Dichotomization of outcomes

• No distinction between death and chronic DoC

• Results may be driven by effects on mortality, prone to bias of decisions to withdraw life-sustaining therapy

Use of ordinal analysis methods, to maximize information extraction and minimize the impact of death as an extreme category in analyses

Account for patients’ and caregivers’ perspectives

Explore the potential of ordinal analyses to enhance statistical power in chronic DoC research

Simulation studies using registry or aggregate data

Observational data suffer from allocation bias due to the unbalance of important covariates between groups

Propensity scores not explored in DoC research, possibly due to a lack of multi-variable prognostic models tailored specifically for DoC

Integrate existing aetiology-based models, models for general ICU patients and models based on previous comorbidities, into new multi-variable models for DoC

Simulation studies on historical cohorts prior to the development of prospective data

Instrumental variable analysis to estimate unmeasured confounding