Frailty predicts 30-day mortality in intensive care patients: A prospective prediction study.

ABSTRACT

BACKGROUND: Frailty is a multidimensional syndrome characterised by a loss of reserve and an increased risk of adverse outcomes. OBJECTIVE: To study the impact of frailty on mortality in unselected intensive care patients, and to compare its discriminatory ability to an established model for outcome prediction in intensive care. DESIGN: A prospective study with a comparison of two prediction models. SETTING: A tertiary mixed ICU, from January 2017 to June 2018. PATIENTS AND MAIN OUTCOME MEASURES: Data on premorbid frailty (clinical frailty scale; CFS), severity of illness (the simplified acute physiology score, third version; SAPS3), therapeutic procedures, limitations of care and outcome were collected in 872 adult ICU patients. A cut-off level of CFS for predicting death within 30 days was identified and unadjusted and adjusted analyses were used to evaluate the association of frailty to outcome. RESULTS: The receiver operating curve, area under the curve of CFS [0.74 (95% confidence interval, 0.69 to 0.79)] did not differ significantly from that of SAPS3 [0.79 (0.75 to 0.83), P = 0.53], whereas combining the two resulted in an improved discriminatory ability [area under the curve = 0.82 (0.79 to 0.86), CFS + SAPS3 vs. SAPS3 alone, P = 0.02]. The correlation of CFS to SAPS3 was moderate (r = 0.4). A cut-off level was identified at CFS at least 5, defining 43% (n=375) of the patients as frail. Frail patients were older with higher SAPS3 and more comorbidities. Treatment in the ICU was more often withheld or withdrawn in frail patients, and mortality was higher. After adjustment for SAPS3, comorbidities, limitations of treatment, age and sex, frailty remained a strong predictor of death within 30 days [hazard ratio 2.12 (95% confidence interval, 1.44 to 3.14), P < 0.001]. CONCLUSION: Premorbid frailty was common in general ICU patients and was an independent predictor of death. Our study suggests that frailty could be a valuable addition in outcome prediction in intensive care.