Edwards Phil, Roberts Ian, Sandercock Peter, Frost Chris
CRASH Trial Co-ordinating Centre, Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, 49-51 Bedford Square London, WC1B 3DP, UK.
Control Clin Trials. 2004 Feb;25(1):31-52. doi: 10.1016/j.cct.2003.08.013.
In large clinical trials where outcome assessment is possible using questionnaires, it may be more cost-effective to mail them to patients than to conduct interviews in-person. However, nonresponse to mailed questionnaires reduces the effective sample size and can introduce bias. We conducted a systematic review and meta-analysis of randomized controlled trials evaluating the effect of questionnaire length on response rates. We searched 14 electronic bibliographic databases, the reference lists of relevant trials, and we contacted the authors of eligible trials to ask about unpublished data. For each trial identified, we used logistic regression to estimate the odds ratio for response per one page increase in the number of pages included in the questionnaire. We pooled the regression coefficients in a random effects meta-analysis. Heterogeneity among the coefficients was assessed using a chi-square test at a 5% significance level. We specified a priori that the reduction in the odds of response per one page increase would be greatest among trials comparing relatively short questionnaires. We used meta regression to examine the relationships between the regression coefficients, the length of the questionnaires used in each trial, and other study characteristics. A total of 38 randomized controlled trials were identified where participants were allocated to questionnaires of differing lengths and where the number of pages used was known. There was significant heterogeneity between the regression coefficients estimated from each trial. In meta regression, most of the heterogeneity was explained by variation in the length of the questionnaires used in each trial. Among trials in which the shortest questionnaire was a postcard, the odds of response were more than halved for each additional page used (0.39; 95% CI 0.34 to 0.45). In the remaining trials, pooled effect sizes were much smaller. In trials of one page compared with either two or three pages, the odds of response per one page increase was 1.01 (95% CI 0.82 to 1.24). For one page compared with four or more pages, and for two or more pages compared with longer alternatives, the odds ratios per one page increase were 0.90 (95% CI 0.83 to 0.98) and 0.98 (95% CI 0.96 to 0.99), respectively. There were no statistically significant associations between trial results and other study characteristics. It appears that response can be increased by using a shorter questionnaire. Moderate changes to the length of shorter questionnaires will be more effective than moderate changes to the length of longer questionnaires. If a choice of follow-up questionnaire exists for a clinical trial, the shorter one should be used. If a new follow-up questionnaire is to be designed, it should be made as short as possible without compromising the data collection requirements of the trial.
在可使用问卷进行结果评估的大型临床试验中，将问卷邮寄给患者可能比亲自进行访谈更具成本效益。然而，问卷未得到回复会减少有效样本量，并可能引入偏差。我们对评估问卷长度对回复率影响的随机对照试验进行了系统评价和荟萃分析。我们检索了14个电子文献数据库、相关试验的参考文献列表，并联系了符合条件试验的作者询问未发表的数据。对于每项纳入的试验，我们使用逻辑回归来估计问卷每增加一页回复的比值比。我们在随机效应荟萃分析中汇总回归系数。使用卡方检验在5%显著性水平评估系数之间的异质性。我们预先设定，在比较相对较短问卷的试验中，问卷每增加一页回复几率的降低幅度最大。我们使用荟萃回归来检验回归系数、每项试验中使用的问卷长度以及其他研究特征之间的关系。共确定了38项随机对照试验，其中参与者被分配到不同长度的问卷，且使用的页数已知。每项试验估计的回归系数之间存在显著异质性。在荟萃回归中，大部分异质性可由每项试验中使用的问卷长度差异解释。在最短问卷为明信片的试验中，每增加一页使用，回复几率减半以上（0.39；95%可信区间0.34至0.45）。在其余试验中，汇总效应量要小得多。在一页问卷与两页或三页问卷的比较试验中，问卷每增加一页回复几率为1.01（95%可信区间0.82至1.24）。对于一页问卷与四页或更多页问卷的比较，以及两页或更多页问卷与更长问卷的比较，问卷每增加一页的比值比分别为0.90（95%可信区间0.83至0.98）和0.98（95%可信区间0.96至0.99）。试验结果与其他研究特征之间无统计学显著关联。似乎使用较短问卷可提高回复率。对较短问卷长度的适度改变比对较长问卷长度的适度改变更有效。如果临床试验有后续问卷可供选择，应使用较短的问卷。如果要设计新的后续问卷，应在不影响试验数据收集要求的前提下尽可能缩短。
