Community surveys of late-life depression: who are the non-responders?

community surveys of depression among older people may be particularly prone to non-response. Information on non-responders is difficult to obtain and often limited to demographics. Therefore, the full extent of response bias is not always known.

to determine factors associated with non-response at each stage of a two-stage survey of late-life depression.

one large general practice (registered population >30000) serving the market town of Melton Mowbray, Leicestershire, UK.

community residents (n=2633) aged 65-74 years registered with the practice.

a two-stage community survey of patients aged 65-74 years. The first stage was an interviewer-administered general health survey including a measure of depressive symptoms. We asked those who screened positive for possible depression to undergo a semi-structured psychiatric interview. We compared use of services and medication by non-responders and responders to both stages using primary-care records. We compared Townsend deprivation scores using data obtained from the 1991 census.

responders to stage 1 were more likely to use both primary [odds ratio (OR) 1.65, 95% confidence interval (CI) 1.38-1.96] and secondary (OR 1.59, 95% CI 1.25-2.02) services and tended to live in more affluent areas (P=0.002). At stage 2, the only difference observed was a lower level of use of tranquillisers or hypnotics among responders (OR 0.27, 95% CI 0.11-0.67).

older people with low levels of contact with health services may be under-represented in community surveys of depression. Investigators should look outside traditional health settings to promote the uptake of response in these studies.