Spatiotemporal AED Location Optimization

Background: Mathematical optimization can be used to plan future AED placement to maximize out-of-hospital cardiac arrest (OHCA) coverage. Many public access AEDs are placed in locations without 24/7 access. AED coverage can be overestimated unless temporal availability is considered. Objective: To develop a new spatiotemporal AED location optimization model that accounts for both spatial and temporal information. Methods: We identified all atraumatic public-location OHCAs occurring in Toronto, Canada from Jan. 2006 – Aug. 2014. We gathered location and operating hours data for 4898 buildings that were used as potential sites for AED placement. We extended a previously published spatial optimization model, which identifies locations to place AEDs that maximize the number of historical OHCAs occurring within 100 m of an AED. The new spatiotemporal model finds AED locations that maximize the number of OHCAs occurring within 100 m of an available AED, considering when the OHCAs occurred (“actual coverage”). We then compared the spatial and spatiotemporal models on actual coverage of out-of-sample OHCAs using 10-fold cross validation. Statistical analysis was performed using McNemar’s test. Results: We identified 2440 atraumatic public-location OHCAs. AED locations chosen by the spatiotemporal model outperformed those chosen by the spatial model by 26.1% in actual coverage (p