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


In summary, in this thorough systematic analysis we found a strong or moderate level of evidence for an association between eleven different risk factors and delirium after on-pump cardiac surgery. These risk factors were age, previous psychiatric conditions, cerebrovascular disease, pre-existing cognitive impairment, type of surgery, perioperative administration of risperidone, blood product transfusion, postoperative oxygen saturation, mechanical ventilation time, atrial fibrillation and renal insufficiency. We found evidence for lack of an association with gender, education, CPB duration, pre-existing cardiac disease or heart failure and CPB duration.

Our review has many strengths. The PRISMA statement was used for the design of the study and the protocol was registered in advance.[20,42] With a very comprehensive search strategy we searched several databases for publications describing potential risk factors, with an extensive time window. We included RCTs as well as cohort studies that used a multivariable approach for data analysis. We excluded studies with populations that may have important differences in delirium pathogenesis, e.g., the inclusion of solely ascending aortic surgery patients in whom certain procedures were applied (such as deep hypothermia, retrograde cerebral perfusion and aortic manipulation) that have influenced the risk of delirium, and that are not applied in other cardiac surgery patients.[43] Off-pump surgery has been shown to influence the risk of postoperative delirium.[44] By excluding studies with these patients we increased homogeneity. Furthermore, the process of screening, selection, data extraction and quality rating was performed in duplicate by two independent reviewers according to a well-defined transparent guideline.

Compared to previous attempts to review risk factors for delirium after cardiac surgery, we found convincing evidence for a smaller number of risk factors.[18,19] In our review we also incorporated negative results, e.g., risk factors that were excluded from the final multivariate model in the different studies suggestive of lack of association. Failure to do so in previous reviews may have resulted in inclusion of ever-mentioned significant risk factors. The lack of consistency or reproducibility can indicate that the observed association resulted from bias, confounding or over-fitting of a statistical model.[18,19] In this more thorough review we had to conclude that for many of these factors strong supporting evidence was lacking.

Additional risk factors found in this review such as perioperative administration of risperidone and postoperative oxygen saturation have either strong or moderate evidence for an association, in contrast to the previous reviews. This is a result of new published studies and the limited search strategy in the previous reviews; due to inclusion of risk factor in the search string and a shorter publication time window.[19] By including patients undergoing off-pump surgery the patient population in these previous reviews was more heterogeneous. As illustrated in Additional file 4 http://www.ccforum.com/content/19/1/346/additional, heterogeneity remains a problem when comparing studies. Differences in outcome measurement days, urgency of surgery, age limitations and adequacy of the screenings tools used led to striking differences in delirium incidence and the detected associations.

Unfortunately, this review also has some limitations. Even though we used a thorough search strategy, studies may have been missed. It is not possible to overcome publication bias. The cutoff scores to identify HQ studies may be arbitrary. Items deemed not applicable were counted as a point, for these were not representative of lack of study quality, but rather the consequence of the tested risk factors. Not including these points in these specific studies would have created different maximal obtainable quality scores per individual study and was therefore not feasible. This review attempted to select homogenous study populations consisting of patients undergoing on-pump cardiac surgery only. We excluded studies reporting either off-pump surgery or solely reporting ascending aortic surgery, however, some of the included studies only reported a small group designated 'other cardiac surgery'. This could have resulted in inclusion of some patients undergoing surgery to the ascending aorta. A limitation of our approach to classify level of evidence for recently suggested risk factors is that these await confirmation by another independent study. We had to group some variables we regarded as similar, such as history of stroke, and history of cerebrovascular or neurological disease.

With regard to future research on delirium in cardiac surgery patients, more emphasis should be put on several methodological issues as outlined in this review. We only included cohort studies using a multivariable approach for data analysis to correct for confounding. However, none of the studies included all identified confounders in the model. The strength of a model largely depends on the variables included, so failure to include key variables results in non-comparable results that are not suitable for statistical pooling. As illustrated by the large variation in reported delirium incidence (2.9 − 54.9 %), selection bias and detection bias may remain an underestimated influence. We recommend incorporation of the aforementioned risk factors as well as the use of the most common assessment tools and a homogeneous study population when designing a new study into this subject in order to create generalizable results.