Faculty of Health and Medical Sciences

Questionnaire design

 

Many types of research involve the design of a specific set of experiments and the use of specific techniques to answer a specific scientific question.

Some projects will involve the use of questionnaires to collect qualitative data in a way that can be analysed scientifically.

Questionnaire design can be tricky, posing focussed questions that clearly reveal the answers to test the strength of a hypothesis. This section considers questionnaires specifically as a guide to the factors that need to be considered to devise an efficient and accurate set of questions to prove your hypothesis.

All work should be done in consultation with your supervisor – this material is a useful guide only.

  1. How to collect your data
  2. Pre-testing, typing and printing
  3. How to conduct a postal survey 
  4. How to code your data for the computer 
  5. Study numbers
  6. Analysing your data

How to collect your data

Data for projects in IMED450 may be collected in a variety of ways, such as analyses of experimental results, reviews of medical records, interviews of patients or other members of the public, or by a self-administered questionnaire.

For most of these latter methods, data need to be collected in such a way that they can be grouped and the responses of different subjects compared. The first step is to design a data collection form (for record review), interview schedule, or questionnaire.

These have basically the same format except that in an interview schedule or questionnaire the questions are set out in more detail. For simplicity, these different types of instruments will all be referred to as ‘questionnaires’. There should be one questionnaire for each patient or respondent. Note that it often takes several drafts before your questionnaire is ready to pilot. Do not become disheartened if this happens - your data and ultimately your final report will be of higher quality.

It is useful to have your supervisor comment on each draft of your questionnaire.

Developing your questionnaire

List the variables

First list the variables that you want to measure, such as:

  • attitudes to abortion
  • history of smoking habits
  • immunisation status.

Then decide which demographic variables are relevant to your research questions, such as:

  • age
  • sex
  • marital status
  • country of birth
  • occupation.
Decide on the questions

You need to formulate the questions to measure your listed variables. Each question should be accompanied by an instruction on how to complete it. Numbered responses are provided on the survey form and the participant is asked to circle the appropriate number(s). We do not usually ask people to tick alternatives because some people use a cross instead and this becomes ambiguous.

Tick boxes are not acceptable as ticks cannot be coded and entered into the computer.

Examples of various sorts of questions are included in the Appendix but, in general, there are two main types:

  1. Open-ended

    A space is left for the respondent to answer in his/her own words. This is quite satisfactory with simple questions such as “What is your age?” “What is your occupation?” With less straightforward questions such as “Why did you give up smoking?” the answers may be difficult to categorise and may miss the point of the question.

  2. Closed (fixed alternative)

    The respondent is asked to choose from a number of fixed alternatives. These may include two possible responses:

    for example,
    Has your child been immunised against measles? (Please circle ONE number).

    1. Yes
    2. No

    or several alternatives:

    for example,
    What is your marital status?(Please circle ONE number)

    1. Never married
    2. Married
    3. Divorced or separated
    4. Widowed

    Alternatives may be arranged on a scale, for example
    in measuring attitudes:

    1. Strongly agree
    2. Agree
    3. Disagree
    4. Strongly disagree

    In all cases of closed questions the alternatives provided must be exhaustive, that is they should include all the possible responses that might be expected. It is often useful to add ‘other (specify)’ at the end of a list of alternatives.

    Secondly, if you are asking the respondent to select one answer only, the alternatives must be mutually exclusive.

    There may be situations where you ask respondents to select several alternatives from a list, such as:

    What were your main sources of information about breast self-examination? (Please circle ONE or MORE numbers)

    1. Newspapers and magazines
    2. TV/Radio
    3. General Practitioner
    4. Friends/relatives
    5. Other  (Please specify).....................................................

It is important to be clear whether you expect one answer or several answers, as these are coded differently for the computer (see later).

For most purposes, questionnaires should have mainly closed questions, with a few open questions.

Requirements for good questions

In formulating questions, ask yourself the following:

  • Are the items clear and unambiguous? (Make sure there are no “double-barrelled” questions in which two concepts are linked by “and” or “or”).
  • Are they relevant?
  • Are they “unbiased”? (Avoid loaded questions which appear to expect a particular answer).
  • Are the questions phrased in language which the respondents will understand?
Question sequence and instructions

The questions should follow a logical sequence. Demographic questions are usually placed at the end of the questionnaire because the accompanying letter leads the respondent to expect questions on a certain topic. However, if the main subject matter of the questionnaire is sensitive, the demographic questions may provide a non-threatening “lead-in”.

There should be a brief covering letter explaining the purpose of the study, requesting co-operation from the respondent, and assuring confidentiality of the results - this is generally placed on a separate front sheet with Department of Public Health letterhead. Specific instructions about the method of answering the question (for example “please circle the appropriate number(s)”) are put with each question.

It is useful to group questions into logical sections, each preceded by a short introduction, such as “Now we would like to ask you about your experience after you left the hospital.” This helps to orient the respondent to a new area of questioning and helps make sense of the questionnaire.

Back to top

Pre-testing, typing and printing

After your questionnaire has been typed you should obtain about 10 copies for pre-testing on a sample of subjects similar to those to be studied.

Next, discuss responses to your pilot study with your Supervisor and modify as required.

Back to top

How to conduct a postal survey

Outward envelopes

Write the name and address neatly on the outward envelope or type labels. Remember, this is the first impression that the respondent receives of your survey. Please put postcodes on all envelopes.

For all except very sensitive topics you should keep a record of the replies for follow-up, so put the individual’s study number in the appropriate place on the questionnaire (or on the reply-paid envelope).  Keep a list of the names and the corresponding numbers and check them off as they are returned.  Be aware of confidentiality.

Reply-paid envelopes

In some cases your supervisor may suggest reply-paid envelopes. Write your project number and supervisor’s surname on every envelope.

Follow-up

After the first posting, you may send a follow-up letter with another copy of the questionnaire to those who have not responded (see your supervisor for a suggested format for this letter).

Back to top

How to code your data for the computer

If you have more than 20 or 30 cases, it is simpler and quicker to use the computer for analysis. This means that your data sheet or questionnaire should be coded in such a way that the data can be entered directly into electronic files. See your supervisor about this to make sure you are on the right track.

Data are usually entered as numbers. These may be numerical codes allotted to categories of the variables, for example, 1.  Male  2.  Female, or the actual numerical values, such as (Age) 65. Only one digit is entered in each column of the electronic file. If a number contains more than one digit, each digit is entered in a separate column. The column or set of adjacent columns is called a “field.”  If the field is, say, two columns wide (for example if numbers run over 10) then single numbers must be preceded by zero to keep the digits in their correct columns - 01, 02, and so on. On the questionnaire, the columns in the computer file are represented by a series of boxes in the right hand margin of the page. Each box represents one column in the computer file. Each field (or set of boxes) must be large enough to accommodate the maximum number of digits in the response. For example, birthweight in grams is in thousands and would occupy a field four (4) columns wide, so four boxes should be allowed, as in birthweight 2486 grams  2 4 8 6

Be aware that it is not permissible to add in extra numbers outside the boxes if you find later that you did not allow enough boxes. This will cause complete confusion for the data entry operator and will involve you in a great deal of extra work.

Each set of boxes is identified at the right hand side by a number. This is the column number corresponding to the last box in each set.

Back to top

Study numbers

Every questionnaire must be given a unique number, called the study number or survey number that corresponds to a particular individual respondent. The study number usually appears at the top of the first page of your questionnaire and is very helpful in the analysis. 

If errors or peculiar codes are found in the computer printout, it is possible to relate them to a particular study number. This avoids having to search all your questionnaires trying to find the mistake. Remember to allow the right number of boxes, for example, for between 100 and 1000 questionnaires your questionnaires would be numbered 001 to 999, and therefore you would need three boxes for the study number.

You should have a master list on which the names and study numbers of your potential respondents are listed. This avoids having to put names on the questionnaires and helps maintain confidentiality.

Coding of closed (fixed alternative) questions

  1. Mutually exclusive items

    for example, Sex: 1. Male, 2. Female

    The respondent is asked to circle the relevant number. There is a single digit in the response and so one box (column) is allowed. The coder merely has to write the circled number in the box provided.

    (Make corrections with liquid paper. The numbers must be clear for the keypunch operator).

  2. Multiple response items

    There may be situations where more than one response is appropriate:

    for example, Have you had any of the following conditions? (list given)
    (Please circle the appropriate number(s))

    In this case you will need to allow one box for each of the alternatives, and put the circled number in the relevant box.  In this example putting the number in the box means “YES” for that particular item and leaving it blank means “NO” (see Question 2 in the attached example questionnaire). It is possible (Comment: missing words)

Back to top

Analysing your data

Your supervisor is your first point of contact regarding the analysis of your data. Often the analysis is specific to the project and may involve specialised statistical tests or programs.

Back to top