BACKGROUND--Predictors of chlamydia and gonorrhoea can be used to increase the cost-effectiveness and acceptability of screening programmes, and allow targeting of control strategies. METHODS--All women attending an STD clinic in 1988-1990 were offered screening for chlamydia and gonorrhoea, and the test results correlated with a wide range of potential predictors using multiple logistic regression. RESULTS--Of 4822 attenders, 3533 (73.3%) were tested for chlamydia over a total of 5430 episodes, yielding 348 (6.4%) positives, and 3510 (72.8%) were tested for gonorrhoea over a total of 5450 episodes, yielding 100 (1.0%) positives. Independent predictors of chlamydial infection were being an STD contact, having endocervical gonorrhoea, being under 25, not having genital herpes, being Aboriginal, using oral contraception, not having a steady partner and having vaginal discharge or dysuria. For gonorrhoea such predictors were being Aboriginal, an STD contact, under 25, tattooed, having vaginal discharge or dysuria, and having had sex outside the state in the past three months. Selective screening criteria for gonorrhoea provided 91% of positives, eliminated the need for 42% of tests and resulted in an increased yield ratio of 1.5 whereas the corresponding outcomes for screening criteria for chlamydia were 91%, 29% and 1.3, respectively. CONCLUSIONS--The diversity of STD epidemiology requires development of empirical screening guidelines for diverse settings. Standardisation of methodology to facilitate comparisons and extrapolation should include investigation of a wide range of variables, available before patient examination, by multivariate analysis, and choice of selective criteria to cover at least 90% of the infected population as well as resulting in a substantially increased yield (preferably an increased yield ratio of at least 1.5).