Background: The timing of the onset of pneumonia and the impact of covariates on its temporal profile is not well described. This knowledge can help in formulation on approaches to prevent post-stroke pneumonia, a condition associated with higher mortality from stroke.
Aims: The aim is to determine the impact of covariates on the temporal profile of pneumonia.
Methods: Retrospective study of admissions to the stroke unit at Monash Medical Center between 15/12/2014-29/6/2016. Survival curve analysis and time-dependent explanations of variable importance were performed.
Results: There 1256 admissions and 866 patients with ischemic and 146 patients with hemorrhagic stroke (age 72.7 ± 14.5, male = 54.0%, National Institute of Health Stroke Scale/NIHSS 7.8 ± 8.0). The frequency of pneumonia for NIHSS[0] was 0 (0/86), for NIHSS[1-5] was 2.35% (10/425), NIHSS[6-10] was 7.83% (17/211), NIHSS[11-15] was 15.18% (12/79), NIHSS[16-20] was 20.48% (17/83), NIHSS[>20] was 24.79% (30/121). The frequency of pneumonia among stroke patients was 86 (9.28%). Risk of pneumonia was 1.5% on day 1, 3.2% on day 2, 4.9% on day 3, 5.5% on day 4, 6.1% on day 5, 6.5% on day 6, 6.9% on day 7. Cox regression show that NIHSS (HR 1.09, 95% CI 1.06-1.13, p<0.001), nasogastric tube/NGT (HR 3.23, 95% CI 1.79-5.82, p<0.001), Charlson comorbidity index (HR 1.13, 95% CI 1.01-1.26, p=0.03), were significantly associated with pneumonia. There was a trend to significance for failed dysphagia screen and NIHSS> 4 (p=0.0538), male (p=0.10), diabetes (p=0.09) but not thrombolysis or ischemic heart disease. The pseudoR2 was 0.585. Collinearity for the covariates was low (maximal variance inflation factor was 1.97). Permutation of covariates show that the largest impact on Brier score (model calibration) occurred with NIHSS, follow by NGT and failed dysphagia screen with NIHSS >4.
Connclusion: After stroke onset, pneumonia occurs early in the first week. Stroke severity, NGT insertion, and higher comorbidity were associated with pneumonia.