9  Limitations

There were some limitations to this current survey and these are summarised below.  

The sample was obtained from three sources: two official sources – hospitals or medical institutions and private clinicians - and the third source, support groups.

The three sample groups were tightly controlled in order to ensure that each formed a similar proportion of the sample. Thus where deficiencies of recruiting points developed in one source, attempts were made to redistribute the questionnaires across others within that segment rather than shifting them to another source.  

Official sources (hospitals or medical institutions)

Because of confidentiality issues, a database of allergy sufferers from which to recruit participants could not be compiled. The researchers therefore divided the sampling across multiple recruiters. This methodology was designed with the dual purpose of maximising representativeness (having a large number of collection points to ensure that any one centre could not unduly influence the findings) and minimising the onerousness for recruiters who were participating (making procedures quick and instructions easy to follow).

To ensure that the selection of eligible participants remained consistent across all recruiters, a set of recruitment criteria were developed for the use of 'official' sources (see Appendix D). This specified the following:

From your records, can you please identify [p#] clients who are ‘at risk’ of anaphylactic reactions to certain foods or food ingredients and who are currently over 1 year of age.   Please select clients who have been seen at the clinic in the past 2 years. We want to include a broad cross-section of the population who are ‘at risk’ of anaphylactic reactions to food or food ingredients, so that we can assess the effectiveness of labelling for all types of allergies.   Please select clients to roughly represent the proportion of each food allergy typically seen at your clinic.   Thus, if for example a third of the clients seen at your clinic have an egg allergy, then that proportion of the total number of questionnaires you have been asked to send out should be sent to clients with an egg allergy.   Clients should be selected at random from your client records for the previous 2 years.   If a client has more than one allergy that may cause anaphylaxis, they should be included as part of the sample for the least common allergy seen at your clinic.

However, since the sampling was through different medical institutions and private clinicians, the final selection of respondents may not have conformed to the recruitment criteria.

The answers respondents gave to question 10 (i.e. Approximately how long ago was the food allergy first identified for the person with the most serious food allergy?) indicate that there were a large number of persons in the survey who had been diagnosed more than two years previously, which was contrary to what was specified in the recruitment criteria [1] .  

This could be explained by a number of hypotheses:

1.       that recruiters may not have sampled strictly according to the criteria provided or lacked the capacity to do so;

2.       that many of the respondents selected had had their allergy identified previously i.e. the recruiter was not the identifying practitioner; or

3.       in many cases, the person with the most serious food allergy was not the person in the household who was selected for participation by the recruiter.

It is impossible to ascertain the degree to which the first two might have contributed to this anomaly. However, in up to 22% of households other serious food allergies were present [2] . It is probable that in some cases, the person with the most serious food allergy was not the person in the household who was selected for participation by the recruiter.

As no control was maintained over the sampling, it may not have been drawn in a uniform fashion and thus, there could be variations in the way participants were selected. In addition, recruiters were instructed to recruit across the broad spectrum of their clients in proportion to their food allergies. This was to ensure that a selection of allergies were included, and not just peanut allergies, which were anticipated to be the largest proportion of all clients. This perhaps may have led to over and/or under representation of some allergies depending on the food allergies predominantly seen by each official source.

As official recruiters from Western Australia could not be engaged, the sample was not as widespread as intended at the outset. Also, the Northern Territory was not included, and the South Island of New Zealand was under-represented, both due to difficulties in locating suitable recruiters. The structure of the final sample versus population is covered in section 3.1.

As the questionnaire packs distributed were in English and most of the respondents indicated that English was the language spoken at home, households where a language other than English was spoken may have been under-represented in this survey.  

The questionnaire asked respondents some retrospective questions, such as when the allergy was first diagnosed. This may have led to recall bias, where the respondent may not have accurately remembered the past. Furthermore, some respondents may have perceived some questions as being intrusive of their medical condition and either not answered the question or provided a response that they deemed acceptable. In these instances, a response bias may have occurred. The questionnaire also asked questions of specific allergens, for example ‘wheat’ rather than ‘all cereals containing gluten’. Therefore, responses provided have been for specific allergens rather that the categories as stipulated instandard 1.2.3of the Code. These measurement biases may have led to over or under estimation of certain responses.

Support groups

In addition to the above, support groups were also used to locate eligible respondents, initially via advertisements in which the selection criteria were outlined and the FREECALL phone number listed. Following this, top-up amounts were mailed-out to random members [3] . Accordingly, a proportion of participants from the support group sample were self-selecting [4] .

Future surveys

It should be noted that in order to maintain comparability, the methodology needs to remain consistent between each wave of the survey. This would ensure that any differences that are observed relate to actual or real changes rather than to changes due to methodological alterations. Because of this, it is recommended that any repeat survey be kept as closely as possible to the way this survey was conducted.

Response rates

The response rates are discussed in some detail in section 2.5.  

Whilst it is important to keep the basic proportions of the sample equivalent across the three sample sub-groups, the survey process could be made easier by attaining more 'in principle' support from the official sources (i.e. hospitals and private clinicians) and relevant ethics committee approvals prior to repeating the survey. This will make the process of recruitment more streamlined and less labour intensive.

Exec Summary  | Section 1 |  2  |  3  |  4  |  5  |  6  |  7  |  8  | 10  |  11  |  Appendixes  |  Full Report pdf 1609 kb

 

---