Association Between Performance on an Interdisciplinary Stroke Assessment Battery and Falls in Patients With Acute Stroke in an Inpatient Rehabilitation Facility: A Retrospective Cohort Study

ObjectiveTo explore the association between demographic factors and functional performance measures of patients with acute stroke in an inpatient rehabilitation facility (IRF) and falls during the IRF stay and to quantify the diagnostic accuracy of functional outcome measures in identifying fallers.DesignRetrospective cohort study.SettingInpatient rehabilitation facility.ParticipantsIndividuals with acute stroke admitted to hospital-based IRF (N=139).InterventionsNot applicable.Main Outcome MeasuresOdds ratios were used to examine the relationship between fall frequency and functional outcome measures (National Institute of Stroke Scale, neglect [Item #11], Berg Balance Scale, Stroke Rehabilitation Assessment of Movement mobility and Stroke Rehabilitation Assessment of Movement lower extremity subscales [STREAM-LE], Montreal Cognitive Assessment, Dynamic Gait Index, and Stroke Impact Scale). Receiver operator characteristic analysis with area under the curve, sensitivity, specificity, and diagnostic odds ratio were used to assess the diagnostic accuracy of each functional outcome measure to distinguish patients who fell vs those who did not fall in the IRF.ResultsA total of 23 patients (16.2%) fell during the IRF hospitalization. Patients who did and did not fall did not differ in terms of age, sex, stroke type, or stroke location. Only the STREAM-LE was associated with falls (odds ratio, 0.93; 95% CI, 0.86-0.99). Area under the curve was 0.67 (95% CI, 0.51-0.82). With a positivity cutoff point of 12, sensitivity and specificity were 73.3% (95% CI, 54.6%-92.2%) and 50.0% (95% CI, 39.9%-59.2%), respectively. The diagnostic odds ratio was 3.4.ConclusionsThe STREAM-LE score at admission to IRF may identify patients with acute stroke who are more likely to fall during their stay. However, the search for measures with greater diagnostic accuracy should continue.