Background and purpose: To determine whether individual Berg Balance Scale (BBS) items or a group of items would have greater accuracy than the total BBS in classifying community-dwelling people with stroke with a history of multiple falls.
Methods: The subjects were 44 community-dwelling individuals with chronic stroke; 34 had one or no falls in the past six months, and 10 had multiple falls. Each BBS item was dichotomized at three points along the scoring scale of 0-4: between scores of 1 and 2, 2 and 3, and 3 and 4. Sensitivity (Sn), specificity (Sp), and positive (+LR) and negative (-LR) likelihood ratios were calculated for all items for each scoring dichotomy based on their accuracy in classifying subjects with a history of multiple falls. These findings were compared with the total BBS score where the cutoff score was derived from receiver operating characteristic curve analysis.
Results: Dichotomized point 3-4 for items B11 (turning 360 degrees), B12 (alternate foot on stool), B13 (tandem stance), and B14 (standing on one leg) all revealed Sn greater than 60%. B14 had the best Sn and Sp (0.90 and 0.50). Combining B11, B12, or B13 with B14 did not improve Sn. Total BBS receiver operating characteristic curve revealed a cutoff score of 52 (Sn = 90% and Sp = 41%).
Conclusion: Using selected items from the BBS may be more time efficient and accurate than the total BBS score for classifying people with chronic stroke living in the community with a history of multiple falls. Prospective study is needed to validate these findings relative to fall prediction.