So caregiver burden and positivity were assessed separately using multivariate models regardless of the results of bivaraite analysis.
Age, education level, and marital status were identified a priori as important variables and were selected for inclusion in the final multivariate models, regardless of whether they met the specified criteria.
A priori, we decided to retain child sex, time-varying HIV status, and assigned study treatment regimen in the multivariate models, regardless of their significance, since these variables are important from a biological perspective.
Having a parent with diabetes contributed strongly to diabetes risk in our present multivariate model, regardless of overall age and other risk factors.
Known risk factors at treatment initiation, such as infections and infestations, chronic obstructive pulmonary disease (COPD), diabetes mellitus, and tobacco use, were considered in the multivariate model regardless of results in the univariate analysis.
Regardless, multivariate models (Table 3) echoed bivariate findings: compared with the non-mood disorder population, individuals with bipolar disorder had about 4.6 times the odds of not working (95% CI 3.52, 6.04), 3.56 times the odds of missing at least two weeks of work (95% CI 2.12, 6.04), and 4.6 times the odds of spending at least 10 missed work days in bed (95% CI 2.75, 7.80).
Since none of our outcomes were rare, we were able to create full multivariate models that included all variables regardless of their statistical significance.
Therefore, in all multivariate models the drinking water variable was retained regardless of its statistical significance.
Subsequently, stage of disease (localised, regional metastasis, or distant metastasis) and smoking (current, former or never) were included in the multivariate models as a core set of covariates, regardless of their confounding effect.
Abdominal obesity was not associated, regardless of the other factors, and MS was not included in the multivariate models since it presented direct collinearity with the other included factors (Table 3).
Additionally, multivariate analysis was repeated by including all clinically relevant covariates, regardless of their significance in the univariate analysis, and CVVHF and all varieties were forced into multivariate models.
