Suggest edit — Ambulance Trajectories

Title

Name

Note

---
title: "Ambulance Trajectories"
type: concept
related: [Forward Search]
source: https://www.jemoka.com/posts/kbhabulance_trajectories/
confidence: high
status: active
---

Problem: current ambulance routing don&rsquo;t optimize significantly on the contextual cases for stroke patients
Stroke hospitals: PSC is smaller than a CSC.
Previous work Routing methods&mdash;
route all patient to nearest PSC, which is worse than route high risk patient to CSC, which is worse than always route to CSC This is counter-intuitive. How do we solve, given a stroke condition, available PSC/CSC locations, traffic, etc., for where and how to route a patient?
Ambulance MDP formulation \(S\): (location, symptom onset, known stroke type, stroke type) \(A\): route to clinic, route to [specific] PSC, route to [specific] CSC will never be downrouted (for instance, if you are at a PSC you will always either stay or go to CSC) \(T(s&rsquo;|s,a)\): location changes distance \(R(s,a)\): &ldquo;probability of patient outcome&rdquo; \(P(success|time)\) (Holodinsky, et. al. 2018) if stroke type is unknown, its a weighted average Solving Forward Search, depth of 2: patient will either get transported or bounced and transported.
Results status quo: people near Stanford hospital/ChanZuck are better MDP: smoother gradient