The timeliness of the EARS algorithms exceeded that of the negative binomial cusum given that the outbreak was detected within the first 7 days (Tables 1 and 2).
More specifically, we used the dates of these confirmed p9 outbreaks to assess both the sensitivity and timeliness of the detection algorithm presented here when using census tract centroids versus individual addresses.
Comparison of the timeliness of the seven cusum algorithms found no significant difference at the 0.005 and 0.001 false alarm levels (Friedman χ2 = 3.8, df = 6, p = 0.71 and Friedman χ2 = 2.1, df = 6, p = 0.91 respectively, Tables 1 and 2).
However, comparison of the adjusted timeliness of the seven cusum algorithms, which allocates the duration of the outbreak as the timeliness result when outbreaks are undetected, found a significant difference at the 0.005 and 0.001 levels (Friedman χ2 = 17.3, p = 0.008 and Friedman χ2 = 25.0, p = 0.0003 respectively, Tables 1 and 2).
Using ICD10-consultations series, the timeliness of CUSUM algorithms that allowed the detection of all outbreaks period ranged between −32.3 and 9.8 days, whereas the specificity ranged from 84.3 to 93.2%.
Using UrgIndex-hospitalisations series, the mean timeliness of CUSUM algorithms that allowed the detection of all outbreaks periods ranged between −58.3 and 18.3 days.
Using ICD10-consultations series, the timeliness of CUSUM algorithms ranged between −8.0 and 27 days, whereas the specificity ranged from 93.9 to 98.2%.
Although we investigated the timeliness of each algorithm for the largest outbreaks in each LGA, there was no one method that consistently detected the outbreaks first.
The mean timeliness of CUSUM algorithms using UrgIndex-hospitalisations series ranged between −10.7 and 14.3 days and the corresponding specificity ranged from 29.4 to 94.8% (Table 2).
Sixth, we computed the sensitivity and timeliness of the three outbreak detection algorithms.
