Use of illness severity scores to predict mortality in interstitial lung disease patients hospitalised with acute respiratory deterioration

Abstract

IntroductionHospitalisations relating to acute respiratory deteriorations (ARD) in Interstitial Lung Disease (ILD) have poor outcomes. Factors predicting adverse outcomes are not fully understood and data addressing the use of illness severity scores in prognostication are limited. ObjectiveTo investigate the use of CURB-65 and NEWS-2 severity scores in the prediction of mortality following ARD-ILD hospitalisation, using prospective methodology and to validate previously determined cut-offs, derived from a retrospective study cohort. MethodsA dual-centre prospective observational cohort study of all adults (≥18y) hospitalised with ARD-ILD in Bristol, UK (n = 179). Gender-Age-Physiology (GAP), CURB-65 and NEWS-2 scores were calculated for each eligible admission.Receiver operating characteristics (ROC) curve analysis was used to quantify the strength of discrimination for NEWS-2 and CURB-65 scores. Univariable and multivariable logistic regression analyses were performed to explore the relationship between baseline severity scores and mortality. ResultsGAP showed some merit at predicting 30-day mortality (AUC = 0.64, P = 0.015); whereas CURB-65 showed modest predictive value for in-hospital (AUC = 0.72, P < 0.001) and 90-day mortality (AUC = 0.67, P < 0.001). NEWS-2 showed higher predictive value for in-hospital (AUC = 0.80, P < 0.001) and 90-day mortality (AUC = 0.75, P < 0.001), with an optimal derived cut-off ≥6.5 found to be sensitive and specific for predicting in-hospital (83% and 63%) and 90-day (73% and 72%) mortality. In exploratory analyses, GAP score addition improved the predictive ability of NEWS-2 against 30-day mortality and CURB-65 across all time-periods. ConclusionNEWS-2 has good discriminatory value for predicting in-hospital mortality and moderate discriminatory value for predicting 90-day mortality. The optimal NEWS-2 cut-off value determined was the same as in a previous retrospective cohort, confirming the NEWS-2 score shows promise in predicting mortality following ARD-ILD hospitalisation.