. 2016 Mar;62(3):e138–e145.

Table 3.

Predicting rural or northern Ontario practice location of practising FPs (97 of 131 CMGs) who finished their PG training in 2011 to 2013 and completed their UG or PG medical education or both at NOSM

PREDICTOR	β	STANDARD ERROR (β)	WALD χ²	P VALUE	OR	95% CI FOR OR
Rural northern Ontario vs other location^*
• Rural background, y	0.15	0.04	11.31	< .01	1.16	1.06–1.27
• Northern Canada background, y	−0.03	0.05	0.30	.58	0.97	0.88–1.08
• NOSM UG only	NA	NA	7.83	.02	1.00	NA
• NOSM PG only	1.84	1.57	1.36	.24	6.27	0.29–136.74
• NOSM UG and PG	2.15	0.77	7.72	< .01	8.62	1.89–39.36
• Constant	−4.17	1.12	13.90	< .001	0.02	NA
Northern Ontario vs other location^†
• Rural background, y	−0.005	0.04	0.02	.89	1.00	0.92–1.07
• Northern Canada background, y	0.04	0.05	.064	.43	1.04	0.94–1.16
• NOSM UG only	NA	NA	30.80	< .001	1.00	NA
• NOSM PG only	2.41	1.21	3.98	.05	11.13	1.04–118.84
• NOSM UG and PG	3.88	0.71	30.15	< .001	48.62	12.15–194.48
• Constant	−1.81	0.99	3.38	.07	0.16	NA
Rural Ontario vs other location^§
• Rural background, y	0.12	0.04	10.06	< .01	1.12	1.05–1.21
• Northern Canada background, y	−0.01	0.05	0.04	.85	0.99	0.91–1.08
• NOSM UG only	NA	NA	5.20	.07	1.00	NA
• NOSM PG only	0.98	1.41	0.48	.49	2.67	0.17–42.35
• NOSM UG and PG	1.40	0.62	5.19	.02	4.06	1.22–13.59
• Constant	−3.13	0.94	11.15	< .01	0.04	NA

CMG—Canadian medical graduate, NA—not applicable, NOSM—Northern Ontario School of Medicine, OR—odds ratio, PG—postgraduate, UG—undergraduate.

Model $χ_{4}^{2} = 19.4$ , P < .01; goodness-of-fit test $χ_{6}^{2} = 1.3$ , P = .97, Nagelkerke R² = 0.31. Correctly classified cases: 84.5% for constant only vs 85.6% for full model. Hosmer-Lemeshow goodness-of-fit test: P > .05 indicated that the model was a good fit with Nagelkerke R² interpreted as a pseudo R². The assumption of linearity in the logit for rural or northern background years was not rejected for any model (P > .11).

^†

Model $χ_{4}^{2} = 54.3$ , P < .01; goodness-of-fit test $χ_{7}^{2} = 6.4$ , P = .50, Nagelkerke R² = 0.58. Correctly classified cases: 59.8% for constant only vs 84.5% for full model.

^§

Model $χ_{4}^{2} = 14.4$ , P < .01; goodness-of-fit test $χ_{7}^{2} = 6.6$ , P = .47, Nagelkerke R² = 0.22. Correctly classified cases: 79.4% for constant only vs 81.4% for full model.