Skip to main content

Table 3 Strength of association (odds ratios, 95% CI), goodness-of-fit statistics, and discrimination for prediction models using a baseline multivariable prognostic model versus baseline pain only, short-term change or repeat score at repeat assessment – long-term perceived recovery as outcome

From: Brief pain re-assessment provided more accurate prognosis than baseline information for low-back or shoulder pain

Study author Prediction model OR (95% CI) Goodness-of-fit statistics c-statistic (95% CI) Comparison of c-statistics (p-value)
Hosmer & Lemeshow test Nagelkerke pseudo R square
Kuijpers Full model (n = 441)
Baseline pain (0–10)
Duration of complaint:
 - less than 5 weeks
 - 6–11 weeks
 - more than 3 m
Concomitant LBP (yes/no)
Shoulder pain score at physical examination (0–14)
Gradual onset (yes/no)

0.89 (0.82 to 0.98)*

1
0.68 (0.40 to 1.18)
0.36 (0.22 to 0.60)*
0.52 (0.32 to 0.84)*
0.96 (0.91 to 1.01)

0.60 (0.39 to 0.95)*

X 2 (8) = 10.45, p = 0.24

0.19

0.71 (0.67 to 0.76)
 
Full model plus 6w Pain Score (n = 441) 0.72 (0.65 to 0.80)* X 2 (8) = 10.89, p = 0.21 0.29 0.78 (0.74 to 0.83) Full model vs. full model plus 6w pain score, p <0.001*
Van der Windt Full model (n = 282)
Baseline Pain (0–10)
Co-existing neck pain (yes/no)
Preceding trauma (yes/no)
Diagnosis Acute Bursitis (yes/no)

0.80 (0.69 to 0.91)*
0.47 (0.27 to 0.82)*
7.44 (1.69 to 32.85)* 2.32
(1.03 to 5.20)*

X 2 (8) = 9.05, p = 0.34

0.17

0.72 (0.66 to 0.78)
 
  Full model plus 4w Pain Score (n = 270) 0.85 (0.76 to 0.95)* X 2 (8) = 15.07, p = 0.06 0.22 0.75 (0.69 to 0.82) Full model vs. full model plus 4w pain score, p = 0.10
  1. *p < 0.05