State of the Science in Pediatric ICU Delirium: An Integrative Review

Abstract

Delirium is a complication of critical illness associated with poor outcomes. Although widely studied in adults, comparatively little is understood about delirium in pediatric intensive care units (ICU). The purpose of this integrative review is to determine the extent and nature of current evidence, identify gaps in the literature, and outline future areas for investigation of pediatric ICU delirium. Eligible articles included research reports of delirium in pediatric ICU samples published in English since 2009. After an extensive literature search and consideration for inclusion/exclusion criteria, 22 articles were chosen for review. Delirium was highly prevalent in the ICU. Delirium episodes developed early in hospitalization, lasted several days, and consisted of hypoactive or mixed motor subtypes. Frequently identified independent risk factors included young age, developmental delay, mechanical ventilation, and benzodiazepine exposure. Pediatric delirium was independently associated with increased length of stay, costs, and mortality. The long-term cognitive, psychological, and functional morbidities associated with pediatric delirium remain largely unknown. Few researchers have implemented interventions to prevent or manage delirium. There was little evidence for the efficacy or safety of pharmacological management. Multicomponent delirium bundles may significantly decrease delirium incidence. Key quality issues among studies included variation in delirium screening, low levels of evidence (i.e., observational studies), and limited ability to determine intervention efficacy in quasi-experimental designs. Although the quantity and quality of pediatric delirium research has rapidly increased, further studies are needed to understand the long-term effects of pediatric delirium and determine the efficacy and safety of interventions for prevention and management.

Delirium is a complication of critical illness known to be associated with poor outcomes. Presenting as a disturbance in attention and awareness accompanied by a change in cognition (e.g., disorientation, perceptual disturbance), delirium develops over a short period of time with symptoms fluctuating throughout the day (American Psychiatric Association [APA], 2013). Although widely studied in adults, little is understood about delirium in pediatric intensive care units (ICU). Prior to 2009, pediatric delirium research consisted mainly of case reports and case series, many with unstandardized screening methods (Hatherill & Flisher, 2010). In 2009, the pediatric Confusion Assessment Method for the ICU (pCAM-ICU; Smith et al., 2011; Smith et al., 2009) became the first widely available assessment tool. The Cornell Assessment of Pediatric Delirium (CAPD; Traube et al., 2014), preschool CAM-ICU (psCAM-ICU; Smith et al., 2016), and Sophia Observational withdrawal Symptom Scale–Pediatric Delirium (SOS-PD; Ista, Te Beest, et al., 2018) became available soon after. The development of validated screening tools led to a rapid increase in the quantity and quality of pediatric delirium research. A recent systematic review identified 14 new research articles published between 2009 and 2015, nine of which were observational cohort studies (Holly et al., 2018). This increase in publications highlights the growing interest in pediatric delirium among researchers and clinicians.

An updated literature review is needed in the rapidly developing field of pediatric delirium research. Integrative reviews provide a summary of the state of the science for a healthcare phenomenon and, in contrast to systematic reviews, allow for the inclusion of experimental and non-experimental research (Whittemore & Knafl, 2005). The purpose of this integrative review is to determine the extent and nature of current evidence, identify gaps in knowledge, and outline future areas for investigation of pediatric ICU delirium.

Methods

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines informed the conduct of the literature search, data extraction, and analysis of articles (Moher et al., 2009). Covidence systematic review software (Veritas Health Innovation, n.d.) was used for article screening, full-text review, and risk of bias assessment. The primary author (LBK) performed database searching, study selection, data extraction, and risk of bias assessment. Through frequent meetings with the secondary author (TMH), inclusion/exclusion criteria were refined, variables of interest were determined, and uncertainties in study selection and risk of bias assessment were discussed and agreed upon.

Eligibility Criteria

Eligible articles included research reports of delirium in human pediatric (i.e., ≤21 years) samples in the ICU (e.g., medical, surgical, cardiac) published in English. Exclusion criteria included (a) case series or reports, (b) mixed study samples that included both adult (i.e., >21 years) and pediatric patients or both ICU and non-ICU admissions, (c) studies of emergence delirium (transient agitation during emergence from anesthesia) or encephalopathies (e.g., febrile delirium, sepsis-associated encephalopathy), (d) reports of the development of pediatric delirium screening tools, (e) reports of clinician surveys on current clinical practices related to delirium, (f) studies in which the method of delirium screening was unspecified, and (g) studies published prior to 2009. This time range reflects the last decade of research, as well as the first emergence of validated screening tools specific to pediatric delirium (Smith et al., 2009).

Search Strategy

A search of the PubMed, Cumulative Index of Nursing and Allied Health Literature (CINAHL), MEDLINE, PsychINFO, and Embase databases was performed in October 2018 and May 2019. A secondary search was conducted by reviewing reference lists of relevant articles. A variety of search terms were used (e.g., delirium, deliria, delirious syndrome) to account for the heterogeneity of delirium terminology, and terms such as pediatric, child, infant, and intensive care unit, pediatric were used to narrow the search to pediatric critical care. See Table S1 for the full search strategy.

Study Selection

See Figure 1 for process of study selection. After importing all eligible articles (n=5,645), Covidence removed duplicates (n=2,431), leaving 3,214 articles for screening. During initial review of the titles and abstracts, irrelevant articles (e.g., emergence delirium, published prior to 2009) were removed, leaving 349 articles for full-text review. Following full-text review, 327 articles were removed based on inclusion/exclusion criteria. No additional articles were identified through a secondary search of the reference lists of relevant articles. Therefore, 22 articles were included in the literature review.

Figure 1.

Process of Study Selection

Note. Adapted from “Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement” by D. Moher, A. Liberati, J. Tetzlaff, D.G. Altman, & The PRISMA Group, 2009, PloS Med, 6(7), p.e1000097.

Data Extraction

Main variables of interest were delirium prevalence, duration, presentation, risk factors, outcomes, and prevention and management. After careful review of the articles and abstracted data, results were summarized in two synthesis tables (Tables 1 & 2).

Table 1:

Synthesis Table of Observational Studies

Citation	Study Design	Sample Size	Age Range (years)	Setting	Delirium Screening		Delirium Prevalence	Delirium Risk Factorsa	Delirium Outcomes
					Tool	Frequency
Alvarez et al., 2018	Cohort	99	0–21	Surgical CICU	CAPD	BID	57.00%	Young Age	↑ MV Duration
								MV	↑ LOS
								Benzodiazepines
Cano Londoño et al., 2018	Cross-Sectional	77	5–14	PICU	pCAM-ICU	-	19.50%	-	-
					DRS
					DEC
Madden, Hussain, & Tasker, 2018	Cohort	88	0–18	PICU	Psychiatric Evaluation	Variable	17.30%	-	-
Meyburg et al., 2017	Cohort	93	0–17	Surgical PICU	CAPD	BID	65.60%	-	↑ MV Duration
									↑ LOS
									↑ Resource Utilization
Meyburg, Dill, et al., 2018	Cohortb	93	0–17	Surgical PICU	CAPD	BID	65.60%	Young Age	↑ Nursing Workload
								Infection
								MV	workload
Meyburg, Ries, et al., 2018	Point- Prevalenceb	47	1–16	Post-PICU	CAPD	-	-	-	No difference in long-term outcomes.
Mody et al., 2018	Cohortb	580	0–21	PICU	CAPD	QD	21.00%c	Deliriumd	-
								MV
								Benzodiazpinesd
Nellis et al., 2018	Cohortb	1,547	0–21	PICU	CAPD	BID	-	RBC Transfusion	-
Patel et al., 2017	Cohort	194	0–21	Surgical CICU	CAPD	QD	49.00%	Young Age
								Developmental Delay
								Severity of Illness	↑ MV Duration
								Cyanotic Heart Disease	↑ LOS
								Low Albumin
Ricardo Ramirez et al., 2018	Cross-Sectional	156	5–14	PICU		-	18.60%	Developmental Delay	-
								Neurological Disease
					pCAM-ICU			Liver Failure
					DRS			Tachycardia
					DEC			MV
								Psychotropics
								Anticholinergics
Silver et al., 2015	Cohort	99	0–21	PICU	Psychiatric Evaluation	QD	21.00%	Young Age	↑ LOS
								Developmental Delay
								MV
Smeets et al., 2010	Cohort	147	1–18	PICU	Psychiatric Evaluation	Variable	33.30%	-	↑ LOS
									↑ Medical Costs
Smith et al., 2017	Cohortb	300	0–5	PICU	psCAM-ICU	QD	44.00%	Benzodiazepines	↑ LOS
Traube et al., 2016	Cohortb	464	0–21	PICU	CAPD	BID	15.90%	-	↑ Medical Costs
Traube, Silver, Gerber et al., 2017	Cohort	1,547	0–21	PICU	CAPD	BID	17.30%	Young Age
								Developmental Delay
								Severity of Illness	↑ MV Duration
								MV	↑ LOS
								Coma	↑ Mortality
								Benzodiazepines
								Anticholinergics
Traube, Silver, Reeder et al., 2017	Point-Prevalence	994	0–21	PICU	CAPD	-	25.00%	Young Age	-
								MV
								Benzodiazepines
								Antiepileptics
								Vasopressors
								Narcotics
								Physical Restraints

Note. BID = twice daily. CAPD = Cornell Assessment of Pediatric Delirium, CICU = cardiac intensive care unit, DEC = Delirium Etiology Checklist, DRS = Delirium Rating Scale, LOS = length of stay, MV = mechanical ventilation, pCAM-ICU = Pediatric Confusion Assessment Method for the Intensive Care Unit, PICU = pediatric intensive care unit, psCAM-ICU = Preschool Confusion Assessment Method for the Intensive Care Unit, QD = once daily.

^aIdentified by multivariable or multivariate analysis.

^bSecondary analysis.

^cPercentage PICU days.

^dDay-prior.

Table 2:

Synthesis Table of Intervention Studies

Citation	Study Design	Sample Size		Age Range (years)	Setting	Delirium Screening		Intervention	Outcomes
		Intervention	Control			Tool	Frequency
Franken, Sebbens, & Mensik, 2018	Quasi-Experimental	213	108	2–18	PICU	CAPD	BID	Multi-Component Delirium Bundle	Screening Compliance: 6.00–9.00%
									No difference in mean CAPD scores.
Joyce et al., 2015	Cohort	50	-	0–20	PICU	CAPD	Once	Quetiapine	3 Pts. with Prolonged QTc
									0 Pts. with NMS
									0 Pts. with EPS
Rohlik et al., 2018	Quasi-Experimentala	80	80	5–17	PICU	pCAM-ICU	BID	Delirium Screening	Screening Compliance: 57.00%
									↑ Screening Compliance
Sassano-Higgins et al., 2013	Cohort	31	28	0–20	PICU	Psychiatric Evaluation	Once	Olanzapine	↓ Delirium Severity
					CICU	DRS	Twiceb
Simone et al., 2017	Quasi-Experimental	1,875	-	0–22	PICU	CAPD	BID	Delirium, Sedation, Early Mobility Protocols	Screening Compliance: > 95.00%
									↓ Delirium Prevalence
Slooff et al., 2018	Quasi-Experimental	13	-	0–13	PICU	SOS-PD	TID	Haloperidol	Median 2 to 3 days to clinical improvement.
									4 Pts. with EPS
									2 Pts. with Over-Sedation
									0 Pts. with ECG Changes

Note. BID = twice daily. CAPD = Cornell Assessment of Pediatric Delirium, CICU = cardiac intensive care unit, DRS = Delirium Rating Scale, ECG = electrocardiogram, EPS = extrapyramidal symptoms, NMS = neuroleptic malignant syndrome, pCAM-ICU = Pediatric Confusion Assessment Method for the Intensive Care Unit, PICU = pediatric intensive care unit, QD = once daily, QTc = corrected QT interval, SOS-PD = Sophia Observational withdrawal Symptom Scale-Pediatric Delirium, TID = three times daily.

^aIncluded a case-control analysis.

^bAssessed 5 days apart.

Risk of Bias Assessment

To determine the methodological quality of included studies, the Joanna Briggs Institute critical appraisal tools were used (Moola et al., 2017). These tools evaluate the possibility of bias and are specific to each study design (e.g., cohort, cross-sectional) included in the integrative review (Whittemore & Knafl, 2005). Each criterion was marked as met, not met, or not applicable for each study (Tables S2-S5). The degree to which each study did or did not meet criteria determined its overall value to the synthesis of results.

Results

A total of 22 studies involving 6,272 critically ill children were identified. Children were recruited from the pediatric ICU (PICU; n=5,920), cardiac ICU (CICU; n=293), or both (n=59). Designs were primarily cohort (n=14) but also included quasi-experimental (n=4), cross-sectional (n=2), and point prevalence (n=2). Delirium screening methods varied across studies, with some researchers employing multiple tools (Cano Londoño et al., 2018; Ricardo Ramirez et al., 2019; Sassano-Higgins et al., 2013). The CAPD was the most commonly used (n=13), followed by psychiatric evaluation (n=4), pCAM-ICU (n=3), Delirium Rating Scale (DRS; n=3; Turkel et al., 2003), Delirium Etiology Checklist (DEC; n=2; Trzepacz et al., 2011), psCAM-ICU (n=1), and SOS-PD (n=1). In longitudinal studies, delirium screening occurred at least twice daily (n=10), once daily (n=4), or at varying time intervals (e.g., once, as needed; n=4). Researchers focused on describing pediatric delirium, identifying risk factors and associated outcomes, and implementing interventions for prevention and management.

Pediatric Delirium

Researchers described the phenomenon of pediatric delirium, including its prevalence, duration, and presentation.

Prevalence.

Estimates of pediatric delirium prevalence ranged from 15.90% to 65.60% (Meyburg et al., 2017; Traube et al., 2016). Prevalence among mechanically ventilated (MV) children was particularly high (53.00–74.00%; Simone et al., 2017; Traube, Silver, Reeder, et al., 2017). In the general ICU population, rates of delirium were between 15.90% and 33.30% (Smeets et al., 2010; Traube et al., 2016), while a higher occurrence was reported in surgical patients (49.00–65.60%; Meyburg et al., 2017; Patel et al., 2017). However, in an international point prevalence study (n=994), postoperative patients were less likely to screen positive for delirium compared to non-surgical patients (Traube, Silver, Reeder, et al., 2017).

Differences in the prevalence of pediatric delirium early and late in hospitalization remain unknown. In multiple samples, children developed delirium shortly after ICU admission, generally within the first three days (Alvarez et al., 2018; Patel et al., 2017; Simone et al., 2017; Traube, Silver, Gerber, et al., 2017). However, in other samples, children who screened positive for delirium had been in the ICU significantly longer than those who screened negative (Traube, Silver, Reeder, et al., 2017). Similarly, in a small sample of PICU patients with delirium (n=13), the median time to diagnosis was 7 days (Slooff et al., 2018).

Duration.

The duration of a delirium episode is defined as the number of days with at least one positive delirium screen. Researchers estimated a median duration of 1 (Smith et al., 2017) to 2 (Traube et al., 2016) days and an average duration of 2 (Patel et al., 2017) days. Approximately a quarter of children (22.00–26.60%) who developed delirium experienced multiple discrete episodes throughout their ICU stay (Alvarez et al., 2018; Traube, Silver, Gerber, et al., 2017).

Presentation.

Delirium has three motor subtypes: hyperactive, hypoactive, and mixed. Hyperactive delirium is characterized by agitation, irritability, and hallucinations and is recognizable due to these overt symptoms. However, hyperactive delirium accounted for a small percentage (5.00–20.70%) of pediatric delirium days (Alvarez et al., 2018; Ricardo Ramirez et al., 2019). Hypoactive delirium, characterized by withdrawal and lethargy, was more common (46.40–55.20%; Ricardo Ramirez et al., 2019; Traube, Silver, Gerber, et al., 2017). A sizeable percentage (19.20–45.20%) of children experienced the mixed motor subtype (Cano Londoño et al., 2018; Traube, Silver, Gerber, et al., 2017), which presents as a combination of hyper- and hypoactive behaviors. These patterns of delirium motor subtypes in children mirror those experienced by adult ICU patients (Peterson et al., 2006). Alternatively, in a small sample of children with delirium (n=13), Slooff and colleagues (2018) found a high percentage of hyperactive cases (61.00%, n=8) and a low percentage of hypoactive cases (8.00%, n=1).

Researchers described distinctive patterns of postoperative delirium (Alvarez et al., 2018; Meyburg et al., 2017). In these samples, approximately half of the children (46.00–49.20%) experienced mild delirium that resolved within 24 hours. Another subgroup of children (31.00–50.80%) experienced more severe delirium lasting over 24 hours. Scores on the CAPD were significantly higher, indicating more delirium behaviors, in those with the severe form (Meyburg et al., 2017). The majority of children with postoperative delirium experienced the hypoactive (52.00%) or mixed (43.00%) motor subtype (Alvarez et al., 2018).

Several researchers studied sub-syndromal delirium (SSD), a subclinical form of delirium in which individuals present with signs of altered mental status without meeting full diagnostic criteria for delirium (Cole et al., 2013). In a small, cross-sectional sample (n=77; Cano Londoño et al., 2018), 19.50% (n=15) of children screened positive for delirium while an additional 14.20% (n=11) exhibited SSD. The two groups were clinically and demographically homogeneous. Children with SSD presented with an altered level of consciousness with an acute, fluctuating onset. Children with delirium experienced significantly more inattention and disorganized thought processes than children with SSD. No significant differences were noted in motor subtype pattern between children with delirium and children with SSD. In a large sample of critically ill children (n=1,875), Simone and colleagues (2017) noted SSD, defined as a single positive CAPD score, in 2.60% (n=48) of children. Among children on MV, there were significant differences between children with delirium and children with SSD; children with delirium were more likely to be female, have a psychiatric history, and experience an increased MV duration and length of stay (LOS).

Risk Factors

Risk factors for pediatric delirium were divided into non-modifiable and modifiable variables. Only variables identified as independent risk factors through multivariable or multivariate analysis were reported.

Non-Modifiable risk factors.

Non-modifiable risk factors are characteristics of the critically ill child that were present prior to admission.

Age.

Young children were highly vulnerable to delirium. In a sample of 99 critically ill children, those 2 to 5 years of age were most at risk, compared to younger (<2 years) or older (>13 years) children (Silver et al., 2015). Other researchers consistently identified age 2 years or less as an independent risk factor (Alvarez et al., 2018; Meyburg, Dill, et al., 2018; Patel et al., 2017; Traube, Silver, Gerber, et al., 2017; Traube, Silver, Reeder, et al., 2017).

Postoperatively, young children exhibited especially high rates of delirium. In a general surgical sample (n=93), Meyburg and colleagues (2017) found that 86.40% of infants less than a year old screened positive for delirium, and age less than one year was identified as an independent risk factor (Meyburg, Dill, et al., 2018). Alvarez and colleagues (2018) noted that in a sample of 99 children recovering from cardiac surgery, each additional month of age decreased odds of delirium by 65.00%. In another cohort of 194 children admitted to the CICU following cardiac surgery, 55.80% (n=53) of children ever delirious during their CICU admission were 2 years of age or less (Patel et al., 2017).

Developmental delay.

Researchers defined developmental delay as severe, pre-existing impairment in the age-appropriate ability of a child to communicate with their caregiver prior to hospitalization (Patel et al., 2017; Silver et al., 2015), or as a severe neurological disability resulting in dependence on others (Traube, Silver, Gerber, et al., 2017). Through use of the CAPD or psychiatric evaluation, developmental delay was identified as an independent risk factor for delirium in multiple pediatric samples (Patel et al., 2017; Ricardo Ramirez et al., 2019; Silver et al., 2015; Traube, Silver, Gerber, et al., 2017). In an attempt to counteract the low specificity of CAPD screening in children with developmental delay (Traube et al., 2014), some researchers confirmed positive screens with an ICU practitioner or psychiatrist prior to diagnosis (Patel et al., 2017; Silver et al., 2015).

Severity of illness.

High severity of illness, defined as a child’s probability of mortality on admission, was an independent risk factor for delirium (Patel et al., 2017; Traube, Silver, Gerber, et al., 2017). Indicators of high severity of illness, such as infection and tachycardia, were also associated with delirium (Meyburg, Dill, et al., 2018; Ricardo Ramirez et al., 2019). Certain diagnoses were identified as independent risk factors for delirium, including cyanotic heart disease, liver failure, and neurological disease (Patel et al., 2017; Ricardo Ramirez et al., 2019).

Modifiable risk factors.

Modifiable risk factors are clinical variables associated with critical illness and pediatric critical care.

Mechanical ventilation.

Multiple researchers identified MV as an independent risk factor for delirium (Alvarez et al., 2018; Meyburg, Dill, et al., 2018; Mody et al., 2018; Ricardo Ramirez et al., 2019; Silver et al., 2015; Smeets et al., 2010; Traube, Silver, Gerber, et al., 2017; Traube, Silver, Reeder, et al., 2017). Alvarez and colleagues (2018) found that MV increased the odds of delirium by 400.00% in a sample of children recovering from cardiac surgery (n=99).

Sedating medications.

In a large cohort of 1,547 critically ill children, those who experienced a sedation-induced coma during their PICU admission were significantly more likely to develop delirium (Traube, Silver, Gerber, et al., 2017). Pediatric ICU clinicians rely heavily on benzodiazepines as a first-line sedation drug (Kudchadkar et al., 2014). Benzodiazepine exposure was identified as an independent risk factor for pediatric delirium (Alvarez et al., 2018; Meyburg, Dill, et al., 2018; Mody et al., 2018; Smith et al., 2017; Traube, Silver, Gerber, et al., 2017; Traube, Silver, Reeder, et al., 2017). In a cohort of young, critically ill children (n=300), those with 1.00 mg/kg of benzodiazepine exposure per 24-hour period, compared to those with no benzodiazepine exposure, had almost three times the odds of developing delirium the following day (Smith et al., 2017). Mody et al. (2018) used marginal structural modeling to determine the causal effect of benzodiazepine exposure in 580 PICU admissions. Controlling for day-prior delirium, benzodiazepines increased the risk for next-day delirium by 333.00%. Furthermore, in a subgroup analysis of children without prior delirium, those who received benzodiazepines were over 4 times more likely to transition to delirium the following day. A dose-response relationship between benzodiazepine exposure and delirium risk has been established (Meyburg, Dill, et al., 2018; Mody et al., 2018). Psychotropic drugs were also identified as an independent risk factor for delirium (Ricardo Ramirez et al., 2019).

Other medications.

Several other medication classes were associated with delirium risk. Anticholinergic drugs were independently associated with delirium (Ricardo Ramirez et al., 2019; Traube, Silver, Gerber, et al., 2017). In a retrospective cohort of 88 PICU patients with prolonged admissions (i.e., LOS ≥15 days), Madden and colleagues (2018) found a high burden of anticholinergic medication exposure. The four most commonly prescribed medications were midazolam, morphine, vancomycin, and steroids. Although these medications have a low potential for adverse effects compared to other anticholinergics, the prolonged, additive exposure presents risk for critically ill children. While Madden and colleagues did assess delirium, their study was not powered to detect a difference in incidence between children with high and low levels of anticholinergic medication burden.

Researchers reported conflicting findings regarding the association between opioids and pediatric delirium. In a large sample of critically ill children (n=994), opioid exposure was independently associated with increased risk (Traube, Silver, Reeder, et al., 2017). However, in a retrospective analysis of 580 PICU admissions, Mody et al. (2018) found that day-prior opioid exposure was not independently associated with next-day delirium. In their international point prevalence study, Traube, Silver, Reeder, and colleagues (2017) identified antiepileptics and vasopressors as additional medication classes significantly associated with delirium.

Physical restraints.

Physical restraints were identified as independent risk factors for pediatric delirium (Traube, Silver, Reeder, et al., 2017). In a sample of 994 critically ill children, those who were physically restrained had 4 times the odds of delirium, compared to those who were not restrained.

Nutrition.

In a cohort of 194 children recovering from cardiac surgery in the CICU, low albumin levels were significantly associated with delirium (Patel et al., 2017). Children with baseline albumin levels greater than 3.00 mg/dl had lower odds of delirium compared to children with levels below 3.00 mg/dl.

Blood products.

In a secondary analysis of a large cohort of critically ill children (n=1,547), Nellis and colleagues (2018) found an independent association between administration of red blood cells (RBC) and delirium. A dose-response relationship was identified, such that each additional 10.00 mL/kg transfused increased a child’s odds of delirium by 90.00%. Importantly, no association was noted between pre-transfusion hemoglobin level and delirium, suggesting that anemia was not the mechanism increasing delirium risk.

Outcomes

Clinical outcomes.

Researchers found significant associations between pediatric delirium and poor clinical outcomes, including increased MV duration, LOS, medical costs, and mortality.

Mechanical ventilation.

Critically ill children with delirium experienced a significantly increased MV duration (Alvarez et al., 2018; Meyburg et al., 2017; Patel et al., 2017; Simone et al., 2017; Traube, Silver, Gerber, et al., 2017). Across samples, children with delirium spent an estimated 20.70 to 72.00 more hours on MV, on average, than children without delirium (Meyburg et al., 2017; Traube, Silver, Gerber, et al., 2017). This association has not yet been validated with a multivariable or multivariate model to control for potential confounders, such as severity of illness.

Length of stay.

Pediatric delirium was associated with an increased ICU and hospital LOS (Alvarez et al., 2018; Meyburg et al., 2017; Patel et al., 2017; Silver et al., 2015; Simone et al., 2017; Smeets et al., 2010; Smith et al., 2017; Traube, Silver, Gerber, et al., 2017). In a sample of children in the CICU (n=194), Patel and colleagues (2017) found that delirium was associated with a nearly 60.00% increase in CICU LOS. In the multivariable model predicting CICU LOS, only MV rivaled delirium as a risk factor for increased LOS. Traube, Silver, Gerber, et al. (2017) noted that in a large cohort of critically ill children (n=1,547), those who were ever delirious experienced a PICU LOS over twice as long as children who were never delirious. This association remained statistically significant even with control for confounding variables, including severity of illness, MV, age, gender, surgical factors, admission diagnosis, developmental delay, and time to first delirium assessment (Meyburg et al., 2017; Patel et al., 2017; Silver et al., 2015; Smeets et al., 2010; Traube, Silver, Gerber, et al., 2017).

Medical costs.

Children with pediatric delirium used more hospital resources and incurred higher medical costs (Meyburg, Dill, et al., 2018; Meyburg et al., 2017; Smeets et al., 2010; Traube et al., 2016). Based on a conservative estimated delirium prevalence of 5.00% and an estimated increase in PICU LOS of 2.39 days for children with delirium, Smeets and colleagues (2010) determined that annually, pediatric delirium was associated with a 1.50% increase in direct medical costs. In a cost analysis of 464 PICU admissions, pediatric delirium was associated with a 85.00% increase in PICU costs, controlling for LOS, age, gender, and severity of illness (Traube et al., 2016). A dose-response relationship was noted in which costs significantly increased as delirium duration increased, even with control for LOS. Children with delirium had significantly higher costs in multiple categories, including nursing, therapy, laboratory, pharmacy, and radiology. Similarly, in a cohort of critically ill children recovering from surgery (n=93), researchers found significant associations between increased level of care, increased nursing workload, and delirium (Meyburg, Dill, et al., 2018; Meyburg et al., 2017).

Mortality.

Few researchers explored the association between pediatric delirium and mortality. In a cohort of critically ill children (n=1,547), mortality rates were significantly higher for children who were ever delirious during their PICU stay (Traube, Silver, Gerber, et al., 2017). This relationship remained significant even after controlling for the probability of mortality on admission. In fact, delirium was a stronger predictor of mortality than the Pediatric Index of Mortality-3 (Straney et al., 2013), a widely used prognostic indicator.

Long-Term outcomes.

The long-term outcomes associated with pediatric delirium remain largely unknown. In a small pilot study (n=47), Meyburg, Ries, and colleagues (2018) conducted follow-up appointments with participants from an earlier study of postoperative delirium in the PICU (n=93; Meyburg et al., 2017; Meyburg, Dill, et al., 2018). At an average of 17.70 months after discharge, children who experienced delirium did not have significantly different cognitive or behavioral functioning, compared to children who did not experience delirium (Meyburg, Ries, et al., 2018). While these initial results suggest a lack of association between delirium and long-term outcomes, the small sample size and prolonged follow-up time may have prevented the detection of subtle differences in functioning that would be more evident in larger samples followed during the first months after discharge.

Prevention and Management

Clinicians and researchers have begun to develop and test interventions for the prevention and management of pediatric delirium, including implementation of routine delirium screening, multicomponent, bundled interventions, and pharmacological management.

Delirium screening.

Researchers had varying levels of success in implementing routine delirium screening. Simone et al. (2017) used multiple educational strategies during implementation in their PICU, including interdisciplinary education, in-person training, unit champions, and monthly compliance reports. Compliance remained above 95.00% throughout the 22-month study. Franken et al. (2019) attempted to improve delirium screening rates in their PICU using an online education module. Three months after the education, a small improvement was seen compared to baseline, but rates remained low (9.00% vs. 6.00%). Rohlik et al. (2018) tracked compliance during implementation of delirium screening in their PICU. Following in-person training and 9 months of implementation, compliance was 51.00%. Three months later, after additional education (i.e., emails, staff meetings, public display of screening rates), compliance improved to 71.00%. Patient-specific factors associated with increased screening compliance included higher severity of illness and developmental delay. Children on MV were less likely to be screened, as were children admitted when the PICU census was high.

Nurses identified multiple barriers to delirium screening (Franken et al., 2019; Rohlik et al., 2018), including a lack of documentation in the electronic medical record (EMR), dislike of pre-set screening times (i.e., 12PM, 12AM), and “busy” patients. Nurses reported difficulty completing assessments and a lack of knowledge regarding how to screen infants, children with developmental delay, or those undergoing MV. Finally, nurses did not trust the medical team to act upon positive screenings, indicating a need for interdisciplinary cooperation in delirium management. Nurses suggested the use of EMR reminders to perform delirium screening, expanding screening to children of all ages, additional education, and development of a delirium treatment protocol.

Multicomponent, bundled interventions.

Based on evidence of the efficacy of bundled, multicomponent delirium interventions in adult populations (Pun et al., 2019), researchers have begun to adapt and implement these interventions in pediatric ICUs (Franken et al., 2019; Simone et al., 2017). Interventions included validated delirium and pain screening tools, identification of potential delirium etiologies, sedation protocols, avoidance of deliriogenic medications, reorientation, environmental modification, progressive mobilization, family presence, and sleep promotion. During sequential implementation of delirium, sedation, and early mobility protocols over a 22-month period, Simone et al. (2017) noted a decrease in delirium prevalence from 19.30% to 11.80%, with a significant decrease in delirium cases over time. While delirium screening compliance was high throughout the study (95.00%), compliance rates for bundle components were not reported. Franken and colleagues (2019) found no difference in average CAPD scores for children admitted in the 3 months following bundle implementation (n=213), compared to a retrospective control group (n=108). Lack of a significant difference in scores could be due to low rates of screening compliance (6.00–9.00%), which resulted in few positive screenings. Compliance rates for bundle components were not reported.

Pharmacological management.

Few researchers systematically studied the efficacy and safety of pharmacological management of pediatric delirium. In a retrospective analysis, Sassano-Higgins and colleagues (2013) compared delirium symptoms between children who received olanzapine (n=31), an atypical antipsychotic, and children who received no antipsychotics (n=28). Both groups had significant improvement in delirium severity. However, controlling for initial severity, children who received olanzapine experienced greater improvement in delirium symptoms than those who did not. Use of an unvalidated method for assessing pediatric delirium severity, retrospective application of the DRS, decreases confidence in these results.

In a quasi-experimental study of 13 PICU patients, Slooff et al. (2018) evaluated the safety of protocolized management with haloperidol, an antipsychotic. Adverse events were noted in 38.00% (n=5) of the children, including four instances of extrapyramidal symptoms and two of heavy sedation. In half of these instances, the adverse effect resolved after lowering the haloperidol dose. In the other instances, haloperidol was discontinued and in two cases biperiden, a haloperidol antagonist, was administered. The researchers noted a median of two to three days to clinical improvement in delirium symptoms, but the lack of a control group prevents the conclusion that haloperidol shortened delirium duration. Joyce et al. (2015) evaluated the safety of quetiapine, an atypical antipsychotic, for delirium management in 50 PICU patients. Of the 16 children who had an electrocardiogram (ECG) performed during the study period, 18.75% (n=3) had a prolonged corrected QT (QTc) interval. For one child, the prolonged QTc resolved without intervention, while the quetiapine dose was decreased in another. The third child expired, unrelated to the ECG change. Given that only 32.00% of the sample received an ECG, it is difficult to determine the true extent of ECG changes experienced by participants.

Risk of Bias Assessment

Risks for bias were noted within and across study designs (Tables S2-S5). Measurement issues were the most common limitation. Delirium screening methods varied across studies. Several researchers used the DRS and DEC (Cano Londoño et al., 2018; Ricardo Ramirez et al., 2019; Sassano-Higgins et al., 2013), although these tools have not been validated in children. Reliance on psychiatric referral for delirium detection likely led to selection bias in multiple studies (Madden et al., 2018; Sassano-Higgins et al., 2013; Smeets et al., 2010), where agitated children with hyperactive delirium were more often referred to psychiatry than children with hypoactive delirium. Multiple researchers assessed for delirium once daily (Mody et al., 2018; Patel et al., 2017; Silver et al., 2015; Smith et al., 2017), although current guidelines recommend screening at least twice daily due to the fluctuating nature of delirium (Harris et al., 2016). This likely led to missed episodes of delirium. Franken et al. (2019) compared average CAPD scores between an intervention and control group to determine efficacy of a delirium bundle. As scores on the CAPD have not been validated as indicators of delirium severity, average CAPD scores are not a meaningful measure. Simone et al. (2017) defined delirium as consecutive, positive CAPD screenings for at least 48 hours, although delirium diagnostic criteria do not specify a required duration (APA, 2013).

Several findings differed across delirium screening tools. Age 2 years or less was consistently identified as an independent risk factor for delirium when using the CAPD. The low reported specificity of this tool may signify a propensity to consider behaviors of young children indicative of delirium when there are alternative explanations (e.g., pain, anxiety; Leroy & Schieveld, 2017; Traube et al., 2014). Alternatively, this finding may reflect the widespread CAPD use among included studies. Smith and colleagues (2017) reported a high delirium prevalence among children aged 6 months to 2 years (53.00%) using the psCAM-ICU, and young age was a significant predictor of increased delirium duration. Although the SOS-PD was used in only one included study (Slooff et al., 2018), several findings contrasted with general conclusions of this review. First, the authors reported an overall delirium prevalence of 0.84% in their PICU, in contrast to the high prevalence noted in other studies. Second, the motor subtype patterns in this sample contrasted with other studies, with the majority of children experiencing hyperactive delirium. These contrasting patterns are similar to those reported in SOS-PD validation studies (Ista, Te Beest, et al., 2018; Ista, van Beusekom, et al., 2018), suggesting systematic differences between the SOS-PD and other screening tools.

Several limitations inherent to observational research were common among cohort studies (Table S2). Since consent was often obtained after ICU admission, it was impossible to ensure that children were delirium-free prior to study participation (n=8). This limits confidence in the identification of risk factors, as the temporal relationship between cause and effect cannot be verified. Multiple researchers limited participant follow-up time (n=7), potentially missing delirium that developed later in hospitalization. Several researchers failed to report the proportion of missing data or participant drop-outs (n=4), or to perform analyses to ensure that participants with missing data were not significantly different from participants without missing data (n=5). Issues with quasi-experimental design (Table S3), including lack of a control group (n=2) and/or lack of multiple measures before and after the intervention (n=2), limited the ability to determine the effect of delirium interventions on clinical outcomes. Researchers were often unable to report baseline delirium prevalence before an intervention because the study site did not routinely screen for delirium at baseline (Rohlik et al., 2018; Simone et al., 2017).

Variations in delirium screening, inherent limitations of observational research, and the limited ability to determine intervention efficacy in quasi-experimental designs indicate that results of this literature review should be interpreted with caution.

Discussion

Pediatric ICU delirium knowledge is expanding rapidly. Previous authors of systematic reviews noted a lack of pediatric delirium research (Hatherill & Flisher, 2010), especially in critical care (Creten et al., 2011). In contrast, 22 studies comprising 6,272 critically ill children were identified in the present review. Of the included studies, 18 were published after 2015, the last search date of the most recent systematic review (Holly et al., 2018). Study designs were primarily observational with quasi-experimental studies the second most common design, indicating a transition from describing pediatric ICU delirium to actively intervening for prevention and management. Results of this integrative literature review will be discussed with consideration for advances in knowledge in comparison with previous systematic reviews, remaining limitations of current knowledge, and directions for future research.

Pediatric Delirium

Delirium was highly prevalent in pediatric ICUs, affecting up to 65.60% of critically ill children. Delirium episodes developed early in hospitalization, lasted several days, and presented as the hypoactive or mixed motor subtype. In previous reviews, authors identified the lack of validated tools for pediatric delirium detection as a major limitation to research (Creten et al., 2011). In contrast, the increased availability of validated tools allowed researchers to provide more reliable estimates of pediatric delirium incidence, duration, and presentation in the current review. Further descriptive studies are needed to clarify the prevalence of delirium early and late in hospitalization and verify the typical pattern of delirium motor subtypes. Variations in delirium screening noted in this review, including use of tools not validated for children, infrequent screening (i.e., less than twice daily), and differences in findings based on instrument choice, suggest further research is needed to evaluate and refine available screening tools.

Previously identified ambiguities in diagnostic criteria for delirium (Hatherill & Flisher, 2010), which led to confusion with related neurological syndromes such as emergence delirium, febrile delirium, and iatrogenic withdrawal syndrome, were also identified in the present review. Researchers reported distinctive patterns of postoperative delirium and SSD, including significant differences in demographic and clinical characteristics, delirium screening scores, and clinical outcomes in affected children. Further studies are needed to determine if postoperative delirium and SSD are separate clinical syndromes, and to clarify diagnostic criteria specific to pediatric delirium (APA, 2013).

Risk Factors

Patient and clinical characteristics repeatedly identified as independent risk factors for pediatric delirium in this review include young age, developmental delay, high severity of illness, MV, benzodiazepine exposure, and anticholinergic medications. These characteristics were previously identified as potential risk factors for pediatric delirium (Hatherill & Flisher, 2010; Holly et al., 2018). New independent risk factors identified in this review include infection, tachycardia, cyanotic heart disease, liver failure, neurological disease, opioid exposure, antiepileptics, vasopressors, low albumin levels, and RBC administration. Risk factors identified in previous reviews and not confirmed as independent risk factors in the present review include male gender (Silver et al., 2015), pre-existing emotional or behavioral problems, and parent stress or anxiety (Meyburg, Ries, et al., 2018).

Although multiple independent risk factors for pediatric delirium were identified, the observational nature of included studies prevents the confirmation of a temporal relationship between each risk factor (i.e., cause) and delirium (i.e., effect). In certain populations (e.g., elective surgery), it may be possible to recruit children prior to hospitalization to ensure that participants are delirium-free at the beginning of the study. Alternatively, advanced statistical methods in which time-dependent risk factors (e.g., medication exposure) are modeled with next-day delirium as the outcome variable can provide further support for the identification of independent risk factors (Mody et al., 2018; Smith et al., 2017). Further studies are needed to confirm newly identified independent risk factors, as well as previously identified risk factors with little current support, including psychotropic medications and physical restraints.

Outcomes

Pediatric delirium was independently associated with increased ICU and hospital length of stay, ICU costs, and mortality. Further research is needed to determine whether pediatric delirium is independently associated with prolonged MV duration. Hatherill and Flisher (2010) noted a potential association between pediatric delirium and increased mortality, which was confirmed in the present review. Researchers should continue to investigate the association between pediatric delirium and mortality, as current evidence is limited. The long-term cognitive, psychological, and functional morbidities associated with pediatric delirium remain largely unknown. Although initial evidence in a small sample (n=47) assessed over a year after PICU discharge suggested that pediatric delirium was not associated with long-term impairments (Meyburg, Ries, et al., 2018), further short- and long-term follow-up of survivors of pediatric critical illness is needed. This area of research is of particular importance given the growing understanding of long-term morbidities associated with pediatric critical illness known as pediatric post-intensive care syndrome (Manning et al., 2018; Watson et al., 2018).

Prevention and Management

There is little current evidence for the efficacy or safety of pharmacological management of pediatric delirium. While interventions identified in prior reviews were primarily focused on pharmacologic management (Creten et al., 2011; Hatherill & Flisher, 2010), only three studies in the current review addressed pharmacological interventions. Although there were improvements in study design (e.g., quasi-experimental) compared to previous evidence (e.g., case series), knowledge remains limited. The sample sizes of included studies were small and two of the three studies focused on safety, rather than efficacy. Without further evidence of efficacy, it remains unclear whether the adverse effects associated with antipsychotic management outweigh potential benefits. High quality experimental studies are needed to determine the safety and efficacy of pharmacological management.

In contrast, results of the present review reflect an increased interest in alternatives to pharmacological management, including three quasi-experimental studies that involved implementation of routine delirium screening and/or a multicomponent, bundled intervention. Researchers had varying levels of success in their attempts to implement delirium screening. Given the low prevalence of routine pediatric delirium screening (i.e., 2.00–7.00%; Kudchadkar et al., 2014; Staveski et al., 2018), further research is needed to identify successful implementation strategies. Although multicomponent, bundled interventions may decrease the incidence of delirium (Simone et al., 2017), further studies are needed. When adapting interventions from the adult population, researchers should ensure that intervention components are developmentally appropriate for the pediatric population.

Implications

Research.

Major areas for future pediatric delirium research include:

• Evaluation and refinement of valid and reliable screening tools for children of all ages.

• Clarification of diagnostic criteria specific to the pediatric population.

• Investigation of postoperative delirium and SSD as potentially unique clinical phenomena.

• Identification of facilitators and barriers to implementation of routine screening.

• High quality experimental studies to determine the efficacy and safety of multicomponent interventions for prevention and management. This includes measurement of baseline delirium prevalence, documentation of protocol compliance, and use of valid and reliable screening tools.

• High quality experimental studies to determine the efficacy and safety of pharmacological prevention and management. This includes recruitment of a large sample, randomization, a control group, and use of valid and reliable screening tools.

• Short- and long-term follow-up of children following ICU discharge to identify long-term morbidities associated with delirium.

Practice.

Pediatric ICU clinicians should consider the following practice changes:

• Implementation of routine delirium screening facilitated by multiple education strategies, unit champions, EMR documentation of delirium screening, and regular compliance reporting.

• Promotion of interdisciplinary cooperation in delirium screening, prevention, and management through interdisciplinary education, discussion of delirium during interdisciplinary rounds, and implementation of evidence-based interventions.

Limitations

Several limitations of the present review should be noted. First, to focus the review scope, reports of the development of pediatric delirium screening tools or clinician surveys of current clinical practices related to pediatric delirium were excluded. This limits the ability of this review to summarize the current state of the science in these areas. Mixed study samples that included both adult and pediatric or both ICU and non-ICU admissions were also excluded, to maintain a focus on pediatric ICU delirium. This may have led to the exclusion of articles with relevant information related to pediatric delirium. Second, only LBK performed database searching, study selection, data extraction, and risk of bias assessment. Although this increased the chance for human error and bias, LBK engaged in frequent discussion with TMH to refine inclusion/exclusion criteria, determine variables of interest, and resolve uncertainties in study selection and risk of bias assessment.

Conclusion

An updated integrative review of the literature was performed to summarize the current state of the science in pediatric ICU delirium research. The purpose of this review was to determine the extent and nature of current evidence, identify gaps in knowledge, and outline future areas for investigation. Pediatric ICU delirium knowledge is expanding rapidly, with 22 studies comprising 6,272 critically ill children published since 2009. Delirium is highly prevalent in the ICU and independently associated with increased length of stay, costs, and mortality. Researchers are beginning to transition away from describing pediatric delirium towards actively intervening in its prevention and management. Although the quantity of pediatric delirium research has rapidly increased, key quality issues were identified, including variations in delirium screening, low levels of evidence (i.e., observational studies), and limited ability to determine intervention efficacy in quasi-experimental studies. Further research is needed to better understand the long-term effects of pediatric delirium and determine the efficacy and safety of interventions for its prevention and management. Decreasing the incidence and duration of pediatric ICU delirium has the potential to improve neurocognitive symptom management and outcomes in survivors of pediatric critical illness.

Table 3:

Critical Appraisal of Cohort Studies

Criteria	Alvarez et al., 2018	Joyce et al., 2015	Madden, Hussain, & Tasker, 2018	Meyburg et al., 2017	Meyburg, Dill, et al., 2018	Mody et al., 2018	Nellis et al., 2018
Were groups similar and recruited from the same population?	Y	N/Aa	Y	Y	Y	Y	Y
Was the exposure measured similarly across all people?	Y	Y	Y	Y	Y	Y	Y
Was the exposure measured in a valid and reliable way?	Y	Y	Y	Y	Y	Y	Y
Were confounding factors identified?	Y	Y	Y	Y	Y	Y	Y
Were strategies to deal with confounding factors stated?	Y	Y	N	Y	Y	Y	Y
Were the participants free of the outcome at the start of the study?	N	N	N	Y	Y	N	Y
Were the outcomes measured in a valid and reliable way?	Y	N	N	Y	Y	N	N
Was follow-up time long enough for outcomes to occur?	N	N	N	N	N	Y	N
Was follow-up complete, or reasons for loss to follow-up described and explored?	Y	Y	Y	N	N	N	Y
Were strategies to address incomplete follow-up utilized?	N	N/Ab	Y	N	N	Y	N
Was appropriate statistical analysis used?	Y	Y	N	Y	Y	Y	Y
Were groups similar and recruited from the same population?	Y	Y	Y	Y	Y	Y	Y
Was the exposure measured similarly across all people?	Y	Y	Y	N	Y	Y	Y
Was the exposure measured in a valid and reliable way?	N	N	Y	Y	N	Y	Y
Were confounding factors identified?	Y	Y	Y	Y	Y	Y	Y
Were strategies to deal with confounding factors stated?	Y	Y	Y	Y	Y	Y	Y
Were the participants free of the outcome at the start of the study?	N	N/Ac	N	N	Y	Y	N
Were the outcomes measured in a valid and reliable way?	Y	N	Y	Y	Y	Y	Y
Was follow-up time long enough for outcomes to occur?	Y	N	Y	Y	Y	Y	Y
Was follow-up complete, or reasons for loss to follow-up described and explored?	Y	Y	Y	Y	Y	Y	N
Were strategies to address incomplete follow-up utilized?	N	N/Ab	Y	Y	Y	Y	Y
Was appropriate statistical analysis used?	Y	Y	Y	Y	Y	Y	Y

Note. Y = met, N = not met, N/A = not applicable.

^aOne-group cohort design.

^bRetrospective chart review.

^cInclusion criteria included presence of outcome.

Table 4:

Critical Appraisal of Quasi-Experimental Studies

Criteria	Franken, Sebbens, & Mensik, 2018	Rohlik et al., 2018	Simone et al., 2017	Slooff et al., 2018
Is it clear what is the “cause” and what is the “effect”?	Y	Y	Y	N
Were the participants included in any comparisons similar?	Y	Y	Y	Y
Were the participants included in any comparisons receiving similar treatment, other than exposure to the intervention?	Y	Y	Y	Y
Was there a control group?	Y	Y	N	N
Were there multiple measures of the outcome pre- and post- intervention?	Y	N	N	Y
Was follow-up complete and if not, were differences in follow-up between groups adequately described and analyzed?	Y	Y	Y	N
Were the outcomes of participants included in any comparisons measured in the same way?	Y	Y	Y	Y
Were outcomes measured in a reliable way?	N	N	N	Y
Was appropriate statistical analysis used?	Y	Y	N	Y

Note. Y = met, N = not met, N/A = not applicable.

Table 5:

Critical Appraisal of Cross-Sectional Studies

Criteria	Cano Londoño et al., 2018	Ricardo Ramirez et al., 2018
Were the criteria for inclusion in the sample clearly defined?	Y	Y
Were the study subjects and the setting described in detail?	Y	Y
Was the exposure measured in a valid and reliable way?	N	N
Were objective, standard criteria used for measurement of the condition?	Y	Y
Were confounding factors identified?	Y	Y
Were strategies to deal with confounding factors stated?	N	Y
Were the outcomes measured in a valid and reliable way?	Y	Y
Was appropriate statistical analysis used?	Y	Y

Note. Y = met, N = not met.

Table 6:

Critical Appraisal of Point-Prevalence Studies

Criteria	Meyburg, Ries, et al., 2018	Traube, Silver, Reeder, et al., 2017
Was the sampling time frame appropriate to address the target population?	N	Y
Were study participants sampled in an appropriate way?	Y	Y
Was the sample size adequate?	N	Y
Were the subjects and setting described in detail?	Y	Y
Was the data analysis conducted with sufficient coverage of the identified sample?	Y	Y
Were valid methods used for the identification of the condition?	Y	N
Was the condition measured in a standard, reliable way for all participants?	Y	Y
Was appropriate statistical analysis used?	Y	Y
Was the response rate adequate, and if not, was the low response rate managed appropriately?	Y	Y

Note. Y = met, N = not met.