gms | German Medical Science

Objective: A high risk of falling after stroke has been reported, both during hospital stay and after patients have been discharged [1], [2]. In order to assist with fall prevention, it is necessary to identify the risk factors for falling and subsequently to take counter-measures against these. Many of the methods that are used by medical institutions to identify patients at risk of falling are, however, based on personal experience, subjective observations and mental evaluations, and there have been only a few reports that have focused on specific investigational methods, standardized judgment criteria for fall prediction and the establishment of an evaluation index [3]. The objective of this study was to identify risk factors for falling in hemiplegic stroke patients so as to develop a predictive model.

Material/Methods: In total, 72 stroke patients (42 males, 30 females; age 67.6±10.3 years) were included in the study. All falls were reported by medical staff using a dedicated fall report and this formed the basis on which the subjects were allocated into either a faller (occasional and repeat) or non-faller group. The following were extracted from the inpatients’ hospital records: occurrence of falls during hospital stay; age; gender; stroke type; time from stroke onset; length of hospital stay (LOS); the affected side of the body (right, left or bilateral); FIM on admission and discharge; BBS on admission; and MMSE. The relationships between falls and these patient characteristics were investigated. Fallers and non-fallers were compared to detect mean differences between the two groups for continuous variables using a Student’s t-test and the χ2 test. The variables that achieved statistical significance were then included in a multivariate logistic regression analysis. Pearson correlation coefficients were calculated to determine the strength of the relationships between variables. Sensitivity and specificity in predicting fall status were calculated. The ability to predict the outcome of falling in stroke patients was assessed by the area under the receiver operating characteristic (ROC) curve. Statistical evaluation of the data was carried out using the SPSS® statistical package (SPSS Inc., Chicago, IL, USA) for Windows^®. A P-value of <0.05 was considered to be statistically significant with the exception of the entry probability for logistic analysis.

Results: The characteristics of the stroke patients and statistical comparisons between the faller and non-faller groups are shown in Table 1 [Tab. 1]. Age and LOS were significantly higher in the faller group compared with the non-faller group, while significantly lower values were recorded in the faller group compared with the non-faller group for time from stroke onset, total FIM on admission and discharge, BBS on admission and MMSE The Pearson correlation coefficients for the variables (Table 2 [Tab. 2]) show a strong correlation between BBS on admission and total and motor FIM on admission and discharge Age, time from stroke onset, LOS, cognitive FIM on admission and discharge, BBS on admission, and MMSE were subjected to logistic regression analysis using the occurrence of falls as a dependent variable. To prevent multicollinearity, the strongly correlated total and motor FIM on admission and discharge were excluded, and the factor selected was BBS on admission. Based on the calculated OR, BBS on admission was found to be significantly related to falls (Table 2 [Tab. 2]). The ROC curve analysis (Figure 1 [Fig. 1]) showed that BBS on admission best differentiated subjects who had fallen from those who had not fallen at a threshold score of 29 points (sensitivity 80%; specificity 78%) and area under the curve of 0.81.

Conclusion: The numerical presentation of predicted fall risk and the establishment of a practical index for use in clinical practice were investigated in stroke patients. With stroke patients in the convalescence to maintenance stage, balance on admission was closely related to the risk of falling.