Earthquake early warning (EEW) is currently deemed a credible approach to seismic resilience enhancement in modern societies, especially if part of a more holistic earthquake mitigation strategy involving other risk reduction tools such as structural upgrading/retrofit. Yet, there remains a strong need to 1) assess the feasibility of EEW in various seismotectonic contexts, considering specific target applications/end users; and 2) develop next-generation decision-support systems relying on interpretable probabilistic impact-based estimates toward more risk-informed decision-making on EEW installation/alert triggering. These challenges are addressed in this paper, which showcases a series of recent significant EEW contributions by the authors. First, we present the results of a state-of-the-art feasibility study for EEW in schools performed across the Patras region of Greece, attempting to spatially combine traditional seismologically-driven EEW decision criteria (i.e., warning time) with proxy risk-oriented measures for earthquake impact (i.e., building fragility and the number of exposed school students). These results show that, under certain conditions, EEW could be effective for the schools in the considered case-study region. We then demonstrate an advanced end-user-centred approach for improved risk-informed decision-making on triggering EEW alerts. The proposed methodology integrates earthquake-engineering-related seismic performance assessment procedures and metrics with multi-criteria decision-making (MCDM) within an end-to-end probabilistic framework. The performance-based earthquake engineering component of such a framework facilitates the computation of various damage/loss estimates (e.g., repair cost, downtime, and casualties) by combining target-structure-specific models of seismic response, fragility, and vulnerability with real-time ground-shaking estimates. Additionally, the incorporated MCDM methodology enables explicit consideration of end-user preferences (importance) towards the estimated consequences in the context of alert issuance. The developed approach is demonstrated using an archetype school building for the case-study region, for which we specifically investigate the optimal decision (i.e., "trigger" or "don't trigger" an EEW alert) across a range of ground-motion intensity measures. We find that the best action for a given level of ground shaking can vary as a function of stakeholder preferences. [ABSTRACT FROM AUTHOR]