New Zealand Statistical Association 2024 Conference

Tori Diamond

University of Auckland

Novel applications of linked administrative data – adding longitudinal capability to the Te Kupenga survey

Can linked administrative data be used to transform New Zealand's only sample survey on Māori wellbeing into a longitudinal study?

This project extends the usefulness of an important survey dataset by linkage to admin data, effectively adding longitudinal capability within a linked administrative data source. This created robust statistical processes to transform an official statistics survey into a nationally representative cohort study. StatsNZ’s Integrated Data Infrastructure (IDI) holds administrative and survey datasets containing a range of variables linkable at the individual level. Te Kupenga is a large nationally representative post-censal survey of the Māori population and is the only survey with culturally informed variables but is often under-utilised in research.

The Te Kupenga survey was used as a foundational cohort linking to outcomes and determinants in different datasets from different time periods. Outcomes included Ambulatory Sensitive Hospitalisations (ASH) (post-2013) and COVID-19 vaccinations (post-2020), while determinants included individual, household and geographic variables. Linking a representative survey to admin data created issues of loss to follow-up and missing data, so the original sample is not maintained after linkage. Loss to follow-up and missingness differed depending on variable selection and time periods. So, new universally applicable weights were not possible. However, we created a robust, generally applicable process for re-weighting survey data to account for missingness and loss to follow-up in admin data.

This project demonstrates the approach for turning a sample survey into a longitudinal cohort using admin data and creates methods that can be used for other official statistics surveys.