Delirium in the NICU: A Point Prevalence Study

Prospective delirium screening took place at a 46-bed Level IV NICU in an urban, academic setting, which admits >1000 newborns annually. The study took place on 8 predetermined study days spanning from March through August 2019; study days were separated by a median of 14 days. We included every term and term-corrected patient (≥37 weeks' gestation, or ≥37 weeks’ postmenstrual age) who was present in the NICU on each study day. Subjects were excluded only if they were deeply sedated (unarousable to verbal or tactile stimulation).

All subjects were screened with the Cornell Assessment of Pediatric Delirium (CAPD) (Fig 1). The CAPD is an observational delirium screening tool widely used in academic PICUs, provides a reliable method for recognizing pediatric delirium at the bedside, and is validated in infants of term and term-corrected age.¹⁵  It is not a point-in-time screen; rather, it reflects an extended period of observation, and is generally scored by the bedside nurse toward the end of her or his shift. Our operational definition of delirium was a CAPD score of 9 or higher, consistent with the widely accepted case-definition for delirium, which consistently tracks with important outcome measures.^{3,5,6,16–18} 

At ∼3 pm on each study day, the principal investigator approached the bedside registered nurse of eligible subjects. The CAPD was thus scored by the bedside nurse after an ∼8 hour period of observation. Nurses were provided with developmental anchor points for the key ages of newborn, 4 weeks, 6 weeks, 8 weeks, and 28 weeks, for bedside reference (Supplemental Table 3).¹⁹  A limited amount of nonidentifiable clinical information was also recorded, including gestational age at birth, presence of known underlying neurologic disabilities (see Table 1 for categories), and type of respiratory support. The principal investigator entered the results directly into a Research Electronic Data Capture survey via a mobile electronic device.²⁰  The institutional review board granted waiver of consent for this observational, minimal risk study.

Patient characteristics were reported as n (%) and median (interquartile range) for categorical and continuous variables, respectively. Delirium was defined as a CAPD score ≥9. Enrollment goal was 150 infants, with a hypothesized delirium prevalence of 20%, allowing for exact 2-sided 95% CIs to be within ±6% of true prevalence. Hypothesized prevalence was determined from the pediatric critical care literature, which suggests an overall delirium prevalence of ∼25% in the PICU.^2,4 

Univariate analysis identified relationships between patient characteristics and delirium and those determined to be statistically (P < .05) or clinically significant were included in a multivariable regression model to calculate adjusted odds ratios. Model fit and selection was assessed using Pearson’s χ² statistic and Akaike’s Information Criteria, respectively. Analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC).

A total of 151 term and term-corrected infants were present in the NICU on the 8 study days. Two infants were unarousable to stimulation and were thus excluded; another 2 infants could not be screened because the bedside nurse was unavailable at time of sampling. Thus, 147 (98.6%) were successfully screened with the CAPD. Of these, 119 subjects (81%) had no underlying neurologic disability. Eighty-one subjects (55%) were born preterm (<37 weeks' gestational age). Thirty-six infants (24%) were on supplemental oxygen and 15 infants (10%) were on positive pressure MV (Table 1).

For the 147 completed CAPD evaluations, the median CAPD score overall was 3 (interquartile range: 1–7). Thirty-three CAPDs were positive, for an overall delirium prevalence of 22.4%. Median prevalence per study day was 20%, with an interquartile range of 17% to 27%. There was a significant association between MV and delirium (66.7% vs 17.4%, P < .001). There was no significant difference in delirium rates between infants born <37 weeks versus ≥37 weeks (Table 2).

A total of 64% of infants with underlying neurologic disability had a positive delirium screen. Among infants without neurologic disability (n = 119), delirium prevalence was lower at 12.6% (P < .001). Subgroup analysis of infants without neurologic disability revealed similar results to the overall cohort, with MV significantly associated with delirium (55.6% vs 9.3%, P = .002). Again, gestational age at birth did not significantly affect delirium rates.

A multivariable regression showed independent associations between delirium status and underlying neurologic disability (adjusted odds ratio [aOR]: 12.3, confidence interval [CI]: 4.5–33.6), and need for MV (aOR: 9.3, CI: 2.5–34.6).

Case reports have identified delirium in the very young infant, but to our knowledge, we are the first with this systematic study to investigate the prevalence of delirium in the term-NICU population. More than 98% of eligible infants were successfully screened for delirium using the CAPD, and 22.4% screened positive. When excluding infants with an underlying neurologic disability (where the CAPD has been shown to have less specificity), delirium prevalence remained high at 12.8%. These results mirror prevalence rates of 12% to 65% reported in different PICU populations.^{2,4,6,17,18,21,22}  Routine screening for delirium has become the benchmark of prevention, because it creates awareness and vigilance within the care team. Both adult and pediatric trials have revealed that rigorous bundled approaches to screening lowers both delirium rate and number of delirium days.^17,23 

Consistent with delirium studies in other pediatric populations, delirium prevalence was higher in infants with underlying neurologic disease.^18,24,25  In the geriatric population, delirium rates are considerably higher in patients with dementia, because they have less cognitive reserve when faced with the stress of serious illness.²⁶  NICU patients with neurologic disorders could be at similarly increased risk. However, in this population, the CAPD has decreased specificity.²⁷  Although a positive CAPD is still most likely delirium, it may instead reflect the underlying static encephalopathy, rather than the acute and fluctuating mental status change that is required for a diagnosis of delirium.¹⁵  Therefore, subgroup analysis was also performed to include only those children with normal neurologic development, where we are more confident that the positive CAPD represents a diagnosis of acute delirium. Even in this “lower-risk” cohort, delirium prevalence was ∼13%.

In our NICU cohort, delirium prevalence was higher in children on MV. This parallels results in other pediatric populations.^4–6,16,18  For example, in an international point prevalence study of delirium in 994 PICU patients, MV was associated with a 70% increase in odds for delirium (aOR: 1.7, CI: 1.1–2.7).⁴  The association between delirium and MV may reflect both severity of illness and the frequent use of benzodiazepines as sedation in this population.^18,28–31  Severity of illness has been shown to correlate with delirium risk in many studies and is not a modifiable risk factor. However, sedative choice is modifiable. Further research will be needed to explore the relationship between specific medication categories and NICU delirium.

Although this pilot study strongly suggests a high prevalence of delirium in term neonates in the NICU, it would be premature to suggest that the CAPD has been adequately validated in this environment. In the original CAPD validation study, 25 infants were included (7 were <1 month of age). Receiver operating characteristic analysis for these infants showed an area under the curve of 0.911, revealing CAPD validity even in extremely young children.¹⁵  However, a recent study by Valdivia et al revealed a decrease in interrater reliability of the CAPD in infancy, specifically with items 2 and 3, which refer to the infant’s purposefulness and awareness.³²  Further study is necessary to better establish validity and interrater reliability in a NICU setting and possibly expand the developmental anchor points for awareness and purposeful movements in extremely young infants.

This pilot study has important limitations. Because it is a single-center study, results may not be widely generalizable across units, given that inpatient NICUs vary in their admission criteria and patient profile. It is an observational study, so we describe only associations between delirium and certain patient populations, and do not presume causality. Our prevalence design allowed us only to describe delirium prevalence on the chosen study days; because delirium fluctuates over a hospital course, our prevalence rate likely underestimates the true delirium incidence in term neonates in the NICU. We were unable to control for a small number of repeated measures in our data; there is a small chance that this affected our multivariable analysis. Our study assessed only certain potential delirium risk factors, yet many others (exposure to sedative medication, location in single room versus open bay, light and noise pollution, use of restraints) have been implicated in the development of delirium in older subjects. These same factors should also be investigated in the neonatal population. Most importantly, although the CAPD has been validated in infants, no study has yet assessed interrater reliability and validity specifically in the NICU setting. The prevalence rates reported should be taken as prevalence of positive delirium screens, rather than a conclusive diagnosis of delirium.

It is likely that delirium occurs frequently in term neonates in the NICU. Future studies are necessary to establish feasibility, validity, and interrater reliability of delirium screening in this population.