Delirium after cardiac surgery: incidence and risk factors^†

Abstract

INTRODUCTION

Neurological complications after cardiac surgery remain an intriguing issue with a complex background. Delirium is a neuropsychiatric syndrome, which found its place in cardiac surgery from the early beginning of cardiopulmonary bypass (CPB) described as postcardiotomy delirium [1]. In more recent years, brain damage after cardiac surgery has been described in terms of stroke and neurocognitive decline, with reference to Type-I and Type-II injury, respectively [2]. Delirium is probably the most common symptom of Type-II injuries. The distinction between delirium and the former nomenclature has not yet been clarified [3].

Postoperative delirium is a clinical reality being common after all types of surgery. Delirium is described as fluctuations in cognition and attention. By definition it is assumed to express an underlying cause and is often transient [4]. The reported incidence of delirium after cardiac surgery varies, exemplified by 8–52% [5–7]. The condition is more common among old patients [8]. The span in incidence between studies may partly reflect various means for how delirium has been screened and diagnosed. Delirium is a psychiatric syndrome that hides various types of pre-existing (predisposing), inflicted brain injuries and side-effects of the medical management (precipitating). Advanced age, previous cerebrovascular disease and diabetes exemplify the commonly referred predisposing risk factors [5]. Suggested precipitating factors are: ventilator time [5], postoperative infections and prolonged operating time [8]. The subdivision of risk factors has a purpose. Predisposing factors may be used for risk evaluation prior to surgery, whereas precipitating factors may be considered for improving the clinical care and avoiding delirium.

This study aimed at identifying potential predisposing and precipitating risk factors behind delirium in consecutive patients, 70 years and older undergoing cardiac surgery.

MATERIALS AND METHODS

RESULTS

In total, 78/142 (54.9%) of the patients were diagnosed with delirium postoperatively, based on the DSM-IV-TR criteria. Delirium was divided into subprofiles: hypoactive (n = 56/78), hyperactive (n = 4/78), mixed (n = 4/78). The remaining patients were non-classifiable (n = 14/78). Risk-factor analyses were not conducted on subgroup levels.

Patient characteristics for predisposing and precipitating variables, respectively, are given in Tables ​Tables11 and ​and2.2. These tables also describe the univariate results with respect to delirium and the variables of interest for multivariable testing are indicated to be further listed in Table ​Table3.3. In multivariable logistic regression, the following predisposing variables were found independently associated with delirium: preoperative pain score on NRS, diabetes, oxygen saturation and combined surgery. These variables had about the same statistical influence (i.e. Wald) in predicting postoperative delirium. The precipitating variables were separately analysed to expose: volume load during operation, postoperative ventilator time and plasma sodium concentration in the ICU. Among these, the volume load during operation had the strongest predict influence.

A somewhat different model was obtained when the predisposing and precipitating variables were collectively analysed (Table ​(Table3).3). In this model diabetes, ventilator time, sodium concentration and volume load during operation remained independently associated with postoperative delirium, together with age and early postoperative temperature. The variable peptic ulcer showed a borderline significance in this model. When this unexpected variable was removed, the model remained essentially unchanged with only a modest decrease in explanatory strength. The highest predictive influence in the model was obtained from the variable volume load during operation.

In Table ​Table4,4, the statistical results are shown for the various models applied on the cohort. All models (e.g. predisposing, precipitating and combined) were highly significant. Nevertheless, the model based on precipitating variables gave a somewhat better explanatory strength than that based on predisposing factors. The strength increases when both predisposing and precipitating factors were collectively considered in the model. This was illustrated by the Nagelkerke R² value explaining 34.6% of the variance. The corresponding ROC curves are shown in Fig. ​Fig.11 with their areas under curve listed in Table ​Table44.

Open in a separate window

Figure 1

ROCs of the tested multivariable models. See text for further details.

DISCUSSION

Delirium was diagnosed in 55% of the patients, 70 years and older undergoing consecutive routine cardiac surgery. Both predisposing and precipitating factors contributed to delirium. However, when predisposing and precipitating were combined, the predictive strength of the model increased notably with more than one-third of the variance explained. In overview, our study confirmed many of the previously suggested risk factors. In addition and of great interest in our study, the variable volume load during the operation was found to be a dominant factor, not previously identified.

The incidence of delirium was higher in our report than in most other studies [7, 15–18]. Partly, this finding may reflect our careful and comprehensive cognitive testing being performed both before and repeatedly after the operation. Studies not using cognitive tests usually report a much lower incidence rate, since many patients with hypoactive delirium are likely to have escaped detection [15, 17, 18]. Also, our observational period of delirium included four postoperative days being longer than in most other studies [7, 18]. A third factor that may have contributed to our results was the selection of older patients and inclusion of surgical procedures other than isolated coronary bypass.

The variable volume load during the operation was strongly associated with postoperative delirium. This variable added together infusions administered in the operating theatre for anaesthetic and CPB purposes. Blood transfusions were separately analysed and blood products had no independent influence on delirium. On average, each patient was given 3.7 l of fluids. The difference between delirious and control patients was small in numeric terms, but the variable was associated with the highest predictive influence. Per litre of extra volume, the risk of developing delirium increased nearly 3-fold. An association between volume load and delirium has not been previously reported. However, Bucerius et al. [5] observed that patients exposed to intraoperative haemofiltration showed an increased risk of postoperative delirium. Moreover, they also reported that patients exposed to beating heart coronary bypass surgery were protected against delirium, implying a mechanistic role of CPB. This comparison must be considered with care because in their study no cognitive testing was performed and their reported incidence rate of delirium was only 8.4% [5].

A high sodium concentration in the first postoperative morning predicted delirium, independently from that of an increased volume load during surgery. The two variables showed no statistical correlation, an observation that talks against an electrolyte disturbance being the result of intraoperative infusions given. The association between sodium concentration and delirium is previously identified [19]. The interpretation regarding sodium in our study is complex because none of the patients was hypernatremic by definition. It must also be brought to mind that both intraoperative volume load and sodium concentrations in the ICU may be surrogate variables for unknown factors. Nevertheless, our study used a wide range of analysed variables which counteracts this potential error.

Additional variables with relevance to delirium were identified. In brief, diabetes increased the risk of delirium 3.5 times, which finds support in previous literature [5, 20]. Age is another well-known risk factor behind delirium [15, 18]. In our study, the risk of developing delirium increased 10% per extra year of age. Here only older patients were analysed, 70 years of age and older. Despite this restrained selection of older patients, age remained independently associated with delirium in this cohort. The selection may also explain the high incidence rate of delirium, compared with many previous studies.

Among the precipitating factors, a prolonged ventilator time was associated with delirium. This observation supports previous suggestions [15]. The risk of delirium increased ∼20% per each extra hour of ventilation. Nevertheless, the variable is complex and a prolonged ventilator time may reflect an underlying cause (e.g. comorbidities, anaesthetic managements or complications at surgery). Also, the variable is bias from local ICU routines (e.g. extubation protocols and differences between individual care givers). Moreover, some studies have reversed the causality to explain a prolonged ventilator time as a result of delirium rather than the other way around [15, 18]. An increased body temperature in the early postoperative period was associated with delirium. Our observation finds scattered support in the literature [21] and is not assumed to reflect an infectious disease, being too early after surgery. Infection is otherwise a well-known trigger of delirium [22]. In contrast to our finding, a low intraoperative body temperature has also been found relevant [16, 19]. Intraoperative temperatures were not recorded in our study. However, the cooling during CPB interferes with this issue with a low CPB temperature reflecting a complex and long-duration procedure probably being more important than the temperature as such. An unexpected finding in our study was that patients reporting gastritis and/or peptic ulcer problems had an increased risk of delirium. This observation showed a borderline significance and must be considered with caution. The multivariable model was also tested without this variable included without affecting the results notably.

This study reports the results from consecutive older patients undergoing cardiac surgery. The study was based on comprehensive assessments including cognitive testing both before and repeatedly after surgery. All assessments were performed by two trained research nurses. Similarly, a large number of variables were systematically extracted for analysis, pre-, intra- and early postoperatively. Unfortunately, pharmacological interactions were not explored. Our study is also limited from its relatively small sample size in relation to the surplus of variables analysed. Therefore, our findings should be validated in future studies. Certainly, contributing factors behind delirium are to be sought among a variety of factors beyond those tested in our study, for example, inflammatory and embolic mechanisms. Also, our results are valid only for the cohort studied of older patients during their period of hospitalization.

In conclusion, delirium was common among older patients undergoing cardiac surgery. Both predisposing and precipitating factors contributed to delirium and it is potentially possible to modify several of these factors in daily clinical practice. The strongest precipitating factor for delirium was an increased volume load during operation. An association between delirium and volume load has not been previously identified, a detail that it may be possible to investigate in a randomized study.

Funding

This work was supported by grants from the Heart Foundation of Northern Sweden; the Erik and Anne-Marie Detlof's Foundation; the Dementia foundation, the Borgerskapet in Umeå Research Foundation; the Medical Faculty at Umeå University Sweden; the University of Umeå; the County Council of Västerbotten, Sweden, and the Strategic Research Programme in Care Sciences, Sweden.

Conflict of interest: none declared.

Acknowledgements

REFERENCES