Diagnostic accuracy of FAST and BEFAST for diagnosing stroke in primary care

Background and relevance

Stroke is a serious medical emergency. It can lead to severe disabilities and death if not detected and treated early. The general practitioner (GP) is ofter the first in the path from differential diagnosis, through diagnosis to treatment, both during office hours or during out-of-hours in primary care services (OHS-PCs). It starts with telephonic ‘triage’ that solely relies on symptoms and signs as expressed by the patient. This determines the urgency for further assessment and the differential diagnosis of stroke or a benign condition. During telephone triage, the FAST (‘Face drooping’, ‘Arm weakness’, ‘Speech difficulty’, ‘Time to call for help’) test or the BEFAST (‘Balance’, Eyes’, FAST) test is often used, and it is part of the Netherlands Triage Standard (NTS), a decision support tool for triage nurses at OHS-PC. In the hospital setting, the diagnostic accuracy (the combination of sensitivity and specificity) of (BE-)FAST showed to be highly variable; from modest to poor. In the primary care setting, the accuracy (BE-)FAST has not been evaluated. In this domain though, the patient population with neurological deficits is more heterogeneous and has a lower prior change of stroke with in general less specific signs and symptoms compared to in the hospital. To complicate things even more, the initial contact is through telephone and with triage nurses that are not medically trained. Visual assessment and physical examination are not possible. This all will result in a lower accuracy of (BE-)FAST compared to the hospital setting.

Objectives

1. to assess the accuracy in terms of sensitivity, specificity, positive predictive value and negative predicted value of (BE-)FAST for stroke diagnosis at the OHS-PC
2. to improve (BE-)FAST predictors for stroke including sex, age, cardiovascular risk factors and respective interactions between these

Design

A cross-sectional diagnostic accuracy and diagnostic modelling study using data from the SAFETY-FIRST study. In this study contains over 1350 telephone triage contacts and consultations with patients suspected of stroke calling nine OHS-PCs in the Netherlands.

Methods

The primary outcome in this project is stroke, defined as the final diagnosis related to the OHS-PC consultation as recorded in the file of the patients’ own general practitioner (GP). The diagnostic accuracy of (BE-)FAST is expressed as the sensitivity, specificity, positive predictive value and negative predictive value with 95% confidence intervals. Subsequently, it will be assessed whether (BE-)FAST can be improved using binary logistic regression to develop models based on (BE-)FAST. First sex and age will be included, allowing for possible non-linearity and interaction terms, followed by additional predictors for stroke including patient characteristics (e.g. history of cardiovascular disease) and signs and symptoms (e.g. onset and duration of symptoms). Besides stroke, the outcome TIA will be evaluated. All models will be validated using internal-external cross validation. Performance of each model will be expressed by discrimination, calibration, the Brier score and the Cox-Snell pseudo-R squared.

Intended results and impact

This project provides insight in the accuracy of (BE-)FAST. With this knowledge improvements of the NTS, currently used at nearly all OHS-PCs in the Netherlands, can be developed and ideally assessed in action-research aiming to optimize the workflow of patients with neurological deficits. These steps can lead to adjustments in the NTS and nationwide implementation, resulting in improved and earlier diagnosis and better outcomes for all patients in the Netherlands suspected of stroke calling the OHS-PC.