Adamson Simon J, Sellman John Douglas, Frampton Chris M A
National Addiction Centre (Aotearoa New Zealand), University of Otago, Christchurch, New Zealand.
J Subst Abuse Treat. 2009 Jan;36(1):75-86. doi: 10.1016/j.jsat.2008.05.007. Epub 2008 Jul 26.
Patient characteristics as predictors of alcohol use disorder treatment outcome were examined on three levels, identifying whether or not variables were significant predictors of drinking-related outcome in univariate analysis, in multivariate analysis, and in multivariate analyses limited to studies including several "key predictors." Also, a model was developed to predict total percentage of variance in treatment outcome accounted for in each study using each of the key predictors and a range of methodological factors. The most consistent univariate predictors were baseline alcohol consumption, dependence severity, employment, gender, psychopathology rating, treatment history, neuropsychological functioning, alcohol-related self-efficacy, motivation, socioeconomic status/income, treatment goal, and religion. When these key predictors were combined into multivariate analyses, baseline alcohol consumption and gender showed substantial reductions in predictive consistency whereas the remaining variables were not greatly affected. The most consistent predictors overall were dependence severity, psychopathology ratings, alcohol-related self-efficacy, motivation, and treatment goal. The two predictor variables most associated with greater variance accounted for in predictive models, when controlling for broader methodological variables, were baseline alcohol consumption and dependence severity. Few predictor variables were examined in more than a third of studies reviewed, and few variables were found to be significant predictors in a clear majority of studies. However, a subset of variables was identified, which collectively could be considered to represent a consistent set of predictors. Too few studies controlled for other important predictor variables. Attempts to synthesize findings were often hampered by lack of agreement of the best measure for predictor variables.
在三个层面上考察了作为酒精使用障碍治疗结果预测因素的患者特征，确定变量在单变量分析、多变量分析以及仅限于包含若干“关键预测因素”的研究的多变量分析中是否为饮酒相关结果的显著预测因素。此外，还开发了一个模型，以预测每项研究中使用每个关键预测因素和一系列方法学因素所解释的治疗结果方差的总百分比。最一致的单变量预测因素是基线酒精消费量、依赖严重程度、就业情况、性别、精神病理学评分、治疗史、神经心理功能、与酒精相关的自我效能感、动机、社会经济地位/收入、治疗目标和宗教信仰。当将这些关键预测因素纳入多变量分析时，基线酒精消费量和性别在预测一致性方面大幅下降，而其余变量受影响不大。总体而言，最一致的预测因素是依赖严重程度、精神病理学评分、与酒精相关的自我效能感、动机和治疗目标。在控制更广泛的方法学变量时，与预测模型中解释的更大方差最相关的两个预测变量是基线酒精消费量和依赖严重程度。在所审查的研究中，超过三分之一的研究考察的预测变量很少，而且在绝大多数研究中很少发现变量是显著预测因素。然而，确定了一组变量，总体上可被视为代表一组一致的预测因素。控制其他重要预测变量的研究太少。由于缺乏对预测变量最佳测量方法的一致性，往往妨碍了对研究结果的综合。
