Sep 102016

10843758_sJon-Emile S. Kenny [@heart_lung]

Truth is like the sun, its value wholly depends upon our being at a correct distance away from it.”

-Hjalmar Söderberg

The evening is invigorating; the Stockholm Marathon has been run and cobblestoned cafés, concealed within winding, narrowed side-streets are in the throes of summer’s end.  I take a coffee and meander about the Grand Hotel finding a bench to sit and think.  It’s both impossible and obvious to envision what this city was like a century ago.  I see Doktor Glas and his peers discussing truth in this same air and space.

And how far away from truth are we in medicine?  How close do large database analyses bring us to what is definite?  Does homogenized statistical analysis distill away meaning from the patient in front of you?

A Tale of Two Extubations

You are the overnight intensivist; it has been a busy shift from 7pm to 10pm and before you are two patients.  The first is a 56 year old man who was intubated early in the morning having been found drunk in a Thompkins Square park.  He has no known medical history other than chronic alcohol abuse.  He was extubated about 2 hours prior to your shift started and he is becoming combative because he wants to leave against medical advice.  His gas exchange and vital signs are normal.  The second patient is a 63 year old man with advanced melanoma intubated 8 hours prior for airway protection following hemorrhagic conversion of a pontine metastasis.  His family has gathered and decided to focus on comfort by having his endotracheal tube removed.

Extubating After 1900h

In a recent and hypothesis-generating propensity-analysis, Gershengorn and colleagues have assessed the association between extubation falling within the hours of 7pm and 7am [termed ‘overnight extubation’] and the rate of re-intubation.  Secondarily, they assessed the relationship between overnight extubation and mortality, ICU length of stay [LOS] and hospital LOS.

Their derivation data set was from the Project IMPACT database which contained information on patients collected between October 1, 2000 and March 29, 2009.

Multivariate logistic regression was utilized to predict the likelihood of extubation at night and then propensity score analysis was utilized to compare quasi cases (those actually extubated at night) to quasi case controls (those with a similar propensity to be extubated at night, but were actually extubated during the day).


Firstly, the actual ‘risk’ of extubation overnight decreased with time; that is, overnight extubations occurred much more frequently in the early 2000s as compared to 2008-2009.  This trend held even for ICUs with overnight intensivists.  Secondly, the variable most highly associated with overnight extubation was duration of mechanical ventilation [MV].  Those who had been receiving mechanical ventilation for < 12 hours were much more likely to have removal of mechanical ventilation in the evening hours.

What will be most-discussed, however, is that their results seem to confirm dogma in most ICUs.  The patients who were extubated at night, as compared to their statistical case-controls did not fare as well.  Specifically, considering patients with duration of MV less than 12 hours, there was no difference in need for re-intubation, but a ~ 1% absolute risk increase of both ICU and hospital-associated death.  In patients requiring mechanical ventilation for more than 12 hours, there was a 2% absolute risk increase in the need for re-intubation and a ~ 5% absolute risk increase of both ICU and hospital-associated mortality.

The authors performed some sub-group analyses to limit confounding variables.  Firstly, they analyzed extubations which were limited to survivors in an effort to exclude palliative extubations.  What they found was that there was still an increase in the risk for reintubation.  Additionally, in the APACHE cohort, a more contemporary but less detailed database, mortality was higher for overnight extubations, but only for those in whom MV was longer than 12 hours.  Lastly, in patients in whom reintubation occurred more than 1 hour following removal of MV [designed to exclude unplanned extubations], mortality and reintubation rates were also higher in those extubated at night.


The first thing that struck me was that age was not specifically mentioned.  Presumably, this variable was part of the propensity calculation, but a dedicated subgroup analysis would have been interesting.  Was ICU mortality higher for overnight extubations for those aged 18-36?  Presumably, this age-group would carry less comorbidity and require intubation for more reversible causes such as intoxications.  If there is something inherent to ICU care in the evening that raises the risk of removal of MV during overnight hours, one would expect it to be present across all ages.

Secondly, I considered another covariate – year of extubation.  The authors note that evening extubation was much more likely at the turn of the century as compared to 2009.  Was overnight extubation a marker of older practices?  And if so, did this reflect previous standard-of-care and older mortality rates?  The data collected from Project IMPACT began only 6 months after 5-7 mL/kg tidal ventilation was introduced for ARDS and over one year before Rivers et. al changed our approach to severe sepsis and septic shock.  Think how the approach to these pathophysiologies have changed and how ICU care has evolved over the last 16 years.  Could it be possible that a patient - receiving large volume ventilation and ad lib sepsis management - extubated at night in the year 2000, have been compared to a similar patient extubated during the day in the year 2009?

Thirdly, the authors explicitly excluded patients with changes in their goals of care following removal of MV.  This was one reason the Project IMPACT database was utilized – because it provided such information.  However, they do note that in patients intubated for less than 12 hours, the adjusted odds ratio of death is greater than that of re-intubation suggesting that these patients either died from cardiac arrest prior to reintubation, or that the extubations in the evening may actually have been palliative.  They cite the low risk of cardiac arrest as evidence against the former.

Finally, we are faced with the possibility that the authors have detected a real signal.  Does removing MV in the evening hours in-and-of-itself raise the risk of death?  Does this mirror data which associates poor outcome in hospitalized patients who suffer cardiac arrest overnight?  If so, is it acceptable as a medical community to simply correct this deficiency by delaying extubations or must we consider this finding unacceptable – that efficacy of care should not change based on hour of the day?

Return to the Cases

Both patients are extubated – the alcoholic before 7 pm, and the patient with a brainstem bleed at 10:30pm; importantly both may be similar in a propensity-based analysis for predicting overnight extubation [similar age, similar duration of MV, single comorbidity].  Of equal importance, their predicted mortality is also quite different.

In totality, I believe that the data of Gershengorn and colleagues is provocative and worth further, contemporary investigation.  Yet we mustn’t be wholly swayed by data which may reflect the truth of older practice, a truth devalued by our distance from it.




Get our weekly email update, and explore our library of practice updates and review articles.

PulmCCM is an independent publication not affiliated with or endorsed by any organization, society or journal referenced on the website. (Terms of Use | Privacy Policy)


Let’s Plan for Extubation in the Morning