CasualtyLogic: a hospital rehearses a major incident twenty times a year
In a mass casualty event, the decisions made in the first hours determine most of the preventable deaths. The tools for rehearsing those decisions run at human speed, one scenario at a time.
Tabletop exercises and simulation mannequins are good training. They are also profoundly limited: one scenario, one pace, one fixed set of variables. A hospital system preparing seriously for a major incident might get through ten or twenty rehearsals in a year. The number of ways a real incident can differ from those twenty is not twenty.
CasualtyLogic runs ten thousand parametric variations of the same incident overnight. Not to replace the exercise, but to find the decision failure modes that a human-paced rehearsal will never surface — the combinations of casualty load, resource availability, evacuation delay and communication breakdown that turn a protocol that works into a protocol that does not.
There is a second problem it addresses, and it is more uncomfortable. The mass casualty protocols in use across health systems today are derived from retrospective analysis of incidents that already happened. They are optimised for the past. Nothing in that process tests how a protocol performs under conditions that have not occurred yet — a novel incident type, a degraded communications environment, a surge that exceeds every historical precedent.
The platform models the physiological deterioration of casualties over time, the decisions responders make under incomplete information and time pressure, and the operational constraints that shape both. That third layer is the one usually left out: evacuation delays, transport capacity, personnel ratios, and multi-site incidents where facilities have to coordinate. Those are not background details. They are frequently the deciding factor.
One capability we think matters disproportionately is counterfactual after-action review. Feed in the decision timeline from a real incident or an exercise, and the platform reconstructs what would have happened along different decision paths. That turns a debrief from a qualitative discussion about what felt wrong into a quantitative account of what cost lives.
Where it stands: technology readiness level three, twelve months of development ahead of it, and no deployment anywhere. It is designed to run on published clinical literature rather than restricted patient data, which means it can be validated without waiting on data access agreements — but validated is exactly what it is not, yet.