Diurnal sedative changes during intensive care: impact on liberation from mechanical ventilation and delirium

Abstract

Objective

To determine whether benzodiazepine and propofol doses are increased at night and whether daytime and nighttime sedative doses are associated with delirium, coma, and delayed liberation from mechanical ventilation.

Design

Single-center, prospective cohort study nested within the Awakening and Breathing Controlled randomized trial

Setting

Saint Thomas Hospital in Nashville, TN from 2004–2006

Patients

Adult patients receiving mechanical ventilation for greater than 12 hours with continuous recording of hourly sedation dosing

Interventions

We measured hourly doses of benzodiazepine and propofol exposure during the daytime (7am–11pm) and nighttime (11pm–7am) for five days. We quantified nighttime dose increases by subtracting the average hourly daytime dose on the preceding day from subsequent average hourly nighttime dose. We used multivariable logistic regression to determine if daytime and nighttime dose increases were independently associated with delirium, coma, and delayed liberation from mechanical ventilation.

Measurements and main results

Among 140 patients, the median APACHE II score was 27 [IQR: 22, 33]. Among those receiving the sedatives, benzodiazepine and propofol doses were increased at night on 40% and 41% of patient-days, respectively. Of 485 patient-days, delirium was present on 160 (33%) and coma on 206 (42%). In adjusted models, greater daytime benzodiazepine dose was independently associated with failed SBT and extubation, and subsequent delirium (p<0.02 for all). Nighttime increase in benzodiazepine dose was associated with failed SBT (p<0.01) and delirium (p=0.05). Daytime propofol dose was marginally associated with subsequent delirium (p=0.06).

Conclusions

Nearly half of mechanically ventilated ICU patients received greater doses of sedation at night, a practice associated with failed SBTs, coma, and delirium. Over the first five days in our study, patients spent 75% of their time in coma or delirium, outcomes that may be reduced by efforts to decrease sedative exposure during both daytime and nighttime hours in the ICU.

INTRODUCTION

Mechanically ventilated patients often receive sedatives to reduce anxiety and ventilator dyssynchrony (1–4). Despite evidence-based recommendations to reduce sedation in the intensive care unit (ICU) (5), moderate to heavy sedation remains common (6–9). In fact, sedation protocols—though widely shown to improve outcomes—are frequently not used (10–12).

Sedatives may lead to many undesirable short- and long-term consequences, including more ventilator and ICU days (13, 14). Benzodiazepines—the most widely used sedative in the ICU (11)—are also an important risk factor for delirium (15, 16). Even months after administration, sedatives may lead to symptoms of depression and post-traumatic stress disorder among ICU survivors (17, 18). As such, many randomized controlled trials examined strategies to reduce sedation and found that protocols and daily interruption of sedatives shorten time on the ventilator and improve outcomes.(14, 19–21).

Little data guide critical care providers on how to administer sedatives at night. A widely held, though anecdotal, belief is that caregivers deliver greater sedative doses at night, a practice that may make patients less arousable the following day (7, 22). One study found that at least one sedative was dosed higher at night than during the day and that nurses were less likely to consider patients over-sedated (7). Though sedation should not be equated with sleep (23, 24), some nighttime caregivers may think sedatives lead to restorative sleep. To date, no studies examined the association between nighttime sedation practice and patient outcomes.

We hypothesized that greater dosing of benzodiazepines and propofol at night promotes delirium, coma, and delays in liberation from mechanical ventilation. To test this hypothesis, we performed a secondary data analysis of hourly sedative dosing among patients enrolled at the largest site of the Awakening and Breathing Controlled Trial (19).

METHODS

We performed a single-center, prospective cohort study nested within the Awakening and Breathing Controlled randomized trial (ClinicalTrials.gov NCT00097630). The parent trial evaluated a paired sedation and ventilator weaning protocol for mechanically ventilated ICU patients (19). Among patients enrolled at Saint Thomas Hospital in Nashville, TN from March 2004 to 2006, we recorded hourly sedative doses over the first five study days. Because of limited personnel and funding, these data were not collected at other centers. Adult patients receiving mechanical ventilation for more than 12 hours were excluded if: 1) admitted after cardiopulmonary arrest, 2) had neurologic deficits that prevented independent living, 3) were of moribund status, 4) received mechanical ventilation for greater than 2 weeks, or 5) were enrolled in a separate trial. We obtained written informed consent from patients or authorized surrogates prior to hospital discharge. The study was approved by Institutional Review Boards at Saint Thomas Hospital and Vanderbilt University (Protocol #030803).

Sedation practice

Prior to trial enrollment, physicians and nurses managed all subjects with patient-targeted sedation. We describe pre-trial and in-trial sedation procedures in the Supplementary Digital Content. The trial protocol did not direct sedation practice at night.

Exposures and outcomes

The primary exposures were the average hourly sedative doses for benzodiazepines and propofol during daytime (7am–11pm) and nighttime (11pm–7am) hours. On study day, changes in sedative doses at night were quantified by subtracting the average hourly daytime dose from the subsequent average hourly nighttime dose. We evaluated daytime and nighttime hourly doses for the first four 24-hour periods after enrollment for each patient, since the majority of patients were mechanically ventilated during this time. The nighttime period was determined a priori to correspond with when we hypothesized most sedation adjustments would occur (25). All hourly sedative doses (including infusions and boluses) were abstracted from the medical record in standardized fashion using ICU flowsheets. Benzodiazepine doses were standardized to lorazepam equivalents (mg/hr), and propofol doses were corrected for weight on admission (mcg/kg/min).

Our primary outcomes—measured each day following sedative exposure—were the delay in discontinuation of mechanical ventilation and the presence of delirium and coma. Since all patients were managed with an SBT protocol (19), we considered each day that no SBT was attempted or was failed to represent a delay in liberation from mechanical ventilation. We defined failed extubation as reintubation or death within 48 hours. Study personnel did not participate in decisions regarding extubation. We rigorously measured both delirium and coma (26). Trained study personnel assessed patients for delirium each day until ICU discharge using the Confusion Assessment Method for the ICU (26). We assessed level of consciousness each day using the Richmond Agitation-Sedation Scale (RASS); we defined coma as no response to verbal or physical stimulation (RASS-5) or response to physical or painful stimulation but no response to voice alone (RASS −4) (27).

Statistical analysis

We report continuous variables using median (interquartile range), and categorical variables as proportions. We report the frequency of our adverse outcomes per patient-day, as our data included repeated measurements for each subject. To determine if sedation dosing was independently associated with adverse ICU outcomes on the day following sedative exposure, we used multivariable logistic regression with generalized estimating equations (GEE) to account for repeated measurements within patients. We interpreted reduced odds of successful extubation or SBT as a “delay”. We constructed models for each outcome and included in all models all four exposures, i.e. the average hourly daytime doses of propofol and benzodiazepines and the nighttime increase/decrease in average hourly propofol and benzodiazepine doses. We further describe the patient-day denominators for models in the Supplementary Digital Content. Because relationships between drug dose and outcomes were potentially nonlinear, we modeled exposure variables using restricted cubic splines. For parsimony, we removed nonlinear terms when there was no clear evidence of nonlinearity (p for nonlinearity >0.20). In every model, we adjusted for randomization arm in the parent study, age, and modified SOFA score on the day of sedative exposure; for models of delirium/coma, we also adjusted for mental status (e.g. normal, delirious, coma) on the day of exposure. Adjustment variables are further described in the Supplementary Digital Content. We modeled a change in average hourly sedative dose and nighttime change in average hourly dose from the 10^th to 90^th percentile of the distribution, rather than a one unit (mg/kg/min or mg/hr) increase. We graphed the probabilities of outcomes adjusted for the median or mode of covariates across plausible exposure values. We performed sensitivity analyses, including using an alternative definition of night (9pm–7am), adjustment for the day and night change in opiate dose, and adjustment for admission diagnosis of sepsis. We used R (v2.11.1 patched) for all analyses.

RESULTS

We evaluated 140 eligible patients (Table 1). Patients were diagnosed with sepsis (N = 60,43%). Of 177 SBTs performed, 73 (41%) were successful; extubation was successful on 39 (20%) of 196 eligible patient-days. Of 485 patient-days on which mental status was evaluated, we observed delirium on 160 (33%) days and coma on 206 (42%). Cumulative sedation dosing increased during the study (Figure 1). We observed the proportion of patients receiving sedatives decreased, whereas the average hourly dose of benzodiazepines and propofol among those receiving the drugs remained relatively constant (Figures E1). On average, the change in nighttime hourly dosing for benzodiazepines and propofol was small (Table 2). Depending the night examined, we observed 33% to 45% of patients receiving benzodiazepines and 30% to 59% of patients receiving propofol had these sedatives increased at night (Table 2).

Table 1.

Cohort characteristics and outcomes ^~

	All Patients (N = 140)
Age, yrs	66 [55 – 75]
Female gender	70 (50)
APACHE II score on admission	27 [22 – 33]
SOFA score	9.0 [7– 12]
Diagnosis of sepsis on admission, N (%)	60 (43)
Days of mechanical ventilation	5.6 [2.6 – 9.6]
ICU length of stay, days	8.6 [5.4 – 13.9]
Intervention arm in parent trial, N (%)	70 (50)
Study outcomes per patient day
Delirium, N (% of 485 days)*	160 (33)
Coma, N (% of 485 days)*	206 (42)
Successful SBT, N (% of 177 days) ^	73 (41)
Successful extubation, N (% of 196 days) #	39 (20)

^~Data presented as median [interquartile range] or N (%) as appropriate

^*Mental status was assessed on all study days prior to death or discharge

^{^}Success of SBTs was determined on days when an SBT was performed

^#Success of extubation was determined on days when an SBT was performed and/or extubation (planned or self) occurred

Abbreviations: APACHE = Acute Physiologic and Chronic Health Evaluation; SOFA Sequential Organ Failure Assessment score

Figure 1.

Cumulative sedative dosing during study period for patients receiving benzodiazepines (Panel A, mg) and propofol (Panel B, mg). Gray ribbons indicate the 95% confidence interval.

Table 2.

Variation in sedative dosing during the night^*

	Study day
Variable	One (N=138)~	Two (N=132)~	Three (N=120)~	Four (N=109)~
Benzodiazepines, N (%) receiving drug	56 (41)	51 (39)	42 (35)	43 (39)
Dose decreased at night, N (%)	23 (41)	25 (49)	21 (50)	17 (40)
Dose unchanged at night, N (%)	8 (14)	7 (14)	7 (17)	7 (16)
Dose increased at night, N (%)	25 (45)	19 (37)	14 (33)	19 (44)
Increase in hourly dose at nighttime, mg/hr^	0.4 [0.2, 0.8]	0.5 [0.1, 0.8]	0.2 [0.2, 0.5]	0.2 [0.1, 0.5]
Propofol, N (%) receiving drug	83 (60)	57 (43)	39 (33)	30 (28)
Dose decreased at night, N (%)	46 (55)	30 (53)	13 (33)	21 (70)
Dose unchanged at night, N (%)	5 (6)	6 (10)	3 (8)	0 (0)
Dose increased at night, N (%)	32 (39)	21 (37)	23 (59)	9 (30)
Increase in hourly dose at nighttime, mg/kg/min^	8 [5, 13]	4 [3, 11]	8.5 [5, 19]	5 [3, 18]

^*Nighttime period corresponds to 11pm – 7am; all dosing expressed as median [IQR]

^~Cohort totals (N) correspond to number of eligible patients, including those no longer receiving a sedative, with complete sedation data (i.e., data collected before ICU discharge or death)

^{^}Change in nighttime dose determined only among patients with increase in dose

Benzodiazepine doses were significantly associated with delays in liberation from mechanical ventilation (Table 3). Both higher average hourly daytime doses of benzodiazepines (p<0.01) and increases in nighttime benzodiazepine dose (p<0.01) were associated with reduced odds of successful SBT on the following day (Table 3, Figure 2). Higher doses of benzodiazepines during the daytime were also associated with reduced odds of successful extubation the following day (p = 0.02; Figure 3). Yet, nighttime changes in dose were not associated with successful extubation on the following day. Neither the average hourly daytime dose of propofol nor changes at night were significantly associated with delay in successful SBTs (p=0.09 & 0.43, respectively) or delayed extubation (p=0.67 & 0.94, respectively).

Table 3.

Changes in nighttime sedation dose and odds ratios of liberation from mechanical ventilation ^#

	Percentiles		Successful SBT		Successful extubation
Variables	10th	90th	OR (95% CI)	P value	OR (95% CI)	P value
Change in sedative dosing at night
Benzodiazepine dose, mg/hr	−0.19	0.25	0.4 (0.2 – 0.8)	<0.01	0.7 (0.3 – 1.7)	0.38
Propofol dose, mcg/kg/min	−7.7	5.7	0.8 (0.5 – 1.4)	0.43	1.0 (0.6 – 1.6)	0.94
Daytime sedative dosing
Benzodiazepine dose, mg/hr	0	2.4	nonlinear*	<0.01	nonlinear*	0.02
Propofol dose, mcg/kg/min	0	40.1	0.4 (0.2 – 1.1)	0.09	1.2 (0.5 – 2.6)	0.67

^#N=192 observations for extubation model, N=173 observations for SBT model

^*Because these associations are nonlinear, the magnitude of the associations cannot be summarized using a single odds ratio but are displayed in Figures 2 and 3.

Interpretive example: The odds ratios (ORs) reflect the change in odds of the outcome that was independently associated with a change in the exposure from the 10^th percentile value to the 90^th percentile value. For example, a nighttime increase in benzodiazepine dose of 0.25 mg/hr, compared to a decrease of 0.19 mg/hr, was independently associated with a 60% reduction in odds of successful spontaneous breathing trial on the day following sedative exposure.

Figure 2.

Probability of successful spontaneous breathing trial (SBT) on the day following sedative exposure, adjusted for parent study treatment group, age, SOFA score, and mental status on the day of sedative exposure, over plausible ranges of (A) daytime hourly average benzodiazepine dose, (B) daytime average hourly propofol dose, (C) the day-to-night change in hourly average benzodiazepine dose and (D) the day-to-night change in hourly average propofol dose. Gray ribbons represent 95% confidence bounds. All plots adjusted to the median (continuous variables) and mode (categorical variables) of model covariates.

Interpretive examples: For patients in whom the average hourly dose of benzodiazpines is 1 mg/hr during the daytime, the probability of a successful SBT on the subsequent day is approximately 0.1 panel A after adjusting for other covariates. For patients in whom the average hourly dose of benzodiazepines is increased by 1 mg/hr at night, the probability of a successful SBT on the subsequent day is approximately 0.15 panel C after adjusting for covariates including daytime benzodiazepine dose.

Figure 3.

Probability of successful extubation on the day following sedative exposure, adjusted for parent study treatment group age, SOFA score, and mental status on the day of sedative exposure, over plausible ranges of (A) daytime hourly average benzodiazepine dose, (B) daytime average hourly propofol dose, (C) the day-to-night change in hourly average benzodiazepine dose and (D) the day-to-night change in hourly average propofol dose. Gray ribbons represent 95% confidence bounds. All plots adjusted to the median (continuous variables) and mode (categorical variables) of model covariates.

Interpretive examples: For patients in whom the average hourly dose of propofol is 30 mcg/kg/min during the daytime, the probability of a successful extubation on the subsequent day is approximately 0.2 panel B. For patients in whom the average hourly dose of benzodiazepines is increased by 1 mg/hr at night, the probability of a successful extubation on the subsequent day is approximately 0.1 panel C after adjusting for covariates including daytime benzodiazepine dose.

The average hourly daytime benzodiazepine dose was associated with delirium on the following day (p<0.01; Table E1). On patient-days when the average hourly daytime dose of benzodiazepines exceeded 2 mg/hr, the adjusted probability of delirium the following day approached 100% for non-comatose patients (Figure 4). Nighttime increases in benzodiazepine dose were also associated with delirium (p=0.05). Higher daytime propofol doses were marginally associated with delirium (p=0.06), whereas nighttime change in propofol dose was not associated with delirium on the following day (p=0.27). The average hourly daytime doses of both benzodiazepines and propofol were associated with coma on the following day (p<0.01 & 0.02, respectively). And, independent of daytime doses, an increase in the dose of either benzodiazepine or propofol at night was associated with coma the following day (p=0.04 & p=0.02, respectively).

Figure 4.

Probability of delirium on the day following sedative exposure, adjusted for parent study treatment group, age, SOFA score, and mental status on the day of sedative exposure, over plausible ranges of (A) daytime hourly average benzodiazepine dose, (B) daytime average hourly propofol dose, (C) the day-to-night change in hourly average benzodiazepine dose and (D) the day-to-night change in hourly average propofol dose. Gray ribbons represent 95% confidence bounds. All plots adjusted to the median (continuous variables) and mode (categorical variables) of model covariates.

Interpretive examples: For patients in whom the average hourly dose of benzodiazpines is 2 mg/hr during the daytime and who are assessable for delirium the following day, the probability of experiencing delirium is approximately 1.0 panel A). For patients in whom the average hourly dose of propofol is increased by 10 mcg/kg/min at night, the probability of experiencing delirium on the subsequent day is approximately 0.1 panel D after adjusting for covariates including daytime propofol dose.

We observed no changes in our results in multiple sensitivity analyses (eTable 2).

DISCUSSION

In this nested prospective cohort study, we confirmed that greater use of benzodiazepines and propofol during intensive care is associated with adverse outcomes among mechanically ventilated patients. Benzodiazepine dose during the day and increases at night were associated with delayed liberation from mechanical ventilation and subsequent delirium. Daytime benzodiazepine dose was associated with delayed extubation and coma. Greater use of propofol during the daytime and increased dosing at night similarly increased the odds of coma. These findings highlight the importance of daytime sedative administration on adverse ICU outcomes as well as the incremental increase in risk among those receiving greater doses at night.

Contrary to anecdotal experience and the results of one observational study (7), sedative doses were not increased at night for most patients. This surprising finding may be due to: 1) routine use of validated sedation scales in participating ICUs may have made nurses more likely to recognize and avoid oversedation at night, 2) broader awareness of the adverse outcomes of oversedation may have influenced local practice, and 3) nurses’ knowledge of an ongoing trial may have heightened sensitivity regarding sedation management. Even though sedation was not increased in some patients, an important minority received more sedation at night, with a broad range in dose adjustments.

The outcomes of critically ill patients are closely tied to liberation from mechanical ventilation (28, 29). When SBTs were coordinated with spontaneous awakening trials in the ABC Trial, patients experienced a 14% survival advantage and 4-day reductions in ICU and hospital lengths of stay (19). Despite structured spontaneous awakening intervention in this trial, we still observed that sedation with higher doses of benzodiazepines on the day before SBTs is associated with reduced odds of successful SBT on the following day. This was not unexpected in light of multiple studies showing that benzodiazepines prolongs ventilator time compared with alternatives, including propofol, remifentanil, and dexmedetomidine (30–32).

Our study also confirms the findings of others (15, 16, 33, 34) showing that benzodiazepines are an independent risk factor for development of delirium during critical illness. It extends these data by showing that benzodiazepines, even when given more than 8 hours before a delirium assessment, contributes to greater odds of delirium. These data build the body of evidence that benzodiazepines are associated with delirium and may therefore lead to longer hospital stays (26, 35), impaired long-term cognitive function (36), and greater mortality (26, 37, 38). Our findings regarding propofol are similar to those observed in earlier work (15) and could be interpreted as a trend suggesting propofol is a risk factor for delirium. Before such a conclusion can be made, however, a much larger study is needed.

Our novel observation was that increased dosing of benzodiazepines and propofol at night is an independent risk factor for coma the following day. This finding contradicts views that heavier sedation at night can be offset by dose reduction (or full interruption) in the morning. Greater nighttime sedation may have been due a mismatch between sedation assessments by ICU staff and patient needs at night (7) or the perception that sedation promotes sleep. In fact, higher doses of benzodiazepines may reduce the quantity of REM sleep and cerebral blood flow (39, 40). As propofol typically has a short half-life and brief emergence times (41), the association between nighttime propofol dosing and subsequent coma highlights the context-sensitive half-time of propofol—the half-times increase with longer durations of therapy.(41, 42)

These findings have implications for sedation protocols and the ICU staff who implement best-practice. Since one-third of ICUs in teaching hospitals do not use sedation protocols (43), this study argues for wider adoption of structured sedation practices to reduce total sedative exposure. Particular new emphasis is needed during nighttime care. As liberation from mechanical ventilation is an important quality measure in ICUs, we demonstrate a close process-outcome relationship between nighttime sedation and patient outcomes that may be modifiable (44, 45). Future efforts to improve protocol adherence may be focused towards nighttime ICU staff, who may be less compliant with certain ICU quality measures (46). Nighttime staff may have reduced opportunity for formal training in ICU assessments for delirium or sedation requirements (47), disrupting the link between patient needs and drug delivery. Our data provides a rationale for greater research into protocol implementation and nursing assessments of nighttime sedation practice (48).

We recognize important limitations to our study. First, our data derives from a single center participating in a randomized controlled trial. Thus, our results may not inform about sedation dosing or its association with outcome beyond the time period (i.e., the first week of mechanical ventilation) or calendar years (e.g., if sedation practices in the ICU have changed markedly in the past 5 years) we studied. We expect greater variation in sedative dosing within ICUs that do not use sedation scales to guide dosing; our results may therefore be conservatively biased. Our results may also be less applicable in ICUs that employ strategies including dexmedetomidine (32, 49), analgosedation (31, 50), or avoidance of sedating medications (50). Second, delirium may have gone undetected in some instances, resulting in misclassification bias. Recent studies suggest the sensitivity of the CAM-ICU is lower in ICUs where training is brief (51) compared with institutions where training is more substantial.(52) But, the instrument detects >80% of delirium when used by highly trained research staff (53–57) as was done in the current study. Third, we used “average hourly sedation dose” as our primary exposure, which may not capture the effects of bolus medications. Boluses were not distinguished from continuous infusions, nor were they protocolized in the parent study (19); the effects of boluses vs. infusions cannot therefore be assessed. Finally, unmeasured confounders—such as nighttime procedures, new delirium, sleep quality, or antipsychotic medications—may account, in part, for associations we observed. These data, in addition to nighttime sedation assessments, could further inform the mechanism underlying the associations we observed.

In conclusion, we found that greater use of benzodiazepines during the day is associated with reduced odds of successful spontaneous breathing trials, extubation, and increased odds of subsequent delirium and coma. Further risk for adverse outcomes is present when benzodiazepine doses are increased at night. We found that daytime propofol doses were associated with subsequent delirium and coma, but only coma was associated with nighttime increases in propofol. These results provide evidence that reductions in total sedative exposure among mechanically ventilated patients, including after dark, may improve patient outcomes.