Risk Factors for Delirium After On-pump Cardiac Surgery

A Systematic Review

Alex NC Gosselt; Arjen JC Slooter; Pascal RQ Boere; Irene J Zaal


Crit Care. 2015;19(346) 

In This Article


Study Identification

The search yielded 1,462 unique references of which a total of 1,279 references were excluded based on a review of the title and abstract, leaving 183 references for full-text review. Of these, 50 solely represented an abstract for a conference presentation, and 10 articles for which the full text was unavailable, had no contact information or lacked a response, and were consequently excluded. Of the remaining 123 articles, 13 did not refer to an appropriate domain, 12 did not use delirium as an outcome, or used a subjective or non-validated delirium assessment tool, 11 used no multivariable approach or randomized design, 45 had a multitude of these reasons, and 8 were excluded for other reasons. Therefore, a total of 34 studies were included for systematic review (see Fig. 1 for flowchart of study selection, and Additional file 3 http://www.ccforum.com/content/19/1/346/additional).

Figure 1.

Flowchart of study selection. Of 2,213 references, 34 studies (26 cohort, 8 randomized controlled trials) were eligible for inclusion

The 34 studies included consisted of 24 cohorts, 2 before/after observational studies and 8 RCTs; of which 4 also used a multivariable cohort approach to identify other risk factors. Three cohort studies were based on the same study population but presented models on different risk factors.[24,25,57]

Study Characteristics

Sample size in the 34 included studies varied from 36 to 4,079 patients, with a variation in delirium incidence of 2.9 − 54.9 %. Several studies limited their study population to patients undergoing coronary artery bypass graft (CABG) and/or valve surgery only; others refrained from defining cardiac surgery or included thoracic aortic surgery. For 28 studies use of CPB was confirmed in all patients, the use of CPB in the remaining 6 studies was uncertain, however, as off-pump surgery was not mentioned either, we retained those for our analysis. In most studies only patients undergoing elective surgery were included; four studies did not mention surgical urgency. The confusion assessment method (CAM) and/or the adjusted version for the ICU (CAM-ICU) were used most extensively (n = 19). In most studies researchers or trained research nurses assessed delirium, and the assessor was not described in one study.[4] One RCT included patients with sub-syndromal delirium, aiming to prevent progression to full-blown delirium.[26] Follow-up time varied with six studies only assessing patients during their ICU stay, six started at postoperative day 2 (mostly to exclude the residual effect of anesthetic drugs) and two did not fully report the duration of follow up (see Additional files 4 and 5 http://www.ccforum.com/content/19/1/346/additional).

Methodological Quality

The results of the quality analysis according to the SIGN checklists are presented in Additional files 6 and 7 http://www.ccforum.com/content/19/1/346/additional. Of 30 cohort studies, 13 (43.3 %) were scored as HQ studies, 13 (43.3 %) as AQ and 4 (13.3 %) as LQ studies. Of the eight RCTs, five were graded as HQ, two as AQ and one as LQ. Several studies did not report delirium assessment preoperatively, indicating the reported frequencies of delirium could refer to the prevalence rather than the incidence. Two studies included urgent surgery, but preoperative mental status assessment was not performed; these studies were deemed to have high risk of performance bias.[27,28]

Several studies had detection bias; 16 due to limited follow up, e.g., only ICU assessments or exclusion of postoperative day 1, or both day-2 and day-3 assessments. In another five studies,[29–33] delirium detection was likely influenced by including ICU/ventilated patients for which the assessment tool (CAM/DOS) was not validated. One study used retrospective chart review using DSM-IV criteria for diagnosis of delirium.[28] Blinding of exposure was deemed not applicable in most cohort studies evaluating basic risk factors, e.g., age. In before − after studies,[27,31] and those studying baseline cognitive/executive functioning,[30,32,34–36] failure to address this issue resulted in possible bias. Five studies investigated the association between delirium and a risk factor that was not properly defined[4,28,37–39] or validated.[31] Most studies had fairly appropriate statistical models, although one was graded lower quality as the final multivariable model included only length of stay and mechanical ventilation time,[4] another failed to include the before − after intervention in the statistical model.[27]

The eight RCTs had a median quality of 9 (range 5–9). One study used CAM in ICU patients and received no point for outcome measurement.[3] Two studies compared dexmedetomidine with morphine or propofol. Due to open-label use of other sedatives and analgesics, the treatment groups were not similar, making conclusions about the intervention cumbersome.[40,41] The investigators were not blinded to the intervention in one study.[41]

Level of Evidence for Identification as a Risk Factor

The results of the best evidence analysis are presented in Table 2. Point estimates of the individual statistically significant variables or those with an adjusted estimate ≤0.5 or ≥1.5 are provided in Additional file 2 http://www.ccforum.com/content/19/1/346/additional.

Eleven variables were found to have strong evidence for an association with postoperative delirium. Amongst the predisposing variables, strong evidence was found for age, previous psychiatric conditions, cerebrovascular disease and pre-existing cognitive impairment. For precipitating factors evidence was strong for type of surgery and perioperative blood product transfusion, and moderate for postoperative renal insufficiency and hypoxemia. Strong evidence was found for the association between postoperative delirium and prolonged mechanical ventilation, and postoperative atrial fibrillation, but cause − effect could not be determined. In two HQ studies a significant reduction was observed in the incidence of delirium in patients postoperatively treated with risperidone, resulting in strong evidence for this association. Additionally, there was evidence for no association between postoperative delirium and gender, education, previous cardiac disease or heart failure, and CPB duration.