The evidence record

We wrote that exercise programmes "reduce fall rates by about 23%... across 108 trials and 23,407 participants."

Both numbers are real, and both appear in the review’s abstract, but they do not belong together. The review covered 108 randomized trials and 23,407 participants in total. The 23% rate reduction is pooled from a smaller subset: 59 studies, 12,981 participants. Attaching the review’s full size to a subset’s effect size overstated the evidence behind the number by roughly double.

We wrote that adding cognitive training to strength-balance exercise "lowered fear of falling and fall rates compared with strength-balance training alone."

It did not. Fear of falling and fall rate both improved significantly within each group (p < 0.001 each), but the paper states plainly there was "no significant interaction between the groups for the falls parameter," and no interaction for fear of falling either.

Our structured data attached the effect size Hedges g = 0.22 directly to simultaneous physical-and-cognitive training.

That number belongs to any combined training against a control, across 41 comparisons in the network meta-analysis, not to simultaneous training specifically. Simultaneous training’s advantage is a ranking within that network, not a separate effect size.

We credited this trial with "some persistence at three-month follow-up."

The paper says the opposite about the motor-training effect: fear-of-falling scores in the combined group drifted from 29.7 back to 30.7 once training stopped (p = 0.031), and the authors write that the motor-training effect "did not last after the end of the treatment."

We described the whole sample as older adults "with concern about falling," and credited the trial with increasing "gait confidence."

The trial deliberately enrolled people both with and without concern about falling, and its findings split along exactly that line: people who arrived worried got the fear-of-falling benefit; people who did not saw their thinking-while-walking improve instead. Neither group got both. And the trial never measured "gait confidence"; it measured step length and gait line.

We described the review’s falls figure, 5 of 30 studies reporting a drop in fall frequency, as "17.8% fewer falls."

That percentage does not exist in the paper. Five-of-thirty is a count of studies that found a reduction, not a pooled effect size. We have corrected it to read as a count everywhere it appears on this site.

We carried "$50 billion" as the annual health care cost of non-fatal falls among older adults.

That is the CDC’s 2015 figure. The CDC’s current figure is $80 billion.