Outcome and quality of life in patients with postoperative delirium during an ICU stay following major surgery

The institutional review board and the ethics committee of the Hospital of São João, Porto, Portugal, approved the study; informed consent was obtained preoperatively from each patient. This prospective cohort study was performed in the post-anesthesia care unit (PACU) of the Hospital of São João. Within the PACU, there is a surgical intensive care unit (SICU) with five beds, where critically ill patients are admitted, closely monitored, and treated.

The subjects were recruited over a 10-month period between November 2008 and August 2009. All of the scheduled postoperative patients considered eligible for the study were Portuguese-speaking adults who were admitted to the SICU for major noncardiac and nonneurological surgeries requiring anesthesia and a postoperative hospital stay of more than 48 hours. Patients who underwent emergency surgery or did not provide informed consent were excluded from the study, as were those with a history of central nervous system disease, Parkinson’s disease, neurological or cardiac surgery, delirium or antipsychotic medication, or drug abuse (including alcohol abuse).

The following variables were recorded on admission to the SICU: age, sex, body mass index (BMI), and preadmission comorbidities, specifically ischemic heart disease, congestive heart failure, cerebrovascular disease, hypertension, renal insufficiency, diabetes, and hyperlipidemia. The anesthesia data collected for each patient consisted of information on American Society of Anesthesiologists Physical Status (ASA-PS), duration and type of anesthesia, amount of crystalloids administered during surgery, and frequency of use of colloids and blood products. SICU data, length of hospital stay (LOS), and mortality were also recorded for all the patients.

The Acute Physiology and Chronic Health Evaluation (APACHE) II score [14] and the Simplified Acute Physiology Score II (SAPS II) [15] were calculated using standard methods. The Revised Cardiac Risk Index (RCRI) was calculated using the classification system reported by Lee et al.[16]. Troponin I levels were also recorded on admission and for each day in the SICU. The mortality data were derived from the records of registered SICU mortality, hospital mortality, and mortality at six months after SICU discharge.

Sedation levels were evaluated using the Richmond Agitation and Sedation Scale (RASS) [17]. The ICDSC was only administered if the level of sedation as assessed by the RASS was between −3 and +4. Each patient admitted to the SICU and included in the study was evaluated prospectively by research staff physicians and one of the bedside nurses to establish a diagnosis of delirium, using the ICDSC [18]. The ICDSC consists of eight items based on the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria and features of delirium, including inattention, disorientation, hallucination (delusion psychosis), psychomotor agitation or retardation, inappropriate speech or mood, sleep/wake cycle disturbances, and symptom fluctuation.

Scores were assigned to each ICU patient by a nurse during every shift. The patients were assessed on the basis of the ICDSC at least once every 8 h for the entire duration of their SICU stay. All patients that scored 4 or higher in the ICDSC at least once were considered to have delirium.

HRQL was assessed using the Medical Outcomes Study Short Form 36 (SF-36) [19], which was completed directly by the patients before surgery and six months after discharge from the SICU.

The SF-36 consists of eight sections or domains, which are the weighted sums of the questions in their section. The eight domains are vitality, physical functioning, bodily pain, general health perceptions, physical role functioning, emotional role functioning, social role functioning, and mental health. Patient records were checked against hospital records after six months to determine whether a patient was still alive before being considered for inclusion in the long-term postoperative follow-up aspect of the study. A formal letter, a validated Portuguese SF-36 self-report form, and a return envelope were sent to all known survivors [20].

An evaluation of functional capacity, based on the ability of the patient to undertake personal and instrumental ADL, was performed before surgery. For all patients, this same evaluation was repeated six months after SICU discharge. A questionnaire that evaluates the functional independence of the individual with respect to performance of personal ADL (P-ADL) and instrumental ADL (I-ADL) was used based on Katz’s Index of Independence in ADL [21] and the Lawton I-ADL [22] scale. The Lawton I-ADL scale is an easily administrable assessment tool that provides self-reported information about the functional skills necessary for living in the community. Deficits in this scale were recorded and a summary score ranging from 0 (dependent) to 7 (independent) was obtained. The Katz ADL scale assesses basic personal activities of daily living, and ranks adequacy of performance in six functions. Dependency while performing each personal activity was evaluated and a summary score ranging from 0 (independence in all activities) to 6 (dependency in all activities) was obtained. The P-ADL functions were bathing, dressing, going to the toilet, transferring from the bed to a chair, continence, and eating. We considered the following activities as I-ADL functions: using the telephone, shopping, housekeeping, food preparation, using public transport, handling finances, and effectively taking responsibility for their own medications.

The patients’ responses were categorized into two groups: able or unable to perform each activity or group of activities. Patients were considered to be dependent if they were dependent with regard to any I-ADL or P-ADL activity. We further evaluated the patients who became dependent in I-ADL and P-ADL after the SICU stay.

The primary outcome was mortality at six-month follow-up. We considered hospital mortality as a secondary outcome. We also considered changes in each domain of the HRQL as outcomes. The changes in HRQL were computed on the basis of the differences between the findings in each of the eight domains of the SF-36 before surgery, and six months after discharge from the SICU. Dependency with regard to P-ADL and I-ADL were also considered as outcomes. We considered that patients became P-ADL-dependent if they were not dependent before surgery but were dependent at the time of the six-month evaluation. Similarly, we considered that patients became dependent for I-ADL if they were not dependent before surgery but were dependent at the time of the six-month evaluation. All data relating to functional status and quality of life before surgery were obtained directly from the patients before surgery.

For each dichotomous outcome - hospital mortality, mortality at the six-month follow-up, and dependency (I-ADL and P-ADL) - a separate multiple logistic regression model was used considering each outcome as a dependent variable and delirium as an independent variable, among other preadmission variables. Some independent variables, such as age and APACHE score, were included in the models as a priori knowledge. Variables correlated with outcomes at P <0.2 (chi-square test or Mann-Whitney U test) were also included in the respective model as independent variables, together with delirium and the other variables selected from a priori knowledge, in accordance with the methods described by Greenland [23].

For each SF-36 domain, a separate multiple linear regression analysis was performed with changes in the SF-36 domain considered as dependent variables and delirium as an independent variable, together with other independent variables predetermined to be included in the models as a priori knowledge. We performed no model reduction in the multivariate analyses, in accordance with methods described by Greenland [23, 24].

On admission to the SICU, 7% of the patients were dependent with regard to at least one I-ADL and 19% were dependent with regard to at least one P-ADL. The patients with delirium were more likely to be I-ADL-dependent (16% vs. 6%, P = 0.002) and P-ADL-dependent (39% vs. 16%, P <0.001) than those without delirium. On admission, Katz and Lawton scores indicated that the patients with delirium were more ADL-dependent than those without delirium (P <0.001and P <0.001 respectively). Six months after SICU discharge, 19% of the patients were dependent with regard to at least one I-ADL, and 40% with regard to at least one P-ADL. The patients with postoperative delirium exhibited more I-ADL dependency (39% vs. 16%, P <0.001) and P-ADL dependency (54% vs. 38%, P <0.001). At this time point Katz and Lawton scores indicated that the patients with postoperative delirium were more I-ADL-dependent, and more P-ADL-dependent than patients without postoperative delirium (P <0.001and P <0.001 respectively).

The results of multivariate analysis for becoming dependent for both instrumental and personal ADL at the six-month follow-up are presented in Tables 4 and 5. Preadmission variables identified by the logistic regression model as significantly correlated with becoming P-ADL-dependent were delirium, and erythrocyte transfusion (Table 4). The only preadmission variable identified by the model as significantly correlated with becoming I-ADL-dependent was COPD (Table 5).

The results of multivariate linear analysis for changes in quality of life in each SF-36 domain are provided in Table 6. The variable identified by the linear regression model as significantly correlated with changes in quality of life in the physical functioning domain was delirium. In the general health perception domain, changes were associated with total RCRI >2, and cerebrovascular disease. In the vitality domain, changes were associated with delirium, female gender, anesthesia (locoregional), cerebrovascular disease, and total RCRI >2. For social function, changes were associated with delirium, hypertension, and the use of fresh frozen plasma and platelets, and in the mental health domain, troponin I at SICU admission was associated with these changes. With regard to physical role function, bodily pain, and emotional role domains, none of the variables investigated correlated significantly with changes in quality of life. Included as independent variables in the models were the following variables: age in years, gender, ASA-PS, BMI, duration of anesthesia, type of anesthesia, locoregional anesthesia, emergency surgery, temperature at SICU admission, troponin I at SICU admission, hypertension, hyperlipidemia, COPD, high-risk surgery, ischemic heart disease, congestive heart disease, cerebrovascular disease, renal insufficiency, insulin therapy for diabetes, total RCRI, crystalloids, colloids, erythrocytes, fresh frozen plasma, platelets, and APACHE II scores. In all models, we found no indication of multicolinearity between variables; in fact, all VIF values were >3.