Olsen J, Sabroe S
Scand J Soc Med. 1984;12(1):55-63. doi: 10.1177/140349488401200109.
Summing up information from interview or questionnaire data into a single number, an index, is not uncommon in social medical research, but the conceptual and mathematical background for this is too often neglected. The so-called "Rasch models" specify circumstances where a single index can be used. The necessary computer programs for doing this are now generally available (10, 11). Two often used questionnaires concerning neurotoxicity due to solvent exposure are tested using a Rasch model. None of the questionnaires were found to be in agreement with the necessary requirements for the model and no single number describes the respondent's degree of neuropsychiatric dysfunction. Much more work is needed before a suitable screening questionnaire is available.
在社会医学研究中,将访谈或问卷调查数据汇总成一个单一数字(即一个指数)的做法并不罕见,但这种做法的概念和数学背景却常常被忽视。所谓的“拉施模型”规定了可以使用单一指数的情况。现在通常可以获得用于此目的的必要计算机程序(参考文献10、11)。使用拉施模型对两份常用的关于溶剂暴露所致神经毒性的问卷进行了测试。结果发现,没有一份问卷符合该模型的必要要求,也没有一个单一数字能够描述受访者的神经精神功能障碍程度。在获得一份合适的筛查问卷之前,还需要做更多的工作。