Abstract
BACKGROUND: Under-identification of Aboriginal and Torres Strait Islander (hereafter referred to as Aboriginal) people can result in inaccurate estimation of health outcomes. Data linkage has improved identification of Aboriginal people in administrative datasets. AIM: To compare three methods of ascertainment of Aboriginal status using only pregnancy data from the Western Australian Midwives Notification System (MNS), to the linked Indigenous Status Flag (ISF) derived by the Department of Health. MATERIALS AND METHODS: This retrospective population-based cohort study utilised logistic regression to determine which demographic characteristics were associated with under-identification, and the effect of ascertainment method on perinatal adverse outcomes. RESULTS: All methods identified a core group of 19 017 (83.0%) Aboriginal women and the ISF identified 2298 (10.0%) women who were not identified using any other method. Under-ascertainment was lowest when a woman's Aboriginal status was determined by ever being recorded as Aboriginal in the MNS data, and highest when taken as it had been recorded for the birth in question. Maternal age <20 years, smoking during pregnancy, pre-existing diabetes, a history of singleton preterm birth and being in the lowest 20% of Socio-Economic Indexes for Areas score were all associated with a higher chance of being identified by the methods using only the MNS. These methods were less likely to identify nulliparous women, and those with maternal age ≥35 years. The method of ascertainment of Aboriginality did not make a significant difference to the adjusted predicted marginal probabilities of adverse perinatal outcomes. CONCLUSION: Unlinked pregnancy data can be used for epidemiological research in Aboriginal obstetric populations.