Fall Predictors In Multiple Sclerosis- An Exploratory Study

Background: Falls are a major health concern in persons with Multiple Sclerosis (PwMS). Due to its multifactorial nature, it is vital to investigate the prime determinants that distinguish high-risk fallers from those with lesser risk.

Objectives: This study aimed to identify the significant contributors from demographics, strength, balance, functional mobility, spasticity, self-report measures and cognition which can predict falls in PwMS. 

Methods: Retrospective data of 55 PwMS (mean[SD] age=56.2[12.2] yrs, EDSS= 5.2[1.5]) from the University of Utah Wellness Clinic was analyzed. The variables included were number of falls recalled in past year; gender; Expanded Disability Status Scale (EDSS); age; knee extensor strength; balance assessed by Berg balance test and functional reach; functional mobility measured by timed-up-go, twenty-five feet walk, stairs test, six-minute walk distance; spasticity by Modified Ashworth Scale (ASH-ave); self-report measures such as MS walking scale-12, Activities-specific Balance Confidence scale (ABC) and modified fatigue impact scale; and cognition determined by paced auditory serial addition test and Symbol Digit Modalities Test (SDMT). A stepwise multiple regression for model selection was performed followed by hierarchical binary logistic regression to examine the effectiveness of prediction in fallers versus non-fallers and in recurrent fallers compared with 0 or 1 fall. 

Results: Of fifty-five, 34 (61.8%) reported 156 falls. Of the fallers, 25 (73.5%) reported recurrent (>1) falls. All predictors accounted for 72.3% of the variance in number of falls. ABC, ASH-ave and SDMT, in the respective order, were found to be the prime predictors, R² =0.61, F (3, 34) = 17.86, p<0.0001, indicating that PwMS with high self-reported balance confidence, low spasticity and high cognition report less falls. The probability of discriminating a faller from non-faller was higher in people with higher ASH-ave, low SDMT and low ABC, χ² (3)=8.38, p<0.05, predictive accuracy (78.7%) and R²=0.23, with ABC as the most significant predictor. Additionally, ABC cut-off score of 60 improved the specificity (66.7%), sensitivity (87.5%) and model accuracy (80.9%). Similarly, the probability of determining a recurrent faller from others approached significance with higher ASH-ave, low SDMT and low ABC, χ² (3)=7.25, p>0.05, accuracy (70.2%) and R²=0.19. ABC cut-off score of 60 improved the specificity (75%), sensitivity (69.6%) and accuracy (72.3%).

Conclusions: Poor self-reported balance confidence, high spasticity and poor cognition significantly increased the number of falls and enhanced the probability of determining a faller from non-faller in PwMS. A similar trend was noted in recurrent fallers. Moreover, ABC emerged as a substantial predictor and a cut-off to enhance prediction has been suggested. These results can be effectively utilized for fall-risk screening and intervention purposes in PwMS.