Background The Berg Balance Scale (BBS) is often used in clinical practice to predict falls in the older adults. However, there is no consensus in research regarding its ability to predict falls. Objective To verify whether the BBS can predict falls risk in older adults. Data source Manual and electronic searches (Medline, EMBASE, CINAHL, Ageline, Lilacs, Web of Science, Cochrane Library and PEDro) were conducted using blocks of words (older adults, falls, BBS, study design) and their synonyms, with no language restrictions and published since 1989. Study selection criteria Prognostic studies or clinical trials were used to assess the BBS and falls history. Data extraction and data synthesis In this narrative synthesis, two independent assessors extracted data from articles and a third reviewer provided consensus, in case of disagreement. The methodological quality was assessed using the Quality In Prognosis Studies tool. Results 1047 studies were found and 8 studies were included in this review. The mean BBS score was high, regardless of the history of falls. Three studies presented cut-off scores for BBS, ranging from 45 to 51 points. Two studies reported a difference in the BBS score between fallers and non-fallers. Studies presented low to moderate risk of bias. Limitations Unfeasible to conduct a meta-analysis due the heterogeneity of included studies. Conclusion The evidence to support the use of BBS to predict falls is insufficient, and should not be used alone to determine the risk of falling in older adults. Systematic review registration number PROSPERO CRD42016032309.