How is pneumonia diagnosed in clinical stroke research? A systematic review and meta-analysis

Diagnosis of pneumonia complicating stroke is challenging, and there are currently no consensus diagnostic criteria. As a first step in developing such consensus-based diagnostic criteria, we undertook a systematic review to identify the existing diagnostic approaches to pneumonia in recent clinical stroke research to establish the variation in diagnosis and terminology. Studies of ischemic stroke, intracerebral hemorrhage, or both, which reported occurrence of pneumonia from January 2009 to March 2014, were considered and independently screened for inclusion by 2 reviewers after multiple searches using electronic databases. The primary analysis was to identify existing diagnostic approaches for pneumonia. Secondary analyses explored potential reasons for any heterogeneity where standard criteria for pneumonia had been applied. Sixty-four studies (56% ischemic stroke, 6% intracerebral hemorrhage, 38% both) of 639 953 patients were included. Six studies (9%) reported no information on the diagnostic approach, whereas 12 (19%) used unspecified clinician-reported diagnosis or initiation of antibiotics. The majority used objective diagnostic criteria: 20 studies (31%) used respiratory or other published standard criteria; 26 studies (41%) used previously unpublished ad hoc criteria. The overall occurrence of pneumonia was 14.3% (95% confidence interval 13.2%-15.4%; I(2)=98.9%). Occurrence was highest in studies applying standard criteria (19.1%; 95% confidence interval 15.1%-23.4%; I(2)=98.5%). The substantial heterogeneity observed was not explained by stratifying for other potential confounders. We found considerable variation in terminology and the diagnostic approach to pneumonia. Our review supports the need for consensus development of operational diagnostic criteria for pneumonia complicating stroke