Does Timing of Delirium Onset Affect Outcomes
in The Critically Ill?
Jennifer P. Booth, PharmD; Daniel S. Eiferman, MD, MBA; Judith A. Tate, PhD, RN; and Claire V. Murphy1*, PharmD
1Department of Pharmacy, The Ohio State University Wexner Medical Center. 410 West 10th Avenue, Columbus, OH 43210, USA.
2Department of Surgery, Division of Trauma, Critical Care, and Burn, The Ohio State University Wexner Medical Center. 410 West 10th Avenue, Columbus, OH 43210, USA.
3College of Nursing, The Ohio State University, 368 Newton Hall, 1585 Neil Avenue, Columbus, OH, 43210, USA.
Submission: March 25 2021; Published: January 27, 2022
*Corresponding author: Claire V. Murphy, PharmD, BCPS, FCCM, Lead Specialty Practice Pharmacist – Critical Care, The Ohio State University Wexner Medical Center, Department of Pharmacy, 410 West 10th Avenue Room
How to cite this article: Jennifer P Booth, Daniel S Eiferman, Judith Tate, Claire V. Murphy. Does Timing of Delirium Onset Affect Outcomes in The
Critically Ill?. A Research Article and Review of The Literature. J Anest & Inten care med. 2022; 11(5): 555821. DOI 10.19080/JAICM.2022.11.555821
Introduction: Intensive care unit (ICU) delirium increases morbidity and mortality, but little is known regarding the role timing of development plays.
Objective: To evaluate the effect of timing of ICU delirium onset and identify risk factors for late-onset delirium.
Methods: Clinical outcomes of adult patients with early- and late-onset delirium were compared. A multivariable logistic regression analysis was performed to identify factors associated with late-onset.
Results: A total of 150 patients were included (108 early-onset; 42 late-onset). No difference was observed in ICU length of stay among survivors. In-hospital mortality was significantly higher for late-onset (40.5 vs. 24.1%, p= 0.046). Surgical ICU admission increased risk of late-onset (AOR 3.71, p=<0.001), while hypertension (AOR 0.40, p=0.021) and prehospital antidepressant use (AOR 0.28, p=0.002) reduced risk.
Conclusion: While less common, late-onset delirium was associated with higher mortality risk. Future research should focus on preventative strategies in patients at risk for late-onset delirium.
Keywords:Delirium, Intensive care units, Critical care, Critical care outcomes, Risk factors, Neurocognitive disorders
Abbreviations:ICU: Intensive Care Unit; LOS: Length of Stay; CAM-ICU: Confusion Assessment Method for the ICU; REDCap: Research Electronic Data Capture; IQR: Interquartile Range
Delirium is a common consequence of critical illness, with prevalence of 31.8-87% for patients admitted to the intensive care unit (ICU) and approximately 80% for mechanically ventilated ICU patients [1-3]. Delirium is associated with longer hospital and ICU lengths of stay (LOS), increased mortality, higher healthcare costs, increased mechanical ventilation requirements, and cognitive impairment after discharge [1-3]. Known risk factors for ICU delirium include age, illness severity, alcohol abuse, hypertension, dementia, and respiratory disease [2-4]. Processes of care during ICU admission that may increase risk of delirium include mechanical ventilation and inpatient administration of benzodiazepines, sedatives, and analgesics [1,2,4].
While there has been a significant focus on the identification of risk factors for delirium, few studies have described the timing of delirium onset. Regal and colleagues reported that 95.8% of delirium studies documented diagnoses without timing of onset . Of available studies that report timing of onset, one study reported a mean time to onset of ICU delirium as 1.5-1.7 days which was corroborated by Lin and colleagues who found that nearly half of delirium onset occurred on the second day after admission to the ICU [6,7]. Overall, the number of studies analyzing the time
to onset of delirium is sparse and with significant limitations.
Therefore, this study aims to evaluate the effect of timing
of ICU delirium onset on patient outcomes and identify any
associated risk factors relative to timing. Understanding the effect
of timing of onset may guide management of ICU delirium and
assist clinicians with appropriate risk stratification including the
development of targeted delirium prevention protocols.
A single-center, retrospective analysis of adult patients with
ICU delirium was conducted at a tertiary care, academic medical
center. Adult patients between the ages of 18 to 89 years who were
admitted to the medical or surgical ICU for at least 48 hours and
subsequently developed ICU delirium between January 1, 2016
and December 31, 2016 were eligible for inclusion. Diagnosis
of delirium was defined by the documentation of a positive
Confusion Assessment Method for the ICU (CAM-ICU) at any time
point during ICU admission. Onset of delirium was defined as the
time from ICU admission to the first documented positive CAMICU.
Early-onset delirium was defined as an onset of delirium less
than or equal to 24 hours of ICU admission, whereas late-onset
delirium was defined as an onset of delirium greater than 24 hours
from ICU admission. Patients were excluded if delirium developed
before ICU admission, CAM-ICU status was documented as
“unable to assess” throughout the first 24 hours of ICU admission,
or if they were pregnant, incarcerated, or admitted to the neurocritical
care unit. If a patient met criteria for evaluation on more
than one hospital admission, only the first admission was included
for evaluation; whereas if a patient had multiple ICU admissions
during the index hospitalization, they were excluded from
Patients who developed early-onset ICU delirium were
compared to those who developed late-onset ICU delirium with
a primary outcome of ICU LOS among survivors. Secondary
outcomes included hospital LOS among survivors, all-cause
hospital mortality, and discharge disposition.
Patients were randomly selected using a random number
generator from an electronic report of ICU admissions during the
study period and were then screened for inclusion and exclusion
criteria. Patients were screened until 150 met inclusion criteria
for evaluation. For patients meeting criteria, de-identified data
was then entered and managed using the secure, web-based
application, REDCap (Research Electronic Data Capture) .
Data collection included date and time of delirium diagnosis,
date and time of hospital and ICU admission and discharge,
demographic information, social history including concurrent
substance (alcohol, tobacco, and illicit drug) use, reason for ICU
admission, ICU type, past medical history, and home medications prior to admission. Past medical history was based on
documentation within the provider notes and included: anemia,
chronic kidney disease, chronic respiratory disease (chronic
obstructive pulmonary disease, asthma), cognitive impairment
(Alzheimer’s disease, dementia, Parkinson’s disease, mental
retardation and developmental disability), diabetes mellitus,
heart disease (coronary artery disease, congestive heart failure),
hypertension, liver disease, mental disorder (attention-deficit/
hyperactivity disorder, anxiety, bipolar, depression, schizophrenia,
obsessive-compulsive disorder, posttraumatic stress disorder),
obesity (body mass index > 30 kg/m2), vision/hearing impairment,
and vitamin/mineral deficiency. Home medications were collected
based on the reconciled prior to admission medication list within
the electronic medical record. Home medications collected
included those that are known to have action on the central nervous
system such as anticholinergics, anticonvulsants, antidepressants,
antipsychotics, baclofen, barbiturates, benzodiazepines, clonidine,
gabapentin, mirtazapine, opioids, pregabalin, steroids, trazodone,
Categorical variables are presented as frequency
(percentages) and were compared using the Pearson chi-square.
Continuous variables were summarized by mean + standard
deviation for normally distributed data or median [interquartile
range (IQR) 25-75%] for non-normally distributed data and were
compared using the student’s t-test or Wilcoxon rank-sum test
respectively. All tests were two-tailed, and a p value < 0.05 was
determined to represent statistical significance. After univariable
analysis, a backwards, stepwise, multivariate logistic regression
was performed to determine independent risk factors for lateonset
ICU delirium. Variables significant at the 0.20 level in the
univariable analysis were included in the multivariable logistic
regression model. All statistical analyses were performed using
SPSS version 24.0 for Windows (SPSS, Inc., Chicago, IL).
This study was granted exempt status by the institutional
review board of The Ohio State University (Study #2018E0478).
A total of 2,550 patients were eligible for screening, from
which 476 randomly selected patients were screened to obtain
150 eligible patients for evaluation (Figure 1). Early-onset ICU
delirium was observed in 108 patients (72%), and late-onset ICU
delirium in 42 patients (28%). The primary reasons for exclusion
were an ICU LOS of less than 48 hours and lack of delirium
Baseline characteristics including age, race, reason for ICU
admission, and current substance abuse were similar between
both groups (Table 1); however, the early-onset group had a
lower proportion of males and were more likely to be admitted
to the medical ICU. Significant differences were observed between
groups with regards to comorbidities and prescribed medications
prior to hospitalization. All of the following were observed at a significantly higher frequency in the early-onset group compared
to the late-onset group: past medical history of hypertension and prior to admission antidepressant, benzodiazepine, gabapentin or
Abbreviation: BMI, body mass index.
The median time to delirium onset from hospital admission
was 43.5 hours (13.3-107.6) for the early-onset group and 88.9
hours (47.3-173.3) for the late-onset group; whereas the median
time to delirium onset from ICU admission was 1.8 hours (0.2-
10.6) for the early-onset group and 49.5 hours (35.5-86.1)
for the late-onset group (Table 2). There was no significant
difference in the primary outcome, ICU LOS among survivors, or
secondary outcomes: hospital LOS among survivors and discharge
disposition. However, in-hospital mortality was significantly higher for the late-onset delirium group (Table 2).
After adjusting for male sex and prior to admission
benzodiazepine use, the following factors were significantly
associated with late-onset delirium (Figure 2): surgical ICU
admission (AOR 3.71, 95% CI 1.62-8.48, p<0.001), past medical
history of hypertension (AOR 0.40, 95% CI 0.17-0.91, p=0.021)
and prior to admission antidepressant use (AOR 0.28, 95% CI
Abbreviations: ICU, intensive care unit; LOS, length of stay; IQR, interq
Early-onset delirium was more prevalent in this cohort of
patients that developed delirium during their ICU admission,
with 72% developing delirium within the first 24 hours of ICU
admission. In-hospital mortality was significantly lower among
this population compared to those with late-onset delirium,
suggesting that the timing of ICU delirium may affect clinical
outcomes. Admission to the medical ICU, past medical history
of hypertension, and prior to admission antidepressant use
were independent predictors of the development of earlyonset
delirium; whereas admission to the surgical ICU was an
independent predictor of late-onset delirium.
To the best of our knowledge, this is the first study to
compare early and late-onset ICU delirium across a heterogenous
population of medical and surgical ICUs patients in order to
identify risk factors and effects on clinical outcomes. Our results
are similar to Kim and colleagues who found that surgical patients
who developed post-operative delayed delirium (>5 days following
surgery) were associated with significantly longer duration of
delirium and hospital LOS as compared to medical ward and postoperative
delirium patients (within 5 days) .
Previous studies have dichotomized delirium at ICU admission
at 24 hours and at 5 days to investigate the timing of delirium
and its effect on outcomes [7,9-12]. Although there is no widely
accepted definition distinguishing early and late-onset delirium,
our interquartile ranges for early-onset (0.2-10.6 hours) and lateonset
(35.5-86.1 hours) suggest the existence of a natural gap at a
point in time between 12 and 36 hours. McCusker and colleagues
defined prevalent and incident delirium at admission in their
prospective cohort study ; however, this definition would not
have been appropriate for our retrospective study as it would not
have captured those who developed delirium shortly after ICU
admission and/or those that were not assessed with CAM-ICU
immediately upon admission. Shu-Min Lin and colleagues only
reported results for an early-onset delirium group, defined as
development within five days of ICU admission ; whereas this
definition would not have been appropriate for the current study
as only four patients in our study developed delirium later than
five days into their ICU stay. In other words, five days is likely to
capture most cases of delirium and is not a useful cutoff for early
versus late-onset. Overall, our results are similar to the available
studies that report timing of delirium onset, in which the majority
of delirious patients in our study presented within the first 48
hours of ICU admission.
Our results suggest that medical ICU patients with more
comorbidity may be predisposed to early-onset delirium, whereas
surgical ICU patients tend to develop delirium later in their ICU
stay. We believe there are several plausible explanations for these
findings. First, late-onset delirium in surgical ICU patients may be
multi-factorial and secondary to post-operative complications,
also explaining the significantly higher in-hospital mortality in
these patients. In other words, development of late-onset delirium
may serve as a marker of deterioration in the surgical ICU. This
is corroborated by a retrospective observational study of postoperative
surgical ICU patients by Lee and colleagues that reported
greater ICU LOS, hospital LOS, and mortality with late brief (for
<1 day after post-operative day 0) and persistent delirium (for
≥1 days), compared to early brief (for <1 day on post-operative
day 0) or no delirium . Next, the inflammatory or septic
state of many medical ICU patients and/or higher comorbidity at
admission may contribute to the pathophysiology of early-onset of
delirium development in these patients [9,14]. Finally, early-onset
delirium was also significantly more common in patients with
antidepressants, benzodiazepines, and gabapentin or pregabalin
as prior to admission medications suggesting that the affects
from these neuropsychiatric medications may play a role in the
development of early-onset delirium .
Currently, there is no recommended pharmacological agent
for the treatment of delirium . Instead, non-pharmacologic
prevention and management strategies remain the focus of ICU
delirium management protocols within ICU liberation bundles.
We believe that recognition of the differences between early
and late-onset delirium may assist clinicians with improved risk
stratification and the development of targeted delirium prevention
protocols for high-risk patients. By targeting prevention efforts
based on risk, patients most vulnerable to the development
of delirium may benefit even when resources are limited.
Recommended prevention strategies include multi-component
non-pharmacologic strategies to optimize mobility, sleep, hearing,
and vision . Our results suggest that early-onset delirium may
be refractory to prevention due the limited amount of time for
ICU protocols and interventions to impact delirium development.
Therefore, the focus of delirium management strategies should be
on resolution of early-onset delirium and prevention of late-onset
Limitations of this study include the retrospective study
design including the reliance of data collection from the electronic
medical record, limited sample size, and lack of a widely accepted
definition to distinguish between early and late-onset delirium
. Patients admitted to the neuro-critical care unit were
excluded due to the challenges associated with assessing delirium
in this population; however, patients with traumatic brain injury
were not explicitly excluded. Additionally, the configuration of
our critical care service separates medical ICU and surgical ICU
patients; however, we recognize many ICUs combine both patient populations. The heterogeneity of these two patient populations
limits the generalizability of our study and the conclusions that
can be drawn from this cohort; however, we believe this also to
be a strength as it allowed us to explore potential differences
between two distinct patient populations with inherently different
timelines throughout their ICU LOS.
Early-onset delirium was common among ICU patients with
delirium. Patients in the medical ICU, with hypertension, or on
antidepressants prior to admission were more likely to develop
early-onset delirium. In-hospital mortality was lower in earlyonset
delirium, suggesting that timing may affect clinical outcomes.
Additional research is needed to analyze medical and surgical ICU
patients separately to further explore the differences between
early and late-onset delirium. Better understanding of the role
of delirium timing may help guide ICU delirium prevention and
management and improve clinical outcomes.