Paramedic identification of stroke mimic presentations : development and preliminary evaluation of a pre-hospital clinical assessment tool

Abstract

Background Stroke mimic (SM) conditions produce stroke-like symptoms through diverse mechanisms. Up to 43% of pre-hospital suspected stroke patients are SM because identification tools prioritise sensitivity over specificity, leading to inefficient use of ambulances and stroke services. No existing pre-hospital SM identification tools could be identified. A pragmatic SM identification tool using easily available information from suspected stroke patients was developed. Methods A systematic literature review and a national paramedic survey generated possible tool content. Independent predictors were isolated by regression analysis of selected variables documented in ambulance records of suspected stroke patients linked to primary hospital diagnoses (derivation dataset, n=1,650, 40% SM). The tool was refined using an expanded dataset (n=3,797, 41% SM), usability testing and professional focus groups. The potential clinical impact was evaluated through basic service efficiency modelling and focus groups. Results The “STEAM tool” combines six variables:  1 point for Systolic blood pressure<90mmHg  1 point for Temperature>38.5oC with heart rate>90bpm  1 point for seizures or 2 points for seizures with known diagnosis of Epilepsy  1 point for Age<40 years or 2 points for age<30 years  1 point for headache with known diagnosis of Migraine  1 point for FAST-ve suspected stroke A score of ≥2 on STEAM predicted SM diagnosis in the refinement dataset with 5.5% sensitivity, 99.6% specificity and positive predictive value (PPV) of 91.4%. External validation (n=1,848, 33% SM) showed 5.6% sensitivity, 99.5% specificity and a PPV of 85.0%. Focus groups with paramedics and hospital clinicians identified benefits and risks to patients ii and clinical services from using STEAM. Conclusions A multi-method approach developed and validated a tool using common clinical characteristics to identify a small proportion of SM patients with a high degree of certainty. The tool appears feasible for pre-hospital use but its impact will depend upon local models of stroke care.