Table 6.
Factors associated with the heterogeneity of prevalence estimates of alcohol dependence among adults in China using backward stepwise meta‐regression for seven outcomes reported in 10 or more studies from 1987 to 2013.a
| Outcome | Covariate | Bivariate β | 90% CI | Stepwise β | 90% CI | R2 (%) b |
|---|---|---|---|---|---|---|
| Current alcohol dependence (%) | Sample size | |||||
| 2nd quartile | –7.0 | –11.3, –2.8 | –7.0 | –11.3, –2.8 | 43.0 | |
| 3rd quartile | –6.8 | –10.7, –3.0 | –6.8 | –10.7, –3.0 | ||
| 4th quartile | –5.5 | –10.5, –0.4 | –5.5 | –10.5, –0.4 | ||
| Provides reasons for non‐participation | ||||||
| Yes | 5.0 | 1.7, 8.2 | ||||
| Used validated tools | ||||||
| Yes | –5.7 | –9.3, –2.1 | ||||
| Male current alcohol dependence (%) | Provides reasons for non‐participation | |||||
| Yes | 8.9 | 2.5, 15.3 | 5.7 | 4.1, 7.4 | 97.0 | |
| Sample size | ||||||
| 2nd quartile | –22.7 | –30.3, –15.1 | –28.4 | –24.6, –13.6 | ||
| 3rd quartile | –24.9 | –32.6, –17.2 | –28.0 | –25.9, –15.0 | ||
| 4th quartile | –20.1 | –29.2, –11.0 | –28.6 | –26.1, –14.9 | ||
| Used validated tools | ||||||
| Yes | –23.4 | –30.7, –16.1 | ||||
| Life‐time alcohol dependence (%) | Sample size | |||||
| 2nd quartile | –2.3 | –5.9, 1.2 | –2.3 | –5.9, 1.2 | 16.9 | |
| 3rd quartile | –4.9 | –8.5, –1.4 | –4.9 | –8.5, –1.4 | ||
| 4th quartile | –5.1 | –8.6, –1.5 | –5.1 | –8.6, –1.5 | ||
| Use of informants | ||||||
| Yes | –3.6 | –6.8, –0.4 | ||||
| Region of China | ||||||
| East | –3.6 | –6.8, –0.3 | ||||
| Male life‐time alcohol dependence (%) | Published in English | |||||
| Yes | 6.1 | 0.2, 12.0 | 6.7 | 1.2, 12.2 | 25.3 | |
| Sample size | ||||||
| 2nd quartile | –2.1 | –7.6, 3.5 | –4.3 | –9.9, 1.2 | ||
| 3rd quartile | –9.6 | –15.9, –3.3 | –7.4 | –14.1, –0.6 | ||
| 4th quartile | –8.1 | –14.4, –1.8 | –8.7 | –14.6, –2.8 | ||
| Female life‐time alcohol dependence (%) | Use of substitute respondents | |||||
| Yes | 0.2 | <0.1, 0.4 | 0.2 | <0.1, 0.4 | 79.9 | |
| Urban life‐time alcohol dependence (%) | Sample size | |||||
| 2nd quartile | –1.9 | –4.9, 1.1 | –1.9 | –4.9, 1.1 | 41.5 | |
| 3rd quartile | –4.8 | –7.5, –2.0 | –4.8 | –7.5, –2.0 | ||
| 4th quartile | –4.4 | –7.3 –1.6 | –4.4 | –7.3 –1.6 | ||
| Use of informants | ||||||
| Yes | –2.1 | –4.4, –0.1 | ||||
| Rural life‐time alcohol dependence (%) | Provides reasons for non‐participation | |||||
| Yes | 8.4 | 3.6, 13.2 | 8.4 | 3.6, 13.2 | 38.5 | |
| Sample size | ||||||
| 2nd quartile | –3.7 | –9.2, 1.8 | ||||
| 3rd quartile | –6.1 | –10.9, –1.2 | ||||
| 4th quartile | –5.8 | –10.9, –0.7 | ||||
| Representativeness of the sample | ||||||
| Yes | 5.5 | 0.9, 10.0 | ||||
Bivariate meta‐regression was conducted for each of the above seven outcome for all 21 covariates [i.e. quality of the study, language of publication, survey date (three categories), region (west versus east), sample size (four categories) and 16 quality assessment items]. The backward stepwise meta‐regression models started with the study characteristics that were associated significantly with heterogeneity in the bivariate analyses listed in the table. Significance level for retention in the final model set at P = 0.10. See Methods for details.
R 2 represents the proportion of between‐study variations associated with the covariates in the final model. CI = confidence interval.