Comparing Sample Surveys of Health with Official Population Statistics: Some Methodological Issues and Empirical Findings

摘要: This article reports an examination of factors that might affect the interpretation computed relative risks medical illness obtained when estimates morbidity derived from a specific subpopulation sample are compared with official statistical bureau surveys parent population. Interpreting significant or nonsignificant risk depends upon how fair comparison is. Three types error considered: (1) those arising subjects, including coverage, selection, response and comparability; (2) measurement processes, data collection instruments, interviewers processing; (3) confounding, which occurs factor for outcome is differentially distributed between exposure groups. An example given two sets estimates, one epidemiological cohort study national random list Australian Vietnam veterans, other by Bureau Statistics area probability sample. Variables measuring prevalence 37 recent chronic conditions were both studies their ratio as in veteran population was greater than 1.0 all except condition, 18 36 30 carried 99% confidence intervals excluding 1.0. Adjustment variables thought to have potentially biased gave varied results. The issue adjustment becomes overadjustment meaning variable its place putative causal pathway.