Prognostic scoring models of mortality in a real-life ICU patient cohort have poor value in guiding decisions concerning treatment restrictions of patients with COPD and acute respiratory failure.

Background

To improve the assessment of which patients would benefit from intensive care treatment and to facilitate end-of-life discussions, we evaluated the discriminative value of several scoring systems predicting the mortality of patients with an Acute Exacerbation of COPD (AECOPD) in our population.

Materials And Methods

We used electronic patient data systems to recover all parameters associated with mortality from patients with AECOPD admitted to our ICU. Several acute physiology scores and prognostic scoring systems (BAP-65, modified DECAF, eMRCD and the CAOS score) were evaluated. The area under the receiver operating characteristic curve (AUROC) was used to test for discriminative quality in predicting in-hospital mortality.

Results

A total of 159 AECOPD admissions were included. Forty-seven patients (29.6%) died in hospital. No significant differences were found in demographic features, severity of chronic illness, physical fitness and co-morbidity between in-hospital survivors and non-survivors. Regarding the severity of acute illness and prognostic models, only Apache II, -IV, CAPS and the CAOS score proved to be significantly different between groups. When tested for discriminative quality, however, none of the acute physiology scores or prognostic models tested proved to be of discriminative value (AUROCs of 0.532-0.670).

Conclusion

We evaluated clinical parameters as well as prognostic and physiology scoring systems predicting mortality in patients with AECOPD treated in our ICU. The results showed poor or no discriminative quality. Caution should be used when implementing prognostic scoring systems for patients with AECOPD not validated for the target population.