A Better Rule Than “More Is Better”
The central question in fall prevention for community-dwelling older adults is not whether comprehensive programs are good or bad. It is when breadth produces enough delivered, patient-important benefit to justify the extra assessment, coordination, burden, and uncertainty. The evidence supports a precise answer: more comprehensive prevention is not always more effective because a wider risk map creates more possible pathways to benefit, but also more interfaces at which treatment can be misclassified, delayed, declined, or incompletely measured. A component earns its place by surviving that whole chain.
The conclusion is about alignment, not size
The first alignment is between architecture and mechanism. A single-component exercise program, a multicomponent package offered to everyone, a needs-based multifactorial program, and a mixed program that combines assessment with universal treatment make different promises. The needs-based program promises better matching: identify risks, then route each person to relevant components. The mixed program promises matching plus a baseline intervention for everyone. A universal package promises broad coverage without relying on a detailed allocation decision.
None of those promises is self-fulfilling. Assessment is a routing mechanism, not the treatment itself. A universal component is not automatically relevant to every participant. A package can be comprehensive in what it asks about and incomplete in what it changes. The architecture matters because it determines who owns the next decision and how many handoffs stand between a detected risk and a completed treatment.
The second alignment is between mechanism and outcome. The USPSTF evidence review found that exercise was associated with reductions in the rate of falls, the risk of having at least one fall, and the rate of injurious falls. It did not find a statistically significant reduction in the individual risk of an injurious fall or in fall-related fractures. Multifactorial interventions reduced the rate of falls but did not show the same consistent pattern for fallers, injurious falls, or fractures. The conclusion is not that exercise prevents every serious consequence or that multifactorial care has no value. It is that the evidence changes when the outcome changes.
This distinction is central to the apparent paradox. A program can reduce frequent events without demonstrating a reduction in rare, severe events. A fall count can improve while the number of people who fall changes little. A broad program can address more risks without showing that it prevents more fractures. If a claim uses a reduction in falls to imply protection against loss of independence, it has crossed an evidence boundary.
The third alignment is between effect estimate and certainty. The Canadian network review expanded the comparison to 219 randomized trials, 167,864 participants, 59 intervention nodes, and eight outcome networks. That broader design found the clearest signals around supervised, long-duration balance/resistance exercise, group tai chi, selected education or cognitive-behavioral therapy, whole-body vibration, and some mixed packages. But only about half of the nodes in each network had at least low-certainty evidence for benefit. For people who fell, 18 of 57 nodes had moderate-certainty evidence; many non-fall outcomes had very-low-certainty evidence because data were sparse.
A larger network therefore improves the map without making every road equally reliable. Its indirect comparisons depend on transitivity and coherence: studies must be comparable in relevant effect modifiers, and direct and indirect evidence must not materially disagree. The network’s moderate certainty is also outcome-specific. A program can have moderate-certainty evidence for reducing fallers and little or very-low-certainty evidence for fractures or long-term-care admission. This is why the network review did not treat a surface ranking as a universal league table.