Table 6

Comparison of multilevel models for change for each health outcome before introducing ICF categories in the model and for the final model
GH Scale MCS WHODAS II Score
Before introducing ICF categories Final model Before introducing ICF categories Final model Before introducing ICF categories Final model
Variance estimates of random effects: mean (SE)
Intercept 267.81** (22.01) 211.96** (17.75) 52.27** (5.63) 35.94** (3.76) 127.63** (12.95) 72.65** (7.38)
Residual 130.02** (6.61) 115.79** (5.90) 57.41** (3.09) 42.24** (2.25) 78.08** (4.15) 55.42** (2.95)
Goodness-of-Fit: mean (SE)
Deviance (−2 Log Likelihood) 10250.08 (3.98) 10110.68 (4.67) 8148.10 (2.60) 7798.55 (4.34) 8965.81 (3.27) 8496.80 (8.56)
Number of Parameters 16 24 15 25 15 21
Chi-square statistic(Degrees of freedom) 11.82** (10) 251.22** (18) 262.92** (10) 612.47** (20) 261.96** (10) 730.97** (16)
AIC 10282.08 (3.98) 10062.68 (4.67) 8118.10 (2.60) 7748.55 (4.34) 8935.81 (3.27) 8454.80 (8.56)

The deviance and Akaike’s Information Criterion (AIC). The smaller the deviance and the AIC the better the model fits the data.

*P < 0.05, **P < 0.001.

Variance estimates of random effects and goodness-of-fit statistics are reported for each model. For assessing goodness-of-fit the models are compared to the corresponding individual growth model. For each parameter its mean and the mean of its standard errors (SE) over the five imputed datasets are reported.

Cieza et al.

Cieza et al. BMC Psychiatry 2013 13:78   doi:10.1186/1471-244X-13-78

Open Data