Questionnaire development and sampling frame

There were a number of issues that needed to be considered in adopting this approach.

These included defining the purpose of the questionnaire, developing a valid and reliable

questionnaire instrument for assessing attitudes, refining, piloting and testing the

questionnaire, gaining a valid and representative sample, ensuring adequate response

rates and assessing non-responders.

The purpose of the questionnaire was to assess patient awareness of pneumococcal

vaccine, to ascertain whether patients were able to identify whether they fell into a risk

group and which risk groups these were, to gauge attitudes to vaccination in general and

pneumococcal vaccination amongst those patients already attending for influenza

vaccination.

A self-administered questionnaire was used as the instrument for measuring attitudes and

gathering other data for this study. A suitable questionnaire was devised after searching

the literature on questionnaire design and pneumococcal vaccination. The questionnaire

was modified after a small pilot study to a sample of twelve patients. There was

accompanying explanation included at the beginning outlining the purpose of the

questionnaire with information at the beginning on pneumococcal vaccination and at the

end for patients who wished to have the vaccination (Appendix 1).

The questionnaire was devised using published guidelines and considering the various

stages of questionnaire design including data content, question selection and wording,

coding, presentation and layout (Stone 1993; Lydeard 1991). It included an initial section

outlining the reasons and indications for pneumococcal vaccination and possible side

effects. Information about the benefits and side effects of pneumococcal vaccine was

included so that the questionnaire itself served as a method of raising awareness in the

target group. Patients were asked to complete questions on general attributes such as age

and sex as well as beliefs such as whether they thought they were in a risk group, prior

knowledge of pneumococcal vaccine and the source of this information.

The second part of the questionnaire consisted of a series of statements reflecting

attitudinal statements were derived from five dimensions or areas of concern identified

from the literature review. The attitude statements, which included underlying beliefs

(cognition), feelings (emotion) and what patients reported that they would do as a result

of their beliefs and feelings, sometimes termed resultant behaviour or action tendency,

were developed using recognised techniques (Oppenheim 1966). These techniques

included assembling a pool of relevant items, constructing a questionnaire instrument

from these, applying this to an appropriate sample, developing the instrument based on

the response to questions and item analysis, assessing reliability and validity and

modifying the instrument on the basis of this if necessary (Proctor 1993).

Because respondents were more likely to reply in the affirmative leading to acquiescent

bias (Martin 1964), paired statements were employed expressing opposite attitudes, i.e. a

'balanced' questionnaire. Some of the statements were simply reversed in wording to

produce their negative counterpart whereas others used recognised negative concepts

identified from the preliminary exploration (Figure 2). This technique had been used

previously in general practice (Pringle et al. 1984) to counter bias due to ‘response

acquiescence’. The five pairs of statements were ordered to avoid paired questions

occurring together giving an item pool of ten questions in all. A Likert-type (Likert 1932)

format with five response codes numbered one to five, ranging from “strongly agree” to

“strongly disagree” with a central “not sure” numbered three was used for each statement.

Care was taken to avoid ambiguous statements, double questions and loaded questions

and the questionnaire was developed with awareness of the various types of respondent

tendency), skewed responses and social desirability bias (Lydeard 1991) whilst

acknowledging that some of these could only be detected at the analysis stage. A suitable

coding frame was developed for the questionnaire responses so that the data could easily

be entered onto a database for analysis.

The sample that was selected was patients attending, or being visited for, influenza

vaccination. This included vaccinations administered by practice and district nurses or

doctors during a single influenza vaccination season. This set of patients was chosen

because of the overlap between risk groups for the two vaccines, because those attending

for influenza vaccine would be generally positive to vaccination and because it was felt

that targeting them would lead to the greatest rise in coverage for pneumococcal vaccine.

An alternative would have been to select a random sample from all patients in the

practice or those above a certain age or those in particular risk categories. An advantage

of this approach would have been that it targeted a wider, more representative group of

patients who might have been eligible for influenza and pneumococcal vaccination. The

disadvantages would have included the additional costs of postage and the risk of a low

response rate that often occurs with postal questionnaires.

Nursing and residential homes or similar institutions were excluded because patients

within these establishments did not fall within the recommendations for pneumococcal

vaccination per se but also because it was known that a proportion of nursing home

patients were unable to complete the questionnaire because of illness or disability, for

Although this sampling strategy might have led to selection bias, i.e. excluding patients

who did not attend for influenza vaccine for whatever reason, this was felt to be a

legitimate pragmatic approach for the purposes of this study as it was the patients who

were sampled who would subsequently be targeted for pneumococcal vaccination.

A small initial sample of twelve patients attending the practice diabetic clinic was chosen

to pilot the questionnaire. Patients attending this clinic were handed the questionnaire

before their appointment and had an opportunity to complete and return the questionnaire

before leaving. This allowed the investigator to check whether questions were understood

and that the format, layout, wording and responses were adequately accounted for by

directly checking with each patient.

After minor modifications the questionnaire was distributed to patients attending or being

visited for influenza immunisation by receptionists and district nurses. The questionnaire

was not strictly anonymised. Patients who were interested in receiving pneumococcal

vaccination in addition to influenza vaccination, after receiving the patient information,

were given an opportunity to record their name on the questionnaire so that they could be

contacted at a later date when the vaccination programme had begun. Those patients who

did not wish to have the additional vaccine or preferred not to give their name could

return the form anonymously.