Country of birth recording in Australian hospital morbidity data : accuracy and predictors

Objective: Country of birth (COB) is an important determinant of health risks and outcomes. Despite the widespread use of hospital morbidity data for research, little is bknown about the quality of COB recording in these data. This study validated the recording of 40 COBs in Australian hospital morbidity data and identified factors associated with the recording accuracy. Methods: Questionnaire data from the 45 and Up Study were linked to records from the NSW Admitted Patient Data Collection (APDC), yielding linked records for 39,559 overseas-born people. Measures of agreement between self-reported COB from the 45 and Up Study ('gold standard') and COB as recorded in the APDC were calculated and logistic regression analysis was used to explore predictors of correct recording of COB. Results: On average, 81% of APDC records for overseas-born people had the correct COB recorded. Sensitivities varied between 61% (Germany) and 99% (Spain). Specificities exceeded 99.5%. Positive predictive values were greater than 90%, except for Hong Kong (72%). Ireland (85%), and Czechoslovakia (89%). Negative predictive values exceeded 98%, except for the United Kingdom (87%). Kappa values indicated good or excellent concordance between self-reported COB and APDC data (kappa 0.73 or greater). Where inaccurate, COB was mainly recorded as 'Australia' or 'Inadequately described'. COB was recorded with greater accuracy in principal referral hospitals, public hospitals and hospitals in metropolitan areas. Accuracy decreased with duration of residence in Australia, increased with age at immigration, and varied according to COB. Conclusion: COB recording accuracy was generally high but varied according to hospital characteristics, country, and measures of acculturation.