National Health Index (NHI) numbers are used in New Zealand (NZ) to identify patients and link their medical records. The NHI dataset contains ethnicity and other demographic data on each patient and so NHIs are important for measuring disparities in health status, in treatment utilisation and outcomes between ethnic groups, and monitoring attempts to reduce disparities. However, previous studies have found inaccuracies in ethnicity data in the NHI dataset1–3 that could introduce significant errors into estimates of healthcare need.4
In the 2006 Census, Pacific people made up 6.9% of the NZ population. This proportion will increase significantly in the future, and Pacific people experience much higher levels of ill-health than other New Zealanders.5 There are a range of reasons why the NHI system might not capture accurate data about Pacific people, such as healthcare staff's lack of understanding of Pacific naming conventions. Our study aimed to investigate the accuracy of ethnicity data in the NHI dataset for Pacific people in Dunedin.
Most previous studies of the accuracy of ethnicity data have compared administrative records. In an earlier study, we developed a novel methodology of working with communities to collect personal details, and asking the Ministry of Health to check recorded ethnicities against the ones provided by the community members.6 In this study, we worked with the Pacific Trust Otago and attended events organised by the Trust. The purpose of the study was explained to people attending the events, and they were asked to participate. We asked them to provide date of birth, gender, full name, any other names they might be recorded as, address, length of residence in NZ and ethnicity. We used both paper and an electronic form for data collection. Participants were also given the option of being contacted if their ethnicity was incorrectly recorded in the NHI database. Participants with incomplete information on the paper-based questionnaire were contacted by telephone to provide the missing information. Data was sent to the Ministry of Health in an attempt to locate NHI numbers for participants and find their ethnicity/ies. We compared these ethnicities with those given by participants. For each group, the data was analysed (using Microsoft Excel) based on linked and matched data with age and length of residence in NZ.
The 120 participants were from a range of Pacific ethnicities and had a mean age of 38 (standard deviation of 18). A NHI number was found for 118 of the participants. Only 93 of the 118 (79%) had self-identified ethnicity correctly recorded in the NHI system. The accuracy of ethnicity matching within the NHI database for Samoan, Tongan, Cook Island, Māori and Other Pacific ethnic groups was 88%, 83%, 91% and 55% respectively. Most (109/118, 92%) participants were recorded as having a Pacific ethnicity. Three were recorded as Māori and six did not have an ethnicity recorded. There was no clear pattern for accuracy of ethnicity data by age or by length of residence in NZ.
The use of an electronic form for data collection, a clearer larger paper form and an additional question asking for other names participants may be known by, is likely to explain why we were able to find a NHI number for a much higher proportion of participants in this study compared to the previous one with Asian people.6
Inability to find a NHI number could be the result of human error in our data collection or incorrect details being recorded in the NHI system. Ethnicity data may be inaccurate for a range of reasons, including healthcare staff members’ reluctance to ask about ethnicity, guessing ethnicity based on appearance or name, poor technique for asking or inaccurate recording (ticking an incorrect box), or patients being unwilling to provide ethnicity information.
We consider the accuracy of ethnicity data to be low: only 79% of participants had their self-identified ethnicity correctly recorded in the NHI system. However, at the higher level of coding, most (92%) people were identified as Pacific. This suggests that researchers and policymakers can be more confident at the higher level of coding, but considerable caution should be exercised in analysis carried out at the level of specific Pacific ethnicities. The study was solely based on Pacific people in Dunedin and may not be representative of the whole Pacific population in NZ.
High-quality ethnicity data is essential for monitoring health status, developing effective policies and services, and reducing ethnic disparities and inequalities, and this study shows that improvements need to be made in the recording of Pacific ethnicity data.