The Gale readResearch

A chart prompt for exercise trimmed older adults' falls by 9%

A randomized trial across 22 primary care practices found that an exercise-recommendation prompt in the electronic record lowered coded falls from 21.2 to 19.3 per 100 patient-years, and left mortality unchanged.

By Gale Staff · October 20, 2026 · JAMA Network Open

Not reviewed by a clinician.

Illustration: Brain health & aging

The short answer

An exercise-prescribing prompt built into the doctor's electronic record modestly reduced falls among older adults at risk, from 21.2 to 19.3 per 100 patient-years, a rate ratio of 0.91 (95% CI 0.88-0.95). Fall-related injuries fell less, from 16.7 to 15.7 per 100 patient-years (rate ratio 0.94, 95% CI 0.90-0.97), and mortality did not differ. The trial was randomized by practice and followed patients for 12 to 23 months, but falls were counted from diagnosis codes in the record, so a fall that never reached a chart is invisible to it, and only 22.1% of intervention patients actually received an exercise prescription.

Who to see

Geriatrician

An exercise prompt in the doctor's record modestly cut falls among older adults at risk; a geriatrician can check your fall risk and talk through exercise.

Find care →

The visit after the screening questions

Picture an older woman at the end of a routine primary care visit who has just answered yes to a short screen about falling. The authors' starting point is that exercise lowers fall risk but is not often put into practice in primary care, so a yes on a screen does not always become a specific plan with exercises matched to the patient.

That gap is what a randomized trial in JAMA Network Open, published online on 16 October 2026 by Nancy K. Latham and colleagues, set out to close. The question it asked is practical rather than biological: if the doctor's electronic record itself suggests tailored exercise for older patients with a positive fall screen, do fewer of them go on to fall?

What the researchers did

The study was a matched-pair cluster randomized clinical trial. The unit that was randomized was the primary care practice, not the patient: 22 practices in an integrated health system in the United States were paired and one practice in each pair was assigned to the new tool. It ran from 9 June 2023 to 8 May 2025, with 12 to 23 months of follow-up. Adults aged 65 or older who had a positive result on a 3-item fall-risk screen at a primary care visit were included.

The intervention was decision support inside the electronic health record, called eSTEPS in the paper. It offered tailored exercise recommendations to the clinician, an optional referral to physical therapy, and materials for the patient in the form of handouts, a website and an app. It is a software prompt and a set of materials, not a drug. The registration identifier is NCT04993781.

The primary outcome was the rate of patients with one or more falls, identified from diagnostic codes in the electronic record (ICD-10). Secondary outcomes were fall-related injuries, identified with validated coding algorithms, and mortality.

What they found

The analysis covered 11,945 participants with a mean age of 77.05 years, of whom 8,044 (67.3%) were female. There were 5,681 patients (47.6%) in intervention practices and 6,264 (52.4%) in control practices.

Coded falls, including falls with injury, occurred at 19.3 per 100 patient-years (95% CI 18.6-20.1) in intervention practices and 21.2 per 100 patient-years (95% CI 20.2-22.1) in control practices. The rate ratio was 0.91 (95% CI 0.88-0.95; P<.001). In absolute terms the gap is 1.9 falls per 100 patient-years, which is this story's subtraction rather than a figure the abstract prints. Put another way, among 100 comparable patients followed for a year, about two fewer would be expected to have a coded fall, if the difference reflects the tool.

Fall-related injuries followed the same direction with a smaller gap: 15.7 per 100 patient-years (95% CI 15.0-16.5) in intervention practices against 16.7 (95% CI 16.1-17.4) in control practices, a rate ratio of 0.94 (95% CI 0.90-0.97; P=.001). That is a difference of 1.0 per 100 patient-years, again by subtraction. The abstract reports that no differences were observed in mortality, and gives no mortality figures to weigh.

The authors call the reduction modest, and the numbers support that word. A 9% relative reduction in falls and a 6% relative reduction in injuries are real in the statistical sense, with intervals that stay clear of 1, and small enough that they will not be noticed by any one patient or family.

Why the uptake figures matter as much as the headline

The adoption numbers sit in the same paragraph of the abstract as the results, and they change how the main result is read. Exercise prescriptions through the tool were issued for 1,255 intervention patients, which is 22.1% of that group. In intervention practices, 215 clinicians (77.9%) used the decision support to deliver the exercise program at least once over the study period. The abstract does not state how many clinicians there were in total; 215 as 77.9% implies roughly 276, which is again a derived figure.

Two things follow from that. First, the comparison is by practice, so every patient in an intervention practice counts in the intervention rate whether or not a prescription was ever issued. A difference measured that way blends patients who got an exercise plan with patients who got nothing different at all, and it answers the question a health system asks, which is whether switching the tool on helps its population, rather than the question a patient asks, which is whether doing the exercises helps me. Second, the effect cannot be traced to exercise itself. The tool also offered physical therapy referral and patient materials, and the abstract does not separate what each part contributed or how many patients used them.

If most patients in the intervention group never received a prescription and the whole group still fell less, one reading is that the effect among those who did was larger than the average suggests. That is a possibility the abstract does not test, and it should be held as one.

The main caveat: falls counted by codes

Falls here are falls that someone recorded and coded. A patient who tumbles at home, is not hurt and tells no one does not appear in the electronic record, and neither does a fall that was treated without a fall diagnosis attached. If the prompt changed how often clinicians asked about or coded falls, the measured rate could move without the true rate moving. In a practice where clinicians are reminded about falls, that could push the count either up or down, and the abstract does not say which. The fall-related injuries outcome used validated coding algorithms, and it moved in the same direction.

Randomizing practices in matched pairs is the strength of the design. With only 22 practices, though, differences between practices can still matter, and the abstract does not describe how well the pairs were balanced. Randomizing at the practice level also cannot blind the clinicians, and the abstract does not say whether anything was done to keep the coding the same in both arms.

Why it matters

Falls are a leading cause of injury in older adults, and the authors open by noting that exercise lowers fall risk but is not often put into practice in primary care. This trial tests a nudge to the clinician inside a tool they already use, and it finds a small benefit across thousands of patients in routine care. A difference of about two falls per 100 patient-years across a large clinic population adds up in absolute numbers. The authors conclude that the results support scalable, workflow-integrated fall prevention. The abstract does not report costs, clinician workload or patient experience, so whether the gain is worth the effort is left open.

What this study can't tell you

  • Whether the benefit comes from exercise: the tool bundled exercise recommendations, optional physical therapy referral and patient handouts, a website and an app, and the abstract does not separate their contributions.
  • How much benefit a patient who actually received an exercise prescription had: only 1,255 of 5,681 intervention patients (22.1%) were issued one, and the comparison counts everyone in an intervention practice.
  • How many falls really occurred: falls were identified by ICD-10 diagnostic codes in the electronic record, so falls that were never reported or never coded are missed, and a change in how often falls were asked about or coded could shift the count.
  • Whether the result carries over to health systems that differ from this one: the 22 practices belonged to a single integrated US health system, and the abstract does not describe its patients, clinicians or electronic record beyond that.
  • What happens beyond 12 to 23 months: follow-up ended there, and the abstract does not say whether the gap between groups widened, held or faded.
  • What the tool cost or what it asked of clinicians: the abstract reports no costs, no time-per-visit figures and no patient or clinician experience measures.
  • Whether any harm offset the benefit: the abstract gives no mortality numbers beyond reporting no difference, and does not describe adverse events from the exercise advice.

The Gale read

This is a trial that asks a modest question and gets a modest, believable answer. Randomizing whole practices in matched pairs suits an intervention that lives in clinic workflow, and the headline numbers are reported the way a careful paper reports them: rates per 100 patient-years with intervals, a rate ratio of 0.91 for falls and 0.94 for injuries, and no claim on mortality. The size of the effect deserves as much weight as its direction. About two fewer coded falls per 100 patient-years is a population effect, visible in a health system's totals and invisible to any single patient, and it should be read as the effect of switching a tool on rather than of a person doing exercises. The two weaknesses worth carrying away are connected: falls are counted from codes, and just over one in five intervention patients received a prescription, so what moved the numbers is not isolated. Nothing in the abstract suggests the result is false, and it also does not show which ingredient does the work. A fair summary is that it makes a case for trying prompts like this one at scale, with better measurement, rather than a case that they have been proven to prevent falls everywhere.

Common questions

does exercise reduce falls in older adults

In this trial, a prompt that recommended tailored exercise to primary care clinicians was followed by fewer coded falls: 19.3 against 21.2 per 100 patient-years, a rate ratio of 0.91 (95% CI 0.88-0.95). The trial tested a decision-support tool with exercise advice, optional physical therapy referral and patient materials, not exercise on its own, and only 22.1% of intervention patients were issued an exercise prescription, so it cannot say how much of the effect was the exercise.

can a doctor's electronic health record prompt prevent falls

It appears to help a little. Across 22 practices randomized in matched pairs, practices using the prompt had a 9% lower rate of coded falls (rate ratio 0.91) and a 6% lower rate of fall-related injuries (rate ratio 0.94) than control practices over 12 to 23 months. The abstract reports no mortality difference, and it counts falls from diagnosis codes, which miss falls that were never recorded.

how much did the fall prevention tool reduce falls in the trial

Falls fell from 21.2 to 19.3 per 100 patient-years, a gap of 1.9 by subtraction, and fall-related injuries fell from 16.7 to 15.7 per 100 patient-years, a gap of 1.0. The authors describe the reduction as modest. Because the comparison counts every patient in an intervention practice, whether or not a prescription was issued, the gap reflects how the tool worked across a clinic population rather than what happened to someone who followed the exercises.

how many older adults had a fall prevention exercise prescription in the trial

Exercise prescriptions through the tool were issued for 1,255 of the 5,681 patients in intervention practices, or 22.1%. In those practices 215 clinicians (77.9%) used the decision support at least once during the study. The abstract does not report how many of the patients who received a prescription went on to do the exercises.

Sources

  1. 1.Latham NK, Hommel E, Kiesman M, Lipsitz SR, Bates DW, Carroll DL, Frost AC, Garabedian PM, Loewenthal JV, Sainlaire M, Samal L, Song W, Wien M, Dykes PC. Clinical Decision Support for Fall Prevention in Primary Care. JAMA Network Open. 2026;9(10):e2639240. doi:10.1001/jamanetworkopen.2026.39240 link

Read the study →

1 sources, numbered by first appearance. General health information, not medical advice. AI-assisted editorial content — every citation independently verified. Editorial policy

More from Gale Staff

The Latest Health Research →

Published October 20, 2026 · Gale Staff

All Gale Staff analyses →