NADRA 2018: Q&A with key note speaker Linsay Gray

Linsay Gray, what is the theme of your presentation at NADRA 2018?

I will be talking about the continuing decline in participation in population-based health surveys. I am delighted to have been invited to present on this universal issue. As well as considering the associated challenges, I will cover potential solutions which are particularly ameneable to implementation in Nordic settings.

What are the reasons behind the decline in participation levels for population-based health surveys?

While the issue of individuals being less and less likely over time to agree to respond positively to survey invitations of any kind is well recognised, the mechanisms at play behind the diminishing participation levels are less well appreciated. Suggested explanations include the growing challenge invovled in making contact with prospective respondents and those that are contacted being more and more likely to refuse, even with offers of incentives; a societal climate that is increasingly hostile to survey participation is considered to be a factor. Whatever the reasons, non-participants tend to be younger or older adults, male, disadvantaged, and, crucially, less healthy.

What are the consequences of this?

As in any study, if the numbers of participants is lower than aimed for, this has a detrimental effect on the power of analysis. Of even greater general concern, when we are systematically missing particular population sub-groups, there is likely to be attendant bias. For instance, if those agreeing to participate are healthier overall than those refusing, then the sample we have is not representative of the general or ”target” population. This impacts the reliability of the use of health surveys as a basis for the development and evaluation of policy. Simply drawing inference on the sample of participants is unlikely to provide us with reliable estimates of any health or health-related measures of interest such as alcohol consumption levels. If, as is typically the case, the participation level of a survey series has changed over time, this is likely to have a distorting effect on detected temporal trends.

What needs to be done?

There are a number of ways of dealing with non-participation. In an ideal world, most or all invited individuals would agree to participate and there would be no issue. In the real world, those conducting surveys invest resources trying to maximise participation with often limited return. There are statistical methods available as options that can be applied to help correct for non-participation bias, which I’ll cover in my presentation. Socio-demographic differences can be dealt with in a relatively straightforward way but there is an underlying assumption that the health of respondents is equivalent to that of the general population within socio-demographic groups, which is rarely the case. I’ll outline methodology that can overcome this issue. The register-based systems in operation mean the Nordic countries are well placed to implement these methods.