics in critical care. We believe that the findings of this study would have been more informative, if these issues had been included into the study design.