Concept demonstrator: Holding site location, ambulance allocation, and relocation decision support tool

Date
2017-03
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Before the start of a shift, the dispatchers at the Western Cape Emergency Control Centre (WC ECC) decide where to place holding sites and how many ambulance to allocate to each holding site. During a shift they decide when and where to relocate ambulances. At present, dispatchers make these decisions based solely on their experience and intuition. In this project a concept demonstrator decision support tool (DST) is developed which produces solutions for the near-optimal placement of holding sites per shift, ambulance allocation, and relocation per hour of that shift based on predicted ambulance demand rates. The DST is developed with the aim of assisting the dispatchers at the WC ECC with holding site placement, ambulance allocation, and relocation decisions. The real-world instance utilised during the development of the concept demonstrator DST consists of six months' historical call data from the City of Cape Town and the Cape Winelands municipalities. Singular spectrum analysis is used to forecast ambulance demand according to incident priority. The extended queuing maximum availability location problem model is adapted to t the real-world instance. The model aims to simultaneously maximise expected ambulance coverage and minimise ambulance relocations by manipulating holding site placement, ambulance allocation, and relocation. The solution method implemented for the model as a whole is the arti cial bee colony algorithm. The DST was solved for four planning week instances, at 95% service reliability. Predicted demand for the planning week is predicted using historical demand that precedes the planning week and a recommended schedule of holding site placement, ambulance allocation, and relocation is generated for the predicted ambulance demand. The performance of this schedule is evaluated using the observed historical demand for the planning week. Di erent approaches for the classi cation of calls { consider all calls to be life-threatening, or calls to be life-threatening or non-life-threatening { as well as for the implementation of the model constraints are considered. The results indicate that the WC ECC can improve ambulance coverage with the current, or even smaller, ambulance eet size if decisions are made with the assistance of the DST that anticipates the probable future ambulance demand. The concept demonstrator DST's solutions' expected percentage coverage compared to the actual percentage coverage exceeds 150%. However, it is invalid to compare these values like-for-like as a signi cant number of real-world factors, including the speci c road conditions at the time of each call, the responsiveness of both the ECC operator handling the call and the ambulance team involved, and the communication connection between the ECC call operator and the ambulance team, in uence the real-world response rate and could not be modelled in the DST. However, even when these factors are taken into account, the discrepancy between the actual and the predicted performance is sufficient to convincingly demonstrate the potential of the concept demonstrator DST to assist theWC ECC in further improving their response time.
AFRIKAANSE OPSOMMING: Voor die aanvang van 'n skof besluit die ambulaansversenders by die Wes- Kaapse noodbeheersentrum (WC ECC) waar om wagstasies te plaas en hoeveel ambulanse om by elkeen te plaas. Tydens 'n skof besluit hulle wanneer en waarheen ambulanse geskuif moet word. Tans, maak die versenders staat slegs op hul eie ervaring en intu sie om hul besluitneming te lei. In die projek is 'n konsep demonstreerder besluitsteunstelsel (DST) gebou wat oplossings vir die naas-optimale plasing van wagstasies per skof, ambulaansplasing en -rondskuiwing per uur van daardie skof bepaal gebaseer op voorspelde ambulaansaanvraag. Die konsep demonstreerder DST is ontwikkel met die doel om die versenders by die WC ECC te help met die besluitneming aangaande wagstasieplasing, ambulaansplasing en -rondskuiwing. Die werklikheidsgeval, waarvoor die DST ontwikkel word, bestaan uit ses maande se historiese oproepdata van die Stad Kaapstad en die Kaapse Wynland munisipaliteite. `Singular spectrum analysis' is gebruik om die ambulaansaanvraag volgens voorvalprioriteit te voorspel. Die uitgebreide `queuing maximum availability location problem' model is aangepas om by die werklikheidsgeval te pas. Die model streef om die maksimum verwagte ambulaansdekking en die minimum rondskuiwingskoste deur middel van verbeterde wagstasieplasing, ambulaansplasing en -rondskuiwing te vind. Die oplossingsmetode wat gebruik is vir die algehele model is die `arti cial bee colony' algoritme. Die DST is vir vier gevalle opgelos met 'n 95% diensbetroubaarheidsvlak. Die ambulaansaanvraag vir die beplanningsweek is voorspel gebaseer op historiese ambulaansaanvraag, wat nie die beplanningsweek se historiese ambulaansaanvraag bevat nie. Daarna is'n aanbevole wagstasieplasing, ambulaansplasing en -rondskuiwing skedule gegenereer vir die voorspelde ambulaansaanvraag. Die skedule is geïmplimenteer vir die beplanningsweek. se historiese ambulaans aanvraag. Die resultate is gebruik om die skedule se prestasie the evalueer. Verskillende benaderings vir die hantering van die oproepe volgens voorvalprioriteit { ag alle oproepe as lewensbedreigend, of ag hulle as lewensbedreigend of nie-lewensbedreigend { en twee implementerings van die ambulaansplasingsbeperking word oorweeg. Die resultate dui aan dat die WC ECC die ambulaansdekking kan verbeter met die huidige, of selfs kleiner, ambulaansvloot as besluite geneem word met behulp van die konsep demonstreerder DST in afwagting van die waarskynlike ambulaansaanvraag. Die DST se oplossings se verwagte persentasie ambulaansdekking oorskry die werklike persentasie ambulaansdekking wat bepaal is vir die historiese oproepdata met 150%. Dit moet inaggeneem word dat hierdie waardes nie dieselfde is nie. Beduidende gevalle van die werklikheidsgeval se faktore, insluitend die spesi eke toestand van die paaie tydens elke oproep, die uksheid van die noodbeheersentrum se telefoonoperateur en die ambulaansbemanning, en die kommunikasie tussen die telefoonoperateur en die ambulaansbemanning, be nvloed die werklike reaksietyd en kon nie gemodelleer word nie. Tog, selfs wanneer die faktore inaggeneem word, is die verskil tussen die waargenome en voorspelde prestasies voldoende om oortuigend die potensiaal van die konsep demonstreerder DST te demonstreer as hulpmiddel vir die WC ECC om hul reaksietye verder te verbeter.
Description
Thesis (MEng)--Stellenbosch University, 2017.
Keywords
UCTD, Emergency medicine, Ambulance service -- Dispatching, Spectrum analysis, Forecasting -- Mathematical models
Citation