Skip to main content

Table 3 Multi-level logistic regression models for referral to the LMP

From: Predictors of primary care referrals to a vascular disease prevention lifestyle program among participants in a cluster randomised trial

Explanatory variables   Empty model Model 11
Patient characteristics   OR (95% CI) OR (95% CI)
SEIFA quintile2 1 - poorest   1.00 (reference)
  2   9.87 (0.21, 471.1)
  3   1.44 (0.12, 17.53)
  4   3.36 (0.09, 122.4)
  5 - richest   7.18 (0.33, 155.1)
BMI <25   1.00 (reference)
  ≥25   2.87 (1.10, 7.47)
Physical activity Active   1.00 (reference)
  Inactive   2.90 (1.36, 6.14)
Stage of change for increasing fruit and vegetables Maintenance   1.00 (reference)
  Contemplation/preparation/action   0.70 (0.28, 1.78)
  Pre-contemplation   0.61 (0.11, 3.49)
Stage of change for decreasing dietary fat intake Maintenance   1.00 (reference)
  Contemplation/preparation/action   1.56 (0.60, 4.05)
  Pre-contemplation   2.88 (0.47, 17.71)
Stage of change for increasing physical activity Maintenance   1.00 (reference)
  Contemplation/preparation/action   2.75 (1.07, 7.03)
  Pre-contemplation   0.83 (0.14, 4.79)
Stage of change for losing weight Maintenance   1.00 (reference)
  Contemplation/preparation/action   1.20 (0.41, 3.48)
  Pre-contemplation   1.59 (0.46, 5.52)
GP advice/referral for physical activity in previous 3 months No   1.00 (reference)
  Yes   1.43 (0.68, 3.00)
Practice characteristics    
Practice location urban   1.00 (reference)
  rural   12.50 (1.43, 109.7)
Practice size: ≥ 3 GPs   1.00 (reference)
  1-3 GPs   16.05 (2.74, 94.24)
Between patient variance (SE3)   1.176 (0.852) 1.045 (0.627)
Intra class correlation   0.263 0.241
Explained variance 4 (%)   - 11.14
  1. *P < 0.05.
  2. Statistically significant p-values (<0.05) are shown in bold.
  3. Multilevel logistic regression1 Model 1: includes all variables found to be significant in univariate analysis. 2 2006 index of relative socio-economic advantage/disadvantage, 3 Standard error, 4 Explained ‘between practice’ variance using the variance in the empty model as reference.