Invasive vs. noninvasive ventilation as the initial ventilatory strategy in very elderly patients admitted to intensive care due to community-acquired pneumonia: a multicenter retrospective cohort study

Autores:B. Besen, O. Ranzani, M. Park
Medical intensive care unit, Hospital das Clínicas HCFMUSP – University of Sao Paulo Medical School, Sao Paulo, Brazil; 2 Pulmonary division, Heart Institute (InCor) – Hospital das Clínicas HCFMUSP – University of Sao Paulo Medical School, Sao Paulo, Brazil
Correspondência: B. Besen
Intensive Care Medicine Experimental 2019, 7(Suppl 3):001479
Data: 2019


The very elderly patients (i.e., those aged ≥ 80 years-old) represent an increasing subgroup of patients admitted to intensive care units (ICUs). Community-acquired pneumonia (CAP) is a common reason for ICU admission and how to best provide initial ventilatory support in these patients is unknown.


To evaluate if the initial ventilatory strategy is associated with hospital mortality in very elderly patients admitted to the ICU.


We did a retrospective multicenter cohort study of very elderly patients admitted to the ICU with a diagnosis of CAP from 2009 through 2012. Data was retrieved from the ICU quality database (Epimed Monitor) and checked for consistency. We used logistic regression to evaluate the primary outcome (hospital mortality), after adjusting sequentially for potential confounders chosen from a causal directed acyclic graph. We imputed missing data with multiple imputation by chained equations using the predictive mean matching algorithm. We evaluated ad-hoc defined subgroup effects with interaction terms in the logistic model. Sensitivity analyses were done to assess model consistency regarding the primary outcome.


6,318 very elderly patients were admitted to the 11 ICUs during the study period. 1,713 admissions were due to respiratory causes, of which 678 patients were admitted for CAP. Mean age was 86.3 years (4.7); 288/678 (42.5%) were male. Mean SAPS 3 score was 63.5 (12.8), with a median SOFA of 4 [2; 7] and a median Charlson comordity score of 1 [1; 3]. 191/678 (28.2%) patients were previouslly bedridden and 253/678 (37.3%) needed
some assistance in activities of daily living. 309/678 (45.6%) did not use any MV strategy (no MV group); 232/678 (34.2%) underwent NIV first; 137/678 (20.2%) underwent IMV first. IMV patients were sicker at ICU admission when compared to both NIV and no MV groups (SAPS 3: 76.5 vs. 62.6 vs. 58.4, p-value < 0.001), with no statistically significant differences in the antecedent characteristics (burden of comorbidities and performance status). Hospital mortality was 90/309 (29 %) for no MV; 114/232 (49 %) for NIV; 90/137 (66 %) for IMV. For the comparison IMV vs. NIV, crude OR (95% CI, p-value) was 1.98 (1.28 – 3.07, p = 0.002). After adjusting for age, sex and antecedent characteristics, the aOR was 2.17 (1.38 – 3.42, p = 0.001). After including nonrespiratory SOFA in the model, aOR was 1.23 (0.74 – 2.03, p = 0.419). After including respiratory variables (Pao2/Fio2; pH; Pco2) in the model, the aOR was 1.1 (0.65 – 1.85, p = 0.720). Interaction terms for the following characteristics showed no evidence of effect modification: heart failure, chronic obstructive pulmonary disease, bedridden status, non-respiratory SOFA ≥ 4, cute respiratory acidosis and Pao2/Fio2 ≤ 150. Complete case analysis (sensitivy analysis) yielded the same results.


The initial ventilatory strategy was not associated with hospital mortality in very elderly patients admitted to the ICU due to CAP, regardless of antecedent characteristics. Acuity variables were responsible for the major part of positive confounding for mortality in this population. This suggests a less invasive approach may be sufficient for most patients to provide the benefit of the doubt during an ICU trial for very elderly patients.