OPTIMIZING FALLS RISK PREDICTION FOR INPATIENT STROKE REHABILITATION: A SECONDARY DATA ANALYSIS

S. Gangar¹, S. Sivakumaran¹, A. Anderson¹, K. Shaw¹, L. Estrela¹, H. Kwok^2,1, R. Davies^2,1, A. Tong², N.M. Salbach^1,3,4

Background: During inpatient rehabilitation, approximately one third of patients post-stroke experience a fall that may result in injury, fear of falling, and healthcare costs. Identifying individuals at risk for falls is essential to ensure timely implementation of falls prevention strategies.

Purpose: To compare sociodemographic and clinical characteristics of fallers and non-fallers; and evaluate the ability of the Berg Balance Scale (BBS) and Morse Falls Scale (MFS) alone and in combination to predict falls in an inpatient stroke rehabilitation setting.

Methods: A longitudinal study involving a secondary analysis of health record data from 818 patients with stroke (mean age: 70.3 years; 52.2% men) admitted to an urban, rehabilitation hospital was conducted. Sociodemographic and clinical characteristics of fallers (defined as having ≥1 fall during hospital stay) and non-fallers were compared. Sensitivity and specificity of BBS and MFS cut-points, alone and in combination, were calculated.

Results: Low admission BBS score and admission to a low-intensity stroke program, but not older age, male sex, English as the preferred language, higher admission MFS score, or type and side of stroke, were associated with falling (p<0.05). A BBS cut-point of 29 had optimal sensitivity (82.4%) and specificity (57.4%). A MFS cut-point of 30 had the highest sensitivity (73.2%), with a specificity of 31.4%. Combining cut-points of 45 (BBS) and 30 (MFS) yielded the highest sensitivity (74.1%) with a specificity of 42.7%.

Conclusion(s): Initial balance function and admission to low-intensity programs are important predictors of falls in inpatient stroke rehabilitation. A BBS cut-point of 29 alone appears superior to using the MFS alone or combined with the BBS to predict falls.

Implications: Findings should be interpreted considering the limitations of a secondary data analysis wherin the quality or accuracy of the data collected may be sub-optimal. Lack of consistency in the data documented may have introduced random or systematic errors and reduced power to reveal associations between predictor variables and falls occurrence and the precision of estimates of diagnostic test properties. Study findings identifying which sociodemographic and clinical characteristics predict falls during inpatient stroke rehabilitation should inform education of rehabilitation teams and future research. The ability to predict falls highlights the BBS as an optimal measure of balance for inpatient stroke rehabilitation settings. Although the BBS appears to be a more accurate predictor of falls than the MFS, timing, duration and training to administer a falls screening tool should be considered before implementation in falls prevention programs.  

Funding, acknowledgements: This work was supported by a personnel award from the Heart & Stroke Foundation.

Keywords: stroke rehabilitation, falls, prediction

Topic: Neurology: stroke

Did this work require ethics approval? Yes

Institution: University of Toronto

Committee: Health Sciences

Ethics number: 37179