Therapeutic Interventions are Important Risk Factors for Delirium in the ICU Setting (cid:0) A Multicenter Observational Study

Delirium is an important independent predictor of negative clinical outcomes in intensive care unit (ICU) patients. The purpose of this study was to investigate the territorial incidence of ICU delirium, its related risk factors, and short-term outcomes in Shandong Province, China, to provide precise information for territorial patient management. A multicenter prospective observational study was conducted. Patients with delirium were dened as any patient with at least one positive CAM-ICU or ICDSC assessment. Demographics, admission clinical data, daily interventions provided to patients and environmental factors were collected.

This study was registered in the Chinese Clinical Trial Registry (ChiCTR1900021360).

Background
Delirium has recently been rede ned in the DSM-5 as an acute disturbance in attention and awareness, with additional disturbances in cognition not explained by a pre-existing neurocognitive disorder and caused by another medical condition [1]. Delirium impairs the patient's ability to receive, process, store and recall information. Delirium is common among intensive care unit (ICU) patients, depending on the type of ICU, acuity of illness, presence of mechanical ventilation, and other risk factors [2]. The incidence of delirium ranges from 16-80% in the ICU [2][3][4]. It can lead to short-term and long-term adverse outcomes in patients. For instance, ICU delirium is associated with nonplanned tube drawing, increased mortality, prolonged hospitalization and mechanical ventilation. It may also increase hospitalization costs and the occurrence of cognitive disorders after discharge from the ICU and even disability after hospitalization [5][6][7][8]. Although delirium is associated with a poor outcome, it has not received su cient attention from clinicians [2,9]. Delirium is often poorly recognized even by experienced physicians, which may lead to inappropriate treatments and poor outcomes [10].
Over 100 different risk factors have been described for delirium. Barr et al. stated that the risk factors that are positively and signi cantly associated with the development of delirium in the ICU are (1) pre-existing dementia, (2) a history of hypertension and alcoholism, and (3) a high severity of illness on admission [11]. Acute illness, medications, and environmental effects alter delirium in different ways.
Each additional point on the APACHE II score at admission increases the risk of delirium by 5-6% [12]. A number of investigations have reported a positive relationship between benzodiazepine use and delirium occurrence in critically ill adults [13][14], while Zaal documented that benzodiazepine use, especially with continuous infusion, causes a dose-dependent increase in the risk of delirium [15]. Environmental factors such as daylight and ICU sound may affect the occurrence of delirium. Critically ill patients admitted to single-bed rooms have a lower prevalence of delirium than those admitted to multibed rooms [16].
Numerous pharmacological and nonpharmacological strategies have been investigated as delirium treatments, but consistent results regarding improved outcomes have not been obtained [17]. None of the studied agents signi cantly improved clinical outcomes, including mortality [18][19][20]. Therefore, delirium prevention strategies, such as the evaluation and modi cation of risk factors, are prioritized by current critical care practice guidelines [21].
The risk factors associated with delirium vary depending on the purpose of the study and the selected cohort. The relatively large variation is probably associated with many factors, such as patient populations, screening instruments for delirium (the confusion assessment method for the diagnosis of delirium in the ICU (CAM-ICU) vs Intensive Care Delirium Screening Checklist (ICDSC)), and diagnosis and treatment habits. Risk factors for delirium have gradually attracted the attention of clinicians and researchers in China. According to our previous investigation [22], knowledge, attention, and evaluations of delirium in the ICUs of Shandong Province were signi cantly de cient, which may explain why epidemiological investigations of risk factors for delirium are rare. Thus, the risk factors for delirium in Shandong Province, a northeast seaside province with a population of approximately 100 million people, must be examined to provide unique endemic evidence for the prevention and control of delirium and to explore the incidence and short-term outcomes of patients with ICU delirium in tertiary hospitals.

Study design and participants
Consecutive patients admitted to the participating ICU centers of tertiary hospitals in Shandong Province were evaluated for eligibility from May 1, 2018 to Jan 31, 2020. Before patient recruitment, written informed consent was obtained from each subject or his/her legally authorized representative (surrogate). Adult patients (≥ 18 years old) who were admitted to the ICU for more than 24 hours and signed the informed consent form were enrolled in the study. Patients were excluded from this study if they had 1) coma, dementia, cerebral infarction or hemorrhage, brain trauma, or mental disorders; 2) a previous history of delirium or other mental disorders and long-term use of psychotropic drugs; or 3) were unable to be assessed because of vision, hearing and physical disabilities. All patients were treated according to ABCDE bundles and spontaneous breathing trials (SBTs). The choice of drug dosage and the duration of administration were determined by the physicians.

Delirium screening
Two of the trained research team members in each participating ICU separately assessed the included patients two times during the day of the investigation. Delirium screening was carried out twice daily at 8:00 AM − 10:00 AM and 14:00-16:00 PM using CAM-ICU and ICDSC. All patients were evaluated for delirium until discharge from the ICU or death. Patients with delirium were de ned as any patient with at least one positive CAM-ICU or ICDSC assessment during our trial.
Data collection and quality control Data obtained included:1) Baseline characteristics: sex, age, smoking history, obstructive sleep apnea hypopnea syndrome, chronic lung disease, diabetes, heart disease, chronic renal insu ciency, chronic liver disease, cerebrovascular disease, and history of hypertension; 2) Disease condition: principal diagnosis, the indication for ICU hospitalization, trauma, days of ICU stay, APACHE II score, and Sequential Organ Failure Assessment (SOFA) score; 3) Vital signs: temperature and mean arterial pressure; 4) Laboratory tests: pH value, partial oxygen pressure, oxygen saturation index, and levels of albumin, bilirubin, creatinine, BUN, blood sodium, blood phosphorus, serum cholinesterase, procalcitonin (PCT), C-reactive protein (CRP), lactate, and hemoglobin; 4) Treatment: sedative drugs and dosage, vasoactive drugs and dosage, arti cial airway, mechanical ventilation, continuous renal replacement therapy (CRRT), central venous catheter, invasive artery, drainage tube, urinary catheter, stomach catheter, nutrition support therapy, restraint, ability to see the clock, single room, and hormone use; 5) Delirium assessment: Richmond Agitation Sedation Scale (RASS), CAM-ICU and ICDSC; and 6) Outcomes (after 28 days): duration of the ICU stay and duration of the hospital stay.
We explored the main risk factors for ICU delirium using all of the related data described above. All data were collected electronically using an electronic clinical report form. Each participating unit had a unique password to log into this online electronic system to ensure con dentiality.
All research team members who executed the assessment of delirium and data collection were experienced professional ICU staff. Each member of the research team was trained on the signi cance, objectives, speci c contents, and methods of the research before and during the study according to the quality control processes executed by the supervisor of the study. Data accuracy and compliance were also guaranteed by the same quality control team and unscheduled meetings of all responsible persons at the participating centers. Every participating site chose a speci c date in each month to enroll and assess all patients who were treated in the ICU.
Data on center management were collected and uploaded to each participating site through electronic data online software. Each site chose one day in each month to execute the assessment of delirium in the morning and afternoon, depending on their convenience, but they had to repeat the assessment on the same day in the following months until the end of the study. On the study day, the two researchers from the same center independently began their assessment and data collection at the same time. Every included patient was assessed, and all the related information was recorded at the same time.

Statistics
Patient characteristics, related covariates and outcomes were summarized using descriptive statistics. X 2 tests were used to verify differences between categorical variables that were possibly related to the development of delirium. The t-test was used for normally distributed data, and the Mann-Whitney rank sum test was used for nonnormally distributed data.
Multivariable logistic regression analysis was used to investigate the risk factors associated with the status of delirium, and we used variable selection. Odds ratios are reported with 95% con dence intervals (95% CIs). All tests were 2-sided, and a p value < 0.05 indicated statistical signi cance. Data handling and analyses were performed with SPSS 25.0(IBM SPSS Statistics 25.lnk. Chicago, IL, USA).

Demographics and Characteristics of Participants
During the 27-month study period, 11 ICU centers participated the study, including 7 teaching hospitals and 4 territorial medical centers. One thousand one hundred ninety-eight patients were screened and 536 noncomatose patients were ultimately eligible for the study, including 333 male patients (62.1%) with a median age of 64 (IQR, 51 to 74) years and an APACHE II score of 13 (IQR, 10 to 19) at the time of admission. The median length of ICU stay was 8 days (IQR, 3 to 17). One hundred eighteen patients (22%) experienced delirium at least once during the ICU stay. The characteristics of all patients are shown in Table 1.  In addition to these routine, widely studied risk factors, we also evaluated daily interventions.
Differences between the delirium and nondelirium group were observed in most of these interventions, including sedatives and analgesics such as midazolam, propofol and butorphanol; enteral nutrition; arti cial airway; nasogastric tube, use of restraint straps, and the ability of patients to see the clock. Results are shown in Table 3. Among the patient's vital signs and laboratory examinations recorded on the day of assessment( shown in Table 4), body temperature and BUN and CRP levels were risk factors for delirium.

Multivariate analysis
Risk factors with p < 0.05 in the univariate analysis were included in the regression analysis. The multivariate logistic regression analysis revealed 6 factors that were retained in the optimized multivariate model, of which enteral nutrition was a protective factor, while midazolam, chronic renal insu ciency, and restraint were risk factors.Shown in Table 5.

Consequences of delirium
The duration of mechanical ventilation in patients with delirium was signi cantly higher than that in patients without delirium [9 (IQR: 4-17) days vs 6 (4-12)days, p < 0.05], and the length of stay in the ICU was longer.Shown in Table6.

Discussion
Shandong Province is one of the most populous provinces in China, with a population of approximately 100 million, and it is the origin of Confucian culture. Delirium is associated with culture because it deeply affects the population at almost every level, especially at the psychological level [23]. Therefore, an investigation of delirium in Shandong Province has special meaning. To the best of our knowledge, this study is the rst Chinese multicenter analysis of delirium.
Approximately 22% of patients experienced delirium at least once during the ICU stay, similar to our previous study (28.75%) [22]. Patients with delirium were likely to have a longer duration of mechanical ventilation, ICU stay, and hospital stay. In the univariate analysis, age, SOFA score, a history of cerebrovascular disease and impaired renal function were risk factors for delirium. Most therapeutic interventions were associated with delirium in the univariate analysis, including enteral nutrition, arti cial airways, nasogastric tubes, use of restraint straps and sedative medications, such as midazolam, propofol and butorphanol. Among the patient's vital signs and laboratory examinations reported on the day of assessment, body temperature and BUN and CRP levels were risk factors for delirium.
Consistent with previous studies, age and disease severity have always been risk factors for delirium [24,25]. Patients with delirium may more frequently be over 65 years of age and have SOFA scores ≥ 2. Unlike previous studies, no difference in APACHE II scores was observed between patients with and without delirium. We found a large number of errors in the APACHE II scoring process, which may explain this result. We are performing additional research to address this discrepancy.
Although hypertension was identi ed as a risk factor for ICU delirium in a systematic review[26], hypertension was not associated with delirium in our study, even in the univariate analysis. Moreover, alcohol abuse is one of the most commonly identi ed risk factors for delirium in Western countries, but in our prior study, we did not nd that an alcohol abuse history affected the development of delirium in critically ill Chinese patients, and alcoholics were seldom affected by delirium. Alcohol abuse may be associated with the culture of the society. Therefore, we did not collect any data on alcohol use in the present study.
These risk factors, such as age and severity of illness, are uncontrollable or even immutable for ICU patients. One of the aims of the current study is to differentiate modi able risk factors and treat these factors in a systemic manner.
Midazolam, butorphanol, and propofol were risk factors for delirium in the univariate analysis. Consistent with other studies, midazolam remained an independent risk factor for delirium in the multivariate analysis [15]. Benzodiazepines and propofol have high a nities for the γ-aminobutyric acid receptor in the central nervous system, and their administration leads to alterations in the levels of numerous neurotransmitters thought to be deliriogenic [27]. Thus, reducing sedative exposure might be an important intervention to improve patient outcomes. However, in our study, 22.0% of patients used midazolam, suggesting that we should reduce the use of long-acting sedatives as much as possible in the future and monitor the sedation status of patients more frequently.
Regarding short-acting agents, several studies have shown that dexmedetomidine may be helpful to minimize the depth and duration of sedation with a potential reduction in the time to extubation and days of delirium in the ICU [28,29]. However, in our study, dexmedetomidine was not correlated with the occurrence of delirium, potentially due to the difference in sedation depth compared with other studies.
Pain is one of the most commonly reported stressors for ICU patients. The negative physiological and psychological consequences of unrelieved pain in ICU patients are signi cant and long lasting. Therefore, su cient analgesia should be ensured in all ICU patients, and potentially painful procedures should be mitigated with a preventive analgesic approach [11,30]. We found that the proportion of analgesics used in this study was small, and butorphanol was a risk factor in the univariate analysis. Little research has been conducted on butorphanol, especially its effects on delirium. Few studies have compared the superiority of fentanyl to butorphanol in ICU analgesia. Because opioid side effects depend on the dose, these drugs must be compared in equianalgesic doses [31]. We hypothesize that butorphanol may affect delirium by affecting gastrointestinal function. eCASH emphasizes the need to reduce total opioid exposure to avoid adverse effects such as feeding intolerance, constipation and ileus [32].
Gastric tubes, visible clocks, enteral nutrition, physical restraints, arti cial airways, and mechanical ventilation were related to delirium in the univariate analysis, consistent with previous research results [22]. Most ICU patients have varying degrees of gastrointestinal dysfunction, and indwelling gastric tubes are common. Few studies have shown the relationship between nutrition and delirium, but our research shows that patients receiving enteral nutrition have a lower risk of delirium than those receiving parenteral nutrition. The speci c reasons for providing early EN are to maintain gut integrity and modulate stress and the systemic immune response to attenuate disease severity. A study focusing on the effect of early enhanced enteral nutrition on patients with a head injury found that enhanced EN appears to accelerate neurological recovery and reduce both the incidence of major complications and postinjury in ammatory responses. However, the mechanism is unclear [33]. Therefore, the study described above reminds us that gastrointestinal function, not nutrition modes, may play an important role in delirium. In patients receiving enteral nutrition, the brain-intestinal axis may regulate central nervous system function. However, the relationship between gastrointestinal function and delirium requires further testing.
Patients treated with mechanical ventilation were more likely to develop delirium in the ICU setting than those without mechanical ventilation, consistent with other studies [34]. In the univariate analysis, the incidence of delirium in patients with tracheal intubation or tracheotomy was higher than that in other patients. The probable explanations are that these patients had a more serious illness and needed larger doses of sedatives and analgesics, and the ndings may also be related to noise, sleep deprivation, patient-care interaction and the mode of mechanical ventilation. Based on current evidence, the causal relationship between these factors is di cult to distinguish, but we should try different methods, including protocols, to help patients wean from ventilation. Measures should be implemented to shorten the mechanical ventilation time as soon as possible.
The ICU environment will also affect the development of delirium in patients potentially because of different noise and light levels due to the ICU architectural design [16,35]. Physical restraints are important means to ensure patient safety, handle agitation and prevent unplanned extubation. However, the incidence of delirium is high when physical restraints are used on patients. Restrictions on the use of restraints in the ICU are required to reduce the occurrence of delirium [36].
Even compared with our own previous study, the risk factors that affect delirium are different. The

Limitations
The study also has several limitations. First, as a drawback of observational studies, some unknown or untestable confounding factor(s) affect our internal validity. Second, the study lasted for 27 months, during which changes in therapeutic interventions and cognition might have affected the results. Third, the routine practice for each participating ICU was not the same, which may lead to bias. Additionally, this study was carried out in Shandong Province, China, where the ICU setting may be different from other provinces in China and other countries, which may limit the external validity of the study. Third, we did not follow the patients to determine the long-term effect of delirium on patient prognosis.
De ned delirium risk factors may help practitioners prevent or implement appropriate therapies to reduce delirium in ICU patients and improve their prognosis. We expect that after the success of the study, more ICU medical staff will recognize, pay attention to and routinely evaluate patients for a delirium diagnosis.
Therefore, the incidence of delirium will soon be gradually reduced, and the prognosis of ICU patients will be improved.

Conclusions
Delirium in the ICU setting was associated with a longer duration of mechanical ventilation and a longer ICU stay. In addition to chronic renal insu ciency, CRP and BUN levels, medical interventions, including midazolam usage, physical restraints and the nutrition support mode, were all important risk factors for delirium in our study. Based on the present study, physicians should pay attention not only to reducing the use of sedatives and analgesics but also to minimizing invasive operations, including nasogastric tubes and physical restraints, on patients. This research is limited to Shandong Province, China, and further research on a larger scale might be needed.

Declarations
Ethics approval and consent to participate The study protocol was reviewed and approved by the Qingdao University Research Ethics Board. Before patient recruitment, written informed consent was obtained from each subject or his/her legally authorized representative (surrogate).

Consent for publication
Not applicable.