Premium
Predicting Falls Using the Stroke Assessment of Fall Risk Tool
Author(s) -
Yang Christine,
Ghaedi Bahareh,
Campbell T. Mark,
Rutkowski Nicole,
Finestone Hillel
Publication year - 2021
Publication title -
pmandr
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.617
H-Index - 66
eISSN - 1934-1563
pISSN - 1934-1482
DOI - 10.1002/pmrj.12434
Subject(s) - medicine , stroke (engine) , odds ratio , receiver operating characteristic , logistic regression , rehabilitation , physical therapy , prospective cohort study , poison control , predictive value of tests , population , physical medicine and rehabilitation , emergency medicine , mechanical engineering , environmental health , engineering
Background Falls in the inpatient stroke population are common, resulting in increased morbidity and slow rehabilitation progress. Falls may result from stroke‐specific neurologic deficits; however, assessment of these deficits is lacking in many fall screening tools. Objective To compare the ability to predict falls of the Stroke Assessment of Fall Risk (SAFR) tool, which includes items related to stroke‐specific neurologic deficits, and the commonly used Morse Fall Scale, which does not include these items. Design Prospective cohort study. Setting Inpatient tertiary stroke rehabilitation unit. Participants Patients (N = 220) with acute stroke. Main Outcome Measures Falls were captured by the medical records from January 2017 to September 2018. Logistic regression analysis evaluated both screening tools for predicting falls by calculating sensitivity, specificity, area under the receiver operating characteristic (AUC‐ROC) curve, and odds ratio (OR). We compared SAFR and Morse mean scores between fallers and non‐fallers using t ‐tests. Results Forty‐eight (21.8%) patients experienced ≥1 fall. SAFR, but not Morse, scores showed a statistically significant difference between fallers and non‐fallers ( P = .001 vs P = .24, respectively). Higher SAFR score was associated with higher odds of falls (OR 1.36, 95% CI [1.12, 1.64]), whereas Morse was not (OR 1.04, 95% CI [0.97, 1.12]). SAFR showed a statistically significant difference in hemi‐neglect between fallers and non‐fallers ( P = .03). Sensitivity and specificity of SAFR were 47.9% and 76.7%, vs 45.8% and 68.0% for Morse, respectively. SAFR positive predictive value and negative predictive value were 36.5% and 84.1%, respectively, similar to Morse (28.6% and 81.8%). The AUC‐ROC was 0.65 for SAFR and 0.56 for Morse. Conclusions SAFR was significantly associated with fall risk and had better discrimination between fallers and non‐fallers than Morse. The neurologic‐specific hemi‐neglect component of SAFR, a component not present on the Morse, was a fall risk factor. Further research evaluating the predictive value of fall scales that include neurologic deficits is needed.