ABSTRACT
BACKGROUND: The Prestroke Independence, Sex, Age, National Institutes of Health Stroke Scale (ISAN), Age, Atrial Fibrillation, Dysphagia, male sex, and National Institutes of Health Stroke Scale (A2DS2), and acute ischemic stroke-associated pneumonia score (AIS-APS) scores were created to predict stroke-associated pneumonia (SAP), one of the most important medical stroke complications. External validation of all such scores in an acute stroke population was the aim of our study. METHODS: Patients with ischemic or hemorrhagic stroke were prospectively enrolled in the multicenter Stroke-Induced Pneumonia in Andalucía project between October 2014 and May 2016. Receiver operating characteristic curves and linear regression analyses were used to determine discrimination ability of the scores. The Hosmer-Lemeshow goodness-of-fit test and the plot of observed versus predicted SAP risk were used to assess model calibration. RESULTS: Among 201 included patients, SAP rate was 15.5% (31). Higher ISAN, A2DS2, and AIS-APS scores were related to SAP (all P < .001). The C statistic was .83 (95% confidence interval [CI], .76-.91) for the ISAN score, .80 (95% CI, .70-.89) for the A2DS2 score, and .82 (95% CI, .74-.90) for the AIS-APS score, suggesting good discrimination. The ISAN and AIS-APS scores showed good calibration (Cox and Snell R2 = .206 and .174, respectively). The A2DS2 score showed the highest sensitivity (87%), and the AIS-APS score showed the highest specificity (92.8%). CONCLUSIONS: In our cohort, the external validation of ISAN, A2DS2, and AIS-APS scores have demonstrated their accurate prediction of SAP and the ability of these scores as screening tools to better manage SAP. The AIS-APS score would be recommendable for the development of future clinical trials.
