Parameter | Data source | Comments/notes/model and data assumptions |
Smoking prevalence in base year 2016 | IHME/GBD | Current smokers: Prevalence as reported by IHME for 2016, by sex and age (http://ghdx.healthdata.org/). Uncertainty: Logistic distribution, SD 0.4 for males and 0.35 for females (matching GBD uncertainty intervals), 1.0 correlated draws by sex and age within each iteration. Former smokers quit <20 years: We estimated the distribution of ex-smokers by year since quitting by:
Note: We did not allow for differential mortality by smoking status. At younger ages, while relative differences in mortality rates by smoking are large, in absolute terms they are not large. Also, the multistate life table model prospectively models proportions in each smoking state beyond 2016; the estimates here are ‘just’ base year estimates for 2016 that overall HALYs gained are not sensitive to. Uncertainty: Nil directly estimated, as a function of uncertainty in ‘current’. Never smokers plus former smokers quit >20 years=1—current—quit <20 years. Uncertainty: Nil directly estimated, as a function of uncertainty in ‘current’. |
Smoking prevalence 2017 and beyond | IHME/GBD | Logistic regression models were fitted to 1980–2016 IHME smoking prevalence data with terms for age, age-squared and year. Coefficients obtained from the fitted models were used to estimate future BAU smoking prevalence. Uncertainty: Propagated from the SEs of coefficients (and their covariance matrix). |
All-cause mortality rates | IHME/GBD | In main life table: Rates by sex and age group for 2016 from IHME were inputted directly. For 2017–2036 (then no further change), annual percentage change (APC) was estimated by sex and age by fitting Poisson regression models on 1990–2016 mortality rates. See online supplementary additional file 6 and 7 and for details. |
All-cause morbidity rates | IHME/GBD | Years of life lived with disability (YLD) for each sex by age group in 2016 from the GBD were divided by that demographic group’s population size, to generate sex-by-age morbidity rate. No time trend was allowed, as changes over time in age-standardised YLD rates are globally minimal.43 These morbidity rates were input to the main life table for calculating HALYs. Uncertainty: ±10% SD log normal distribution. |
Disease-specific incidence, prevalence, case fatality and remission rates | IHME/GBD |
Base year 2016: Raw estimates for disease-specific incidence rates, prevalence, mortality rates (ie, disease deaths divided by total population—not just prevalent cases) and case fatality rates (mortality rate divided by prevalence) were drawn or calculated from IHME data. (Remission rates were set to zero for all diseases other than lung cancer; remission rates for lung cancer were estimated to give a 10% relative survival, that is, remission rates=1/9×case fatality.) Some implausible patterns were found with GBD data, with disease incidence rates being lower than disease mortality rates (see online supplementary additional file 8). Therefore, we used an epidemiological tool (DISMOD II)31 to further process and ensure coherence for 2016 base year estimates. Trends from 2016 to 2031. We used Poisson regression on incidence rates and case fatality rates for 1990–2016 GBD data to estimate APCs to include in the multistate life table model (online supplementary additional file 3). Multistate life table calibration. Good calibration was achieved for the above input parameters (online supplementary additional file 5). Uncertainty: Base year in 2016, rates all ±5% SD (log normal distribution for incidence), correlations 1.0 between sexes for all diseases. APC trends all ±0.5 percentage points SD, normal, correlations 1.0 between sexes. |
Disease-specific morbidity | IHME/GBD | Each disease was assigned a sex and age group-specific disability rate, calculated as that disease’s YLD in 2016 divided by the prevalent cases. Uncertainty: ±10% SD log normal. |
Disease-specific incidence rate ratios by smoking status | IHME/GBD | Rate ratios for current versus never smokers were taken from IHME (see online supplementary additional file 10),1 for each tobacco-related disease included in the model. Uncertainty: That given by IHME for current smoker rate ratios (and shown in online supplementary additional file 10). |
Baseline average pack price | WHO | SBD20 Solomon Islands dollar (SBD) (US$2.53) in 2017.25 www.who.int/entity/tobacco/global_report/2017/table-9-2.xls Uncertainty: Nil |
Baseline excise tax | Tobacco Atlas/WHO FCTC report | 19.8% of retail price in 2016.25
Uncertainty: Nil |
APC, annual percentage change; BAU, business as usual;FCTC, Framework Convention on Tobacco Control; GBD, Global Burden of Disease; HALY, health-adjusted life years; IHME, Institute for Health Metrics and Evaluation;SBD, Solomon Islands dollar; UI, uncertainty interval; YLD, years of life lived with disability.