Table 2.
Cox proportional hazards regression analyses: time-to-airway disease among people immigrating to British Columbia, Canada, 1985–2015.
| Statistical Analysisa | N | Adjusted HR | 95% CI |
|---|---|---|---|
| Aim 1: analyzing post-TB airway disease risk | |||
| Covariate-adjusted (main analysis: respiratory TB vs controls) | 1 005 328 | 2·08 | 1·91 – 2·28 |
| Sensitivity analyses | |||
| Covariate-adjusted (removed ETOH, substance dependence, psychoses, and depression)b | 1 005 328 | 2·11 | 1·93 – 2·30 |
| Covariate-adjusted (van Walraven-weighted Elixhauser comorbidity score)c | 1 005 328 | 2·06 | 1·89 – 2·26 |
| Covariate-adjusted (bronchiectasis and fibrosis added to the airway disease definition) | 1 005 283 | 2·18 | 2·00 – 2·38 |
| Covariate-adjusted (removed respiratory TB patients with pleural samples; n = 55) | 1 005 273 | 2·10 | 1·92 – 2·30 |
| Different TB definitions | |||
| Covariate-adjusted (all forms of TB vs controls) | 1 006 271 | 1·75 | 1·63 – 1·88 |
| Covariate-adjusted (non-respiratory TB vs controls)d | 1 004 733 | 1·36 | 1·20 – 1·53 |
| Age/sex-adjusted (pleural TB vs non-pleural TB) | 1141 | 0·87 | 0·57 – 1·32 |
| Aim 2: assessing potential unmeasured confounding | |||
| PS methods | |||
| PS decile-adjusted (main covariates) | 1 005 328 | 2·27 | 2·08 – 2·49 |
| hdPS decile-adjusted (main covariates + empirical covariates) | 1 005 328 | 2·28 | 2·09 – 2·50 |
| LASSO-hdPS decile-adjusted (main covariates + LASSO-refined empirical covariates) | 1 005 328 | 2·26 | 2·07 – 2·47 |
| Adjustment for smoking behavior proxy variables | |||
| Covariate-adjusted subdata analysis (main covariates + tobacco use variable)e | 31 063 | 1·53 | 1·37 – 1·71 |
| Covariate-adjusted (main covariates + personal health risk proxy variable) | 1 005 328 | 2·03 | 1·85 – 2·22 |
BC = British Columbia; BCCDC = British Columbia center for Disease Control; CI = confidence interval; ETOH = ethanol dependence; hdPS = high-dimensional propensity score; HR = hazard ratio; LASSO = least absolute shrinkage and selection operator; N = analytic sample size; PS = propensity score; TB = tuberculosis.
Notes:
aCovariate-adjusted models included as adjustment variables: age at index, sex, income quintile at index, educational qualification upon immigration, immigration class, TB incidence in country of birth, weighted Charlson comorbidity score, year of arrival in BC, ETOH, substance dependence, psychoses, and depression.
bDue to lower than expected numbers, we removed these three variables from the set of adjustment variables.
cWe replaced the Charlson comorbidity score with the van Walraven-weighted Elixhauser comorbidity score and removed ETOH, substance dependence, psychoses, and depression variables as individual binary covariates because they are included in the van Walraven Elixhauser comorbidity score.
dNon-respiratory TB included everyone who was a confirmed case of TB but did not have a smear or culture positive respiratory sample in the laboratory datasset of the BCCDC TB Registry.
eThis analysis used a lower number of cohort members because it includes only people who were in the BCCDC TB Registry's Person dataset, which is a client list used by Provincial TB Services to track people who are tested or treated for active or latent TB within BC.