Methods & Data
The programme combines qualitative evidence about real-time dispatch work with observational analyses of large-scale EMS data.
The methodological objective is system understanding under uncertainty—especially heterogeneity and non-linear patterns—rather than automation.
Data (high-level)
We use routinely collected EMS operational metadata (e.g., call/dispatch classifications, timestamps, response-time intervals) and on-scene clinical
classifications (e.g., ESS and triage labels where applicable), analysed in aggregate and with privacy-by-design principles.
Exploratory modelling and interpretability
Where machine learning is used, the emphasis is on interpretability tools (e.g., PD/ICE-style analyses) to map non-linear associations and interaction
patterns and to characterise heterogeneity. Conventional performance metrics are reported only to contextualise model difficulty and should not be
interpreted as evidence for deployable triage or dispatch automation.
Response time as a distribution
Safety-relevant performance is evaluated using distributions (including tail delays) and context-dependent variability. Queue dynamics and prioritisation
are treated as central to interpretation.
Ethics and governance
Register-based analyses follow applicable ethical approvals and disclosure control. Outputs are presented at levels suitable for responsible sharing.
Adaptive EMS Allocation 