Stratum-specific likelihood ratios of the General Health Questionnaire in the community: help-seeking and physical co-morbidity affect the test characteristics

Abstract

Background. In evidence-based medicine, stratum-specific likelihood ratios (SSLRs) are now being increasingly recognized as a more convenient and generalizable method to interpret diagnostic information than an optimal cut-off and its associated sensitivity and specificity. We previously examined the SSLRs of the General Health Questionnaire (GHQ) in primary care settings. The present paper aims to examine if these SSLRs are generalizable to the community settings.

Methods. The Composite International Diagnostic Interview (CIDI) and the GHQ were administered on a representative sample of the Australian population in the Australian National Survey of Mental Health and Well-Being. We first compared the SSLRs of GHQ in urban Australia with the estimates that we had previously obtained from the developed urban centres in the WHO Psychological Problems in General Health Care study. If the SSLRs in the community were found to differ significantly from those in the primary care, we sought for explanatory variables.

Results. The SSLRs in urban Australia and in the urban centres in the WHO study were significantly different for three out of the six strata. When we limited the sample to those with physical problems who visited a health professional, however, the SSLRs in the Australian study were strikingly close to those observed for primary care settings.

Conclusions. Different sets of SSLRs apply to primary care and general population samples. For general population surveys in developed countries, the results of the Australian National Survey represent the currently available best estimates. For developing countries or rural areas, the results are less definitive and an investigator may wish to conduct a pilot study.