Background: Dementia and institutionalization are known long-term complications of stroke, but the extent to which post-stroke disability – as captured by the modified Rankin Scale (mRS) in trials – predicts these outcomes remains unclear. Quantifying this relationship can inform prognostication and cost-effectiveness analyses.
Methods: In consecutive 3-month survivors of ischaemic stroke in a population-based prospective cohort (Oxford Vascular Study;2002-2014), we established post-stroke dementia (excluding pre-stroke dementia) and tracked institutionalization (admission to nursing or residential care home) through overlapping methods of cognitive testing and interview-based assessments of patients and carers, and ongoing searches of primary care and hospital records. We related 3-month mRS to 5-year dementia and institutionalization using age/sex-adjusted Cox and logistic regressions (repeated excluding premorbid mRS>2), and compared the area under the receiver operating characteristics curves (AUC) for logistic models versus age/sex alone.
Results: Among 1,425 survivors, 3-month mRS independently predicted 5-year dementia and institutionalization whether or not those with premorbid disability were excluded, with mRS of 3-5 carrying higher hazards of each outcome than mRS 0-2; aHR(dementia)=3.37, 95%CI 2.63-4.32, aHR(institutionalization)=6.33, 4.53-8.85, p<0.0001. However, each step up the mRS ladder was associated with a jump in hazard for each outcome; e.g. aHR(dementia) vs mRS=0 for mRS=1: 0.91, 95%CI 0.48-1.75, p=0.78; mRS=2: 1.56, 0.84-2.92, p=0.16; mRS=3: 2.86, 1.54-5.30, p=0.001; mRS=4: 4.61, 2.46-8.63, p<0.0001; mRS=5: 9.79, 5.13-18.7, p<0.0001. Logistic models including the mRS, particularly in its ordinal form, predicted 5-year dementia or institutionalization better than age/sex alone (e.g. AUC for dementia with only age/sex: 0.72, 95%CI 0.69-0.75; with 3-month mRS: 0.79, 0.76-0.82, p<0.0001).
Conclusions: 3-month mRS strongly predicts post-stroke dementia and institutionalization, both of which should be considered in cost-effectiveness analyses of acute treatments. The incremental rise in hazard with each step up the mRS ladder further supports using ordinal analysis of the mRS in trials.