RxPG - How to Design Questionnaire for Use in Epidemiologial Research and Surveys

Questionnaire are used in all forms of epidemiological studies to measure exposure both to possible causal agents ( eg. exposure to alcohol) and to confounding variables ( e.g. gender). The questionnaire may also be used to measure outcomes in cross-sectional studies ( e.g. disease status). Some technical aspects are of immense importance for designing of the useful questionnaires. The following principles need to be considered in designing questionnaire.

1. Content: The questionnaire should ask the minimum amount about a subject’s experiences that will provide sufficient information to investigate the research question; as such it should be as brief as possible with the question justified in terms of the objectives of the study.

2. Types of question:

a. Open: These types of questions allow the respondents to answer in their own terms.

b. Closed: These questions specify the details of possible answers. Epidemiological questionnaires should include majority of the closed questions to reduce the possibility of interview, response and interpretation biases and to facilitate processing. The following points should be remembered.

i. Open questions should be used for non-categorical data ( e.g. age, date of birth).

ii. In close questions in which only one out of a lot of responses is allowed, the final option should always be ‘Other, please specify’, unless the investigator believes that all possible options are mentioned.

iii. In closed questions, when multiple responses are allowed, each response should be coded as a separate variable.

iv. While using mainly closed questions, respondents should be invited to make comments either in writing or to the interviewer.

3. Wording of questions: Wording should avoid any suggestion that a particular answer is preferred, e.g. ‘leading questions’ should be avoided. Each question should contain only one idea. Questions must be in simple, non technical language avoiding the use of jargons.

4. Question sequence: Questions should follow a logical sequence, as such the sequence that the respondents might expect to follow. On each subject, questions should proceed from the general to the specific. When a response to a general question makes further responses on that subject unnecessary, ( e.g. a life-time non smoker need not answer questions about present and past smoking) a branching of the question sequence may be introduced.

5. Questionnaire layout: This is of prime importance for both interviewer and self-administered questionnaires to arouse and maintain interest and for correct completion. An introduction, explanatory notes and instructions and examples of how to response to a particular question should be in place. They should occupy pages which are separate from the first question, except where instructions and examples may immediately precede a new type of question.

6. Method of administration: Questionnaires can be self- administered or interviewer administered. The advantages of self administered questionnaires are its lower cost and absence of interviewer bias while disadvantages include partial completion, unresolved or unidentified problems of question interpretation and the need for literacy skills of the respondents. Interviewer administered questionnaire have the advantages of ensuring completion, clarification of meaning (though this may introduce bias !) and the support which facilitates cooperation. The main disadvantages are in terms of cost, time and interviewer bias.

7. Coding: It is important at the outset to decide on the coding scheme to be used for each variable. Most responses are best coded and entered into the computer as numbers rather than letters, with the exception of variables such as name or place of birth. A coding guide should be constructed which provides key to the meaning of the numerical responses ( e.g. 1= male, 2=female), and this can be used to clarify output in analysis programmes such as SPSS ( Statistical Package for the Social Science), which allows for the specification of ‘value labels’.