Abstract: The purpose of this research is to investigate the effects and influential factors of computerized psychological treatments for depression. This research adopted the methods of meta-analysis and meta-regression analysis to carry out the literature search through four major data-bases as PubMed, PsycINFO, Embase and Web of Science, 50 literatures were included in the meta-analysis eventually, including 42 randomized controlled trials and 67 samples for effective meta-analysis. The sample for the research is 7920 participants, 4208 participants for computerized intervention group and 3712 participants for the control group, respectively. The results suggest that: (1) the overall effects of computerized psychological treatments for depression is 0.53, which is a medium effectiveness, the effectiveness for the subjects with 3 months tracking is 0.14, while the one with 6 months tracking is 0.16. (2) Significant differences are revealed in the subgroups analysis for the age group, depression severity, support method and measurement scales: With a more close look at the subjects, we can find out that the effectiveness for the youth group is relatively small as 0.24, while the effectiveness for adult group is relatively great as 0.59; the treatment effectiveness for the severe depression group is significantly higher than non-severe group (severe depression: d=0.73; non-severe depression: d=0.48, p=0.003). Significant differences are revealed in the treatment effectiveness for the 4 patterns of support (Face to face support group: d=0.58; Email support: d=0.70; Phone support: d=0.46; no support, d=0.40, p=0.038); Significant differences of the effectiveness are also revealed in the treatment effectiveness for the two different kinds of treatments in the measurement scales (BDI: d=0.63; CES-D: d=0.36, p=0.008); the differences are not significant in the analysis of the 3 subgroups as intervention pattern (network: d=0.54; stand-alone PC: d=0.38, p=0.13), intervention orientation (CBT: d=0.52; PST: d=0.48, p=0.55) and analytical methods (full treatment: d=0.53; intention for treatment: d=0.53, p=0.97). (3) In the total sample, the year of publication significantly affects the effectiveness of treatment (B=-0.0273, p<0.001); drop rate does not significantly affect the effectiveness of treatment for the total sample (B=0.0028, p=0.12). However, with a more complete categorization of the different samples in the measurement, we find out that drop rate significantly affects the treatment effectiveness for the CES-D samples (B=0.0106, p <0.001); intervention unit number significantly affects the effectiveness of treatment for the CES-D sample (B=0.0689, p=0.01), nevertheless, it plays no significant affects for the total samples(B=0.0016, p= 0.86) and BDI samples (B=-0.0313, p=0.08). (4) Publication bias may be seen in this study, but it is unreasonable to overthrow the existing conclusions. Funnel plot suggests that the publication bias may exist, the result from the Begg & Mazumdar’s rank correlation test shows that z=2.52, p<0.01; Egger's regression is significant (t=5.27, df=65, p<0.001), therefore, the two statistical results are significant. But the fail-safe number in the study is 7488. It means that 7488 contrary studies are required to be conducted to overthrow the existing meta-analysis conclusions. Conclusion: The computerized psychological treatments for depression manifests medium effectiveness; the age group, severity of depression, support method, measurement scale and the year of publication are of regulatory role for the overall effective of treatment. Therefore, the research on computerized psychological treatments for depression in the future should pay attention to these regulatory variables for their effectiveness to treatments.

Key words: depression, computerized psychological treatments, effectiveness, meta-analysis, meta-regression analysis, moderating effect

摘要： 运用元分析和元回归分析的方法考察抑郁症计算机化治疗的效果及其影响因素。来源于50篇文献，42项RCT研究的67个样本满足了元分析标准（N=7920）。结果发现：（1）整体效果量为0.53，三个月追踪效果量为0.14；6个月追踪效果量为0.16；（2）在年龄段、抑郁严重程度、支持方式和测量量表四个亚组分析中，其效果量存在显著差异。干预方式、干预取向和分析方法对效果量的影响不显著；（3）出版年份显著影响治疗效果量，脱落率和干预单元数对整体效果量的影响不显著。结果表明：抑郁症的计算机化治疗具有中等的效果量；年龄段、抑郁严重程度、支持方式、测量量表和出版年份对其整体效果量有调节作用。将来抑郁症计算机化治疗的研究应重视上述调节变量对治疗效果的影响。

关键词: 抑郁症, 计算机化治疗, 效果, 元分析, 元回归分析, 调节效应