Section of General Internal Medicine, Minneapolis VA Health Care System, One Veterans Drive (111-0), Minneapolis, MN, 55417, USA.
Center for Care Delivery and Outcomes Research, Minneapolis VA Health Care System, One Veterans Drive (152), Minneapolis, MN, 55417, USA.
BMC Med Res Methodol. 2022 Dec 1;22(1):308. doi: 10.1186/s12874-022-01783-7.
Altering cover letter information to reduce non-response bias in trauma research could inadvertently leave survey participants unprepared for potentially upsetting questions. In an unsolicited, mailed survey, we assessed participants' change in affect post-survey after altering key cover letter information and promising different incentives. We tested direct and indirect effects of participants carefully reading the cover letter on changes in their affect post-survey.
In a 3X2X2 randomized, factorial trial, 480 male and 480 female, nationally representative Veterans who were applying for posttraumatic stress disorder disability benefits were randomized to receive one of 12 different cover letters. The cover letters provided general versus more explicit information about the survey's trauma content and how their names were selected for study; we also promised different incentives for returning the survey. The main outcome was change in affect post-survey. We examined five potential moderators: combat or military sexual trauma exposure, posttraumatic stress disorder or serious mental illness diagnosis, and recency of military service. Mediators between reading the cover letter carefully and post-survey affect included how participants rated the cover letters' information and whether they thought the cover letters prepared them for the survey's content. A Bonferroni corrected alpha of 0.003 was the threshold for statistical significance.
One hundred ninety men and 193 women reported their pre-and post-survey affect. Across all study conditions, out of 16 possible points, the net change in affect post-survey was less than a quarter-point for men and women. Mean changes in post-survey affect did not differ statistically significantly across any of the study factors (ps > 0.06); nor were there statistically significant interactions between any of the study factors and the 5 moderators after accounting for multiple comparisons (ps > 0.02). After controlling for pre-survey affect, reading the cover letter carefully had small effects on changes in post-survey affect, with larger associations seen in the women compared to men. Mediators' effects were often in opposite directions for men and women.
General descriptions of a survey's trauma content appear ethically defensible. Research on cover letters' impacts on survey participants' emotional reactions and how those impacts differ by gender is needed.
在创伤研究中改变求职信信息以减少无应答偏倚，可能会无意中使调查参与者对可能令人不安的问题毫无准备。在一项未经请求的邮寄调查中，我们评估了参与者在改变关键求职信信息并承诺不同激励措施后调查后的情绪变化。我们测试了参与者仔细阅读求职信对他们调查后情绪变化的直接和间接影响。
在一项 3X2X2 随机、析因试验中，我们随机分配了 480 名男性和 480 名女性退伍军人，他们正在申请创伤后应激障碍残疾津贴，以接受 12 种不同求职信中的一种。求职信提供了关于调查创伤内容的一般信息和更明确的信息，以及他们的名字如何被选中进行研究；我们还承诺为返回调查提供不同的激励措施。主要结果是调查后的情绪变化。我们研究了五个潜在的调节因素：战斗或军事性创伤暴露、创伤后应激障碍或严重精神疾病诊断，以及最近的兵役。仔细阅读求职信和调查后情绪之间的中介因素包括参与者如何评价求职信的信息以及他们是否认为求职信为调查内容做好了准备。Bonferroni 校正后的 0.003 为统计学意义的阈值。
190 名男性和 193 名女性报告了他们的调查前后情绪。在所有研究条件下，男性和女性的调查后情绪变化不到四分之一点。在任何研究因素（p>0.06）下，调查后情绪变化的平均差异在统计学上均无显著差异；在考虑到多次比较后，任何研究因素与 5 个调节因素之间也没有统计学显著的相互作用（p>0.02）。在控制了调查前的情绪后，仔细阅读求职信对调查后情绪变化有较小的影响，女性的影响大于男性。男性和女性的中介因素的影响往往相反。
对调查创伤内容的一般描述在伦理上是合理的。需要研究求职信对调查参与者情绪反应的影响以及这些影响如何因性别而异。
