Abstract TMP77: To Treat or Not to Treat? A Survey Exploring Factors Influencing rt-PA Treatment Decisions for Minor Stroke

Abstract

Introduction: The 2015 updated US FDA package insert for rt-PA allows for a greater degree of clinical judgment regarding contraindications; thus there is flexibility in the decision to treat or not to treat. Hypothesis: Identifiable clinical factors influence rt-PA treatment decisions for patients with minor stroke. Methods: An Expert Panel (N=10) selected 7 factors from which to build a series of survey vignettes: NIHSS (1-5), NIHSS area of primary deficit (visual, language, weakness), baseline functional status (fully independent, mild-moderate disability), previous ischemic stroke (IS) (Yes < 6 wks, Yes ≥ 6 wks or No), previous ICH (Yes < 6 mos, Yes ≥ 6 mos, No), recent use of anticoagulation (AC) (Yes < 48 hrs, Yes ≥ 48 hrs, No), and temporal pattern of symptoms in 1 st hr of care (stable, improving). We used a fractional factorial design (150 vignettes) to provide unconfounded estimates of the effect of all 7 main factors, plus first-order interactions for NIHSS. Surveys were emailed to 5 national organizations of neurologists and/or emergency physicians and investigator colleagues. Physicians were randomized to 1 of 10 sets of 15 vignettes, presented in random order. Physicians reported the subjective likelihood of giving rt-PA on a 0-5 scale; scale categories were anchored to probabilities 0, 20, 40, 60, 80, and 100%. A conjoint statistical analysis was applied. Results: Responses from 194 US physicians yielded 156 with complete vignette data: 74% male, mean age 46 (range 27-76), 80% neurologists. Treatment mean probabilities for individual vignettes ranged from 5-95%. Treatment probability increased from 24% for NIHSS=1 to 41% for NIHSS=5. The conjoint model accounted for 25% of total observed response variance. In contrast, a model accounting for all possible interactions accounted for 30% variance. Four of the 7 factors accounted jointly for 58% of total relative importance within the conjoint model: previous ICH (18%), recent AC (17%), NIHSS (13%), and previous IS (10%). Conclusions: Four main variables jointly account for only a small fraction (< 15%) of the total variance related to deciding to treat with IV rtPA, reflecting high variability and complexity. Future studies should consider other variables, including physician characteristics.