Scottish Health Survey – telephone survey – August/September 2020: technical report

1.2 Sample Design

The sample specification for the SHeS 2020 telephone survey was designed by the Scottish Government. A random sample of addresses (11,000 addresses) was selected from the Postcode Address File (PAF), using a stratified design. Participating households included in the survey were those from which a participant or participants took part after contacting ScotCen to opt in. All adults aged 16 and over within these households were eligible for interview and were given the opportunity to take part once initial telephone contact had been made. See section 1.6.4 for more on the sample profile by age and sex.

For the telephone survey, an opt-in approach was used, whereby potential participants were initially contacted by letter to invite them to participate and only re-contacted for interview if they advised ScotCen that they were willing to take part. This typically leads to higher levels of non-response than is typical for SHeS due to the lack of an interviewer visit to encourage participation.

While there is a degree of self-selection bias in the traditional SHeS survey model, this is greater for an opt-in survey. If bias is known to only relate to demographic factors, such as sex, age and area deprivation, to reduce bias, these can be corrected for through weighting (see section 1.7 for more on the weighting approach). However, as there is the potential for opt-in bias to relate to factors such as someone’s health or health behaviours, independent of the demographic factors considered as part of the weighting process, telephone survey estimates are likely to differ, to some extent, from the true population figures.

For the 2020 telephone survey, it is possible that factors such as the presence or absence of a long-term health condition and/or mental health/wellbeing may have influenced willingness to participate, whereby those who perceive themselves to be in better physical and/or mental health may have been more likely to opt in. The size of such an effect cannot be measured from the data collected. There may also be a degree of self-selection bias relating to other topics included in the survey, such as smoking, drinking and general health, similar to and possibly compounding the bias due to mode effects previously (see section 1.1.4).