Predicting the chance of weaning dysphagic stroke patients from enteral nutrition: a multivariate logistic modelling study.

Aim: The aim of this article was to develop a simple predictive model of dysphagia outcome for stroke patients. The study enrolled patients recovering from first-ever stroke (supratentorial lesions) staying in a long-term rehabilitation hospital. On admission, all patients were being fed via nasogastric or percutaneous endoscopic gastrostomy (PEG) tube.

Methods: Functional Independence Measure (FIMTM) scores were assessed on admission. FIM-motor and cognition score, age, days after onset on admission were set as explanatory variables. Target criteria were defined as dichotomous categories; completely oral feeding or any need for nutrition via tube feeding. Multivariate logistic analysis was performed on these data.

Results: Thirty patients were enrolled: age range was 50-94 (median 75) years; FIM-motor scores spanned 13 to 17 (median 13), FIM-cognition scores spanned 5 to 19 (median 9); and days from stroke onset to transfer to long-term rehabilitation care ranged from 15 to 64 (median 43.5) days. Of these patients, 12 were weaned back to oral feeding and 18 were not. Multivariate logistic regression modelling was successful on these data (P=0.0003, R2=0.518; Logit=0.770xFIM-motor+0.089xFIM-cognition 0.070xdays after onset 0.255xage+10.222). Estimated probability for return to oral feeding is nearly 10% when logit equals -2, 50% when logit equals 0, nearly 90% when logit equals 2.

Conclusions: A logit formula factoring in age, FIM scores, and days after stroke onset can readily predict oral feeding outcome. Further studies are needed to assess external validation to establish wide clinical applicability of this prediction model.