Table 4 Results of the full hierarchical geostatistical models* for mental disorders due to psychoactive substance use and neurotic disorders for all people aged 40–69 years in Malmö, Sweden, in 2001.
| Substance related disorders | Neurotic disorders | |||
|---|---|---|---|---|
| Index | 95% CI | Index | 95% CI | |
| Fixed effects (odds ratios) | ||||
| Income of 250 closest inhabitants (v fourth quartile)† | ||||
| Third quartile | 0.99 | (0.75 to 1.29) | 1.21 | (0.84 to 1.73) |
| Second quartile | 1.56 | (1.18 to 2.07) | 1.73 | (1.23 to 2.45) |
| First quartile | 2.08 | (1.52 to 2.83) | 2.03 | (1.44 to 2.89) |
| Violent crimes, 500 m radius (v fourth quartile)† | ||||
| Second quartile | 1.40 | (1.04 to 1.90) | – | |
| Third quartile | 1.45 | (1.05 to 2.03) | – | |
| Fourth quartile | 1.52 | (1.07 to 2.17) | – | |
| Random effects | ||||
| σs2 (neighbourhood variance) | 0.11 | (0.05 to 0.24) | 0.06 | (<0.01 to 0.19) |
| φ (rate of correlation decay) | 0.0024 | (0.0005 to 0.0083) | 0.0061 | (0.0023 to 0.0089) |
| 3/φ (range of spatial correlation in metres) | 1266 | (360 to 5149) | 488 | (339 to 1308) |
| DIC‡ | 7132 | 3635 | ||
*In the full models, the different contextual effects were adjusted for each other and for individual covariates (age, gender, marital status, educational level, and income). †The crime variable varied in space on a larger scale than contextual deprivation: separate ecological geostatistical models with (a) the contextual income factor, and (b) the violent crime variable as outcomes (at the administrative neighbourhood level) showed that the spatial range of correlation was 3613 metres and 6036 metres, respectively, for these two contextual factors. ‡The deviance information criterion (DIC) can be used to compare different models estimated for a similar mental health outcome, but not to compare models for different outcomes. The lower the DIC, the better the fit of the model.