Table 2

 Measures of association between individual and area characteristics and the outcome and measures of variation and clustering in the utilisation of private providers in the county of Scania, Sweden, 2000, obtained from multilevel logistic models*

Empty modelModel with individual level variablesModel with the area level variable
CrI, credible interval; ICC, intraclass correlation; IOR, interval odds ratio; MOR, median odds ratio; OR, odds ratio; PCV, proportional change in variance; SE, standard error. *Multilevel models were estimated with the Markov Chain Monte Carlo method implemented in MLwiN (version 1.2.). †The proportional change in variance expresses the change in the area level variance between the empty model and the individual level model, and between the individual level model and the model further including the area level covariate. The ICC concept has been discussed in detail in a related publication (this paper is the fourth of a series of paper) and we refer the reader to these publications for further details.3–5 ‡As discussed in the text, in a model including explanatory factors one different ICC is computed for each combination of the explanatory factors.
Measures of association (OR, 95% CI)
Individual level variables
    Female (v male)1.57 (1.44 to 1.70)1.56 (1.44 to 1.70)
    Age (in 10 years unit)1.13 (1.11 to 1.15)1.13 (1.11 to 1.15)
    High (v low) educational achievement1.25 (1.13 to 1.38)1.24 (1.12 to 1.38)
Area level variable
    High (v low) percentage of highly educated inhabitants1.95 (1.45 to 2.62)
Interval odds ratio (IOR)[0.75 to 5.05]
Sorting out index82%
Measures of variation or clustering
Area level variance (SE)0.388 (0.080)0.379 (0.078)0.275 (0.059)
PCV†−2.3%−27.4%
MOR (95% CrI)1.81 (1.62 to 2.06)1.80 (1.62 to 2.04)1.65 (1.50 to 1.84)
ICC (latent variable method)0.1050.1030.077
ICC (simulation method)0.0820.070–0.080‡0.044–0.061‡