Background We undertook a systematic review of all published stroke identification instruments to describe their performance characteristics when used prospectively in any clinical setting.

Methods A search strategy was applied to Medline and Embase for material published prior to 10 August 2015. Two authors independently screened titles, and abstracts as necessary. Data including clinical setting, reported sensitivity, specificity, positive predictive value, negative predictive value were extracted independently by two reviewers.

Results 5622 references were screened by title and or abstract. 18 papers and 3 conference abstracts were included after full text review. 7 instruments were identified; Face Arm Speech Test (FAST), Recognition of Stroke in the Emergency Room (ROSIER), Los Angeles Prehospital Stroke Screen (LAPSS), Melbourne Ambulance Stroke Scale (MASS), Ontario Prehospital Stroke Screening tool (OPSS), Medic Prehospital Assessment for Code Stroke (MedPACS) and Cincinnati Prehospital Stroke Scale (CPSS). Cohorts varied between 50 and 1225 individuals, with 17.5% to 92% subsequently receiving a stroke diagnosis. Sensitivity and specificity for the same instrument varied across clinical settings. Studies varied in terms of quality, scoring 13–31/36 points using modified Standards for the Reporting of Diagnostic accuracy studies checklist. There was considerable variation in the detail reported about patient demographics, characteristics of false-negative patients and service context. Prevalence of instrument detectable stroke varied between cohorts and over time. CPSS and the similar FAST test generally report the highest level of sensitivity, with more complex instruments such as LAPSS reporting higher specificity at the cost of lower detection rates.

Conclusions Available data do not allow a strong recommendation to be made about the superiority of a stroke recognition instrument. Choice of instrument depends on intended purpose, and the consequences of a false-negative or false-positive result.