Characteristic | HR (95% CI) | p Value |
Model 1: baseline variables | ||
Age at ARDS diagnosis, per year ≤40 years | 0.99 (0.95 to 1.04) | 0.79 |
Age at ARDS diagnosis, per year >40 years | 0.97 (0.93 to 1.02) | 0.28 |
Charlson Comorbidity Index, per point | 0.75 (0.56 to 0.99) | 0.05 |
Functional Comorbidity Index, per point | 0.93 (0.72 to 1.20) | 0.55 |
Model 2: ICU and discharge variables | ||
Mechanical ventilation, per day ≤5 days | 0.66 (0.54 to 0.81) | <0.001 |
Mechanical ventilation, per day >5 days | 1.02 (0.99 to 1.05) | 0.22 |
Discharge to rehabilitation or other healthcare facility | 0.41 (0.21 to 0.78) | 0.01 |
Model 3: final multivariable model | ||
Charlson Comorbidity Index, per point | 0.77 (0.59 to 0.99) | 0.04 |
Mechanical ventilation, per day ≤5 days | 0.67 (0.55 to 0.82) | <0.001 |
Mechanical ventilation, per day >5 days | 1.02 (0.99 to 1.05) | 0.20 |
Discharge to rehabilitation or other healthcare facility | 0.49 (0.26 to 0.93) | 0.03 |
*HR calculated using Fine and Grey regression models, with a HR <1 indicating a slower time to return to work. To build multivariable models, covariates having univariable associations at p≤0.20 with the primary outcome (online supplementary eTable 1) were included in multivariable models. To prevent overfitting of the multivariable regression models, submodels were created, involving baseline patient variables (model 1) and ICU and hospital discharge variables (model 2). In these sub-models, multicollinearity was observed between APACHE II score, mean daily SOFA score, duration of mechanical ventilation and hospital LOS and addressed by omitting the non-significant variables (APACHE II, SOFA and hospital LOS) from the final model. Remaining variables from models 1 and 2 with an association of p≤0.20 were subsequently included in the final multivariable model (model 3).
APACHE, Acute Physiology and Chronic Health Evaluation; ARDS, acute respiratory distress syndrome; LOS, length of stay; SOFA, Sequential Organ Failure Assessment.