Table 3.
Results of multivariable analyses1 using a disease control program database* to determine the hazard ratios associated with herd type, season of diagnosis, and year of diagnosis with the amount of time it took, in weeks, to eliminate porcine epidemic diarrhea virus.
| Variable | Hazard ratio | p-value | 95% CI | Partial likelihood ratio (p-value) |
|---|---|---|---|---|
| Herd type: | ||||
| Farrow-to-feeder† | ||||
| Farrow-to-wean | 3.36 | 0.044 | 1.03, 10.93 | <0.001 |
| Wean-to-finish | 0.61 | 0.53 | 0.13, 2.83 | |
| Farrow-to-finish | 0.80 | 0.70 | 0.25, 2.54 | |
| Finisher only | 1.07 | 0.89 | 0.39, 2.98 | |
| Nursery only | 2.33 | 0.15 | 0.74, 7.36 | |
| Season: | ||||
| Winter† | ||||
| Spring | 1.40 | 0.20 | 0.84, 2.31 | <0.001 |
| Summer | 5.04 | <0.001 | 2.74, 9.27 | |
| Fall | 7.32 | <0.001 | 3.12, 17.18 | |
| Year: | ||||
| 2014† | ||||
| 2015 | 0.42 | 0.002 | 0.25, 0.72 | <0.001 |
| 2016 | 1.62 | 0.10 | 0.91, 2.89 | |
| 2017 | 2.15 | 0.21 | 0.64, 7.15 | |
The Area Regional Control and Elimination program (ARC&E) database was used to collect diagnostic data on porcine epidemic diarrhea virus herd status of Ontario swine herds on a weekly basis using the Animal Health Laboratory (AHL) at the University of Guelph. Premises that volunteered to participate in the program from January 2014 to October 2017 were included in the study. Descriptive survival analysis statistics are described above.
Referent categories.
SE, standard error; CI, confidence interval.
A Cox's proportional hazard model was constructed to investigate the effect of several predictor variables including herd type, season of diagnosis and year of diagnosis upon the time to eliminate PEDV from the premises in a multivariable model. The time to event (i.e., elimination) was identified as the time in weeks for a premises to change from confirmed or presumed PED-positive to confirmed or presumed PED-negative. A failure occurred if the premises eliminated PEDV. In the current study, the Hosmer-Lemeshow test indicated that the model fits the data (p = 0.46). Also, Harrell's C concordance statistic computed (0.72) found that the model had good overall predictive ability.