In low-resource settings, a stepped care approach is necessary to screen and provide care for pregnant women with mental health problems. This study sought to identify screening items that were most robust at differentiating women experiencing psychological distress and requiring counselling [assessed by screening with the Edinburgh Postnatal Depression Scale (EPDS) and a Risk Factor Assessment (RFA)] from those with a psychiatric disorder as diagnosed by a psychiatrist. Case records of women in an antenatal mental health service in Cape Town were reviewed. Composite scores and individual items on screening scales (EPDS, RFA) of participants who qualified for counselling (n = 308) were compared to those of participants who were diagnosed with a psychiatric disorder (n = 58). All participants with a psychiatric disorder were diagnosed with either depression or anxiety disorders. These participants had higher mean scores on the EPDS and RFA than those who qualified for counselling (p < 0.01). Logistic regression and ROC analyses suggested that the best items to distinguish women with depression or anxiety from those qualifying for counselling were 'I have felt sad and miserable', 'I am not pleased about being pregnant' and 'I have had serious depression, panic attacks or problems with anxiety before' (sensitivity 0.655, specificity 0.750 for this combination of three items). A small number of items may be useful in screening for mental illness in pregnancy which requires higher levels of care. Such screening may contribute to a more efficient stepped care approach.