Risk and Protective Factors Analysis for Postoperative Delirium in Elderly Critically Ill Patients: A Multicenter Retrospective Study


 Objective: To study the incidence and risk factors of postoperative clinical delirium in critically ill elderly patients in order to devise strategies for prevention and treatment.Methods: We collected clinical data from 451 critically ill patients in the intensive care unit from four hospitals in Shanghai Pudong New Area. Postoperative delirium was measured using the Confusion Assessment Method for the ICU (CAM-ICU), and then the incidence of postoperative delirium was calculated. We used univariate analysis and multivariate logistic regression analysis to determine the risk factors associated with postoperative delirium in critically ill elderly patients.Results: Among the 451 critically ill patients, 56 had delirium after operation. The incidence of postoperative delirium was 12.4%. Univariate analysis showed that the incidence of postoperative delirium was significantly higher (P <0.05) in patients who had a past history of chronic obstructive pulmonary disease (COPD), underwent emergency surgery, were not administered intraoperative Propofol, had postoperative hypertension, were administered postoperative methylprednisolone, and did not have postoperative analgesia. Multivariate logistic regression analysis showed that age (OR=1.111, 95% CI, 1.056-1.168), surgical timing (OR=1.154, 95% CI, 1.058-7.200), and use of methylprednisolone postoperatively (OR=8.030, 95% CI, 1.050-61.408) were risk factors. Use of Propofol intraoperatively (OR=0.315, 95%CI, 0.124-0.800) was protective factors for postoperative delirium in elderly critically ill patients.Conclusion: Adavanced age, COPD history, emergency surgery, postoperative hypertension, use of methylprednisolone, and no postoperative analgesia are risk factors for postoperative delirium in critically ill elderly patients. Medical staff should be educated in evaluating delirium in postoperative patients and conducting risk assessment in order to prevent and treat the disease earlier, and reduce the incidence of delirium.


Introduction
Delirium is de ned as an acute and reversible disturbance of consciousness and attention that presents during a short period (hours or days) and has a uctuating course. Delirium is usually accompanied by an unpleasant mood and cognitive dysfunction (1). Clinically, delirium is divided into three subtypes: activity-increasing, activity-decreasing, and mixed-typed. The activity-increasing subtype manifests as excitement, agitation, restlessness, unstable emotion, or aggressive behavior. Activity-decreasing subtype exhibits apathy, drowsiness, impoverishment, and reduced responsiveness. Mixed-type, on the other hand, is characterized by simultaneous or sequential appearance of the activity-increasing and activitydecreasing subtypes (2,3).
Generally, delirium causes serious social burdens, such as long-term cognitive impairment, higher mortality rate, prolonged intensive care unit (ICU) stay and hospitalization, and increased costs. Early intervention is expected to improve prognosis, however, there is currently no effective pharmacologic intervention for this disease. Primary prevention with nonpharmacologic multicomponent approaches have gained widespread acceptance as the most effective strategy for delirium (4,5). Therefore, the aim of this multicenter study was to investigate the incidence and risk factors of postoperative delirium in critically ill elderly patients in the ICU.

Study design
The study was a multicenter prospective cohort study. The data were collected from four hospitals in the Pudong Area, Shanghai (Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Dongfang Hospital of Tongji University, Shanghai Seventh People's Hospital, and Gongli Hospital of Pudong New Area). All patients in our cohort had critical illness between August 2015 to July 2017. Delirium was diagnosed according to the criteria of the confusion assessment method for the ICU (CAM-ICU) approved by the PAD guidelines. Postoperative delirium was measured 3 times a day (8:00, 16:00, and 24:00). We analyzed the total incidence of postoperative delirium across different operations to unveil potential risk factors. Each patient or caregiver signed informed consent. This study was approved by the Ethics Committee of the Renji Hospital, and registered at the China Clinical Trial Registry (Registration Number: ChiCTR-OOC-16008154, registered 25 March 2016).

Study populations
ICU patients aged ≥ 65 years from August 2015 to July 2017 were included in this study. Patients who signed informed consent, intended to stay in ICU postoperatively > 24h and were evaluated as APACHII ≥ 15, ASA < IV were eligible for inclusion. Patients were excluded from analysis for the following reasons: (1) pre-operative cognitive impairment or pre-existing delirium symptoms; (2) patients used psychiatric medications during the past six months; (3) patients underwent brain surgery (craniocerebral trauma); (4) the operation time < 1h; (5) patients were vision or hearing impaired, or unable to communicate; (6) patients' estimated survival time < 24h; (7) patients or their family members refused to participate in this study.

Data collection
A validated daily record list was used to collect the demographic data (age, sex, BMI, and education), history of any current illnesses, use of medication, and other relevant information from each participant. Preoperative risk factors including ASA classi cation, site of surgery, nutritional status, preoperative medication, laboratory results regarding electrolyte examination and arterial blood gas examination were collected by the researchers. Intraoperative risk factors, including intraoperative medication, hypothermia, surgery time, blood loss, blood transfusion volume, intraoperative blood pressure, oxygenation index, arterial blood gas index and arrhythmia were collected by the anesthesiologists. Then, each patient was evaluated for postoperative ICU-related risk factors, including postoperative mechanical ventilation time, pain score, analgesia, sedation, body temperature, postoperative medication, indwelling catheter duration, restraint, quality of sleep, ward conditions, postoperative blood pressure, and blood transfusion. Sleep quality was assessed by self-report from each patient. The participants were then assessed using CAM-ICU to identify the development and subtype of delirium.
Patients were followed for 28 days after the operation and assessed the postoperative outcomes such as duration of ICU stay, duration of postoperative hospitalization, and hospital expense.

Delirium assessment
In this study, a 2-step method as the CAM-ICU criteria was used to identify delirium. The rst step was to assess the patients' consciousness via the RASS score. Patients were excluded if their RASS score was -4 (responsive only to physical stimulus) or -5 (unresponsive to physical and verbal stimulus) as it was di cult to determine how conscious the patient was due to unresponsiveness. If the patient was in coma, then they were ineligible for the CAM-ICU assessment and recorded as "unable to evaluate". If the RASS score was ≥ -3 (+4: aggressiveness; +3: extreme restlessness; +2: restlessness; +1: irritability; 0: sober and calm; -1: drowsiness; -2: mild sedation; -3: moderate sedation), then they proceed to step 2 of the CAM-ICU assessment. The second step contains assessing conscious content from 4 diagnostic characteristics: (1) A violent change or uctuation in the state of consciousness during the past 24 hours, (2) inattention, (3) altered level of consciousness, and (4) confusion in thinking. Each feature was carefully assessed in order to determine whether a patient had delirium. Patients were diagnosed with delirium when feature (1) and (2) plus either feature (3) or (4) were present. Any score other than 0 is a positive value for characteristic (3).
After the patient was admitted to the ICU, CAM-ICU assessments were performed 3 times a day (8:00, 16:00, and 24:00) for 7 consecutive days while in the ICU or until the patient left the ICU. All researchers were professionally trained and are strictly evaluated according to CAM-ICU criteria.

Statistical analysis
Continuous variables were expressed as means + SD or median and interquartile ranges (IQRs). Categorical data are shown as proportions (number and percentage). The t-test or Wilcoxon test were used to compare the continuous variables in patients with and without delirium. T-test was used when both groups of data were normally distributed. In case of abnormalities, the Wilcoxon test was used.
Categorical variables were compared using χ 2 test (Pearson χ 2 test) or the Fisher exact test. Wilcoxon test was used to compare the variables after sequencing. To examine the effects of various factors on the development of delirium, univariate and multivariate stepwise logistic regression analyses were performed. The P values for entry and removal were .05 and .10, respectively. P < 0.05 was considered statistically signi cant. All data were statistically analyzed via SPSS 21.0 software.

Incidence of delirium
During the 2-year study interval, 623 patients were admitted to the ICU, among which 451 (72%) met the inclusion criteria and were enrolled in the cohort while 172 (28%) met exclusion criteria (Fig.1). Among all patients, delirium was identi ed in 56 cases (12.4%), the subtypes of which were shown in Table 1 and Table 2.  The variables that were potentially related to delirium development were divided in the following 3 categories: preoperative, intraoperative, and postoperative. All the variables included in the study were analyzed via univariate statistical analysis and presented in Tables 3-7.
Patients with previous history of COPD showed signi cantly increased incidence of delirium (P<0.01), while other perioperative factors showed no statistically differences, as shown in Table 3. Next, we evaluated nursing-related factors after the patient entered the ICU, Table 4 showed that there was no statistical differences between different nursing-related factors with regards to presence of delirium.  Table 5 showed that the use of indwelling catheter was not signi cantly related to postoperative delirium. Patients' perioperative conditions were closely related to the incidence of delirium. In particular, undergoing emergency surgery (P<0.01), postoperative hypertension (P<0.05), and postoperative analgesia (P<0.05), were found to be signi cantly related to the incidence of postoperative delirium (Table 6).  Table 7 showed that intraoperative administration of propofol and ICU medication of methylprednisolone (P<0.01) were risk factors for the development of postoperative delirium. We performed multivariate statistical analysis by using factors that were found to be statistically signi cant in the univariate analysis as independent variables and delirium as the dependent variable. These variables, that were signi cantly more prevalent in delirious patients compared to non-delirious patients, were incorporated into a stepwise logistic regression analysis. According to our statistical analysis, the best regression model identi ed age, surgical timing, and use methylprednisolone medication as risk factors for postoperative delirium (Table 8).

Discussion
This study showed that the incidence of postoperative delirium in elderly critically ill patients was 12.4%, which was accord with most recent studies(6). Postoperative delirium incidence differs greatly among surgery types. For instance, the incidence for hip surgery was 4-53% (7), for ENT surgery was 12%(8), and for spine surgery was 0.84% (9). Furthermore, postoperative delirium incidence also differs in different periods. For example, the incidence for cardiac surgery was reported to be 32% in 1999(10), 23-26% in 2008 (11) but 13.3% in recent studies (12). Therefore, its necessary to analyze the risk or protective factors for postoperative delirium according to the latest study.
In this study, activity-increasing subtype accounted for 53.6% (30/56), activity-decreasing subtype accounted for 26.8% (15/56), and mixed-subtype comprised 19.6% (11/56) of all patients that suffered from postoperative delirium according to CAM-ICU, which is speci cally designed for non-psychiatrists to evaluate ICU patients. Activity-increasing subtype is easier diagnosed by physicians due to its characteristic features of excitement or aggression, while activity-decreasing delirium is often missed or misdiagnosed as sedation or depression due to the patients' non-responsiveness (13). Among the misdiagnosed patients, 94% of them comprised the activity-decreasing subtype (14). The relatively low incidence of delirium in this study may be partly due to the presence of missed diagnosis of activitydecreasing subtype.
This study showed that old age is a risk factor for delirium, and while age increases, the number of risk factors for delirium increase gradually as well (15). Besides, our research demonstrated that the incidence of postoperative delirium in patients with a history of COPD was signi cantly higher than patients without COPD. COPD patients suffer from cerebral ischemia, hypoxia, metabolic disorders due to primary cardiopulmonary insu ciency, and secondary injury of brain function likely induces postoperative delirium. Patients that underwent emergency operation showed abnormal stress state and disordered hypothalamus pituitary adrenocortical axis (HPA axis) activation. Under emergency operation, 5serotonin, norepinephrine, and dopamine increase, while acetylcholine levels decrease, causing metabolic disturbance. These pathological changes like aggravate brain injury and induce delirium (16).
Uncontrollable factors, such as advanced age, COPD history, emergency surgery, are unmodi able in the clinical work. So what are controllable factors? How does intraoperative and postoperative medication use affect the incidence of delirium in ICU patients. Sedatives and analgesics, such as lorazepam, midazolam, pethidine, and morphine are most likely to cause mental confusion, which may be due to their long duration of actions and increased risk of drug accumulation with altered organ function (e.g., renal and hepatic insu ciency), compared with propofol, dexmedetomidine and fentanyl (17). Propofol is the most widespread narcotic medicine for anesthesia maintenance and induction(18). The advantages of this medicine include quick onset, short recovery time, and less incidence of adverse effects such as postoperative nausea and vomiting. Our studies adjusted some common confounding factors such as the patient's age, demographic characteristics, comorbidities, use of other drugs, type of surgery, baseline cognitive status, etc. Our studies showed that the incidence of delirium in patients receiving low-dose propofol perioperatively was lower compared to patients that did not receive propofol, consistent with other previous studies that a low amount of propofol could successfully treated delirious patients(19).
Although some studies have proposed speci c interventions to reduce the rate of delirium, others have considered evidence-based prevention strategies to assess whether they can further reduce the rate of delirium if applied together in a consistent manner. Usually, pain is the protective response to the nociceptive stimulus. However, continuous postoperative pain could cause negative emotional reactions of anxiety, stress and fear, leading to physical dysfunction. The observational prospective longitudinal cohort study by Alexandra R Feast found a link between pain and delirium, suggesting that pain may be a risk factor for delirium. Because pain and delirium are found to continue and develop during hospitalization, pain and delirium need to be regularly evaluated to effectively manage pain and delirium(20).