A Bayesian conditional probability simulation framework for pre-hospital stroke transport decisions in the tenecteplase era: decision boundaries, global sensitivity, and stochastic dominance

Acute ischaemic stroke triage between Drip-and-Ship (DS), Mothership (MS) and Mobile Stroke Unit (MSU) strategies has been formalised over the past decade by conditional-probability decision models. Two recent developments motivate an updated treatment: the shift of the intravenous thrombolysis standard from alteplase to tenecteplase, and the maturation of randomised evidence on DS-versus-MS triage and on MSU effectiveness. A Bayesian conditional-probability simulation framework is presented that combines random-effects synthesis of the modern trial corpus for intravenous thrombolysis (alteplase, tenecteplase) and endovascular thrombectomy with a log-normal decomposition of onset-to-reperfusion into strategy- and pharmacology-specific process-time components. The prehospital large-vessel-occlusion probability is updated from validated screening scales, and the optimal strategy together with a 95% posterior credible interval is computed at every coordinate of the decision space spanned by the inter-facility transit-time differential and the LVO posterior probability. Sobol first-order and total-effect indices decompose outcome variance over the sampled input space, and first- and second-order stochastic dominance between strategies is tested under the same posterior. Under the production calibration the tenecteplase paradigm reallocated 14.1% of the decision plane relative to alteplase, concentrated in the low-probability, long-differential region; the prehospital large-vessel-occlusion posterior accounted for the dominant share of outcome variance (Sobol first-order index 0.84-0.93); and no coordinate of the decision plane reached 95% posterior agreement on a single optimal strategy. The framework is open and reproducible and depends only on published trial-level evidence. It is presented as a decision-analytic and hypothesis-generating tool for structuring the tenecteplase-era triage question rather than as a validated operational triage rule, and prospective calibration against region-specific stroke-system data would be required before any deployment.