Patient-specific prediction of long-term outcomes will change stroke rehabilitation for the better

Stroke is the most common cause of neurological disability and yet our ability to predict long-term outcomes remains poor. The paper by Selles et al1 used upper limb outcomes from 450 patients with first-time ischaemic stroke to take a refreshingly different approach to the prediction problem. First, allowing repeated clinical measures to contribute to the prediction acknowledges what most clinicians already know—that rate of clinical change is helpful in prognostication. Second, rather than predict outcome at a single future time point, they have created likely recovery trajectories (with CIs) for individual patients. Third, access to this predictive model is freely available online, so that stroke services around the world can more accurately begin to make predictions of individual recovery. Although the current approach concerns upper limb recovery, the principle should apply to all domains.

Current methodological approaches to prediction rely on linear and logistical regression models requiring a single…

Read article at journal's website

Related Articles


Your email address will not be published.