Prior Information in Behavioural Capture-Recapture Methods: Demography
Influences Injectors' Propensity to be Listed on Data Sources and their
Drugs-related Mortality
Ruth King, Sheila Bird, Steve Brooks, Sharon Hutchinson and Gordon Hay
University of St. Andrews
Summary
Bayesian analysis is presented of Scotland's four
capture-recapture data-sources for 2000 by which to estimate numbers of
current injecting drug users by region (Greater Glasgow, elsewhere in
Scotland), sex (male, female) and age-group (15-34 years, 35 years or older).
Secondary analysis goal is estimates and credible intervals for the
demographic influences on Scotland's drugs-related death rates per 100
current injectors. Use of uninformative and then informative priors is
contrasted. The latter take into account that demography influences Scottish
injectors' propensity to be listed on different data-sources; also, that
external information is available about: the total number of injectors, via
an internationally informative prior for injectors' drugs-related
death-rate coupled with Scotland's 1006 drugs-related deaths in 2000-02; and
the male to female ratio of injectors internationally. Incorporation of
informative priors altered the models with highest posterior probability. In
secondary analysis, there was confirmation of higher drugs-related mortality
in older injectors, for those resident elsewhere than Greater Glasgow and
female injectors' lower drugs-related death rate was not sustained beyond 34
years of age. The authors recommend that demographic influences be
accommodated in behavioural capture-recapture estimation, especially when it
is a prelude to secondary analysis, as here of drugs-related death rates.
Medical Subject Headings (MeSH) (keywords):
Bayes theorem; data collection;
epidemiologic methods; models, statistical; mortality; prevalence; substance
abuse, intravenous.
Appeared as King, R., Bird, S. M., Brooks, S. P., Hutchinson, S. J. and
Hay, G. (2005) "Prior Information in Behavioural Capture-Recapture Methods:
Demography Influences Injectors' Propensity to be Listed on Data Sources and
their Drugs-related Mortality".
American Journal of Epidemiology 162 pp 694-703