Spread of porcine circovirus associated disease (PCVAD) in Ontario (Canada) swine herds: Part II. Matched case-control study

Table 3 Final multivariable model obtained using assumptions of the Gaussian distribution or exact method

Variable	Odds Ratio	95% Conf. Interval	P ^a	Overall P^b
Final model obtained based on the maximum likelihood method using assumption of Gaussian distribution
Ownership-adjusted high-risk local exposure (OWHRLE; No high-risk local exposure = ref)				<0.001
High-risk local exposure within ownership	8.4E+6	(not estimated)	0.982
High-risk local exposure outside of ownership	1.32	(0.54, 3.22)	0.542
Easting (100 kms)	1.65	(1.22, 2.21)	0.001	<0.001
Herd PRRSv status (positive)	2.25	(1.18, 4.28)	0.013	0.008
N obs = 680; n groups = 170, LR chisq = 29.4, df = 4, P < 0.001, AIC = 450
Adjusted estimates of the high-risk local exposure obtained using exact logistic regression ^c
Ownership-adjusted high-risk local exposure (OWHRLE; No high-risk local exposure = ref)				<0.001
High-risk local exposure within ownership	^d 25.65	(3.40, +Inf)	<0.001
High-risk local exposure outside of ownership	1.27	(0.45, 3.30)	0.640
Easting binary (>600 km east = 1)	1.81	(0.53, 5.62)	0.378	0.378
Herd PRRSv status (positive)	1.93	(1.02, 3.91)	0.04	0.04
N obs = 680; n groups = 170, Model score = 19.98, P < 0.001

^a P-values are based on the Wald test for estimates of a model obtained using maximum likelihood method and based on Score test for estimates obtained using exact method.
^b Overall P-values are based on the likelihood ratio test for estimates of a model obtained using maximum likelihood method and based on Score test for estimates obtained using exact method.
^c Estimates of the local spread where obtained from the exact logistic regression after adjusting for the herd PRRS status and easting as a binary variable. Adjustment for the easting as a continuous explanatory variable could not be evaluated.
^d Best median unbiased estimates.

ISSN: 1746-6148