Measuring errors and non-participation in a nation-wide study of stillbirth.

BACKGROUND: An observational study in which data are being collected by an anonymous questionnaire is an alternative when an intervention cannot be investigated by a controlled clinical trial. If used, measuring errors and non-participation may decrease validity. METHODS: We compared data retrieved with an anonymous questionnaire in a nation-wide study on stillbirth with information from the Swedish Medical Birth Register. In the original study population with 759 women, four groups were formed comprising participants and non-participants among subjects (stillbirth in 1991) and controls (live birth the same year). To assess misclassification, we compared information provided by the questionnaire to the data for the corresponding group in the Swedish Medical Birth Register. Comparing background information on participants and non-participants elucidated misrepresentation. RESULTS: The figures from the questionnaire and the register were almost identical for hypertension, albumin and sugar in the urine during pregnancy, while anemia and tiredness were reported more often in the questionnaire. Similar results were found in the two data sources regarding the type and frequency of analgesia during the delivery, birth order, gender and weight of the index child. On average, all women reported a larger number of gestation weeks in the questionnaire than was indicated in the register-based information. A higher percentage of the participants than the non-participants, both among subjects and controls, were born in Sweden, had Swedish citizenship and were married or cohabiting. CONCLUSION: Information on many variables derived from an anonymous questionnaire can be used in a scientific context. The validity of a study in the investigated population would probably be enhanced if it were restricted to subjects born in Sweden.