Table 1

Baseline and business-as-usual (BAU) parameters

ParameterData sourceTrend, uncertainty and scenario analyses
Demography
PopulationStatistics New Zealand (SNZ) population estimates for 2018 by sex, age group and ethnicityUncertainty: nil
BAU—epidemiological parameters
All-cause mortality rates (ACMRs)SNZ mortality rates by sex, age and ethnicity for 2020Trends in ACMR were estimated using data from the GHDx database IHME. The annual percentage change in the age-standardised all-cause mortality rates from 1990 to 2019 was −1.9% for sexes combined. Retaining the original BODE3 model assumption of a 0.5% point greater APC for Māori (due to long run trends of closing ethnic inequalities in mortality), we arrived at APCs for ACMR from 2020 to 2035 of: Māori=−2.0%; non-Māori=−1.5%. They were uniform by age. No trends applied beyond 2035. Uncertainty: nil.
All-cause morbidity ratesNZ Burden of Disease Study (NZBDS)25Data on years of life lived with disability (YLD) were obtained from the NZBDS for each sex and age group in 201625 and divided by the population in each sex by age by ethnic group to generate morbidity rates. No time trend was allowed.
Disease-specific incidence, prevalence and case fatality rates (CFRs)NZBDS25For each tobacco-related disease, coherent sets (by sex, age and ethnicity) of incidence rates, prevalence, CFRs and remission rates (zero for non-cancers, the complement of the CFR for cancers to give the expected 5-year relative survival) were estimated using the software DisMod II. Cancer incidence and CFR APC trend using Poisson regression historical trends of incidence and CFRs of diseases. The APCs included as inputs to the PMSLT model out to year 2035 and held constant beyond (future prevalence changes dynamically with model). It was assumed that the APCs were constant by ethnicity. Uncertainty: starting in 2020, rates all ±5% SD, correlations 1.0 between four sex by ethnic group categories for all diseases. APC all ±0.5% SD normal, correlations 1.0 between four sex by ethnic groups for all diseases.
Disease-specific morbidityNZBDS25The sex and age-specific disability rates were calculated as disease’s YLD obtained divided by the prevalent cases. The same disability rate was assumed by ethnicity (ie, those with disease are assumed to have same severity distribution across ethnicity). Uncertainty: ±5% SD (beta distribution).
Tobacco smoking and vaping
Smoking (daily)NZ Health SurveyLogistic regression of NZ Health Survey data for years 2011–2019 was undertaken to ‘predict’ the prevalence of daily smoking (at least one cigarette per day) for years 2020–2040. This ‘prediction file’ was then reanalysed from a sex by ethnicity by 5-year age group perspective (ie, 72 separate sex by age by ethnicity cohorts) to generate future BAU smoking prevalence—and a yearly (cohort ageing) rate of decline—that was then used in the exposure model.
Vaping (daily e-cigarette use)NZ Health SurveySame as above for smoking, but for ‘vaping’ at least daily.
Association of smoking and vaping with disease incidence rates
Smoking–disease incidence rate ratiosRelative risks of disease incidence for the association of current (or ex-smoker) with never smoker were sourced from NZ linked census cancer33 and census mortality34 (censuses include smoking question) and data from the Cancer Prevention Study II for respiratory diseases.35 Attenuation over time since quitting for ex-smokers was modelled using equations and coefficients from Hoogenveen et al.36Standard errors of regression coefficients as described in online supplemental appendix C and tables S20 and S21.
  • APC, annual percentage change; BODE3, Burden of Disease Epidemiology, Equity and Cost-Effectiveness; GHDx, Global Health Data Exchange; IHME, Institute of Health Metrics and Evaluation; PMSLT, proportional multistate life table.