Eighteen studies were included in the review. Sixteen studies (17 cohorts) gave data on a per patient basis and were included in the quantitative meta-analysis (11,378 patients, range 44 to 5,489). The overall quality of the included studies was considered to be moderate to good. Two of the 18 studies were subject to spectrum bias, seven did not define a fall, 16 did not report blinding the assessors of falls and two studies did not report on withdrawals.
Overall pooled sensitivity was 67% (95% CI 52 to 80) and specificity was 57% (95% CI 45 to 69). Positive likelihood ratio was 1.58 (95% CI 1.34 to 1.86) and negative likelihood ratio was 0.57 (95% CI 0.43 to 0.75). In the SROC analyses, 95% confidence and predictive regions were broad; this indicated heterogeneity between studies and reduced the precision of the pooled estimate.
Further results were reported for analyses for low or high prevalence of falls, studies that avoided spectrum bias and studies that provided a definition of a fall.