Abstract TP63: Predictors Of Rapid Swallow Recovery And Avoidance Of Percutaneous Endoscopic Gastrostomy Placement In Preparation For Inpatient Rehabilitation

Abstract

Background: Enteral tube feeding is required for adequate nutrition in post-stroke patients with dysphagia. Nasogastric tube (NGT) feeding is used short-term, while percutaneous endoscopic gastrostomy (PEG) is utilized when dysphagia persists. Furthermore, most post-acute care facilities, including inpatient rehabilitation (IR), do not accept patients with NGT. We assessed factors associated with quick recovery of swallowing function, to identify which patients may avoid unnecessary PEG placement during hospitalization and reduce associated complications. Methods: Using a prospectively collected registry of stroke patients admitted in our system of 5 IRs, we found 806 patients with stroke (1/2018 to 6/2019). At discharge from hospitalization or IR, 83 patients had either a PEG or NGT. We divided patients into two groups based on whether PEG was used for nutrition at discharge from IR or if dysphagia improved and they were able to resume a PO diet. We performed predictive modeling with RIDGE regression analysis using age, sex, marital status, initial NIHSS, infarct location, change in dysphagia outcome and screening scale (DOSS) at first two evaluations, and days hospitalized. Results: Median age was 65 (IQR 60-71) years and 28 (34%) were female. Twenty-eight patients (34%) required nutritional support upon IR discharge. Mean initial NIHSS for patients who passed swallow testing prior to IR discharge was 15 (7.3), similar to patients who failed swallow testing prior to discharge 13.7 (8.0) (p=0.50). Patients who passed swallow testing prior to IR discharge had comparable median length of hospitalization to those who did not (14.9 and 14.3, respectively, p=0.84). Predictive modeling indicated that age, sex, marital status, initial NIHSS, location of infarct, change in DOSS and length of hospitalization can predict passing swallow evaluation satisfactorily (AUC=0.673). Of the 12 patients with NGT at discharge from hospital, no patients required PEG placement. Conclusion: Our model shows that age, sex, marital status, initial NIHSS, location of infarct, change in DOSS and length of hospitalization can satisfactorily predict passing swallow evaluations. However, additional factors involved in the process will be explored to improve the model.