{"resourceId":"scand-j-trauma-ems-travel-time-models-uppsala-ground-helicopter-20261102","versions":[{"version":"external-8fefb690f9cc301e10ca81a3085900a8cd21ee46d267e055c22c7042803311f9","resource":{"id":"scand-j-trauma-ems-travel-time-models-uppsala-ground-helicopter-20261102","title":"Uppsala study: regression and gradient-boosting EMS travel-time models validated on later data explain 70-79% of ground variance but only 38-46% for helicopters","organization":"Uppsala University and Uppsala University Hospital (Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine)","sector":"EMS dispatch / ambulance versus helicopter selection","geography":"Uppsala region, Sweden","publishedAt":"Published 2 November 2026, Volume 34, article 170","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://link.springer.com/article/10.1186/s13049-026-01722-1","evidenceClass":"academic-research","outcomeClass":"mixed","topics":["operating-model","infrastructure"],"finding":"Researchers used 2022-2024 dispatch and flight-log data from Region Uppsala (24,756 priority-1 incidents, 527 with helicopter response) to adjust open-source routing estimates for ground ambulances, and great-circle distances for helicopters, with linear regression and gradient boosting. In validation data collected after the methods were finalized, the best ground models explained R2 0.79 (travel to incident) and 0.70 (travel to hospital); helicopter models reached R2 0.46 and 0.38. Mean absolute percentage error was 18-20% across techniques and intervals. Regression and gradient boosting performed similarly, with a slight edge for boosting. The authors say open-source code could help dispatchers estimate relative travel times and judge whether a helicopter saves time.","sledRelevance":"Concrete, openly published decision-support input for dispatchers choosing between ground and air resources. It is classical prediction rather than generative AI and was not tested in live dispatch, so relevance to AI programs is partial.","evidence":"Opened the full open-access article on Springer Nature Link via HTTPS curl (User-Agent SLED-research-bot/1.0), read abstract, study setting, results and limitations, and saved the text. Supplementary model specifications were not opened.","architectureImplications":"A routing engine plus a small calibration model fed by historical dispatch and flight logs; output is an estimate shown to dispatchers, not an automated decision.","governanceImplications":"Travel-time estimates that steer scarce helicopter use need documented error bounds and a clear rule that clinicians and dispatchers keep final authority.","securityPrivacyImplications":"Training uses incident locations and times from dispatch records; de-identification and access controls are needed before reuse.","caveats":"Single Swedish region; priority-1 missions transported to one primary hospital only; extreme long travel times tended to be under-estimated; helicopter volume was small (527 responses); no patient-outcome or live-dispatch evaluation; Swedish road network and agency structure.","streamIds":["emergency-services"],"roles":{"sales":"Sales framing: this is a bounded, evidence-backed dispatcher aid, not an AI promise. Ground travel-time models held up on later data (R2 0.79 and 0.70), while helicopter estimates were weaker (0.46 and 0.38) and errors still ran 18-20%. Offer a baseline accuracy audit of current ETA tools using local dispatch history. Do not claim faster arrival, better patient outcomes or cost savings; the study did not measure live use, and it covers one Swedish region and high-priority missions only.","engineering":"Engineering view: the method layers regression or gradient boosting over open-source routing output, trained on dispatch-system and flight-log data. Check data quality (documented ambulance starting points), CAD integration latency and handling of long-trip under-estimation. Proof of value is a replay on local missions measuring error, especially in the tail, before any dispatcher display. Helicopter estimates rest on great-circle distance and are noticeably less reliable, so treat them as indicative only.","delivery":"Delivery plan: assemble local historical missions, rebuild the estimators, and validate on a later period as the authors did. Dependencies are dispatch data owners, the air-medical programme and the medical director. Acceptance should be agreed error thresholds by mission type and a documented dispatcher override. Main risks are poor starting-point data, under-estimated long trips, and treating estimates as clinical decisions without outcome evidence."},"roleDetails":{"sales":{"customerProblem":"Dispatchers must decide whether a costly, scarce helicopter will beat a ground ambulance with limited travel-time information.","stakeholders":["EMS medical director","Air-medical programme manager","Dispatch centre manager","Regional EMS administrator"],"discoveryQuestions":["How are ground and helicopter ETAs estimated at tasking today?","Do you track the error of those estimates?","Who decides when a helicopter is launched?"],"valueHypothesis":"Locally calibrated travel-time estimates may help dispatchers judge relative time benefit, as interpretation of one regional validation.","potentialEngagement":"Four-to-six-week audit of current ETA accuracy and prototype calibration using local dispatch history.","unsupportedClaims":["That helicopter or ambulance tasking improves patient outcomes","That accuracy transfers outside Uppsala","That helicopter estimates are reliable"]},"engineering":{"fit":"Relevant to dispatch centres tasking both ground and air units and holding multi-year mission logs.","architectureIntegration":"Open-source routing service plus regression or boosting calibration reading CAD data, returning estimated ground and air times to the dispatcher screen.","prerequisites":["Clean mission and flight logs","Documented unit starting points","Routing engine with local road data"],"constraints":["Single region and priority-1 only","Under-estimation of the longest trips","Weaker helicopter model"],"security":"Treat incident locations and times as sensitive; restrict model-training data access and log outputs shown to dispatchers.","proofOfValue":["Hold-out replay on a later local period","Error reported by distance band and mission type","Tail-error review of long trips"]},"delivery":{"work":"Rebuild and validate travel-time estimators on local data and design the dispatcher display and override.","dependencies":["Dispatch and flight-log data owners","Routing engine setup","Medical director sign-off"],"ownership":"Dispatch centre manager with the air-medical programme lead.","skillsAdoption":"Dispatchers need to understand estimate error and when to ignore it.","governanceCheckpoints":["Data-use review","Validation results review","Go or no-go before live display"],"acceptanceCriteria":["Agreed MAPE and tail-error thresholds met on a later period","Documented override path","Dispatcher feedback reviewed after trial"],"risks":["Poor starting-point data","Over-reliance on estimates","Under-estimated long trips"]}},"retrievedAt":"2026-11-10T03:57:06Z","enrichedAt":"2026-11-10T03:59:17Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Not addressed by the source.","procurementImplications":"Agencies can ask vendors for validation on post-training data and for error measures such as MAPE; open-source code lowers lock-in risk.","operatingModelImplications":"Intended to inform relative ground-versus-helicopter judgments at tasking; the study did not test changes in dispatcher behaviour or patient outcomes.","peerReview":"peer-reviewed","sourceVerification":{"openedUrl":"https://link.springer.com/article/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"}}}]}