Table 5

Multivariate analyses testing associations with the outcome variable. Significant findings are in bold.

Models
Variables tested
OR and 95% CI
• Sensitivity
• Specificity
• Numbers correctly classified
• Area under the ROC

"Best" Swedish model re-tested, according to previous study
Leg pain
1.6 (1.2–2.1)
41%, 87%, 71.5%, 72%


Duration of pain past yr
1.1 (0.8–1.6)



Better at 2nd visit
4.7 (3.4–6.6)


"Full" Swedish model, i.e. including significant variables that had been included in previous study
Leg pain
1.5 (1.1–2.0)
47%, 83%, 71%, 72%


Duration of pain past yr
1.2 (0.8–1.7)



Better at 2nd visit
4.7 (3.4–6.6)



Duration of pain at base-line
1.0 (0.7–1.3)



Immediate improvement
1.3 (1.0–1.7)


Improved Swedish model, i.e. removing irrelevant variables from the model above
Leg pain
1.6 (1.2–2.2)
41%, 87%, 72%, 71%


Better at 2nd visit
4.8 (3.5–6.8)


Final minimal Swedish model, i.e. retaining the "best" variable
Better at 2nd visit
4.9 (3.6–6.8)
68%, 69%, 69%, 69%

Full Finnish model, i.e. allowing for the three new variables included in the present study
Leg pain
1.4 (1.0–2.0)
52%, 83%, 73%, 73%


Duration of pain past yr
1.1 (0.8–1.6)



Better at 2nd visit
5.0 (3.5–7.1)



Duration of pain at base-line
0.9 (0.7–1.3)



Immediate improvement
1.2 (0.9–1.6)


Malmqvist et al. Chiropractic & Osteopathy 2008 16:13   doi:10.1186/1746-1340-16-13