Name analysis to classify populations by ethnicity in public health: Validation of Onomap in Scotland

Abstract

Health inequalities between ethnic minorities and the general population are persistent. Addressing them is hampered by the inability to classify individuals' ethnicity accurately. This is addressed by a new name-based ethnicity classification methodology called 'Onomap'. This paper evaluates the diagnostic accuracy of Onomap in identifying population groups by ethnicity, and discusses applications to public health practice. Onomap was applied to three independent reference datasets (birth registration, pupil census and register of Polish health professionals) collected in Britain and Poland at individual level (n = 260,748). Results were compared with the reference database ethnicity 'gold standard'. Outcome measures included sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). Ninety-five percent confidence intervals and Chi-squared tests were used. Onomap identified the majority of those in the British participant group with high sensitivity and PPV (>95%), and low misclassification (<5%), although specificity and NPV were lowest in this group (56-87%). Outcome measures for all other non-British groupings were high for specificity and NPV (>98%), but variable for sensitivity and PPV (17-89%). Differences in misclassification by gender were statistically significant. Using maiden name rather than married name in women improved classification outcomes for those born in the British Isles (0.53%, 95% confidence interval 0.26-0.8%; P < 0.001) but not for South Asian or Polish groups. Onomap offers an effective methodology for identifying population groups in both health-related and educational datasets, categorizing populations into a variety of ethnic groups. This evaluation suggests that it can successfully assist health researchers, planners and policy makers in identifying and addressing health inequalities.