The 3-m Backwards Walk and Retrospective Falls: Diagnostic Accuracy of a Novel Clinical Measure

Several measures of fall risk have been previously developed and include forward walking, turning, and stepping motions. However, recent research has demonstrated that backwards walking is more sensitive at identifying age-related changes in mobility and balance compared with forward walking. No clinical test of backwards walking currently exists. Therefore, this article describes a novel clinical test of backwards walking, the 3-m backwards walk (3MBW), and assessed whether it was associated with 1-year retrospective falls in a population of healthy older adults. Diagnostic accuracy of the 3MBW was calculated at different cutoff points and compared with existing clinical tests.

This study was a retrospective cohort study including residents of a retirement community without a history of neurological deficits. Demographics, medical history, and falls in the past year were collected, and clinical tests included the 3MBW and the Timed Up and Go (TUG), the 5 times sit-to-stand, and the 4-square step test. Frequency distributions and t tests compared baseline characteristics of people who reported falling with people who did not. Diagnostic accuracy (sensitivity and specificity) was calculated for a series of cutoffs for the 3MBW, the TUG (≥8, 10, and 13.5 seconds), 5 times sit-to-stand (≥12 and ≥15 seconds), and 4-step square test (>15 seconds). Receiver operating curve analyses were used to define 3MBW optimal cutoffs, and the difference between the overall area under the curve (AUC) was statistically tested. SPSS 24.0 and MedCalc 17.1 were used for all analyses.

Fifty-nine adults with a mean (SD) age of 71.5 (7.6) years participated, with 25 people reporting falls in the past year. The mean (SD) time for the 3MBW was 4.0 (2.1) seconds. People who fell had a significantly slower 3MBW time (4.8 vs 3.5 seconds for people who did not fall, P = .015), but not a significantly slower 4-step square test (9.5 vs 8.1 seconds, P = .056), TUG (9.3 vs 8.0 seconds, P = .077), and 5 times sit-to-stand (12.5 vs 10.3 seconds, P = .121). The highest overall AUC for any measure was for the 3MBW at 3.5 seconds (0.707, 95% confidence interval = 0.570-0.821; sensitivity = 74%, specificity = 61%), which was significantly higher than the TUG at 8 seconds (AUC = 0.560, P = .023) and 13.5 seconds (AUC = 0.528, P = .011), the 4-step square test (AUC = 0.522, P = .004), but not significantly higher than the TUG at 10 seconds (P = .098) and the 5 times sit-to-stand at 12 (P = .092) or 15 seconds (P = .276). On the 3MBW, more than 75% of people who were faster than 3.0 seconds did not report any falls, and 94% of people who did not report falling were faster than 4.5 seconds. Of the people who were slower than 4.5 seconds, 81% reported falling.

In a study of healthy older adults, the 3MBW demonstrated similar or better diagnostic accuracy for falls in the past year than most commonly used measures. People walking faster than 3.0 seconds on the 3MBW were unlikely to have reported falling, whereas people slower than 4.5 seconds were very likely to have reported falling. Further validation of the 3MBW in prospective studies, larger samples, and clinical populations is recommended.