Falls After 65: What Actually Prevents Them

Few events in later life are priced as poorly in advance as a fall after 65. A pooled analysis of 22 cohorts of women and 17 of men who broke a hip found that in the first three months after the fracture, the risk of dying from any cause ran 5.75 times as high as in age- and sex-matched controls for women, and 7.95 times as high for men, and that the excess never fully returned to baseline in the years that followed. That is the rare, catastrophic end of the distribution. The common end is smaller and slower: a fall with no fracture at all, followed by a new caution on the stairs, a handrail gripped where it was not gripped before, a walk not taken. The person moves less, the legs weaken, the balance worsens, and the next fall becomes likelier. What goes wrong here seldom looks like a single bad day; it looks like a spiral that began with one.

The question worth settling before any of that happens is which of the things sold and recommended to prevent falls — the exercise classes, the grab bars, the vitamin D, the new glasses, the balance gadgets — have actually been tested, and which of them work.

The honest answer is unusually good news. Falling is one of the best-studied problems in all of aging research (the exercise question alone has more than a hundred randomized trials behind it), and the evidence sorts cleanly. A specific kind of exercise prevents falls with high certainty. Fixing the home works, but only for people already at elevated risk. A medication review targets some of the largest single risk factors on the list. Timely cataract surgery has a randomized trial behind it. And at least one popular intervention, the annual vitamin D mega-dose, made falling more likely in the trial that tested it properly.

One in three, every year

At least one in three community-dwelling adults over 65 falls each year; the figure opens the Cochrane review on the subject. Most falls do no lasting damage, but the ones that do are disproportionately expensive in the currency that matters, which is independence. The hip-fracture meta-analysis rewards reading in full, because the excess risk does not end when the bone heals. By its life-table estimates, a white woman who breaks a hip at 80 carries excess annual mortality, beyond women her age who did not fracture, of 8 percent at one year, 11 percent at two, 18 percent at five, and 22 percent at ten. For a man fracturing at 80 the excess runs 18, 22, 26, and 20 percent at the same marks. Men fare worse than women in the first years after fracture, and neither curve ever rejoins the unfractured population. That ledger still omits the quieter costs that follow even uninjured falls: fear of falling, activity restriction, and the slow surrender of confidence.

Walking speed turns out to be an unnervingly good summary of an older person's prospects. A pooled analysis of nine cohort studies, 34,485 adults 65 and older followed for 6 to 21 years, found that gait speed, combined with age and sex, predicted survival as accurately as a model that added chronic conditions, smoking, blood pressure, body-mass index, and hospitalization history. Every 0.1 meter per second of additional walking speed was associated with a 12 percent lower hazard of death (hazard ratio 0.88), and survival rose across the full observed range of speeds. At age 75, predicted ten-year survival ranged from 19 percent to 87 percent in men, and from 35 percent to 91 percent in women, across that range. Walking speed is not destiny. It is closer to a vital sign, and, unlike most vital signs, it is one that can be measured with a hallway and a phone.

Exercise works, but "exercise" is the operative word

The 2019 Cochrane review pooled 108 randomized trials with 23,407 community-dwelling participants (average age 76; 77 percent women). Exercise of any tested kind reduced the rate of falls by 23 percent compared with control (rate ratio 0.77, 95% CI 0.71–0.83) and the number of people who fell at all by 15 percent, both graded high-certainty evidence. On the review's illustrative baseline of 850 falls per 1,000 people per year, that 23 percent equals roughly 195 falls prevented per 1,000 people annually. Two further findings deserve more attention than they usually get. First, the benefit was not confined to those already at high risk: subgroup analyses found no evidence that the effect differed between trials that recruited people at elevated fall risk and trials that did not. Second, the downstream outcomes point the right way, with appropriate humility; exercise may reduce fall-related fractures by 27 percent (risk ratio 0.73) and falls requiring medical attention by 39 percent (risk ratio 0.61), both graded low-certainty because far fewer trials measured them. Reported harms, where trials tracked them, were predominantly non-serious and musculoskeletal, the ordinary currency of any training program.

The type of exercise mattered more than the headline suggests. Programs built on balance and functional training — standing with a narrowed base, weight shifts, stepping over objects, sit-to-stand work — cut the fall rate by 24 percent on their own, with high certainty. Programs that combined balance work with resistance training did better still: a 34 percent reduction in fall rate (moderate certainty). Tai chi reduced the number of people falling by 20 percent (high certainty for that outcome). The review is candid about what it could not show: for programs that were primarily resistance training alone, dance, or walking, the effect on falls was uncertain. Walking is excellent for many things, and its mortality curve is examined elsewhere on this site, but a daily stroll does not train the specific skill of not falling. Balance is trained the way anything is trained, by working near its edge.

The finding holds even at the frail end of the range. The oldest trial in the canon randomized 233 women aged 80 and over, in New Zealand general practices, to an individually tailored home program of strength and balance retraining, the program that became the Otago Exercise Programme, or to social visits. Over a year the exercise group had 88 falls to the control group's 152, a fall rate of 0.87 versus 1.34 per person-year, and the hazard of an injurious fall dropped 39 percent (relative hazard 0.61). Measured balance improved alongside, with a between-group difference of 0.43 on the trial's balance score at six months. These were not athletes; they were women in their ninth decade doing individually tailored strength and balance work at home, prescribed through their own general practices.

For the question of staying mobile at all, there is the LIFE trial, which randomized 1,635 sedentary adults aged 70 to 89, all with measurable physical limitation, to a structured moderate-intensity program of aerobic, resistance and flexibility work (twice weekly at a center, three to four times at home) or to health-education workshops. Over an average of 2.6 years, major mobility disability, meaning the loss of the ability to walk 400 meters, occurred in 30.1 percent of the exercise group versus 35.5 percent of the education group (hazard ratio 0.82). Persistent mobility disability fell further, 14.7 versus 19.8 percent (hazard ratio 0.72), a 28 percent lower hazard. Serious adverse events were common in both arms of this deliberately vulnerable population, reported by 49.4 percent of the exercise group and 45.7 percent of the education group, a difference that did not reach significance, which is itself informative: the risk of structured movement, even here, ran about even with the risk of workshops. The numbers are modest and the outcome is hard, in exactly the population most often assumed to be past helping.

Tai chi, tested head-to-head

Most fall-prevention trials compare something with nothing. A 2018 trial in JAMA Internal Medicine did the more useful thing and compared two credible programs directly. Researchers screened 1,147 community-dwelling adults across seven Oregon cities and enrolled 670 aged 70 and older (mean age 77.7; 65 percent women) who had fallen in the past year or had impaired mobility, randomizing them three ways: a therapeutically tailored tai ji quan program (Tai Ji Quan: Moving for Better Balance), a multimodal exercise program mixing balance, aerobics, strength and flexibility, or stretching, each delivered as two 60-minute classes a week for 24 weeks.

At six months the stretching group had accumulated 363 falls. The multimodal group had 218, a 40 percent lower rate (incidence rate ratio 0.60). The tai ji quan group had 152, a 58 percent lower rate than stretching (IRR 0.42) and 31 percent lower than the multimodal program itself (IRR 0.69). A secondary analysis of the same trial points at a mechanism: the tai ji quan training, uniquely among the three arms, reduced the cost of walking while doing something else at the same time, the so-called dual-task cost, and that reduction mediated the physical, cognitive and fall benefits measured at 12-month follow-up. A plausible reading, though not one the trial proves, is that everyday falls happen mid-distraction, in walking done while carrying, turning or talking, and that training which couples movement with attention rehearses something close to the real failure mode.

The home, for the right person

Grab bars, non-slip strips, better lighting, removed rugs: the 2023 Cochrane review of environmental interventions (22 trials, 8,463 participants, average age 78) delivers one of the most instructive splits in the literature. Overall, home fall-hazard reduction probably cuts the fall rate by 26 percent. Stratified by who received it, though, the effect resolves into two different realities. In people selected for elevated risk (a fall in the past year, recent hospitalization, needing help with daily activities) the reduction was 38 percent (rate ratio 0.62), high-certainty evidence. In people not selected for risk there was no evidence of benefit, a rate ratio of 1.05 (95% CI 0.96–1.16), also high certainty. Against the higher-risk group's baseline of 1,847 falls per 1,000 people per year, the intervention was worth an estimated 702 fewer falls; against an unselected baseline, none detectable.

The review is equally frank about what the home fixes did not do. Health-related quality of life probably saw little or no important difference, and fall-related fractures and hospitalizations showed little or no difference in the available low-certainty data. Education about hazards, delivered without the hands-on assessment, had evidence too thin to judge, and trials of home modifications aimed at task enablement, the ramps and rails installed for daily function rather than hazard removal, had simply not measured falls at all. The intervention that works is specific: a trained assessor, a walk-through, targeted fixes, and behavioral strategies agreed with the person who actually lives there.

The practical reading is that a home fall-hazard assessment of the kind the trials actually delivered — hazards identified, environmental safety adaptations made, behavioural strategies agreed — is one of the highest-value moves available to someone who has already fallen or is visibly unsteady, and close to a null intervention for a robust 66-year-old whose house merely contains rugs. Prevention budgets, like training loads, work best when they follow risk.

The medicine cabinet before the floor

Some of the largest single fall-risk multipliers in the literature are prescriptions. A systematic review and meta-analysis series on fall-risk-increasing drugs, pooling adjusted estimates from observational studies of adults 60 and older, reported the following odds ratios for falling: antidepressants 1.57 (SSRIs specifically 2.02), antipsychotics 1.54, benzodiazepines 1.42 (long-acting 1.81), opioids 1.60, antiepileptics 1.55, and polypharmacy itself, 1.75. The details resist easy narratives. SSRIs carried a higher pooled fall-risk estimate than the older tricyclics (2.02 versus 1.41), and short-acting benzodiazepines still came in at 1.27. The same group's analysis of other drug classes flagged long-term proton-pump-inhibitor use and opioid initiation as consistent signals, while NSAIDs (OR 1.09) and analgesics as a broad class (OR 1.42, confidence interval crossing 1) showed no significant association. The authors are appropriately blunt about heterogeneity: the pooled numbers summarize a large and differently designed study base, and stratifying by population and setting did not make the variation disappear.

Two cautions cut both ways here. These are associations from observational data; people on antidepressants differ from people not on them, the review authors note that prescription bias could not be accounted for, and no meta-analysis fully launders that. And yet the effect sizes rival anything on this page, and they attach to something uniquely revisable. None of it is a reason to stop a medication unilaterally, since abrupt discontinuation carries hazards of its own. It is, rather, a standing reason geriatric practice emphasizes the regular medication review: the complete list, in front of the prescriber, with the explicit question of whether each drug is still earning its place at this dose and this age.

Eyes: surgery helped; new glasses alone did not

Vision belongs on the list, but with more precision than a general reminder to have the eyes examined. In a randomized trial of 306 women over 70 with cataract, expedited first-eye surgery (about four weeks) versus a routine 12-month wait cut the rate of falling over the following year by 34 percent (rate ratio 0.66), and fractures fell from 8 percent to 3 percent. The visual mechanics were unambiguous: corrected binocular acuity improved by 0.25 logMAR units, and only 8 percent of the operated group saw worse than 6/12, versus 37 percent of controls. The gains also ran wider than eyesight; activity, anxiety, depression, confidence, visual disability and handicap all improved. One precision is worth keeping: the rate of falling fell, while the proportion who fell at least once did not (49 percent operated versus 45 percent control), so surgery made falls rarer and less consequential rather than abolishing them. In this trial the operation delivered within weeks rather than after a year was itself the fall-prevention intervention; the waiting time carried the fall cost.

The broader category supplies the contrast. In the environmental-interventions review, vision-improvement interventions as a group showed little or no effect on falls (rate ratio 1.12, low certainty), and the review flags an open question about whether people need extra advice or precautions while adapting to changed eye prescriptions. The lesson is characteristically unglamorous: fixing a disease that blurs vision prevents falls; new glasses, by themselves, have not been shown to.

The vitamin D cautionary tale

For years, vitamin D was the easiest fall-prevention advice to give. Then a placebo-controlled trial did it properly, at scale, and found the opposite. In 2,256 community-dwelling women 70 and older at elevated fracture risk, a single annual oral dose of 500,000 IU of cholecalciferol, a convenient once-a-year regimen designed to solve the adherence problem, produced more falls than placebo (83.4 versus 72.7 per 100 person-years; rate ratio 1.15) and more fractures (171 versus 135; rate ratio 1.26). The excess falling clustered tellingly: in a post hoc analysis, the rate ratio was 1.31 in the three months right after each dose, easing to 1.13 across the following nine (P = .02 for the difference between periods). Blood sampling in a 137-person substudy showed what the dose was doing. Median 25-hydroxyvitamin D started at 49 nmol/L, with fewer than 3 percent of sampled participants below 25 nmol/L, then spiked to roughly 120 nmol/L a month after dosing and was still elevated a year later.

The trial does not answer the separate question of correcting a measured deficiency; it barely tested one, since almost nobody enrolled was deficient. What it indicts is the mega-dose reflex, the wish for more of a good thing, less often, without measuring anything. On current randomized evidence, high-dose vitamin D is the one item on the standard fall-prevention list that has been shown capable of making the problem worse.

Measurement, done honestly

Two numbers are worth knowing, with their limits attached. The Timed Up and Go, in which a person rises from a chair, walks three meters, turns, returns and sits, flags elevated risk at 13.5 seconds or longer. A meta-analysis of its predictive value found decent specificity (0.74) but poor sensitivity (0.31) at that cutoff: a slow time means something, a fast time rules little out, and the authors conclude that it should not be used alone to identify high-risk individuals. Gait speed carries more information, and it carries it continuously; in the pooled cohorts, survival rose across the entire observed range of speeds, with each additional 0.1 meter per second counting. A companion tool on this site, the Movement Lab, offers timed versions of these and related tests, single-leg stance and chair rises among them, against published norms, on an ordinary floor, with a phone timer.

What the evidence cannot say yet

The list of open questions is short, and worth keeping in view. Whether resistance training alone, without a balance component, prevents falls is genuinely uncertain, as are dance and walking programs, and nobody has run an adequate trial of flexibility or endurance work against falls at all. The drug data are observational, and prescription bias cannot be fully removed: sicker people take more medicines and also fall more. The tai ji quan result, striking as it is, comes from one trial in one region with a mostly white study population; replication elsewhere would firm it up. And screening tools deserve their fine print read aloud: the TUG's own meta-analysts concluded it "should not be used in isolation." The pattern across all of it, though, is consistent enough to act on. The interventions that train or restore the systems that keep a body upright work, and the ones that merely gesture at the problem do not.

What a sensible year looks like

A sensible year, assembled from the trials rather than from enthusiasm, might run as follows. It holds, twice a week, a session that genuinely challenges balance — a tai ji quan or evidence-based balance class where one exists, a structured home program of the Otago type where one does not — with simple strength work attached. Once a year, the complete medication list goes in front of the prescriber; the observational fall data are one reason geriatric practice keeps insisting on that review. After a fall, or once walking feels unsteady, the structured home-hazard assessment earns its place, because that is the population in which it prevented a third of falls. Where a cataract is dimming an eye, the randomized evidence says the wait itself carries a fall cost, worth weighing with the ophthalmologist rather than defaulting to the queue. The vitamin D megadose stays off the list, and whether testing and correcting a low level makes sense is a clinician conversation the trial above does not settle. A timed walk belongs in the year somewhere too, not because the stopwatch prevents anything, but because it tells the truth early, while the spiral is still cheap to interrupt.

Related on Magellan: VO₂max after 50 · Protein, muscle and mTOR · The 10,000-step threshold.

Educational, not medical advice.