Part Two
There were 1513 simulated outbreaks generated in Part Two of the study. Of these, there were 1328 outbreaks with >= 2 detected IPs. Descriptive statistics for these outbreaks are shown in Table 4. There were 114 ‘extreme’ outbreaks that exceeded the size and/or duration of the UK 2001 FMD epidemic, specifically > 2030 IPs or > 221 days (Anderson Inquiry 2002).
Of the 1328 outbreaks with >= 2 IPs, the EDI trigger fired 816 times (61.4% of outbreaks). There were two versions of the EDI trigger, which differed with respect to the EDR component, with one using a 4-day EDR and slightly lower thresholds for triggering than that specified for the 5-day EDR (refer Table 3). Statistical analyses (not reported) indicated there was minimal difference between the overall performance of the two versions, so the results below are based on the combined data.
The day the trigger fired, and whether it was triggered by the cumulative count of IPs exceeding the stated thresholds or EDR are shown in Figure 1. The trigger mostly fired between Days 11 – 21, with the largest proportion of firings on Day 11 (see Figure 1). There were no trigger firings due to IP numbers after Day 21, although there were diminishing numbers of EDR-based trigger firings right through to Day 35.
Table 5 shows the results from the final logistic regression of thetrig variable. The variables catdens , sheepdens andpigdens were all highly collinear with fmdens so were dropped from the model. It can be seen that farm type of the primary case, time to first detection and farm density in the area around the primary case were all associated with the trigger firing. The odds of the trigger firing were higher for grazing / dairy support farms (GRADRY) and pastoral livestock (PLVSTCK) farms relative to dairy farms (DAIRY), whereas lifestyle farms (LIF) were protective (reduced the odds of the trigger firing). There were not enough pig farms selected as the primary case for the role of pig farms to be assessed. Increasing time to first detection and increasing farm density around the primary case were associated with increased odds of the trigger firing.
The third quartiles for the number of IPs and duration amongst the iterations where spread occurred were 49 IPs and 55 days respectively. Iterations that exceeded these values were categorized as large and long outbreaks respectively.
Performance of the EDI trigger was assessed by 2x2 contingency tables for IPs (Tables 6-8) and durations (Tables 9-11).
Final logistic regression models for variables associated with large and long outbreaks are shown in Tables 12 and 13 respectively. The most influential variable was the trig variable, with the odds of a large or long outbreak being 347 or 33.3 times greater respectively if the trigger fired. Besides this, increasing vet andsurvfpd values reduced the odds of a large outbreak, and increasing ft and survfpd values reduced the odds of a long outbreak.