Indirect Area Estimates of Disease Prevalence: Bayesian Evidence Synthesis with an Application to Coronary Heart Disease
Keywords:
Coronary heart disease, deprivation, ethnicity, power prior, prevalence, mortality, small area.Abstract
The incidence of many chronic illnesses, including as cancer, diabetes, asthma, mental illness, and coronary heart disease, varies greatly across regions due to
substantial correlations between risk factors and socioeconomic and ethnic groupings. When evaluating health care demands and comparing health care provision (such as
surgical intervention rates) to health needs, variations in prevalence are crucial. In order to provide indirect prevalence estimates for geographic areas where prevalence
statistics are not otherwise accessible, this work focuses on predicting the prevalence of coronary heart disease using a Bayesian technique to synthesise information of
various sorts. Regression techniques may be used to get gradients by age, sex, ethnicity, wide area, and socioeconomic level. One source is information on prevalence risk
gradients from national health survey data, which often only offer regional identifiers (for confidentiality reasons). A number of health surveys are often accessible, and by
using data on prevalence gradients from previous surveys, one may think about pooling strength across surveys (e.g. via a power prior technique). The second source of
data is population totals by age, sex, ethnicity, etc. from intercensal population estimates or censuses, to which prevalence rates derived from surveys are added. Since there
is a significant link between the frequency of heart disease and death from some major chronic illnesses across regions, data on area mortality is another possible source of
information. Using (a) data from the Health Surveys for England for 2003 and 1999, (b) population data from the 2001 UK Census, and (c) area mortality data for 2003, a
case study examines the growth of estimates of coronary heart disease prevalence in 354 English regions.