Accepted Manuscript

Associations between different sedatives and ventilator-associated events, length-of-

stay, and mortality in mechanically ventilated patients

Michael Klompas, MD, MPH, Lingling Li, PhD, Paul Szumita, PharmD, Ken Kleinman,
ScD, Michael V. Murphy, BA, for the CDC Prevention Epicenters Program

PII: S0012-3692(15)00067-7
DOI: 10.1378/chest.15-1389
Reference: CHEST 66

To appear in: CHEST

Received Date: 9 June 2015

Revised Date: 2 September 2015
Accepted Date: 1 October 2015

Please cite this article as: Klompas M, Li L, Szumita P, Kleinman K, Murphy MV, for the CDC Prevention
Epicenters Program, Associations between different sedatives and ventilator-associated events, length-
of-stay, and mortality in mechanically ventilated patients, CHEST (2015), doi: 10.1378/chest.15-1389.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Abstract Word Count: 250

Manuscript Word Count: 2,547

Associations between different sedatives and ventilator-associated events, length-of-

PT
stay, and mortality in mechanically ventilated patients

RI
Michael Klompas MD, MPH1,2, Lingling Li PhD1, Paul Szumita PharmD2, Ken Kleinman ScD1,

SC
and Michael V. Murphy BA1 for the CDC Prevention Epicenters Program

U
1 Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health
AN
Care Institute, Boston, MA

2 Department of Medicine, Brigham and Womens Hospital, Boston, MA

M
D

Corresponding author:
TE

Michael Klompas, Department of Population Medicine

Harvard Medical School and Harvard Pilgrim Health Care Institute

EP

133 Brookline Avenue, 6th Floor, Boston, MA 02215

Email: mklompas@partners.org
C
AC

Conflicts of Interest: The authors have no conflicts to declare.

Funding: Centers for Disease Control and Prevention

Running Title: Sedatives and outcomes in ventilated patients

Prior Presentation: Society of Critical Care Medicine Annual Meeting; Phoenix, AZ;

January 18, 2015

1
 ACCEPTED MANUSCRIPT

Abstract

Background: Current sedation guidelines recommend avoiding benzodiazepines but

express no preference for propofol versus dexmedetomidine. In addition, there are limited

PT
data on how well randomized controlled trials on sedatives generalize to routine practice

RI
where conditions tend to be more varied and complex.

SC
Methods: We gathered daily sedative exposure data from all patients on mechanical

ventilation for 3 days over a 7-year period in a large academic medical center. We

U
AN
compared hazard ratios for ventilator-associated events (VAEs), extubation, hospital

discharge, and hospital death amongst benzodiazepines, propofol, and dexmedetomidine

M

using proportional subdistribution hazard models with competing risks. We adjusted all

analyses for ICU type, demographics, comorbidities, procedures, severity of illness,

D

hypotension, oxygenation, renal function, opioids, neuroleptics, neuromuscular blockers,

TE

awakening and breathing trials, and calendar year.

EP

Results: We evaluated 9,603 consecutive episodes of mechanical ventilation.

C

Benzodiazepines and propofol were associated with increased VAE risk whereas
AC

dexmedetomidine was not. Propofol was associated with less time to extubation compared

to benzodiazepines (HR for extubation 1.4, 95% CI 1.3-1.5). Dexmedetomidine was

associated with less time to extubation compared to both benzodiazepines (HR 2.3, 95% CI

2.0-2.7) and propofol (HR 1.7, 95% CI 1.4-2.0) but there were relatively few

2
 ACCEPTED MANUSCRIPT

dexmedetomidine exposures available for analysis. There were no differences between

any two agents in hazards for hospital discharge or mortality.

Conclusions: In this large, real-world cohort, propofol and dexmedetomidine were

PT
associated with less time to extubation compared to benzodiazepines but

RI
dexmedetomidine was also associated with less time to extubation compared to propofol.

These possible differences merit further study.

U SC
AN
M
D
TE
C EP
AC

3
 ACCEPTED MANUSCRIPT

Introduction

Current guidelines recommend lighter levels of sedation to manage ventilated patients and

preferentially using non-benzodiazepines, such as propofol or dexmedetomidine,

whenever possible.1 A recent meta-analysis of 6 randomized controlled trials with 1,225

PT
patients did find that non-benzodiazepines were associated with less time on mechanical

RI
ventilation and shorter intensive care length-of-stay compared to benzodiazepine-based

regimens.2-8 The meta-analysis pooled together dexmedetomidine and propofol for

SC
comparison to benzodiazepines, however, because there were too few trials and patients to

consider these agents individually. The analysis therefore could not answer whether there

U
AN
are important differences between propofol and dexmedetomidine in comparison to

benzodiazepines or in comparison to one another.

M

Furthermore, there are few data on how well findings from randomized controlled trials of
D

sedation generalize to routine care settings where many of the assumptions of randomized
TE

controlled trials are not met. Routine practice needs to cater to all mechanically ventilated

patients, not just the select subset eligible for inclusion in trials. In addition, routine
EP

practice frequently involves day-to-day changes in sedatives and concurrent exposures to

C

multiple sedatives whereas randomized controlled trials usually only compare two agents
AC

at a time and assume consistent exposures to other agents for the full course of sedation.

Given these knowledge gaps, we sought to evaluate associations between different

sedatives and patient outcomes within a large, diverse cohort of unselected patients. We

employed statistical methods to accommodate exposures to multiple agents at the same

4
 ACCEPTED MANUSCRIPT

time as well as day-to-day changes in patients exposures. We compared the impacts of

different agents upon hazards for ventilator-associated events (VAEs), extubation, hospital

discharge, and hospital mortality.

PT
Methods

RI
We retrospectively identified all episodes of invasive mechanical ventilation lasting 3

calendar days in Brigham and Womens Hospital in Boston between July 1, 2006 and

SC
December 31, 2013 using a prospective database of ventilated patients maintained by the

hospitals respiratory therapy department. We abstracted daily dichotomous exposures to

U
AN
benzodiazepines, propofol, and dexmedetomidine for all patients using the hospitals

electronic medication administration record. We identified VAEs electronically and

M

classified them as ventilator-associated conditions (VACs), infection related ventilator-

associated complications (IVAC), or the combined outcome of Possible or Probable

D

Pneumonia using CDC definitions.9 Note that the total VAC rate is synonymous with the
TE

total VAE rate.

EP

We created proportional sub-distribution hazards models with competing risks to estimate

C

the impact of daily benzodiazepines, propofol, and dexmedetomidine exposures on VAEs,

AC

time to extubation, time to hospital discharge, and death.10 Time to extubation and time to

hospital discharge are equivalent to duration of mechanical ventilation and hospital length

of stay respectively. The models for VAEs and time to extubation included individual terms

for each sedative for each of the three days prior to the event of interest as well as a

cumulative function variable for each sedative for the period between the start of

5
 ACCEPTED MANUSCRIPT

mechanical ventilation through the fourth day prior to the outcome of interest. This

strategy was designed to give emphasis to patients most recent medication exposures

while still allowing for the possibility that more remote exposures might have a cumulative

and sustained effect on outcomes. For the analysis of hospital discharge versus hospital

PT
death, we adjusted for the cumulative number of days of benzodiazepines, propofol, and

RI
dexmedetomidine exposures between the first day of mechanical ventilation and the

second to last day of mechanical ventilation. We excluded medication exposures from the

SC
day of VAE onset in the VAE model and the day of extubation from all other models in order

to avoid biasing against medications used to manage VAEs, facilitate extubation, or palliate

U
AN
terminal disease.
M

We calculated hazard ratios as the effect of receiving the medication of interest versus a

comparator during each of the three days prior to VAE onset or extubation, plus one extra
D

day in the period from intubation until four days prior to the outcome of interest. For
TE

simplicity, these results can be thought of as reflecting the impact of four days exposure to

the agent of interest versus four days of a comparator while controlling for patients day-
EP

to-day exposures to all other sedatives and multiple potential confounders (see below). We
C

adjusted the hazard ratios for discharge and death to reflect the cumulative impact of four
AC

days exposure to the agent of interest versus four days of a comparator in order to facilitate

comparisons with the hazard ratios for VAEs and extubation.

We adjusted all models for severity of illness by calculating each patients predicted

probability of death at the time of initiation of mechanical ventilation using a score

6
 ACCEPTED MANUSCRIPT

optimized for our dataset.11 This mortality-prediction score included variables for age, sex,

race-ethnicity, intensive care unit type (general medicine, cardiac medicine, cardiac

surgery, thoracic surgery, general surgery, neuroscience), comorbidities (congestive heart

failure, valvular disease, pulmonary vascular disease, chronic lung disease, diabetes, kidney

PT
disease, liver disease, lymphoma, solid cancer, metastases, alcohol abuse, chronic pain,

RI
drug abuse, psychiatric disease), time from hospital admission to initiation of mechanical

ventilation, laboratory tests (sodium, creatinine, glucose, white blood cell count,

SC
hematocrit, platelet count, alanine aminotransferase, total bilirubin, albumin, international

normalized ratio), medications (vasporessors, antibacterials, antifungals, amiodarone,

U
AN
anticonvulsants, corticosteroids, bronchodilators, vitamin K), and procedures (orthopedic

surgery, coronary artery bypass grafting, valve surgery, heart and lung transplants,
M

vascular surgery, digestive tract surgery, splenectomy, and cranial surgery). The area

under the receiver-operator-curve for predicting death using this score in our dataset is
D

0.88.11
TE

We further adjusted for additional clinical factors that may have influenced sedative
EP

choices and outcomes by adding separate fixed covariates for age, race, sex, unit type,
C

additional comorbidities (including alcohol abuse, drug abuse, and chronic pain), recent
AC

cardiac surgery, recent cranial surgery, opioid exposure prior to intubation, and

benzodiazepine exposure prior to intubation. We also included daily variables for

continuous vs intermittent sedative delivery, vasopressors, severe hypoxemia (PaO2:FiO2

ratio 100), daily minimum creatinine, spontaneous awakening trials, spontaneous

breathing trials, opioids, neuroleptics, and neuromuscular blockers. We adjusted for

7
 ACCEPTED MANUSCRIPT

possible temporal trends using the calendar year from each episode of mechanical

ventilation.

We conducted two sensitivity analyses to test the robustness of our findings. We analyzed

PT
cardiac surgery and noncardiac surgery patients separately given the large differences in

RI
patient characteristics and management strategies for these two populations as well as the

fact that 57% of dexmedetomidine exposures in our population were in cardiac surgery

SC
patients. We also did an analysis restricted to patients managed almost exclusively with

one type of sedative (at most one day of exposure to alternative sedatives).

U
AN
All analyses were conducted using SAS Version 9.3 (SAS Institute, Cary, NC). The study was
M

reviewed and approved by the Brigham and Womens Hospital Institutional Review Board

(IRB#2011P001657).
D
TE

Results

There were 9,603 consecutive episodes of mechanical ventilation and 86,714 ventilator-
EP

days during the study period. Patients characteristics are presented in Table 1. The
C

median duration of mechanical ventilation was 6.0 days (interquartile range 4-10 days).
AC

Approximately 66% of patients received at least one day of benzodiazepines, 62% received

at least one day of propofol, and 12% received at least one day of dexmedetomidine.

Sedatives were often used concurrently (Table 2): all 3 agents were given on 10% of

ventilator days and two different agents were given on 46% of ventilator days.

Benzodiazepines were administered continuously on 74% of benzodiazepine days and

8
 ACCEPTED MANUSCRIPT

intermittently on 26% of benzodiazepine days. Cardiac surgery accounted for the majority

of dexmedetomidine exposures hospital wide (57%); the majority of cardiac surgery

patients received dexmedetomidine (62%).

PT
Associations between sedating agents and ventilator-associated events

RI
Adjusted hazard ratios for VAEs are shown in Table 3. Benzodiazepines and propofol were

both associated with increased hazards for VAEs compared to regimens without these

SC
agents (HR 1.4, 95% CI 1.1-1.7 and HR 1.3, 95% CI 1.1-1.6 respectively). On direct

comparison, however, there were no differences between benzodiazepines versus propofol

U
AN
in hazards for VAEs (HR 1.0, 95% CI 0.8-1.2). Propofol was also associated with increased

hazards for IVAC (HR 1.6, 95% CI 1.2-2.2) and the combined outcome of Possible or
M

Probable Pneumonia (HR 1.5, 95% CI 1.0-2.2). Dexmedetomidine, by contrast, was not

associated with increased risk for VAEs compared to regimens without dexmedetomidine.
D

There were trends towards fewer VAEs with dexmedetomidine versus benzodiazepines
TE

(HR 0.7, 95% CI 0.5-1.2) and with dexmedetomidine versus propofol (HR 0.8, 95% CI 0.5-

1.2) but these were not statistically significant.

C EP

Associations between sedatives and time to extubation, hospital discharge, and death
AC

Hazard ratios for the associations between each sedative and extubation, hospital

discharge, and hospital death are presented in Table 4. Benzodiazepines were associated

with a hazard ratio for extubation of less than one suggesting that it decreases the daily

probability of extubation compared to patients unexposed to benzodiazepines (HR 0.89, 95%

CI 0.83-0.96, P=.002); in other words, benzodiazepines extend time to extubation.

9
 ACCEPTED MANUSCRIPT

Propofol by contrast was associated an increased hazard for extubation (HR 1.2, 95% CI

1.1-1.3, P<.0001) suggesting that it decreases time to extubation.

Relative to one another, propofol was associated with an increased hazard for extubation

PT
(HR 1.4, 95% CI 1.3-1.5) suggesting that propofol is associated with less time to extubation

RI
compared to benzodiazepines. There were no significant differences between propofol and

benzodiazepines on hazards for hospital discharge or death.

SC
Dexmedetomidine was associated with higher hazards for extubation compared to

U
AN
benzodiazepines (HR 2.3, 95% CI 2.0-2.7) and propofol (HR 1.7, 95% CI 1.4-2.0) but there

were no differences between dexmedetomidine and either of these agents on hazards for
M

hospital discharge or death.

D

Sensitivity analyses
TE

On sensitivity analyses restricted to cardiac surgery patients and noncardiac surgery

patients respectively dexmedetomidine was associated with higher hazards for extubation
EP

compared to propofol in both cardiac surgery (HR 1.8, 95% CI 1.4-2.4, P<.0001) and non
C

cardiac surgery patients (HR 1.4, 95% CI 1.1-1.8, P=.02). Likewise, amongst the subset of
AC

patients exposed almost exclusively to one type of sedative with no more than one day of

exposure to another sedative (N=2,854 for benzodiazepines, N=2,475 for propofol, N=139

for dexmedetomidine), results were consistent with the primary analysis. Propofol was

associated with higher hazards for extubation compared to benzodiazepines (HR 1.6, 95%

CI 1.4-1.7) and dexmedetomidine was associated with higher hazards for extubation

10
 ACCEPTED MANUSCRIPT

compared to both benzodiazepines (HR 2.4, 95% CI 1.8-3.3) and propofol (HR 1.6, 95% 1.1-

2.1).

Discussion

PT
In this large observational study, we found significant differences between

RI
benzodiazepines, propofol, and dexmedetomidine in risks for VAEs and duration of

mechanical ventilation. Benzodiazepines and propofol were associated with increased

SC
hazards for VAEs compared to regimens without these agents. Propofol and

dexmedetomidine were associated with less time to extubation compared to

U
AN
benzodiazepines. Dexmedetomine was associated with less time to extubation compared

to propofol. There were no differences between any of these agents in hazards for hospital
M

discharge or death.
D

These findings provide a window into the impacts of different sedatives on outcomes under
TE

real-world conditions. These are often more complex and varied than the controlled

conditions required by most randomized controlled trials. Our study included all patients
EP

on mechanical ventilation for 3 days not just a select subset. We were able to
C

accommodate and adjust for exposures to multiple different sedatives on the same day as
AC

well as for day-to-day changes in exposures to different sedatives.

It is possible that some of the differences we observed reflect residual confounding by

indication and/or severity of illness. We adjusted our analyses, however, for an extensive

array of clinical variables that might have influenced clinicians sedative choices and

11
 ACCEPTED MANUSCRIPT

patients outcomes. We further adjusted for year of mechanical ventilation in order to

account for possible temporal trends in sedative choices and patient outcomes.

We did not include data on depth of sedation and therefore it is possible that some of the

PT
observed associations were due to differences in depth of sedation rather than differences

RI
in the drugs per se. The goal of our study, however, was not to identify absolute differences

between agents under perfectly controlled laboratory conditions but rather to observe

SC
whether there are important differences between agents under routine practice conditions.

If it is easier to achieve less sedation with one agent compared to another, then this is

U
AN
arguably a point in favor of that agent. Indeed, in the MIDEX and PRODEX studies, patients

randomized to dexmedetomidine had significantly higher Richmond Agitation and Sedation

M

Scores compared to patients randomized to midazolam and propofol, perhaps accounting

in part for dexmedetomidines shorter time to extubation in both trials.4 Notably, however,
D

other studies have found differences between sedatives independent of depth of sedation.
TE

In the SEDCOM study for example, dexmedetomidine was associated with less time to

extubation compared to midazolam despite similar levels of sedation in both arms of the
EP

study.2 Nonetheless, recognizing that clinicians sometimes purposefully choose deep

C

sedation for some patients, we did control for severity of illness, use of neuromuscular
AC

blockers, poor oxygenation, vasopressors, and other potentially relevant confounders.

Many of our findings mirror signals from randomized controlled trials, especially with

regard to time to extubation. Our findings that propofol and dexmedetomidine were both

associated with less time to extubation compared to benzodiazepines echo trials by Carson,

12
 ACCEPTED MANUSCRIPT

Weinbroum, Pandharipande, Riker, Ruokenen, and Jakob.2-7 The large number of episodes

of mechanical ventilation in our dataset, however, allowed us to extend conceivable but

underpowered signals from randomized controlled trials. There is only one published

study, for example, directly comparing dexmedetomidine and propofol in noncardiac

PT
surgery patients.4 Jakob et al. found that patients randomized to dexmedetomidine were

RI
extubated sooner than patients randomized to propofol (median 69 vs 93 hours, P=.04).

Our study affirms that this difference may indeed be significant (HR for extubation 1.8, 95%

SC
CI 1.4-2.4, P<.0001 in cardiac surgery patients and 1.4, 95% CI 1.1-1.8, P=.02 in non

cardiac surgery patients). Notably, however, we did not find any differences between

U
AN
propofol and dexmedetomidine in hazards for hospital discharge or mortality.
M

Our findings are tempered by important limitations. This was a single center retrospective

analysis and hence some of our findings may be attributable to residual confounding and /
D

or idiosyncratic local practice patterns. We only measured whether each patient was
TE

exposed or not exposed to each agent on any given day but not total dose or adjusted dose

per kilogram body weight. The model we used to adjust for severity of illness was
EP

developed amongst many of the same patients used for the current study and hence it may
C

have been overtrained to this population. Relatively few patients were exposed to
AC

dexmedetomidine compared to benzodiazepines and propofol, and most dexmedetomidine

exposures were in cardiac surgery patients. Nonetheless, our findings persisted on

sensitivity analysis restricted to noncardiac surgery patients.

Conclusions

13
 ACCEPTED MANUSCRIPT

In sum, we found that sedatives vary in their associations with VAEs and time to extubation

but not in their associations with time to hospital discharge or mortality. Both propofol

and dexmedetomidine were associated with less time to extubation compared to

benzodiazepines. This mirrors the conclusions of current guidelines which recommend

PT
light sedation with non-benzodiazepines whenever possible. We also found, however, that

RI
dexmedetomidine was associated with less time to extubation compared to propofol and

may therefore be a preferable agent for selected patients. While this observation does

SC
mirror the limited randomized controlled trial data available, it needs to be interpreted

with caution in light of the many limitations of our study. Further study is warranted.

U
AN
Finally, our analysis helps affirm that findings from highly controlled studies extend to

routine practice despite the greater complexity and diversity of patients and their
M

treatments outside of trials. Better understanding of the differential risk and benefit

profiles of different sedatives can help clinicians refine their prescribing patterns and
D

further improve outcomes for critically ill patients.

TE
C EP
AC

14
 ACCEPTED MANUSCRIPT

Acknowledgements:

MK takes responsibility for the content of the manuscript including the data and analysis.

This study was funded by the Centers for Disease Control and Prevention. The funding

source had no role in the conception, design, interpretation, or presentation of the study.

PT
MK and PS conceived of the study; MK, LL, PS, and KK developed the analysis plan; MK,

RI
MVM, and PS gathered the raw data for the study; MVM cleaned and analyzed the data; LL

performed all statistical analyses; all authors helped write the manuscript and revised for

SC
critical content.

U
AN
M
D
TE
C EP
AC

15
 ACCEPTED MANUSCRIPT

References

1. Barr J, Fraser GL, Puntillo K, et al. Clinical practice guidelines for the management of

pain, agitation, and delirium in adult patients in the intensive care unit. Crit Care

Med. 2013;41(1):263-306.

PT
2. Riker RR, Shehabi Y, Bokesch PM, et al. Dexmedetomidine vs midazolam for sedation

RI
of critically ill patients: a randomized trial. JAMA. 2009;301(5):489-499.

3. Ruokonen E, Parviainen I, Jakob SM, et al. Dexmedetomidine versus

SC
propofol/midazolam for long-term sedation during mechanical ventilation. Intensive

Care Med. 2009;35(2):282-290.

U
AN
4. Jakob SM, Ruokonen E, Grounds RM, et al. Dexmedetomidine vs midazolam or

propofol for sedation during prolonged mechanical ventilation: two randomized

M

controlled trials. JAMA. 2012;307(11):1151-1160.

5. Carson SS, Kress JP, Rodgers JE, et al. A randomized trial of intermittent lorazepam
D

versus propofol with daily interruption in mechanically ventilated patients. Crit Care
TE

Med. 2006;34(5):1326-1332.

6. Pandharipande PP, Pun BT, Herr DL, et al. Effect of sedation with dexmedetomidine
EP

vs lorazepam on acute brain dysfunction in mechanically ventilated patients: the

C

MENDS randomized controlled trial. JAMA. 2007;298(22):2644-2653.

AC

7. Weinbroum AA, Halpern P, Rudick V, Sorkine P, Freedman M, Geller E. Midazolam

versus propofol for long-term sedation in the ICU: a randomized prospective

comparison. Intensive Care Med. 1997;23(12):1258-1263.

16
 ACCEPTED MANUSCRIPT

8. Fraser GL, Devlin JW, Worby CP, et al. Benzodiazepine versus nonbenzodiazepine-

based sedation for mechanically ventilated, critically ill adults: a systematic review

and meta-analysis of randomized trials. Crit Care Med. 2013;41(9 Suppl 1):S30-38.

9. Magill SS, Klompas M, Balk R, et al. Developing a new, national approach to

PT
surveillance for ventilator-associated events. Crit Care Med. 2013;41(11):2467-

RI
2475.

10. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a

SC
competing risk. J Am Stat Assoc. 1999;94(446):496-509.

11. van Mourik MS, Moons KG, MICU Registry, Murphy MV, Bonten MJM, Klompas M.

U
AN
Severity of disease estimation and risk-adjustment for comparison of outcomes in

mechanically ventilated patients using electronic routine care data. Infect Control
M

Hosp Epidemiol. 2015;36(7):807-815.

D
TE
C EP
AC

17
 ACCEPTED MANUSCRIPT

Table 1. Patients characteristics

Table 2. Distribution of sedative exposures amongst episodes and ventilator days.

Table 3. Adjusted risks for ventilator-associated events associated with different sedatives.

PT
Table 4. Adjusted risks for time to extubation, time to hospital discharge, and hospital death

RI
associated with different sedatives.

U SC
AN
M
D
TE
C EP
AC

18
 ACCEPTED MANUSCRIPT

Table 1. Patients characteristics

All Episodes Benzodiazepines Propofol Dexmedetomidine
(N=9,603) (N=6,374) (N=5,941) (N=1,184)
Age
Mean (SD) 61.1 (16.5) 60.0 (16.5) 59.7 (16.9) 61.2 (15.8)
Sex
Male 5502 (57%) 3722 (58%) 3492 (59%) 766 (65%)

PT
Race
White 7401 (77%) 4950 (78%) 4578 (77%) 995 (84%)
Black 822 (8.6%) 520 (8.2%) 504 (8.5%) 59 (5.0%)
Asian 235 (2.5%) 138 (2.2%) 151 (2.5%) 22 (1.9%)

RI
Hispanic 350 (3.6%) 245 (3.8%) 231 (3.9%) 25 (2.1%)
Other 795 (8.3%) 521 (8.2%) 477 (8.0%) 83 (7.0%)

SC
Intensive Care Unit Type
Medical 2785 (29%) 2020 (32%) 1377 (23%) 39 (3.3%)
General Surgery 2159 (23%) 1591 (25%) 1447 (25%) 208 (18%)
Neuroscience 1391 (15%) 582 (9.2%) 1077 (18%) 30 (2.6%)

U
Cardiac surgery 1090 (12%) 593 (9.4%) 886 (15%) 672 (57%)

AN
Thoracic surgery 1021 (11%) 791 (13%) 472 (8.1%) 186 (16%)
Cardiac medical 1013 (11%) 726 (12%) 607 (10%) 41 (3.5%)
Comorbidities
Coronary artery disease 2098 (22%) 1389 (22%) 1234 (21%) 443 (37%)

M
Congestive heart failure 2155 (23%) 1401 (22%) 1347 (23%) 444 (38%)
Cerebrovascular disease 1906 (20%) 1061 (17%) 1201 (20%) 245 (21%)
Chronic lung disease 1241 (13%) 879 (14%) 740 (13%) 190 (16%)

D
Diabetes 1187 (12%) 689 (11%) 765 (13%) 173 (15%)

TE
Kidney disease 1177 (12%) 797 (13%) 639 (11%) 143 (12%)
Cancer 1895 (20%) 1418 (22%) 987 (17%) 183 (15%)
Liver disease 324 (3.4%) 232 (3.7%) 200 (3.4%) 16 (1.4%)
Alcohol abuse 461 (4.8%) 383 (6.0%) 339 (5.7%) 58 (4.9%)
EP
Drug abuse 255 (2.7%) 196 (3.1%) 202 (3.4%) 30 (2.5%)
Charlson Score
Mean (standard deviation) 3.6 (2.7) 3.7 (2.8) 3.3 (2.6) 3.4 (2.4)
C

Ventilator-associated events
Ventilator-associated conditions (VAC) 1308 (14%) 1044 (16%) 892 (15%) 181 (15%)
AC

Infection-related ventilator-associated complications (IVAC) 522 (5.4%) 434 (6.8%) 392 (6.6%) 88 (7.4%)
Possible pneumonia 172 (1.8%) 144 (2.3%) 127 (2.1%) 29 (2.5%)
Probable pneumonia 157 (1.6%) 129 (2.0%) 126 (2.1%) 30 (2.5%)
Outcomes
Median days of mechanical ventilation (interquartile range) 6 (4-10) 7 (4-12) 6 (4-11) 6 (4-11)
Median days of hospitalization (interquartile range) 20 (12-32) 20 (13-33) 20 (12-32) 25 (16-41)
Hospital death 2544 (26%) 1811 (28%) 1418 (24%) 189 (16%)

19
 ACCEPTED MANUSCRIPT

Table 2. Distribution of sedative exposures amongst episodes and ventilator days

Episodes Ventilator
Days
(N=9,603) (N=86,714)
Benzodiazepines Alone 2,193 (23%) 18,149 (21%)
Propofol Alone 1,656 (17%) 10,091 (12%)
Dexmedetomidine Alone 49 (0.5%) 245 (0.3%)

PT
Benzodiazepines + Dexmedetomidine 174 (1.8%) 1,582 (1.8%)
Benzodiazepines + Propofol 3,324 (35%) 36,515 (42%)
Dexmedetomidine + Propofol 278 (2.9%) 1,601 (1.8%)

RI
Benzodiazepines + Propofol + Dexmedetomidine 683 (7.1%) 8,784 (10%)

U SC
AN
M
D
TE
C EP
AC

20
 ACCEPTED MANUSCRIPT

Table 3. Adjusted hazard ratios for ventilator-associated events associated with different sedatives. Hazard ratios reflect the
relative impact of four days of exposure to regimens containing the agent of interest versus regimens without the stated
comparator while taking into account day-to-day exposures to all other sedatives.
Possible or
VAE* IVAC Probable
Pneumonia

PT
Hazard P Hazard P Hazard P
Ratio** Ratio** Ratio**
(95% CI) (95% CI) (95% CI)

RI
Agent versus no agent comparisons
Benzodiazepines vs regimens without benzodiazepines 1.4 (1.1-1.7) .002 1.3 (0.9-1.7) .10 1.2 (0.8-1.8) .28
Propofol vs regimens without propofol 1.3 (1.1-1.6) .003 1.6 (1.2-2.2) .0009 1.5 (1.0-2.2) .03

SC
Dexmedetomidine vs regimens without dexmedetomidine 1.0 (0.7-1.6) .92 1.0 (0.5-1.8) .90 0.9 (0.4-1.9) .69
Agent versus agent comparisons
Propofol vs benzodiazepines 1.0 (0.8-1.2) .84 1.3 (0.9-1.7) .12 1.2 (0.8-1.8) .31

U
Dexmedetomidine vs benzodiazepines 0.7 (0.5-1.2) .24 0.7 (0.4-1.5) .41 0.7 (0.3-1.7) .40
Dexmedetomidine vs propofol 0.8 (0.5-1.2) .27 0.6 (0.3-1.2) .13 0.6 (0.2-1.3) .20

AN
Abbreviations: IVAC infection-related ventilator-associated complications; VAE ventilator-associated events

* The total VAE count is synonymous with the total ventilator-associated condition (VAC) count.

M
** Hazard ratios >1 suggest an increased risk for ventilator-associated events

All models adjusted for age, race, sex, calendar year, intensive care unit type (medicine, neuroscience, cardiac medicine, cardiac surgery, thoracic

D
surgery, general surgery), time from hospital admission to initiation of mechanical ventilation, comorbidities (congestive heart failure, valvular disease,
pulmonary vascular disease, chronic lung disease, diabetes mellitus, renal disease, liver disease, lymphoma, solid malignancies, alcohol abuse, drug

TE
abuse, chronic pain, and psychiatric disease), initial laboratory values (sodium, creatinine, glucose, white blood cell count, hematocrit, platelet count,
alanine aminotransferase, total bilirubin, albumin, international normalized ratio), medications (antibacterials, antifungals, amiodarone,
anticonvulsants, corticosteroids, bronchodilators, vitamin K, benzodiazepine exposure before intubation, opioid exposure before intubation),
EP
procedures (orthopedic surgeries, coronary artery bypass graft surgery, valve surgery, heart and lung transplants, vascular surgery, digestive tract
surgery, splenectomy, cranial surgery), daily exposures to vasopressors, opioids, neuroleptics, and neuromuscular blockers, daily exposure to
continuous vs intermittent sedatives, daily partial pressure of oxygen:fraction of inspired oxygen ratio, daily minimum creatinine, daily spontaneous
C

awakening trials, and daily spontaneous breathing trials.

AC

21
 ACCEPTED MANUSCRIPT

Table 4. Adjusted hazard ratios for extubation, hospital discharge, and hospital death associated with different sedatives.
Hazard ratios reflect the relative impact of four days of exposure to regimens containing the agent of interest versus regimens
without the stated comparator while taking into account exposures to all other sedatives.
Hazard Ratio Hazard Ratio Hazard Ratio
for for for
Extubationa Hospital Hospital
Dischargeb Deathc

PT
P P P
(95% CI) (95% CI) (95% CI)
Agent versus no agent comparisons

RI
Benzodiazepines vs regimens without benzodiazepines 0.89 (0.83-0.96) .002 1.01 (0.98-1.04) .64 1.00 (0.96-1.05) .91
Propofol vs regimens without propofol 1.22 (1.13-1.31) <.0001 0.98 (0.95-1.02) ,35 1.03 (0.97-1.10) .28
Dexmedetomidine vs regimens without dexmedetomidine 2.05 (1.77-2.38) <.0001 0.97 (0.88-1.07) .56 0.98 (0.82-1.17) . 81

SC
Agent versus agent comparisons
Propofol vs benzodiazepines 1.37 (1.26-1.49) <.0001 0.97 (0.93-1.02) .28 1.03 (0.96-1.11) .42
Dexmedetomidine vs benzodiazepines 2.30 (1.96-2.71) <.0001 0.96 (0.87-1.07) .49 0.98 (0.81-1.18) .80

U
Dexmedetomidine vs propofol 1.68 (1.43-1.98) <.0001 0.99 (0.89-1.10) .84 0.95 (0.78-1.15) .58
a Hazard ratios >1 suggest an increased likelihood of extubation over time, i.e. less time to extubation

AN
b Hazard ratios >1 suggest an increased likelihood of discharge over time, i.e. less time to hospital discharge
c Hazard ratios >1 suggest an increased likelihood of death over time, i.e. higher mortality risk

M
All models adjusted for age, race, sex, calendar year, intensive care unit type (medicine, neuroscience, cardiac medicine, cardiac surgery, thoracic
surgery, general surgery), time from hospital admission to initiation of mechanical ventilation, comorbidities (congestive heart failure, valvular disease,
pulmonary vascular disease, chronic lung disease, diabetes mellitus, renal disease, liver disease, lymphoma, solid malignancies, alcohol abuse, drug

D
abuse, chronic pain, and psychiatric disease), initial laboratory values (sodium, creatinine, glucose, white blood cell count, hematocrit, platelet count,
alanine aminotransferase, total bilirubin, albumin, international normalized ratio), medications (antibacterials, antifungals, amiodarone,

TE
anticonvulsants, corticosteroids, bronchodilators, vitamin K, benzodiazepine exposure before intubation, opioid exposure before intubation),
procedures (orthopedic surgeries, coronary artery bypass graft surgery, valve surgery, heart and lung transplants, vascular surgery, digestive tract
surgery, splenectomy, cranial surgery), daily exposures to vasopressors, opioids, neuroleptics, and neuromuscular blockers, daily exposure to
EP
continuous vs intermittent sedatives, daily partial pressure of oxygen:fraction of inspired oxygen ratio, daily minimum creatinine, daily spontaneous
awakening trials, and daily spontaneous breathing trials.
C
AC

22