Fig. 3.

Adjusted associations between comorbidity groups and outcomes.

Differences in hospitalisation rates among comorbidity groups were investigated using negative binomial models to estimate incidence rate ratios (IRR) with 95% confidence intervals. Overall difference in time to death among comorbidity groups were investigated using Royston-Parmar-Lambert flexible parametric survival models to estimate hazard ratios with 95% confidence intervals. All models were adjusted for adjusting for age, gender, ethnicity, socioeconomic status, place of diagnosis (primary care of hospital), cardiovascular medications, smoking, alcohol, comorbidities, BMI, blood pressure, cholesterol and haemoglobin concentrations.