Asking questions about women's reproductive health: Validity and reliability of survey findings from Istanbul

In countries where population-based data on health problems are scarce, the extent of reproductive morbidity can be estimated from replies in structured interviews as a complement or as an alternative to reports from physician's examination and laboratory tests. We examined the sensitivity and specificity of detected morbidity based on these replies as compared to medical diagnoses and explored the consistency of replies when the questionnaire was administered twice, by two types of interviewers in different environments. Data were collected in a cross-sectional survey in Istanbul. The presence or absence of five morbidities, reproductive and urinary tract infections (RTI and UTI), menstrual disorders, pelvic relaxation and anaemia was determined by algorithms based on the replies, and by the physician's diagnosis. Except with anaemia, questionnaire replies were more specific than sensitive in detecting morbidity, probably partly due to many morbid conditions being accepted as normal. Specificity exceeded 80% for home reports of menstrual disorders (93.0%), pelvic relaxation (95.7%), RTI (abnormal discharge and pain) (81.2%) and UTI (80.7%), with the corresponding figure for anaemia at 41.7%; the best sensitivity results were for anaemia (58.3%), RTI (abnormal discharge only) (49.3%) and menstrual disorders (45.4%) with figures for pelvic relaxation and UTI reaching only 17.3 and 13.0% Reliability between the interviews (assessed by the kappa coefficient), was highest at 66.1% for pelvic relaxation and lowest at 39.9% for menstrual disorders. Reliability varied between the two lay interviewers, suggesting the interviewer and the interview conditions are important. Questionnaire-based information on this type of morbidity is most useful for ascertaining perceived ill-health and only of limited use for the corresponding medically defined conditions.