| Title |
Condition-adjusted performance measurement of response time from operational logs using a fixed-threshold exceedance probability |
| DOI |
https://doi.org/10.5573/ieie.2026.63.10.115 |
| Keywords |
Computer applications; Response time measurement; Fixed-threshold exceedance probability; Condition adjustment; Operational log analysis |
| Abstract |
This study constructs a condition-adjusted method for measuring temporal performance using a fixed-threshold exceedance probability, and evaluates the long-term change, using 266,161 operational log records of the nationwide fire response system for 2019-2025. The system response time is defined as the recorded interval from the dispatch time to the on-scene arrival time. This interval is an observable system output, but the composition of incidents and of responding stations also changes, so raw year-to-year comparisons of summary statistics are insufficient. The analytical sample was drawn from 270,174 nationwide fire incident records of the National Fire Agency for 2019-2025, retaining those with an available station identifier that passed the pre-specified temporal-consistency screening. A fixed performance-reference threshold of 14.27 minutes, the 95th percentile of the pooled 2019-2022 reference period, was used, and the probability of exceeding this threshold was taken as the primary outcome. The raw exceedance probability changed from 4.5214% in 2019 to 6.9339% in 2025, a difference of +2.4125 percentage points. Adjusting for fire type, place, distance category, month, day of week, hour and weather with station fixed effects, the 2025 versus 2019 difference was +2.4585 percentage points (95% CI +1.9743 to +2.9428) in a linear probability model, and +2.3436 percentage points (+1.8869 to +2.8530) under logistic regression standardization to the reference period composition. The positive direction was retained across all sensitivity specifications, including exclusion rules, a common station sample and three distance specifications, while the magnitude varied by specification. The positive change also persisted in the sample restricted to stations observed in all seven years, indicating the value of operational logs for longitudinal monitoring of temporal-performance changes not attributable solely to incident and station composition. |