Risk Prediction of Cognitive Decline after Stroke.

BACKGROUND AND PURPOSE: Cognitive decline is one of the major outcomes after stroke. We have developed and evaluated a risk predictive tool of post-stroke cognitive decline and assessed its clinical utility. METHODS: In this population-based cohort, 4,783 patients with first-ever stroke from the South London Stroke Register (1995-2010) were included in developing the model. Cognitive impairment was measured using the Mini Mental State Examination (cut off 24/30) and the Abbreviated Mental Test (cut off 8/10) at 3-months and yearly thereafter. A penalised mixed-effects linear model was developed and temporal-validated in a new cohort consisted of 1,718 stroke register participants recruited from (2011-2018). Prediction errors on discrimination and calibration were assessed. The clinical utility of the model was evaluated using prognostic accuracy measurements and decision curve analysis. RESULTS: The overall predictive model showed good accuracy, with root mean squared error of 0.12 and R2 of 73%. Good prognostic accuracy for predicting severe cognitive decline was observed AUC: (88%, 95% CI [85-90]), (89.6%, 95% CI [86-92]), (87%, 95% CI [85-91]) at 3 months, one and 5 years respectively. Average predicted recovery patterns were analysed by age, stroke subtype, Glasgow-coma scale, and left-stroke and showed variability. DECISION: curve analysis showed an increased clinical benefit, particularly at threshold probabilities of above 15% for predictive risk of cognitive impairment. CONCLUSIONS: The derived prognostic model seems to accurately screen the risk of post-stroke cognitive decline. Such prediction could support the development of more tailored management evaluations and identify groups for further study and future trials.

Inverse problem approach to regularized regression models with application to predicting recovery after stroke.

Regression modelling is a powerful statistical tool often used in biomedical and clinical research. It could be formulated as an inverse problem that measures the discrepancy between the target outcome and the data produced by representation of the modelled predictors. This approach could simultaneously perform variable selection and coefficient estimation. We focus particularly on a linear regression issue, Y ∼ N ( X ß , σ I n ) , where ß ∈ R p is the parameter of interest and its components are the regression coefficients. The inverse problem finds an estimate for the parameter ß , which is mapped by the linear operator ( L : ß âŸ¶ X ß ) to the observed outcome data Y = X ß + ε . This problem could be conveyed by finding a solution in the affine subspace L - 1 ( Y ) . However, in the presence of collinearity, high-dimensional data and high conditioning number of the related covariance matrix, the solution may not be unique, so the introduction of prior information to reduce the subset L - 1 ( Y ) and regularize the inverse problem is needed. Informed by Huber's robust statistics framework, we propose an optimal regularizer to the regression problem. We compare results of the proposed method and other penalized regression regularization methods: ridge, lasso, adaptive-lasso and elastic-net under different strong hypothesis such as high conditioning number of the covariance matrix and high error amplitude, on both simulated and real data from the South London Stroke Register. The proposed approach can be extended to mixed regression models. Our inverse problem framework coupled with robust statistics methodology offer new insights in statistical regression and learning. It could open a new research development for model fitting and learning.