Table 4

Multivariable logistic regression models of baseline characteristics associated with subgroup membership (Subgroup 1 ‘good outcome’ (n=330) used as the reference group)

	Subgroup 2 ‘moderate outcome’ (n=257)			Subgroup 3 ‘poor outcome’ (n=110)
	OR (95% CI)	p Value	ORshrunk	OR (95% CI)	p Value	ORshrunk
Characteristic
Age, years				0.94 (0.88 to 1.00)	0.04	0.94
Body mass index, kg/m²	1.12 (1.07 to 1.18)	<0.0001	1.12
NRS for knee pain (0–10)	1.35 (1.22 to 1.49)	<0.0001	1.33	1.81 (1.51 to 2.16)	<0.0001	1.72
Bony tenderness				2.63 (1.38 to 5.02)	0.003	2.41
Active knee flexion range, degrees				0.96 (0.93 to 1.00)	0.03	0.97
Hip pain	1.54 (1.07 to 2.22)	0.02	1.51	2.34 (1.22 to 4.48)	0.01	2.17
Osteophytosis				5.68 (2.57 to 12.55)	<0.0001	4.85
Comorbidity count ≥3	1.76 (1.10 to 2.81)	0.02	1.72	3.28 (1.62 to 6.64)	0.001	2.95
SF-36 vitality (0–100)	0.98 (0.97 to 0.99)	0.002	0.98	0.96 (0.94 to 0.98)	0.0001	0.96
PCI resting (5–20)				1.16 (1.02 to 1.31)	0.02	1.14
Model quality parameters
Slope	0.96			0.91
AUCshrunk	0.74			0.90
Nagelkerke's R2shrunk	0.22			0.55
p Value Hosmer–Lemeshow statistic	0.25			0.83

AUCshrunk, area under the receiver operation characteristic curve after adjustment for overoptimism using bootstrapping techniques; Nagelkerke's R2shrunk, Nagelkerke's R squared after adjustment for overoptimism using bootstrapping techniques; NRS, numeric rating scale; ORshrunk, OR after adjustment for overfitting using bootstrapping techniques; PCI, Pain Coping Inventory; SF-36, Short-Form 36 Health Survey; Slope, shrinkage factor obtained after bootstrapping, well-calibrated models have a slope of 1.