Table 6.

Multivariable logistic regression model of risk factors for feline degenerative joint disease in 6-year-old cats

		B	SE	Wald χ2	df	Sig	OR (95% CI)
Included in the model
Neuter status at 6 months of age	Neutered						Reference
	Entire	0.678	0.226	8.974	1	0.003	1.97 (1.264–3.071)
Outdoor access	No outdoor access						Reference
	Outdoor access	0.513	0.283	3.298	1	0.069	1.671 (0.96–2.907)
BCS (6 years old)	Not overweight						Reference
	Overweight/obese	0.485	0.184	6.964	1	0.008	1.624 (1.133–2.328)
Trauma incidence	None						Reference
	Yes	0.613	0.179	11.759	1	0.001	1.846 (1.3–2.62)
Intercept		−1.745	0.284	37.61	1	0	0.175
Not included in the model
BCS (4 years old)	Not overweight						Reference
	Overweight/obese	−0.023	0.285	0.007	1	0.935	0.977 (0.559–1.707)

B is the unstandardised regression weight which reflects the change in the logit of the outcome variable (status = case or control) associated with a one-unit change in the predictor variable, and SE is the standard error of B. Wald χ² is the test statistic for each predictor variable and is associated with different degrees of freedom (df). Sig refers to the P value associated with each Wald χ² statistic. The odds ratio (OR) is a measurement of likelihood for each predictor variable and is presented with its 95% confidence interval (CI). The multivariable logistic regression was built using a backwards elimination method, and removal of variables was undertaken based on minimising the log-likelihood ratio statistic (–2LL), rather than the P value associated with the Wald χ² statistic. The model χ² is the difference between the model –2LL and the baseline –2LL; this was χ² (4) = 30.348 (P <0.001). Cox and Snell’s measure was 0.040 and Nagelkerke’s measure was 0.057; these provided an approximate effect size measure for the model

BCS = body condition score