{"resourceId":"sjtrem-uppsala-ml-ground-helicopter-ems-travel-time-prediction-20261102","versions":[{"version":"external-ebfd5c13761c2f770b1cce429cd79cd27c6cf9763d8a5cc732ddaa107e80a6bb","resource":{"id":"sjtrem-uppsala-ml-ground-helicopter-ems-travel-time-prediction-20261102","title":"Uppsala study validates open-source routing plus gradient-boosting and regression models to predict ground and helicopter EMS travel times for dispatch decisions","organization":"Uppsala University; Uppsala University Hospital ambulance, HEMS and emergency dispatch services","sector":"Emergency medical services dispatch and resource allocation","geography":"Uppsala region, Sweden","publishedAt":"2026-11-02","publicationDate":"2026-11-02","eventDate":null,"sourceName":"Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine","sourceLabel":"Development and validation of methods to predict ground- and helicopter-based emergency medical service travel times","sourceUrl":"https://sjtrem.biomedcentral.com/articles/10.1186/s13049-026-01722-1","evidenceClass":"academic-research","outcomeClass":"mixed","topics":["data-security","operating-model","infrastructure"],"finding":"Spangler et al. (published Nov. 2, accepted early version) estimated ground EMS travel times with the open-source OSRM routing tool, then adjusted them with linear regression and gradient boosting trained on historical travel times; helicopter times used great-circle distance adjusted by historical data. On validation data collected after the methods were finalized, the best ground models explained much of the variation in drive time to scene and to hospital (R2 0.79 and 0.70), but helicopter flight times much less (R2 0.46 and 0.38). Mean average percentage error was 18-20% across all techniques and intervals. The authors release open-source code and say it could help dispatchers judge the time benefit of HEMS versus ground units.","sledRelevance":"Helicopter-versus-ground decisions are a costly dispatch choice for US state EMS offices, regional dispatch centers and air medical programs. This is a prospectively validated, open-source machine-learning approach rather than a vendor claim. Transferability caveat: single Swedish region and road network; US agencies would need to retrain on their own historical CAD/AVL times. Interpretation: an ~18-20% error band suggests decision support, not automated assignment.","evidence":"Peer-reviewed, accepted article (early version, subject to edits); abstract-level results inspected, full methods tables not reviewed. Validation on data collected after method finalization. Ethics approval by the Swedish Ethics Review Authority. No patient outcome or dispatcher-use evaluation reported.","architectureImplications":"Open-source routing engine (OSRM) plus ML correction layer trained on historical CAD/AVL travel times; helicopter estimates from great-circle distance with historical adjustment. Could run on-premises alongside CAD.","governanceImplications":"Use as decision support requires documented error bands and a rule that dispatchers or medical directors keep the HEMS activation decision.","securityPrivacyImplications":"Training uses historical incident locations and times, which are operationally sensitive; patient data needs are minimal but location records require access controls.","caveats":"Single Swedish region; early accepted version; abstract-level inspection; helicopter predictions weak (R2 0.38-0.46); no evaluation of dispatcher decisions or patient outcomes.","streamIds":["emergency-services"],"roles":{"sales":"For sales, the framing point is costly dispatch decisions backed by validated, local data. Uppsala researchers showed ML-adjusted routing explains most variation in ground EMS drive times (R2 0.79/0.70) but helicopter times far less (0.46/0.38), with 18-20% average error. A bounded conversation with state EMS offices and air medical programs is about whether their own CAD/AVL history supports similar validation. The evidence cannot support claims of faster patient care or lower HEMS costs.","engineering":"For pre-sales engineering, the pattern is an open-source routing engine with a gradient-boosting correction layer trained on historical travel times, plus a distance-based helicopter estimate. Engineers should check CAD/AVL timestamp quality, map data currency, and how predictions surface in the CAD screen. Constraints: one Swedish region, early accepted version, weak helicopter accuracy, and no dispatcher-use evaluation, so local retraining and prospective validation are prerequisites.","delivery":"For delivery, the work is data preparation, local model retraining and a shadow-mode comparison of predicted versus actual travel times before dispatchers see estimates. The EMS medical director and dispatch operations manager would own use rules. Proposed acceptance criteria include locally measured error within an agreed band and documented display of uncertainty. Risks include over-trusting point estimates, stale road data and weak helicopter predictions."},"roleDetails":{"sales":{"customerProblem":"Dispatchers must choose helicopter or ground EMS without reliable estimates of which arrives and transports faster.","stakeholders":["EMS medical director","Dispatch center manager","Air medical program","State EMS office","CAD administrator"],"discoveryQuestions":["How clean are your historical CAD/AVL travel-time records?","How are HEMS activation decisions made today?","What prediction error would dispatchers accept?","Can predictions display inside CAD?"],"valueHypothesis":"Interpretation: locally validated travel-time predictions could make HEMS-versus-ground choices more consistent; outcome benefit is unmeasured.","potentialEngagement":"A data-readiness review and shadow-mode validation of travel-time predictions on one region's history.","unsupportedClaims":["Faster patient arrival","Reduced HEMS costs","Improved patient outcomes","Accuracy outside Sweden"]},"engineering":{"fit":"Fits regional dispatch centers coordinating ground and air medical resources.","architectureIntegration":"OSRM routing with gradient-boosting and regression correction on historical times; helicopter great-circle estimates adjusted by history; output fed to CAD.","prerequisites":["Historical CAD/AVL timestamps","Current road network data","HEMS base and flight records"],"constraints":["Single Swedish region","Helicopter R2 only 0.38-0.46","Early accepted version subject to edits"],"security":"Restrict access to historical incident location data and model outputs; log model versions used in dispatch.","proofOfValue":["Measure local MAPE on held-out recent data","Run shadow mode against actual travel times","Compare helicopter predictions to flight logs"]},"delivery":{"work":"Prepare data, retrain locally, validate in shadow mode, then pilot as advisory display.","dependencies":["CAD vendor integration","GIS team","Air medical provider data"],"ownership":"EMS medical director with dispatch operations manager.","skillsAdoption":"Dispatchers learn to read predictions with uncertainty ranges and keep final activation authority.","governanceCheckpoints":["Medical director approval of use rules","Shadow-mode results review","Periodic recalibration review"],"acceptanceCriteria":["Proposed: local ground MAPE at or below an agreed threshold","Proposed: uncertainty shown with every estimate","Proposed: quarterly recalibration documented"],"risks":["Over-trust in point estimates","Stale road data","Weak helicopter accuracy","Model drift"]}},"retrievedAt":"2026-11-04T04:04:33.000Z","enrichedAt":"2026-11-04T04:08:05.000Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Gives dispatchers a quantified time comparison rather than relying on experience alone; requires training on interpreting error ranges.","procurementImplications":"Open-source code lowers licensing cost; agencies still need data engineering and validation effort. Interpretation: buyers can ask CAD vendors whether travel-time predictions are validated on local data.","operatingModelImplications":"Supports a more explicit, evidence-based HEMS activation step in dispatch protocols.","sourceVerification":{"responseId":"curl-sjt-2026-11-03","openedUrl":"https://sjtrem.biomedcentral.com/articles/10.1186/s13049-026-01722-1","referenceExcerpt":"Mean average percentage error was 18-20% for all investigated estimation techniques and time intervals.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}