The purpose of this two-centre pilot research study is to establish the feasibility of conducting a full trial that seeks to determine if a cardiac anesthesia policy that uses alternatives to benzodiazepine medications is better at preventing delirium after cardiac surgery when compared with a cardiac anesthesia policy that uses benzodiazepine medications.

ICU data suggests that benzodiazepines are linked to delirium, and minimizing their use has been incorporated into ICU practice guidelines. Cardiac surgery patients are at elevated risk of delirium. However, benzodiazepine administration still constitutes a common part of cardiac anesthesia practice, related to the previous absence of alternate medications and techniques to ensure hemodynamic stability and prevent intraoperative awareness. However, with the advent of new medications and processed EEG monitoring, benzodiazepine use is no longer ubiquitous, with both utilization and dose varying among practitioners. These multiple approaches to cardiac anesthesia care are haphazard, which speaks to the clinical equipoise between them. As such, the investigators have decided to standardize our approach to cardiac anesthesia care, and to evaluate the impact of changes in approach on postoperative delirium, a serious problem for both individual patients and health care systems.

Studying strategies of care and their impact on delirium: There is a great need to understand if a strategy of limitation of intraoperative benzodiazepine use during cardiac surgery can reduce postoperative delirium. Cardiac surgery is done in specialized institutions performing high volumes of surgery in order to reduce complications, increase efficiency, and obtain superior outcomes. The surgical care of patients in these high volume cardiac surgery centres is undertaken using standardized procedures that optimize outcomes. The investigators would like to determine if a policy limiting the use of intra-operative benzodiazepine will reduce postoperative delirium, a question which is asked from the perspective of clinical effectiveness. The only way to answer this question is to randomize between two institutional approaches to benzodiazepine use: a policy of standard use in all patients compared to a policy of limited benzodiazepine use in select patients only. The cluster randomized trial is a methodologic approach designed to test different institutional policies. In the cluster crossover trial, each institution is randomized to use one policy or the other and then, after a treatment and suitable wash-out period, the institution crosses over to the other policy. This approach is methodologically rigorous and tests the effects of the change in practice as it would actually be used in the clinical setting. Therefore, the optimal approach to answer the question of intraoperative benzodiazepine strategy is a cluster crossover trial, which randomizes cardiac surgery centres to each of two standard approaches to intraoperative benzodiazepine use, with crossover within each site.

An institutional policy of limiting intraoperative benzodiazepine use during adult cardiac surgery, compared to a policy of 'ad libitum' intraoperative benzodiazepine use, will reduce the rate of post-operative delirium.

Our goal is to undertake a randomized cluster crossover trial, with randomization at the level of the hospital. Before such a trial can be conducted, we need to establish feasibility. We therefore propose to undertake a pilot study at two hospitals, to establish the feasibility of the trial protocol. This pilot trial will be conducted over approximately 6 months, in 4 study periods. Because of variation in the number of procedures performed within a given month, the duration of each period will be determined by the amount of time required to manage 120 patients according to the selected policy. There will be a 1-week washout period between each study period to enable staff transitions between institutional policies of cardiac anesthesia care. At the outset of the study, we will randomize the pilot site to any of 2 possible alternating sequences of benzodiazepine-minimized (A) or routine benzodiazepine administration (B) (i.e. ABAB or BABA). We call this study the B-free Pilot Trial. We will use data from this pilot in the larger study unless substantial protocol modifications are needed.

Rationale: At least 1 CAM assessment will be required to assess the primary outcome of the large trial, with one CAM per ICU ideal to assess delirium (as it has a fluctuating course).

Protocol compliance by individual practitioners such that 80% of patients are not administered any intraoperative benzodiazepines during the benzodiazepine minimization period, and 80% of patients receive intraoperative benzodiazepines during the ad libitum benzodiazepine period. This corresponds to 192/240 patients treated per protocol within each treatment arm.

Rationale: In this trial we are comparing two different institutional policies. However, in reality, anesthesiologists are individuals who may make individual patient management decisions. In addition, there are concrete reasons why some patients should not be treated according to one or the other policy. We will thus use the pilot study to evaluate the proportion of patients managed according to the policy being used during a given crossover period in order to be certain that, in the large trial, there will be significant differences in benzodiazepine administration between the two periods.

Incidence of intraoperative awareness of no more than 2% during the benzodiazepine minimization period (i.e. 5/240 patients) at one of the two pilot sites (HGH). This is defined as the recall of intraoperative events, and will be ascertained through the use of a standardized questionnaire widely used in research and clinical practice to ascertain patient awareness (see Appendix I: Procedure for Assessment of Intraoperative Awareness).

Rationale: There is a historical belief that benzodiazepines decrease the risk of intraoperative awareness, even though there is no evidence to support this belief.2,4,5 Given that (1) there is no evidence to support the preventative effect of benzodiazepines on intraoperative awareness and (2) the assessment of awareness is not a routine part of clinical practice in all sites in Canada, we would like to establish a rate of intraoperative awareness in patients treated during the benzodiazepine minimization period that is not higher than upper boundary of the 95th percent confidence interval for the pooled incidence reported in three large, recent studies (i.e. 2%). Achievement of this objective during the pilot study will justify not formally assessing for intraoperative awareness at all sites during the multi-centre trial.

Relationship between the incidence of postoperative delirium in hospital-collected, administrative data and Canadian Institute for Health Information (CIHI) data of ≥ 85%. This will be defined as the incidence of postoperative delirium in the CIHI discharge abstracts database divided by the incidence of postoperative delirium reported in hospital administrative data.

Rationale: We believe that the incidence of postoperative delirium is most reliably and accurately recorded in hospital-collected administrative data, using the CAM. However, collecting the outcome of delirium using this data source will be more work-intensive and costly than collecting the outcome of delirium using the CIHI discharge abstracts database, which we believe will underrepresent the true incidence of postoperative delirium. As such, we will collect data from both of these sources during the pilot trial. If the reported incidence of postoperative delirium in the CIHI discharge abstract database is at least 85% of that reported in hospital administrative data, we will use this as the primary data source in the large trial (and will modify the secondary outcomes of duration of delirium, ICU LOS, and hospital LOS to ensure they reflect data fields collected by CIHI). If this threshold is not exceeded, we will obtain the extra funding and human resources required for local data collection in the large trial.

Inclusion Criteria:

• All patients undergoing cardiac surgery during the study period at the Hamilton General and St. Boniface General Hospitals.

Exclusion Criteria:

-