Objectives: To conduct a systematic review of mortality and factors independently associated with mortality in older patients admitted to ICU.

Data sources: MEDLINE via PubMed, EMBASE, the Cochrane Library, and references of included studies.

Study selection: Two reviewers independently selected studies conducted after 2000 evaluating mortality of older patients (≥ 75 yr old) admitted to ICU.

Data extraction: General characteristics, mortality rate, and factors independently associated with mortality were extracted independently by two reviewers. Disagreements were solved by discussion within the study team.

Data synthesis: Because of expected heterogeneity, no meta-analysis was performed. We selected 129 studies (median year of publication, 2015; interquartile range, 2012-2017) including 17 based on a national registry. Most were conducted in Europe and North America. The median number of included patients was 278 (interquartile range, 124-1,068). ICU and in-hospital mortality were most frequently reported with considerable heterogeneity observed across studies that was not explained by study design or location. ICU mortality ranged from 1% to 51%, in-hospital mortality from 10% to 76%, 6-month mortality from 21% to 58%, and 1-year mortality from 33% to 72%. Factors addressed in multivariate analyses were also heterogeneous across studies. Severity score, diagnosis at admission, and use of mechanical ventilation were the independent factors most frequently associated with ICU mortality, whereas age, comorbidities, functional status, and severity score at admission were the independent factors most frequently associated with 3- 6 and 12 months mortality.

Conclusions: In this systematic review of older patients admitted to intensive care, we have documented substantial variation in short- and long-term mortality as well as in prognostic factors evaluated. To better understand this variation, we need consistent, high-quality data on pre-ICU conditions, ICU physiology and treatments, structure and system factors, and post-ICU trajectories. These data could inform geriatric care bundles as well as a core data set of prognostic factors to inform patient-centered decision-making.