Table 2.

List of confounders and covariates used in propensity score processing

Variable	Groups/levels	Comments
VARIABLES INCLUDED IN FINAL PROPENSITY SCORE MODEL
Prioritised ethnicity	Māori Pacific Asian Other (Middle Eastern, Latin American, African, and Other) European/NZ European	Variable used to stratify groups prior to calculation of propensity scores
Nativity	Born in New Zealand Born outside New Zealand
Age group	Ten-year age bands (15–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75+)	Treated as categorical variable
Gender	Female Male
NZDep2013 Quintile [54]	Quintiles 1–5	Treated as categorical variable
Highest educational qualification	Less than upper secondary Upper secondary Tertiary Other	“Other” category included qualification responses that could not be coded to an existing response category.
Employment status	Currently working Looking for work Not in labour force Not in labour force (aged 65+) Other	“Not in labour force” group split into those below and above retirement age (65+). “Other” category included responses that could not be coded to an existing response category.
VARIABLES CONSIDERED BUT NOT INCLUDED IN FINAL PROPENSITY SCORE MODEL
Baseline self-rated health	Excellent/Very good/Good Fair/Poor	Health measure: not included as analysis adjusts for baseline
Baseline mental distress (K10 scale)	Continuous variable (score 0–40)	Health measure: not included as analysis adjusts for baseline
Smoking status	Current smoker (at least monthly) Not current smoker	Health measure: not included (also substantial missing data)
Index of Multiple Deprivation [55] (IMD)	Decile for total IMD score	Overlap with NZDep quintile; did not improve balance
IMD [55] Employment Subscale	Decile for employment measures in local area	Overlap with NZDep quintile (individual level employment also already included)