Presenter of 1 Presentation
Prognostic Modeling and Algorithms for Arm Impairment
Abstract
Abstract Body
Prediction upper limb outcome is important for health-care profession in order to make realistic treatment goals, to select patients that benefit most of an evidence-based intervention and to determine adequate discharge policy and long-term management. For researchers, adequate prediction is provisional to select patients in stroke recovery and rehabilitation trials. Unfortunately, most of current prognostic models are based on classic multivariable regression analyses and did fail to take the time-dependency of clinical covariates into account early post stroke. In addition, most prognostic models did predict recovery at group level instead of individual level. In the current lecture, I will show two recently developed mixture and mixed-effects models that may predict, respectively, upper limb motor recovery following the Fugl-Meyer Upper Extremity score, and, upper limb capacity, following the Action Research Arm Test. Both dynamic recovery models will overcome the main hurdles in prognostic research. First, the models are able to take the time-dependency of clinical covariates into account. Second, both models are able to categorize patients in sub-groups that exhibit different progression rates which may range from no recovery to full recovery within the first 3 months after stroke. Finally, both models are able to predict subjects at an individual level, including the 95%-prediction intervals of uncertainty within the first 6 months post stroke. The derived prognostic algorithms may serve as a starting point for developing an ICT-infrastructure within stroke services allowing to calculate and visualize expected individual recovery profiles of patients with an upper limb motor impairment.