Delirium in a Latin American intensive care unit. A prospective cohort study of mechanically ventilated patients

Objective To establish the prevalence of delirium in a general intensive care unit and to identify associated factors, clinical expression and the influence on outcomes. Methods This was a prospective cohort study in a medical surgical intensive care unit. The Richmond Agitation-Sedation Scale and Confusion Assessment Method for the Intensive Care Unit were used daily to identify delirium in mechanically ventilated patients. Results In this series, delirium prevalence was 80% (N = 184 delirious patients out of 230 patients). The number of patients according to delirium psychomotor subtypes was as follows: 11 hyperactive patients (6%), 9 hypoactive patients (5%) and 160 mixed patients (89%). Multiple logistic regression modeling using delirium as the dependent outcome variable (to study the risk factors for delirium) revealed that age > 65 years, history of alcohol consumption, and number of mechanical ventilation days were independent variables associated with the development of delirium. The multiple logistic regression model using hospital mortality as the dependent outcome variable (to study the risk factors for death) showed that severity of illness, according to the Acute Physiology and Chronic Health Evaluation II, mechanical ventilation for more than 7 days, and sedation days were all independent predictors for excess hospital mortality. Conclusion This Latin American prospective cohort investigation confirmed specific factors important for the development of delirium and the outcome of death among general intensive care unit patients. In both analyses, we found that the duration of mechanical ventilation was a predictor of untoward outcomes.


INTRODUCTION
Delirium is a form of brain-organ dysfunction associated with significantly higher mortality, longer times on mechanical ventilation, extended time in the intensive care unit (ICU) and increased hospital length of stay (LOS) as well as cognitive impairments at one year after hospital discharge. (1)(2)(3) Surveys in Latin American, (4) Brazil (5) and Uruguay (6) revealed that most of the surveyed intensivists did not use a tool for delirium evaluation in their practices.
However, the Society of Critical Care Medicine (SCCM) is conducting the ICU Liberation Collaborative and the ABCDEFCare Bundle to implement the Pain, Agitation, and Delirium Guidelines (www.icudelirium.org and www.iculiberation.org).
These mechanisms of quality improvement can be adapted for use in Latin American countries; however, prospective data are required to understand the epidemiology of this area of ICU care. We also collected outcomes and a myriad of clinical data at baseline to conduct statistical modeling and analyze the following two main questions: What are the risk factors for delirium? What are the risk factors for hospital mortality?
The aim of this study was to establish the prevalence of delirium in a general ICU in Uruguay and to identify predictors of delirium and mortality.

METHODS
This study was conducted between April 20, 2014 and April 20, 2015. The setting was a 25-bed general ICU of the Hospital Pasteur, in Montevideo, Uruguay. A general ICU accepts medical, coronary, surgical and trauma patients.
Hospital Pasteur is a non-university facility, although it has a university medical residencies program. This study design was a single-center prospective cohort study. The inclusion criteria were patients admitted to the ICU and were 18 years of age and older who required mechanical ventilation (MV) for more than 48 hours. The exclusion criteria were as follows: use of noninvasive mechanical ventilation, patients transferred from another medical facility, patients being transferred to another medical facility that prevented adequate follow-up, ICU readmission in less than 48 hours after discharge and patients with severe neurological and neuropsychiatric pathology.
Patients with severe neurological and neuropsychiatric pathology were defined as those with a disability that makes communication impossible between patients and the health team; therefore, questions would have to be formulated using the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU).

Education and training in delirium monitoring instruments
From January 1 st to March 31 st , 2014, training workshops for the Richmond Agitation-Sedation Scale (RASS) (7,8) and the CAM-ICU (9)(10)(11) evaluation were conducted during every shift for doctors, residents, and nurses.

Delirium definition and diagnosis
We used the Diagnostic and Statistical Manual of Mental Disorders 5 th edition (DSM-V) (12) and a validated Spanish version of the CAM-ICU (10,11,13) as diagnostic tools to follow a prospective cohort of ICU patients. A delirium diagnosis was determined when the CAM-ICU was positive.
We defined delirium as a disturbance or fluctuation in mental status from baseline accompanied by inattention and either altered level of consciousness or disorganized thinking. (11,12) There must be evidence from the history, physical examination, or laboratory findings that the disturbance is a physiologic consequence of another medical condition, substance intoxication or withdrawal. Figure 1 shows the diagnostic flow chart used in our investigation. The first step was to assess consciousness status using the RASS. If the patient was alert, the status was an RASS score equal to or greater than -3, following which the CAM-ICU was performed. Accordingly, an acute change in mental status from baseline or a fluctuating mental status during the previous 24 hours as well as inattention were then evaluated. If these parameters were altered and the patient had an RASS score different from 0, a delirium diagnosis was confirmed (as per the CAM-ICU algorithm -http://www.icudelirium.org/docs/ CAM_ICU_flowsheet_Spanish.pdf ).
The RASS was evaluated in the morning before CAM-ICU data were collected and during periods of hyperactivity as well as when there were agitation periods registered by the nursing staff.
The CAM-ICU was performed daily in the morning shift and, in cases of agitation, during other shifts.
Delirium was categorized as follows: Hyperactive patients were CAM-ICU positive and agitated, restless or emotionally labile (i.e., RASS +1 to +4). Hypoactive delirium was characterized when patients were CAM-ICU positive and the RASS was between 0 and -3. Such patients had decreased responsiveness, apathy, lethargy, reduced motor activity, incoherent speech and lack of interest in interactions due to their inattention. If the patient presented a combination of both hyperactive and hypoactive delirium, it was considered a mixed delirium. Hallucinations or delusions are common in both forms but are not diagnostic criteria for delirium.
Any positivity noted within a 24-hour period was considered a "positive" delirium day.
Prospectively, we collected information on age, sex, diagnosis, medical history (alcohol intake, illicit drugs abuse, smoke, non-severe neurological psychiatric diseases), duration in the ICU and in-hospital LOS, Acute Physiology and Chronic Health Evaluation II (APACHE II) scores, total MV days, MV days prior to the first episode of delirium, RASS, Glasgow Coma Scale, CAM-ICU, sedative and analgesic drugs and number of days of use, use of neuroleptic drugs and intravenous steroid use.
Patients were also categorized according to the ICU admission diagnosis: (1) medical or surgical patients and (2) according to the ICU's admission diagnostic codes, which include 27 diseases.
RASS and CAM-ICU monitoring were performed while patients were under invasive mechanical ventilation.
All study variables, including outcome (death or recovery), were recorded until ICU discharge.
At hospital discharge, only the outcome and length of stay were registered.

Statistical analysis
Demographic and clinical variables were summarized using descriptive statistics. Continuous variables were described using the mean and standard deviation (SD) or median and interquartile range, depending on data distribution. Normal distribution was assessed using the Kolmogórov-Smirnov test. A comparison between delirium and non-delirium groups for continuous variables was performed using Student's t-test for independent samples with normal distribution, the Mann-Whitney U test for samples in which the distribution was not normal, and the chi-square test or Fischer's exact test for qualitative variables. P-values less than 0.05 were considered significant.
To establish risk factors associated with delirium development and death at ICU discharge, a logistic regression analysis was performed.
Variables that showed p < 0.20 in the univariate analysis and those considered clinically relevant were elected to compose the multivariate analysis. The results of the multivariate analysis are expressed as odds ratios with 95% confidence intervals.
State Software (State Corp LP, version 13) was used for all calculations.

RESULTS
From April 20, 2014 to April 20, 2015, 1104 patients were admitted to the ICU. From these patients, 290 fulfilled the inclusion criteria, and 230 patients were included in the study ( Figure 2). In table 1, the cohortbaseline characteristics are shown for all patients according to delirium status (never delirium versus ever delirium).
We found significant differences between the delirium group and the non-delirium group regarding analgesia and sedation days that were noted in the clinical outcomes including ICU LOS (p = 0.01) and hospital LOS (p = 0.01) as well as hospital mortality (p = 0.01) ( Table 1).
Two logistic regression models were performed with the first one using delirium as a dependent variable and the second one using mortality as a dependent variable.
For the first model, using delirium as a dependent variable, the variables included were as follows: age over 65 years, a history of alcohol consumption, psychiatric disorders, post-surgical pathology, gender and number of mechanical ventilation days prior to the first episode of delirium (Table 2). Multivariable analysis revealed that age over 65 years, a history of alcohol consumption, surgical pathology and number of mechanical ventilation days prior to the first episode of delirium were independent variables associated with delirium development (Table 2). For the second model, using ICU mortality as a dependent variable, the variables included were as follows: delirium, number of analgesia days, sedation days and MV days, use of intravenous steroids, APACHE II scores, acute coronary syndrome, shock and sepsis. Multivariable analysis revealed that APACHE II values, MV for more than 7 days, acute coronary syndrome and number of sedation days were independent variables associated with ICU mortality (Table 3).

DISCUSSION
We prospectively evaluated incidence, outcomes, and predictors of delirium in mechanically ventilated critically ill patients in a general ICU (medical, coronary, surgical and trauma) using CAM-ICU as a diagnostic tool. In this cohort, the prevalence of delirium was 80% (N = 184). Older age, alcohol consumption, extended in-hospital LOS, MV for more than 7 days, use of analgesia and sedation and in-hospital mortality were associated with delirium development. The multivariate regression model using delirium as a dependent variable identified age > 65 years, a history of alcohol consumption, perioperative period and MV for more than 3 days as independent variables.
The other multivariate regression model using ICU mortality as a dependent variable recognized number of sedation days, mechanical ventilation for more than 7 days, APACHE II scores and acute coronary syndrome as independent variables.
Delirium prevalence was commensurate with previous literature. (2,14,15) According to Salluh's meta-analysis of 44 articles (N = 16,595 patients), only 8 papers revealed a prevalence of delirium greater than 60%. (1) In one paper, Cruz et al., (16) stated that there is a broad range of delirium incidence that may change according to the type of studied patients, the interpretation of clinical findings and the diagnostic screening tools.
If confirmed in other regional ICUs, this high delirium rate is a clarion call for quality improvement efforts in Latin America designed to reduce the burden of acute brain dysfunction in these ICU patients.
Furthermore, this cohort identified that the duration of mechanical ventilation was the most consistent independent predictor of both delirium and high mortality. This provides a target for future work such as the implementation of the ABCDEF bundle and the ICU Liberation Collaborative, since these evidencebased approaches have been shown to improve outcomes including a reduction in duration of mechanical ventilation and delirium and mortality. (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29) Several cohort studies (1,28) have shown that in six-month and one-year periods, the risk of mortality increased by 10% per day that an ICU patient was delirious. Moreover, it is known that the duration of delirium is an independent predictor of cognitive impairment.
Mehta et al. (30) reported 3.6 days for delirium onset and a duration of 2 days, while Klein Klouwenberg et al. (31) reported a duration of 3 days and Shehabi (15) reported a duration of 4 days. The latter author found a doseresponse increase in mortality with increasing durations of delirium from 0 to 1, 2, and 3 delirium days. We used a similar interval from 1 to 3 days and more than 4 days and arrived at consistent findings.
Delirium psychomotor behavior was categorized as hyperactive in 11 patients (6%), hypoactive in 9 patients (5%) and mixed in 160 patients (89%). In Peterson's study, (32) similar findings occurred including a mixed type, which was the most common type (54.9%). Thus, most of the delirium observed in the ICU is hypoactive or mixed, which requires routine monitoring for detection given that 75% of hypoactive delirium is missed. (33)(34)(35) Hyperactivity and agitation occur frequently in ICU patients. Almeida et al. (33) found that delirium was an independent risk factor for agitation in the first 7 days after ICU admission.
Risk factors for delirium could be predisposing or precipitating risk factors. Older age is a well-known predisposing factor widely recognized and closely related to other predisposing factors such as multiple comorbidities including underlying cognitive impairment. Our finding that age over 65 years is an independent delirium factor is in accordance with these previous data. McNicoll et al. (36) found four preexisting factors for delirium in older patients admitted to medical ICUs (dementia, administration of benzodiazepines before ICU admission, elevated creatinine levels and low arterial pH). In Uruguay, the elderly population is growing; consequently, there is an increase in admission rate for elderly patients to the ICU. Due to this aging population, the number of patients at risk of delirium and neuropsychiatric disorders is large and should serve to initiate all the quality improvement that is possible.
As explained in Pitrowsky's (37) study, alcohol consumption and gender are two non-modifiable factors. In our study, both male gender and alcohol consumption were significant for the occurrence of delirium according to the univariate analysis. Alcohol is associated with delirium, which agrees with most of the findings included in the literature that identify alcohol as a predisposing risk factor for delirium. Male gender was associated with alcoholism in our univariate analysis.
The development of delirium in the ICU was an important predictor of increased mortality, (2) and clinical practice guidelines conclude that delirium is associated with high mortality in adult ICU patients. Guidelines also report with high quality evidence that delirium is associated with prolonged ICU and hospital LOS in adult ICU patients. Our results agree with these findings. Regarding other outcomes, our data showed that delirium was associated with significantly increased ICU and hospital LOS. In Salluh's (1) meta-analysis, 28 studies reported that ICU LOS was significantly longer in patients with delirium. Regarding hospital LOS, 22 studies reported that the LOS was significantly longer in patients with delirium.
In Salluh's paper, (1) 10 studies reported that MV duration was a predictor of delirium development. In our series, the number of MV days was 3.5 days for the nondelirium group and 7 days for the delirium patients, which was very close to the findings by van den Boogaard et al. (38) but dissimilar from the average difference of 1.79 days in Salluh's (1) meta-analysis. The logistic regression model using either delirium or mortality as dependent variables exhibited number of MV days as independent predictors of the unwanted outcome. However, it is true that more MV days might merely indicate sicker patients, and the higher mortality rate might intuitively reflect this higher severity of illness. However, numerous other investigators have adjusted for severity of illness and found that delirium is an independent predictor of mortality, which was similar to our findings. Thus, these data complement prior work and for only the second time, provide data from Latin America about the association between delirium and mechanical ventilation and mortality.
We did not find an association between analgesia and sedation days and delirium onset. Multiple publications have shown that benzodiazepines are associated with higher rates of delirium (39,40) and that by reducing benzodiazepines, delirium prevalence is decreased. (16)(17)(18)(19)(20)22) Although there were significant differences between delirious and non-delirious patients, regarding the number of analgesia and sedation days prior to the start of delirium, multiple logistic regression models using delirium as the dependent variable revealed no significant relationship. It is noteworthy that our ICU policy on analgesia and sedation is restrictive; consequently, the difference in the average number of sedation days (1 day versus 2 days) is far less than that reported in the literature. Previous randomized trials that have compared daily sedation interruption with control strategies have helped form a basis for the current ABCDEF Care Bundle and the ICU Liberation Collaborative. (41) It is clear that different approaches towards sedation reduction are possible, and these should be embraced at the local level to increase compliance and overall success. In the end, however, this is likely a key element for future clinical attempts and investigations to reduce mechanical ventilation duration in Latin America with the overarching goal to improve the epidemiology of delirium and hospital mortality.
This investigation has important limitations that must be addressed. First, the ICU in this study was present in one region of a major city in Uruguay, which is strong methodologically, yet these data may not be generalizable to other regions of Latin America. Second, the modeling used was a univariate/multivariate approach rather than an a priori establishment of the clinically relevant covariates, which is preferred by many biostatisticians. In addition, many variables were handled as dichotomous thresholds (e.g., delirium ever/never, age > 65) rather than as continuous or time-varying covariates. Even considering this limitation, however, the results were very consistent with prior work and thus carry face validity. Some variables did not reach significance due to a small sample size. It is well known that studies with a small number of participants are subject to bias, even if the number to reach statistical significance in the sample size calculation is adequate.
The main findings of this cohort will likely be used to initiate better quality improvement in Latin American ICUs where many clinicians have been perhaps overly satisfied with the status quo. Having had very few data from the entire continent other than from Brazil, it has been easy to assume that no problem existed; however, these 80% delirium rates cannot be ignored. One of our study strengths was the health team training in RASS and CAM-ICU since these tools were not routinely used before the protocol implementation. Routine use of RASS and CAM-ICU, which is recommended as usual care (but poorly implemented) in the Pain, Agitation, and Delirium Guidelines, allowed us to identify that 80% of these ventilated Uruguayan ICU patients suffered from delirium, which is consistent with previous cohorts but certainly higher than preferred. The findings are complementary to those of other regions of the world and represent only the second prospective dataset from Latin America in this field.
Lastly, future work should expand the size and regions studied as well as utilize interventions and long-term cognitive and physical outcomes, neither of which was measured in this investigation.

CONCLUSION
This prospective cohort investigation provides novel data from Latin America to further our understanding of the epidemiology of delirium and mortality in intensive care unit patients in this region. Routine use of the RASS and CAM-ICU allowed us to identify an 80% delirium rate in which risk factors for delirium development were age older than 65 years, alcohol consumption, perioperative period and number of mechanical ventilation days. Delirium was associated with extensive intensive care unit and hospital length of stay; and in-hospital mortality.
Hospital mortality risk factors were number of sedation days, illness severity graded as APACHE II measurements and the presence of acute coronary syndrome.

ACKNOWLEDGMENTS
Without the support of the Hospital Pasteur ICU health team, this study could not be performed.
We would like to acknowledge Engineer Rafael Alonso, Associate Professor of Quantitative Methods of the Medical School of the Universidad da Republica.
Métodos: Trata-se de um estudo prospectivo de coorte em uma unidade de terapia intensiva clínico-cirúrgica. Avaliamos os pacientes diariamente, com a Richmond Agitation-Sedation Scale e a Confusion Assessment Method for the Intensive Care Unit, com o objetivo de identificar delirium nos pacientes mecanicamente ventilados.