Risk factors for acute delirium in critically ill adult patients: a systematic review

,


Introduction
Delirium is a syndrome characterized by disturbances of consciousness, attention, cognition, and perception [1].Delirium has multiple aetiologies, but the predisposing risk factors most frequently cited are older age, cognitive impairment, severity of illness, and iatrogenic causes [2,3].Delirium has an acute onset.Symptoms luctuate over a 2 -hour period [ , 5] .Al th o u ghi t sp r e se n ta ti o ni sty p i call ya s soc ia t edwi th symptoms of hyperactive delirium (restlessness, agitation) [ ], two other subtypes exist, "hypoactive" and "mixed" [1].Hypoactive delirium is characterized by lethargy, reduced activity, and apathy [5], whereas mixed delirium features characteristics of both hyperactive and hypoactive deliriums.Although associated with poor clinical outcomes, delirium is typically reversible [6,7].his has implications for management of the critically ill patient; not only is the patient's life threatened by the primary illness, but also the efects of delirium may cause long term sequelae, principally cognitive deicits, and functional decline [8].Hypoactive and mixed deliriums oten go unrecognized despite being more common than hyperactive delirium [3,8], resulting in undertreatment and poorer outcomes [8,9].Such factors present a challenge to clinicians to identify factors and possibly to prevent delirium in critically ill patients.

Methods
h i ss y s t e m a t i cp a p e ri sa na b r i d g e dv e r s i o no faf u l l online publication available at the Joanna Briggs Institute Library of Systematic Reviews (http://connect.jbiconnectplus.org/JBIReviewsLibrary.aspx) [31].

. Inclusion and Exclusion
Criteria.his paper considered studies including randomised controlled trials, nonrandomised controlled trials, and before and ater studies.In their absence, cohort and case control studies were considered for inclusion.Participants were adults (aged 21 years and above) presenting with delirium (hyperactive, hypoactive, andmixed)intheintensivecareunit(ICU).Synonymssuch as ICU psychosis and ICU syndrome were included.Critically ill patients not in the ICU (e.g., those in the general ward) were excluded.

Search Strategy.
A three-step search strategy was utilised.An initial search was undertaken using the search terms "factors, " "delirium, " and "critical care." A comprehensive search strategy was then developed using identiied keywords and MeSH headings (Table 1).Finally, the reference lists of all identiied studies were examined for additional studies relevanttothereview .Publishedandunpublishedliterature from 1990 to 2012, limited to the English language, was searched using ten databases.

Search Results
. Twenty-two studies were included in the paper (Figure 1; Table 2).he studies were conducted in medical, surgical, and cardiac intensive care units.Twenty studies were prospective and two retrospective cohort studies.
Fiteen studies used the Confusion Assessment Method fortheIntensiveCareUnit(CAM-ICU),withtheRichmond Agitation Sedation Scale (RASS) for the diagnosis of delirium.he remaining studies used other delirium assessment tools: the Diagnostic Statistical Manual-IV (DSM-IV), Confusion Assessment Method (CAM), Intensive Care Delirium Screening Checklist (ICDSC), Nursing Delirium Screening Scale (Nu-DESC), and Delirium Rating Scale (DRS).Only one of the studies used randomized sampling [25], whilst the remainder predominantly used large cohort (range from 20 to 1367 patients) convenience sampling.Due to the heterogeneous nature of the included studies, indings are presented in a narrative review.
2.4.Assessment of Methodological Quality.Studies were identiied for relevance via title, abstract, and keywords.Two independent reviewers assessed content relevance.Full texts of eligible studies were retrieved and reviewed using the appropriate critical appraisal instruments from the Joanna Briggs Institute (JBI) [31].

. Patients Admitted to the Medical Intensive Care Unit.
Peterson et al. [25] examined delirium and its motoric subtypes in a medical ICU (MICU).Data on demographics (age, gender, and race), Acute Physiology and Chronic Health Evaluation-II (APACHE-II) scores, and intubation or extubation were collected from 61 randomised participants.Delirium assessments were extensive and rigorous, generating 7,323 CAM-ICU and 21,931 RASS assessments.Results show that patients aged 65 years and older ( = 156) experienced hypoactive delirium more frequently (71.8% versus 57.%) than younger patients ( = 458), and older age was strongly associated with hypoactive delirium.Mixed type (hyper-, hypoactive) delirium was the most common (5 .9%)amongst other subtypes.
In contrast, Lin et al. [20] examined risk factors for earlyonset delirium in mechanically ventilated MICU patients.However, "early onset" was not deined in the study, and no time measures were recorded.Data was obtained from the medical records of 1 3 patients (including APACHE-II scores, patient's medical history, and alcohol use).Data collection was rigorous; the questionnaires used were previously pilot tested, and research procedures were standardised to ensure reliability.A stepwise logistic regression revealed hypoalbuminemia and presence of sepsis on admission as signiicant factorsinthedevelopmentofearlyonsetdelirium.

Patients
Admitted to the Surgical Intensive Care Unit.Robinson et al. [27] recruited 1 patients who were listed for surgery and required postoperative ICU admission.A pilot study was conducted to assess interrater reliability using the CAM-ICU.A high interrater reliability (kappa statistic > 0.96) ensured internal validity of the results.It w a ss h o w nt h a tp r e o p e r a t i v ev a r i a b l e ss u c ha so l d e ra g e , hypoalbuminemia, impaired functional status, preexisting dementia, and preexisting comorbidities were signiicantly associated with delirium [27] .hi ss u p p o r t st h ei n d i n go f Peterson et al. [25] who showed that preexisting dementia was the most signiicant risk factor for the development of postoperative delirium.Examining the course of delirium in older SICU patients, Balas et al. [15] recruited 117 participants.It contrasts with Robinson et al. [27] in that the Informant Questionnaire onCognitiveDeclineintheElderly(IQCODE)wasusedto assess the presence of dementia.he IQCODE is a validated tool in which dementia is assessed by obtaining information from a surrogate.It was found that older adults admitted to the SICU were at high risk of developing delirium.18. % of the participants had dementia on admission, 28.3% of the participants developed delirium in the SICU, and 22.7% of the participants developed delirium in the post-SICU period.he study used descriptive statistics only.Furthermore, the efects of dementia were not explored.
Angles et al. [13] examined risk factors for delirium ater major trauma in patients admitted to the trauma intensive care unit.Results from this group are reported because the majority of trauma patients require emergency surgery.he study had a small number of participants (=5 9 ).It was s h o w nt h a taG C So f1 2o rl e s s ,h i g h e rb l o o dt r a n s f u s i o n s , and higher multiple organ failure score were signiicantly associated with delirium.
In a study examining the efect of hypoxia on cognition, Guillamondegui et al. [18] recruited 97 ICU patients with multiple traumas without evidence of intracranial haemorrhage.Data such as age, race, length of ICU stay, and injury severity score was recorded, and oxygen saturation was measured.Using the CAM-ICU, 57% of patients were "positive" for delirium.Ater adjusting for injury severity score, oxygen saturation, blood transfusions, and blood pressure, it was revealed that the number of ventilator days and ED pulse rate were signiicantly associated with delirium.

Patients
Admitted to the Cardiac Intensive Care Unit.Afonso et al. [10] created a predictive model for postoperative delirium in 112 cardiac surgical patients.Surgery included coronary artery bypass grat (CABG), valve replacement, and aortic surgery. he incidence of delirium was 3 %.I n c r e a s e da g ea n di n c r e a s e dd u r a t i o no fs u r g e r yw e r et h e most signiicant risk factors for postoperative delirium.
Detroyer et al. [16] also examined postoperative delirium in 10 patients focusing on anxiety and depression as risk factors for postoperative delirium.Unlike Afonso et al. [10] the type of surgical procedure was not recorded.Prolonged intubation time and a low intraoperative lowest body temperature were the most signiicant predictors of delirium.
Similar to Afonso et al. [10], Bakker et al. examined predictors of delirium ater cardiac surgery in 201 patients.A Mini-Mental Status Examination (MMSE) was conducted to assess "global cognitive functioning" [1 ] in the participants before surgery, and medical records were evaluated.In the inal logistic regression model, lower MMSE scores, higher creatinine levels, and longer extracorporeal circulation time were independent predictors of delirium.Mortality during the irst 30 days ater surgery was signiicantly higher in   delirious patients (1 % versus 0%) as compared to nondelirious patients, and adverse events ater surgery were more frequent.
In a retrospective study by Andrejaitiene and Sirvinskas [12] examining risk factors for early postcardiac surgery delirium, participants (=9 0 ) were studied as two distinct groups: light-to-moderate delirium and severe delirium.However, the criteria determining severity of delirium were not described.he term "early" was not deined.In addition, there is no comparator group, casting ambiguity on the "true" incidence of delirium ( .17%).As such, the assertion that delirium caused prolonged hospital stay cannot be justiied.It was shown that administering a dose of fentanyl above 1.mg increased the possibility of developing severe delirium.Longer aortic clamping time was also noted as an independent predictor of severe delirium.New atrial ibrillation (AF) episodes also occurred more frequently in patients with severe delirium than those with light-to-moderate delirium.
h es t u d yb yS c h o e ne ta l .[ 28] aimed to examine preoperative and intraoperative cerebral oxygen saturation and its association with postoperative delirium in patients undergoing on-pump cardiac surgery.231 participants were recruited.Cerebral oxygen saturation was assessed using cerebral oximetry, detecting "imbalances in the cerebral oxygen supply/demand ratio" [28].Older age, lower MMSE scores, neuropsychiatric disease, and lower preoperative cerebral oxygen saturation scores were independent predictors for postoperative delirium.However, the patient's sedatives, which may have a profound efect on the development of delirium, were not recorded.

Pharmacological Factors.
Pandharipande et al. [23] examined sedatives and analgesics as risk factors for "patients' transition to delirium." One hundred and ninety-eight mechanically ventilated patients admitted to medical or coronary ICUs were recruited.Using a Markov regression model, it was found that lorazepam was an independent risk factor for daily transition to delirium.Other sedatives and analgesics, such as midazolam, fentanyl, morphine, and propofol, were not signiicant, although they were "associated with trends towards signiicance" [23].
In a follow-up study, Pandharipande et al. [2 ]i n v e stigated the efects of sedatives and analgesics in patients admitted to the surgical ICU (SICU) and trauma ICU (TICU).One hundred mechanically ventilated patients were recruited.Midazolam was found to be a strong risk factor for transition to delirium.However, opiate exposure was inconclusive in that opiates such as fentanyl were a risk factor Agarwal et al. [11] recruited eighty-two adult ventilated patients in burns ICU.Benzodiazepines were found to be independent risk factors for the development of delirium.Results suggest that benzodiazepines were a strong risk factor for the transition to delirium.In comparison to the study by Pandharipande et al. [2 ], opiates and methadone appeared to have protective efects, being associated with a lower risk of delirium.
he association between nurse-administered midazolam and incident delirium was examined by Taipale et al. [30]ina prospective observational study.122 participants undergoing cardiac surgery were recruited.In this ICU setting, there were no formal sedation protocols other than the physician's standing orders and sedatives which were administered pro re nata (PRN) by nurses.his study was notable in the creation of study variables when the diagnosis of delirium did not match those of the physicians' (overall agreement = 71.3%);this had not been done previously.here was also a detailed accounting of recruitment, and measures were taken to enhance reliability of CAM-ICU assessments between researchers.Results showed that, for every additional milligram of midazolam administered, patients were 7-8% more likely to develop delirium.

Evaluation by Other Instruments.
In the medical ICU, three studies were reviewed.Eden et al. [17]a p p l i e df o u r previously studied predictive models designed to predict susceptible ICU patients.his study used the CAM and DSM criteria for delirium diagnosis.Unlike other studies, this study has a small sample size; it has an elderly sample of ten delirious and ten control patients only.Fourteen independent variables were operationalised and incorporated into data collection tools.A composite of these predictive models was synthesized and showed that co-morbidity, presence of infection, a blood urea nitrogen/creatinine ratio of 18 or more, and age were the most signiicant variables, with a sensitivity of 100% and a speciicity of 90%.Ranhof et al. [26]conductedtheirstudyinasubintensivecareunitforthe elderly, recruiting 01 patients.he researchers also used the CAMtodiagnosedelirium.Deliriumwasfoundin29.2%of the patients, of which 13.7% developed delirium in the ICU.Heavy alcohol use, polypharmacy (7 or more drugs), and the use of bladder catheter were predictors of delirium.Ouimet et al. [22]examineddeliri umin820I CUpa tien tsusingthe ICDSC.A history of hypertension, alcohol use (similar to the p r e v i o u ss t u d yb yR a n h o fe ta l .[ 26]), higher APACHE II score, and administration of sedative and analgesic drugs was associated with delirium.
In the surgical ICU, one study was reviewed.Shi et al. [29]conductedastudyinaChineseICUexaminingboththe incidence and risk factors of delirium in 16 patients ater noncardiac surgery. he researchers used the Nu-DESC, a delirium screening tool validated in the Chinese population. he results showed the predictive factors of delirium to be increasing age, history of previous stroke, high APACHE II score on SICU admission, and high serum cortisol level on the irst postoperative day.
In the cardiac ICU, three studies were reviewed.Hudetz et al. [19] examined the incidence of delirium in patients undergoing valve surgery with or without CABG as com-paredtopatientsundergoingCABGalone.Forty-four" education balanced" patients were recruited from the ICU of one veteran afairs medical centre.he ICDSC was used to diagnose delirium before surgery and ive days ater surgery .Postoperativedeliriumoccurredmorefrequentlyin patients undergoing valve surgery with or without CABG as opposed to CABG alone.Uguz et al. [1]c o n d u c t e das t u d y which measured the incidence of delirium as it relates to acute myocardial infarction (AMI) as opposed to surgical procedures.Two hundred and twelve patients who were admitted to the coronary intensive care unit were recruited and assessed using DSM-IV criteria and the DRS.Independent predictors of delirium were advanced age, higher level of serum potassium at admission, and experience of cardiac arrest during MI. he retrospective study by Norkiene et al. [21]h a dav e r yl a r g es a m p l es i z e( = 1367).he researchers studied the precipitating factors for delirium ater CABG and screened for delirium using the DSM criteria.Eight factors were independent predictors of delirium, which w e r ea g em o r et h a n6 5y e a r s ,p e r i p h e r a lv a s c u l a rd i s e a s e , a EuroSCORE (European System for Cardiac Operative Risk Evaluation) more or equal to 5, preoperative intraarterial blood pressure support, blood product usage, and postoperative low cardiac output syndrome.

Discussion
From the studies reviewed, there are a variety of candidate factors associated with delirium in the setting of the intensive care unit.Some are common across all settings, whereas others are exclusive to the type of setting.For example, the importance of valve surgery as a risk factor for delirium [10] is of key importance in a cardiac ICU but lacks importance in the medical ICU, where one is more likely to see cases of sepsis, acute respiratory failure, and renal disease.
In the medical ICU, older age, sepsis, co-morbidity, and heavy alcohol use were the most commonly cited risk factors.Older age is considered a highly signiicant risk factor for delirium due to a reduced synthesis of cerebral neurotransmitters [32].Fluctuations in the neurotransmitter levels lead to impairment in neurotransmission, resulting in increased susceptibility to delirium in older patients.he mechanism by which sepsis causes delirium is poorly understood; however several theories have been postulated; these include brain activation by inlammatory mediators, oxidative stress, and blood-brain barrier breakdown [33].It is possible that all these theories are valid; the manifestation of delirium is likely multifactorial, precipitated by cytokine pathways resulting in the derangement of neurological function.he presence of co-morbidity is not easily explained, although it might be expected that efects on increasing physiological burden may play a part.Heavy alcohol use is known to be associated with delirium tremens, a form of delirium caused by withdrawal of alcohol [3 ].
In the surgical ICU, older age, presence of co-morbidity (including previous history of stroke and dementia), and high APACHE-II score are the most cited risk factors.With a higher APACHE-II score, there is a greater physiological stress with concomitant increase in risk for delirium.
In the cardiac ICU, there were no factors which stood out more signiicantly than others (other than older age and lower MMSE scores).All other factors are likely to be equally signiicant.A study examining all these factors in a composite model is required to determine the most signiicant factors causing delirium in the CICU.
With regard to pharmacological factors, benzodiazepines were identiied as a signiicant risk factor for ICU delirium.Benzodiazepines increase the efect of the neurotransmitter GABA, resulting in increased sedation and hypnosis [23]. he efect on GABA may cause an imbalance in the action and quantity of the other neurotransmitters, causing symptoms to manifest as delirium.In addition, benzodiazepines may cause behavioural disinhibition and aggression [2 ], symptoms similar to hyperactive delirium.
In this paper, two retrospective cohort studies were included in a majority of prospective studies.In comparison, prospective studies are preferred to retrospective studies as patients are available for accurate assessment and examination; in a retrospective review, it is not possible to conirm the patient's condition.A retrospective review further compounds a problem inherent in delirium: diagnosis.Physician's d i a g n o s e sm a yb es u b j e c t i v e ;a ss u c h ,o n ep h y s i c i a nm a y view a patient as delirious whilst another might regard it as preexisting dementia.he propensity for misinterpretation and incorrect diagnosis may be signiicant in clinical settings which do not use standardised criterion such as the CAM-ICU to determine diagnosis.hough the methodology and results of retrospective studies may be apocryphal, they are included in this paper for the sake of completeness.

. Implications for Practice and Research
(i) By creating a predictive model for delirium, clinicians may be able to identify patients at risk of developing delirium and implement preemptive measures.his can be further developed into an ICU-speciic model.For example, a patient in the medical ICU will have a diferent set of risk factors, such as the presence of sepsis, co-morbidity, and alcohol use, from a patient in the cardiac ICU.A protocol based on this model will assist the nurse in monitoring patients at higher risk for developing delirium, identifying modiiable risk factors to prevent or reduce the severity of delirium.
(ii) Clinicians should prescribe benzodiazepines judiciously, moderated by an understanding of the patient's mental status and propensity for developing delirium.Conversely, the precipitation of delirium in a patient prescribed benzodiazepines must be considered in the context of the patient's condition and not attributed to pharmacological reasons alone.
(iii) An alternative to using benzodiazepines as sedatives may be haloperidol.van den Boogard et al. [35] found that haloperidol prophylaxis resulted in lower delirium incidence and more delirium free days as compared to the control group.However, the results still need to be veriied via a reliable randomised controlled trial.
(iv) Randomised control trials should be conducted to investigate the eicacy of other possible sedatives such as dexmedetomidine or opioids in comparison to benzodiazepines.
(v) Strength of studies could be further improved by increasing sample sizes, recruiting from more than one hospital and examining diverse factors in order to synthesise stronger evidence.Future studies may examine the efects of biomarkers on delirium in depth, possibly isolating key biomarkers in the pathway leading to delirium.
(vi) An examination of all the factors examined in the recent literature may be conducted, in order to create a composite model for predicting delirium.his predictive model can be used in future in tandem with research which examines interventions to reduce the incidence of delirium.
4.2.Limitations.his paper was limited by the parameters set in the search strategy; any relevant studies prior to 1990 were not included, possibly inluencing the review indings.I twasalsolimitedbypotentialreportingbias,as" published studies tend to overreport positive and signiicant indings" [36].Only studies written in English were included, possibly excluding relevant studies in other languages.Variability in the results may be attributed to the diference in sample sizes.Diferent study objectives, such as measuring pre-and postoperative variables and biomarkers, may have inluenced the results of the studies.

Conclusion
Old age is a common risk factor for delirium in critically ill adult patients.In both medical and surgical ICUs, risk factors of older age and co-morbidity are signiicant, whilst heavy alcohol use and higher APACHE II scores are signiicant in medical and surgical ICUs, respectively.In the cardiac ICU, a variety of factors were signiicant, such as age and lower MMSE scores.Benzodiazepines are singled out as a signiicant risk factor for delirium.
f o rd e l i ri u mi nth eS I C U ,b u tn o ti nth eT I C U .I na d d i ti o n , o p i a t e ss u c ha sm o r p h i n ew e r el i n k e dt oal o w e rr i s kt o delirium.

Table 2 :
Studies included in the paper.