Evaluating the effectiveness of a telemedicine-based STEADI implementation in primary care on fall outcomes: the STEADI options randomized controlled trial

Abstract

Background and Objectives:

Falls are a leading cause of injury and death among older adults (65 years and above). The Centers for Disease Control and Prevention’s Stopping Elderly Accident, Deaths, and Injuries (STEADI) initiative aims to reduce falls by screening to find at-risk individuals, evaluation, and then addressing modifiable risk factors. This study evaluated the effectiveness of implementing a STEADI-based multi-component fall prevention approach via telemedicine in primary care.

Research Design and Methods:

A randomized controlled quality improvement trial (STEADI Options) was conducted during the COVID-19 pandemic (July 2020–December 2021). Primary care professionals (PCPs) were randomized to either standard-of-care (SOC) or intervention. The intervention included a fall risk assessment by a research nurse, who shared STEADI-based recommendations with PCPs. Recommended interventions were administered at the discretion of PCPs. Outcomes included self-reported falls (primary), medically treated falls, fall-related service utilization (physical therapy, occupational therapy, eye care, podiatry), and medication changes, assessed over one year. Analyses followed an intent-to-treat approach.

Results:

Among 660 participants (353 SOC; 307 intervention), the cohort was predominantly female (67.7%), aged 65–79 years (73.6%), and non-Hispanic White (56.1%). Intervention participants had higher average Stay Independent Screener (SIS) scores (6.3 vs. 5.2; p < 0.001). Overall, 5.5% reported a fall in the past 12 months. No significant differences were observed between groups in fall-related outcomes (e.g., medically treated falls odds ratio [adjusted for SIS]: 0.741; 95% CI: 0.411–1.336).

Discussion and Implications:

Although no significant differences were found, limited recruitment, low baseline fall rates, and unmeasured adherence may have impacted results. Further research is needed to evaluate the feasibility and effectiveness of telemedicine-based STEADI interventions in primary care.

In 2023, 15 million or one in four older adults (aged 65 years and above) in the United States reported having fallen (Centers for Disease Control and Prevention, 2023a). Each year, falls among older adults result in approximately 3.6 million emergency department (ED) visits, 1.2 million hospitalizations, and about 41,000 deaths (Centers for Disease Control and Prevention, 2023b, 2023c). Healthcare spending on nonfatal older adult falls was estimated at $80 billion for 2020 with $53 billion paid for by Medicare alone (Haddad et al., 2024).

To address the burden of falls, the Centers for Disease Control and Prevention developed a clinical fall prevention initiative, Stopping Elderly Accident, Deaths, and Injuries (STEADI), which offers healthcare providers resources to implement clinical fall prevention in their outpatient practice. The STEADI algorithm is available online (www.cdc.gov/steadi). STEADI includes three core elements: (a) Screening older adults annually to identify those at increased risk of falling using Three Key Questions (3KQ) (“Fallen in the past year,” “Feels unsteady when walking or standing,” or “Worries about falling”) or a 12-item Stay Independent Screener (SIS) with statements similar to 3KQ embedded within (“I have fallen in the past year,” “I steady myself by holding on to furniture while walking at home,” and “I am worried about falling”); (b) Assessing older adults at increased risk of falls to identify modifiable risk factors (e.g., use of high-risk medications; impaired gait, strength, and balance; decreased visual acuity); and (c) Intervening by recommending effective strategies for each identified risk factor to reduce fall risk (e.g., manage medications, referral to physical therapy or community-based exercise programs such as Tai Chi, referral to ophthalmology or optometry) (Eckstrom et al., 2021).

While STEADI is a multifactorial fall prevention intervention and there is evidence that multifactorial interventions reduce falls by up to 24% (Gillespie et al., 2012; Hopewell et al., 2018), there have been limited large-scale efforts to evaluate the effectiveness of STEADI-based fall prevention in primary care settings. One study found that implementing the STEADI initiative in primary care reduced fall-related hospitalizations among older adults at risk who received a fall plan of care but did not measure rates of falls or fall injuries (Johnston et al., 2019). In a study comparing a pharmacist-led abbreviated STEADI medication review to usual care, the intervention identified medication risks but showed no significant impact on medication prescribing or fall outcomes (Blalock et al., 2020). Another study that used the RE-AIM framework to evaluate the systemwide implementation of the STEADI fall-risk screening initiative across 34 outpatient clinics, found high initial reach and adoption but declines over time and variability across clinics and professions (Vincenzo et al., 2024). Beyond STEADI, the Strategies to Reduce Injuries and Develop Confidence in Elders (STRIDE) trial tested a nurse-led multifactorial fall prevention approach in 86 primary care practices. It reduced self-reported fall injuries but showed no significant effect on serious fall injuries, hospitalizations, or mortality (Ganz et al., 2022).

The primary objective of this project was to implement STEADI-based fall prevention within primary care offices of a large health system, integrate STEADI into the health system’s electronic health record (EHR), and evaluate the impact of STEADI on older adult falls and fall injuries. In collaboration with NORC at the University of Chicago and the Emory Healthcare system, the STEADI Options randomized quality improvement trial (STEADI Options) was developed to examine the impact of multifactorial STEADI-based fall prevention on health outcomes compared to standard of care (SOC). Due to the COVID-19 pandemic and resulting low recruitment rates, the protocol was adapted in 2020 to allow for a telemedicine-based delivery of the intervention (Rein et al., 2024). During the COVID-19 pandemic, telemedicine rapidly expanded as a means of delivering healthcare services, enabling continued access to care while minimizing virus exposure risks for both patients and providers (Koonin et al., 2020; Mann et al., 2020).

Purpose

The researchers aimed to examine the effect of a telemedicine STEADI-based fall prevention program versus SOC on reported falls in the year following the initial fall prevention assessment visit. In addition, they examined differences in secondary outcomes, including fall-related service utilization (referral to physical therapy, occupational therapy, eye care, or podiatry services) or medication use (prescription and non-prescription), following the index visit to help explain the mechanisms for any observed differences in fall outcomes.

Study design overview

The STEADI Options trial, previously detailed (Rein et al., 2024), was implemented as a telemedicine adaptation of the STEADI initiative in five primary care clinics between July 2020 and December 2021. Primary care professionals (PCPs) within these five primary care clinics were randomized each week, based on a coin toss, to either provide STEADI-based fall prevention or SOC annual wellness screening and general disease state management to their patients scheduled for an outpatient visit. Other intervention arms that were initially developed but were discontinued included the Medication Management Arm and Physical Therapy Intervention arm (https://clinicaltrials.gov/study/NCT05390736#participation-criteria). These intervention arms were abandoned due to low overall recruitment and adjusting the study design to limit to two arms (one full STEADI intervention and one SOC) due to small sample size. Patients of PCPs randomized to the STEADI-based fall intervention arm received multifactorial fall risk assessment via telemedicine (Zoom or phone) conducted by the research nurse prior to the visit (see online supplemental material for a full description of fall risk assessment). Although the initial protocol specified videoconferencing for assessments, telephone-based assessments were used as an alternative when technology barriers arose. The nurse communicated STEADI-based recommendations to PCPs through a dedicated electronic chart note. Fall risk factors and recommendations for action (e.g., refer to physical therapy for gait, strength, and balance deficits; reduce or stop medications) were communicated to their PCP. PCPs received group orientation by clinic prior to the start of the study on procedures, what to expect, and expectations for acting on STEADI recommendation. Patients assigned to SOC did not receive any STEADI-based fall risk assessment prior to their office visit and were only asked to participate in study data collection by NORC. SOC included a day of care visual assessment on the way to the room to determine the patient’s risk of falling during the clinical visit, as well as general disease state management. Additional assessment of risk factors and addressing identified risk factors was determined by the PCP. The Emory University Institutional Review Board considered the STEADI Options trial protocol to be quality improvement and not human subjects’ research. Survey data collection for the study was deemed human subjects research and the IRB reviewed and approved those protocols. This study was registered with ClinicalTrials.gov (NCT05390736).

Study population

The study population consisted of adults aged 65 years and older with a non-acute primary care visit scheduled at one of the five participating primary care clinics. Inclusion criteria consisted of being cognitively able to answer screening questions in English; having someone physically present with the patient who could assist or get assistance if necessary, during the telemedicine assessment visit; as well as having access to a computer, tablet, phone with internet, or telephone. Older adults without a valid phone number and those who opted out of receiving health system text messaging were excluded. Older adults who answered “Yes” to any of the statements used as proxy to the 3KQ or scored four or more on the 12-item SIS were considered at increased risk of falls and invited into the study. The SIS was completed either by older adults via text/email link prior to appointment, or through remote screening by phone call with a clinical research interviewer. A total of 3,722 patients were screened, 3,002 were excluded, and 720 patients were enrolled in the study. Of those, 60 patients who were initially assigned to the Medication Management Arm or the Physical Therapy Intervention arm were discontinued from the study due to insufficient recruitment. A total of 307 were assigned to the STEADI intervention arm and 353 were assigned to SOC (Figure 1). Of those assigned to the intervention arm, n = 72 were consented and completed preevaluation assessment but did not complete assessment (no show). These were included in intent-to-treat analysis.

Figure 1.

CONSORT diagram for eligibility and enrollment of the Stopping Elderly Accidents Death and Injuries (STEADI) options trial participants, 2020–2021.

Electronic health record data collection

Data from study participants’ EHRs were extracted at two time points. First, baseline health and demographic characteristics were collected within 12 months prior to the screening visit. These variables were used as covariates in multiple regression models. Second, fall-related health service referrals physical therapy, occupational therapy, eye care, podiatry, and visits were collected within 12 months after the patient’s scheduled office visit with primary care. These referrals and visits served as a record of events that occurred during the study period. Baseline data included information on select comorbid conditions that could be associated with fall risk: cognition problems (e.g., Alzheimer’s disease, dementia), Parkinson’s disease, cardiac arrhythmia, orthostatic hypotension, depression, and incontinence (Table S1; see online supplementary material). Patient demographic information such as age, sex, race and ethnicity, and marital status (single, married, partnered, or unknown) were drawn from EHR data. Medically treated falls were identified based on the International Classification of Diseases, Tenth Revision, Clinical Modification diagnosis codes indicating a fall or codes to specify a fall injury (Table S1; see online supplementary material) in any diagnostic field; manual chart review was conducted by a clinical research coordinator to confirm whether a fall was indicated in the chart. Fall events were coded as “confirmed” if the case notes made clear reference to a fall. Falls were coded as “ambiguous” if they contained diagnostic codes associated with a fall but made no reference to a specific fall (i.e., no clear documentation for a medically treated fall). Information about medications was extracted for both time periods (i.e., pre- and post-index visits). Other data elements like study arm assignments, SIS score, and whether a STEADI assessment was conducted were drawn from study tracking files.

Survey data collection

In addition to injurious falls identified in the EHR, patients were asked to self-report falls that occurred without injury or for which they did not seek medical attention or were attended to by a medical provider outside of the Emory system. Patient surveys were collected at baseline and during three follow-up surveys fielded at 4, 8, and 12 months (four waves of data collection) to help minimize recall bias of fall events. In the baseline survey, patients were asked questions about their current medications (prescription and non-prescription sleep aids, opioid and non-opioid pain relievers, tricyclic antidepressants, benzodiazepines, and antipsychotics) and any fall prevention services they were currently receiving. Subsequent surveys asked about falls (including number of falls and any injurious falls), medications, and any fall-related preventive services they received at Emory or externally since last surveyed.

Missing survey data

Not all patients completed one or more surveys; therefore, weighting was used to adjust the survey respondent sample to account for differential nonresponse. For patients who completed some survey collection waves, but not others, imputation was used to fill in missing reports. Both nonresponse weighting and imputation models accounted for differences in survey participation and fall reports between study arms and utilized information about the SIS score, medical conditions, medication, and fall events from the baseline EHR data. Multiple imputed values were drawn and the additional variance due to imputation was included in standard error estimates.

Data analysis

All data sources were merged using the unique study identifier assigned to each patient. The primary outcome measures consisted of “any” and “number of” measures for falls that occurred in the year after the index visit date as reported in either the EHR or the patient surveys, falls recorded in the EHR, falls reported in patient surveys, falls resulting in injury (self-reported) or for which medical attention was sought, and falls reported in surveys that did not cause injury or for which medical attention was not sought. The researchers examined secondary outcomes to understand the mechanisms for potential fall mitigation, looking at whether the STEADI intervention was associated with fall-related service utilization and medication use. The prescription and non-prescription medications of interest, including 8 classes of medications known to increase fall risk (Rein et al., 2024).

Assuming a 50% baseline risk of falling based on the expected performance of a screening tool to detect fall risk (Ye et al., 2020), this study was designed to have 80% power to detect a relative risk (RR) of 0.82 for reported falls, requiring a sample of 500 in each arm, or RR of 0.78, requiring a sample of 350 in each arm, or alternatively a RR of 0.75, requiring a sample of at least at least 250 people per arm with a recruitment goal of 500 in total (Rein et al., 2024). Target enrollment was planned at n = 500 in each arm, including the discontinued study arms.

The statistical analyses followed an intent-to-treat approach, which measures the difference in outcomes between the randomly assigned groups, regardless of whether the patient received the STEADI assessment intervention. The researchers modeled count outcomes using Poisson regression and reported the exponentiated coefficients, which can be interpreted as incidence rate ratios. Binomial outcomes were modeled using logistic regression and similarly reported exponentiated coefficients, which can be interpreted as odds ratios. Model specifications included using the treatment assignment alone (unadjusted models) and two specifications of multiple regression to address sample imbalance: first, including the SIS score as a linear measure and second, further adjusting for age (linear), sex, race (indicator for non-Hispanic or Latino White and non-Hispanic or Latino Black), marital status, and health conditions and medication use prior to study enrollment.

Results

There were 353 older adults assigned to SOC (control arm) and 307 assigned to the STEADI intervention arm based on randomization of their clinic provider. The mean age was 75.8 (Standard Deviation [SD]=6.8, results not shown). The sample was majority female (67.7%), under 80 years of age (73.6%), and non-Hispanic or Latino White (56.1%) (Table 1). The average SIS score was higher in the intervention arm (6.3; SD = 3.0) compared to SOC arm (5.2; SD = 2.9) (p < 0.001). A fall was reported by 5.5% of participants in the previous 12 months.

Table 1.

Demographic and health characteristics of older adults (65 years and above) treated in a primary care setting where telemedicine based (Stopping Elderly aAcidents, Deaths, and Injuries (STEADI) fall prevention was implemented, 2020–2021.

	Total		Standard of care control		STEADI intervention		Difference between arms (p-value)
	Number (n = 660)	Percenta	Number (n = 353)	Percenta	Number (n = 307)	Percenta
Characteristic
Age
65–69	128	19.4	62	17.6	66	21.5	0.338
70–74	193	29.2	103	29.2	90	29.3
75–79	165	25.0	99	28.0	66	21.5
80–84	93	14.1	47	13.3	46	15.0
85 and older	81	12.3	42	11.9	39	12.7
Sex
Male	213	32.3	118	33.4	95	30.9	0.496
Female	447	67.7	235	66.6	212	69.1
Race/ethnicity
White, Not Hispanic, or Latino	370	56.1	207	58.6	163	53.1	0.321
Black, Not Hispanic, or Latino	181	27.4	95	26.9	86	28.0
Asian, Not Hispanic, or Latino	12	1.8	7	2.0	5	1.6
Hispanic or Latino	8	1.2	5	1.4	3	1.0
Other races/unknownb	89	13.5	39	11.0	50	16.3
Marital status
Singlec	88	13.3	51	14.4	37	12.1	0.348
Married	315	47.7	175	49.6	140	45.6
Partnered	139	21.1	71	20.1	68	22.1
Unknown	118	17.9	56	15.9	62	20.2
SIS
SIS score	5.7	(SD = 3.0)	5.2	(SD = 2.9)	6.3	(SD = 3.0)	<0.001
Medical history (at baseline) d
1+ falls (in past 12 months)	36	5.5	17	4.8	19	6.2	0.439
Cognition problems	6	0.9	4	1.1	2	0.7	0.516
Parkinson’s disease	10	1.5	4	1.1	6	2.0	0.389
Cardiac arrythmia	2	0.3	1	0.3	1	0.3	0.921
Orthostatic hypotension	13	2.0	7	2.0	6	2.0	0.979
Depression	33	5.0	19	5.4	14	4.6	0.629
Incontinence	40	6.1	23	6.5	17	5.5	0.599
Medication (at baseline) d
Tricyclic antidepressants	32	4.8	23	6.5	9	2.9	0.032
Benzodiazepines	142	21.5	79	22.4	63	20.5	0.551
Antipsychotics	25	3.8	17	4.8	8	2.6	0.136
Antispasmodics	48	7.3	28	7.9	20	6.5	0.479
Muscle relaxants	55	8.3	27	7.6	28	9.1	0.501
Antihistamines	111	16.8	61	17.3	50	16.3	0.723
Hypnotics	53	8.0	31	8.8	22	7.2	0.441
Opioids	207	31.4	111	31.4	96	31.3	0.945

Note. SIS = Stay Independent Screener; STEADI = Stopping Elderly Accidents, Deaths and Injuries initiative.

^aNot all percentages reported will add up to 100%.

^bThe race/ethnicity category “Other race/unknown” includes people who identified as “Native American,” “Pacific Islanders,” “Bi-racial, multi-racial, or unable to classify,” and those who chose not to identify.

^cThe marital status category “single” includes individuals who may have been widowed, divorced, or never married and who are not currently married or with a significant other/life partner.

^dCategories are not mutually exclusive as participants may have a number of chronic conditions or medications at baseline.

Overall, 31.4% of study participants were taking an opioid medication, 21.5% were taking a benzodiazepine, and 15% were taking both opioids and benzodiazepine (54 in intervention arm and 46 in SOC; results not shown). Older adults assigned to the SOC arm were more likely to be taking a tricyclic antidepressant during the baseline period (6.5%) compared to intervention arm (2.9%) (p = 0.032). There were no other significant differences between the study arms on baseline demographic and health characteristics.

There was a total of 416 reported falls (237 in the intervention arm and 179 in the SOC) and 660 confirmed falls recorded in the EHR (353 in the intervention arm and 307 in the SOC; results not shown). The self-reported fall rate in the intervention arm was 46.5% higher (RR = 1.465; 95% CI: 1.018–2.110) than the SOC arm in unadjusted models (Table 2). However, after adjusting for SIS score (model 1) and SIS score, demographic characteristics, and EHR falls at baseline (model 2) these differences were no longer significant (RR = 1.098, 95% CI: 0.792–1.524 and RR = 1.091, 95% CI: 0.797–1.493, respectively).

Table 2.

Fall related outcomes among older adults (65 years and above) treated in a primary care setting where telemedicine based Stopping Ederly Accidents, Deaths, and Injuries (STEADI) fall prevention was implemented, 2020–2021.

Outcome	Treatment effect only (95% CI)	Treatment effect adjusted for patient characteristics. (95% CI)
		Model 1	Model 2
Number of falls	Incidence rate ratios
No. of self-reported falls (survey)	1.465	1.098	1.091
	(1.018–2.110)*	(0.792–1.524)	(0.797–1.493)
No. of self-reported medically treated falls (survey)	1.121	0.868	1.066
	(0.360–3.486)	(0.331–2.276)	(0.431–2.637)
No. of confirmed falls (recorded in EHR)	0.809	0.670	0.733
	(0.428–.530)	(0.355–1.264)	(0.383–1.402)
No. of confirmed and ambiguous falls (recorded in EHR)	0.798	0.674	0.716
	(0.451 – 1.415)	(0.380 – 1.195)	(0.395 – 1.299)
No. of self-reported falls (survey) or confirmed or ambiguous fall (recorded in EHR)	1.256	1.007	1.066
	(0.916–1.722)	(0.743–1.365)	(0.747–1.366)
Individuals with at least one fall	Odds ratios
At least 1 self-reported fall (survey)	1.288	1.073	1.143
	(0.827–2.006)	(0.672–1.715)	(0.703–1.857)
At least 1 self-reported medically treated fall (survey)	1.496	1.174	1.402
	(0.565–3.963)	(0.452–3.048)	(0.517–3.802)
At least 1 confirmed fall (recorded in EHR)	0.894	0.756	0.812
	(0.473–1.690)	(0.393–1.452)	(0.418–1.574)
At least 1 confirmed or ambiguous fall (recorded in EHR)	0.862	0.741	0.771
	(0.485–1.535)	(0.411–1.336)	(0.422–1.410)
At least 1 fall self-reported (survey) or recorded in EHR	1.185	1.018	1.041
	(0.827–1.695)	(0.693–1.495)	(0.706–1.534)

Note. EHR = electronic health record; SIS = stay independent screener.

Adjustment model 1 covariates: SIS score.

Adjustment model 2 covariates: SIS score, age, sex, race/ethnicity, marital status, EHR falls (baseline).

Number of falls is the counts of the number of surveys reported falls, number of people with 1+ survey reported falls include imputed falls and represent the average number of falls across multiple imputations.

Survey estimates are based on the 431 patients who responded to the survey and have been weighted to correct for differential survey nonresponse. For patients with missing survey waves (i.e., those who did not respond fully), outcomes have been imputed based on patient demographics, study arm, SIS Score, fall history, and health conditions from the EHR.

^*p < .05.

There were no significant differences in other fall related outcomes in the intervention arm in any of the unadjusted or adjusted models. The treatment group had consistently lower odds of experiencing at least one confirmed fall recorded in the EHR, with a 10.6% reduction in the unadjusted analysis (OR = 0.894, 95% CI: 0.473–1.690), and 24.4% reduction after adjusting for SIS (model 1) (OR = 0.756, 95% CI: 0.393–1.452), and an 18.8% reduction in model 2 (OR = 0.812, 95% CI: 0.418–1.574), although none of these differences achieved statistical significance due to wide confidence intervals that included unity.

There were also no significant differences in any of the secondary outcome measures including fall-related service utilization or medication use reported by older adults in the intervention arm compared to the SOC arm in either the unadjusted or adjusted models (Table S2; see online supplemental material). Estimates were generally near the null, with wide confidence intervals.

Discussion

In this study of a primary care-based, telemedicine-delivered STEADI intervention, the researchers found no effect of the intervention on any fall-related primary outcomes. There were no significant differences between the STEADI intervention arm and the SOC arm on any of the secondary outcomes. There was a nonsignificant 24.4% reduction in EHR-confirmed falls in the intervention group after adjusting for SIS. While study protocols were successfully modified to conduct the trial by telemedicine during a pandemic, the number of participants enrolled was lower than intended, which reduced the power needed to observe significant findings.

Among the study population, about 1 in 20 reported a fall in the previous year at baseline. This is lower than expected among older adults where national estimates are one in four older adults reporting a fall each year (Centers for Disease Control and Prevention, 2023a). In addition, although all participants were deemed at risk for future falls using the 3KQs or SIS score of 4 or more, the average SIS score was a modest 5.7 (SD = 3.0). These data may indicate that this study population were not in the highest risk group for falls where fall prevention interventions are most effective.

Multifactorial fall prevention interventions have been shown to be more effective in reducing serious fall injuries when compared to enhanced usual care (Tricco et al., 2017). However, similar to this study, the STRIDE study, a large-scale, pragmatic trial conducted in a real-world practice setting that aimed to reduce falls in older adults, did not show a significant reduction in injurious falls (Bhasin et al., 2020). The findings of no significant reduction in falls may have been due to the complexity of implementing the multifactorial interventions, which involves multiple risk assessments and tailored prevention plans, resulting in variability in how interventions get implemented across sites, or adapted by the patient, potentially weakening the overall effectiveness (Reckrey et al., 2021).

In response to the 2022 World Guidelines for Falls Prevention (Montero-Odasso et al., 2022), the American Geriatrics Society workgroup recommended updating the STEADI initiative by reframing fall risk as a chronic condition, prioritizing patient-centered goals, and implementing a multidisciplinary, longitudinal assess-intervene workflow that includes additional assessments to address comprehensive risk factors (Johnson et al., 2025). These changes also include regular and longitudinal follow-ups to enhance identification and management of multifactorial fall risks over multiple visits, ensuring ongoing, team-based care for older adults at risk of falls (Johnson et al., 2025).

In a study comparing multifactorial trials, categorized into (a) those with direct interventions by investigators and (b) those offering only advice or referrals, none of the 14 trials relying solely on advice or referrals demonstrated a significant benefit to the intervention (Tinetti & Kumar, 2010). In contrast, in trials with direct interventions, or where an investigator carried out or ensured completion of at least one component (e.g., balance training or medication reduction), 8 out of 11 showed improved fall-related outcomes (Tinetti & Kumar, 2010). The core components of the STEADI initiative are to screen, assess, and intervene. To maximize the effectiveness of fall prevention, a fourth core component—structured follow-up—should be incorporated to ensure completion of recommended interventions and to support ongoing risk reduction. This approach is supported by the American Geriatrics Society, which emphasizes that follow-up and comprehensive management of identified risk factors are essential to the success of fall prevention programs (Nicholson et el., 2024).

Medically treated falls had a baseline risk much lower than assumed for power calculations and would have necessitated a larger sample size to detect a difference. The initial study was designed at 80% power to detect an RR of 0.82 with an aim to recruit 500 participants in each of the study arms. The final sample included only 353 older adults assigned to SOC and 307 assigned to intervention. Previous studies have reported a substantial impact on patient recruitment during the pandemic into clinical trials due to patient reluctance to enroll, especially among certain populations such as older adults; difficulty in implementing trial protocols remotely, and a shift in the research focus (Sathian et al., 2020). This study, initially designed for in-person screening and assessment of patients for fall risk, was adapted in about 5 months during the COVID-19 pandemic for telemedicine delivery (Rein et al., 2024). The intent was that all participants in the intervention group would receive the study assessment via videoconferencing. However, because many participants did not possess the knowledge, technology, or internet connection to support virtual interaction during assessments at that time, the study protocol was further modified to allow for phone assessments, which may have resulted in ineffective evaluation of some fall risk factors (Rein et al., 2024).

In addition, the effectiveness of the intervention may have been influenced by the conditions of the COVID-19 pandemic, whereby patients may have chosen not to follow through on recommended referrals (e.g., physical therapy) due to social distancing guidance, and whereby health systems had limited access to certain services (e.g., ophthalmology; Rein et al., 2024). While there are particular challenges to achieve target enrollment for remotely delivered trials (Sathian et al., 2020), continued efforts to adapt and innovate in telemedicine delivery is worthwhile for several reasons (e.g., ensuring adequate representation of hard-to-reach [e.g., rural] populations, older adults with impaired mobility, and individuals without access to transportation to centralized trial assessments).

Survey responses for reported falls are subject to recall bias. Recall bias in reported falls occurs when individuals inaccurately remember or report past falls, leading to potential misrepresentation of data. Individuals may overestimate or underestimate the frequency or severity of their falls due to memory lapses, particularly over longer recall periods (e.g., quarterly vs. annual report). Recall bias can impact the validity of studies on fall prevention and interventions, as inaccurate reporting may skew results. Alternative methods for detecting falls that are not medically treated (such as wearable devices [Warrington et al., 2021], capturing daily fall data [Antcliff et al., 2022], or keeping a daily fall diary [Hannan et al., 2010]) may be needed to produce unbiased estimates.

Previous studies have highlighted older adults’ facilitators and barriers to participation in fall prevention strategies. Facilitators to participating in fall prevention strategies include increased awareness of fall risks, access to educational programs, support from healthcare providers, and a safe home environment (Fernandes et al., 2021; Kiami et al., 2019). However, barriers may include a lack of knowledge about fall prevention, or belief that falls are inevitable, fear of falling, limited mobility, financial constraints, and a lack of physical fitness (Fernandes et al., 2021; Kiami et al., 2019). In addition, social isolation, social awkwardness of participating in group-based exercises, lack of knowledge, stigma, and lack of access to healthcare services can further hinder participation in fall prevention strategies (Fernandes et al., 2021; Gardiner et al., 2017).

This study adds to the existing literature by providing the first randomized controlled trial data on the effectiveness of implementing a STEADI-based fall prevention approach via telemedicine in primary care. While previous studies have demonstrated that the STEADI initiative can reduce fall-related hospitalizations when implemented in traditional in-person primary care settings, and that multifactorial interventions are recommended, this study specifically evaluates the feasibility and impact of delivering STEADI protocols remotely.

Integrating telemedicine with STEADI protocols helps overcome barriers to in-person care, extends access, and supports ongoing fall prevention, especially for patients with mobility or transportation challenges (Chan et al., 2021). Successful implementation depends on EHR integration, staff training, and clinical leadership to manage workflow issues (Casey et al., 2017; Johnston et al., 2023). Telehealth interventions, particularly when combined with exercise, have been shown to reduce fall risk and improve balance, although impacts on strength, cognition, and quality of life are less clear (Chan et al., 2021).

In practice, these findings support incorporating telemedicine-enabled STEADI protocols into primary care, emphasizing multidisciplinary collaboration and ongoing staff education (Casey et al., 2017; Johnston et al., 2023). Future research should focus on large-scale trials to assess long-term outcomes, cost-effectiveness, and scalability, as well as explore personalized interventions using technology.

Limitations

Several limitations should be considered when interpreting these findings. First, this study was conducted during the COVID-19 pandemic, which may have affected recruitment, participant engagement, and the generalizability of results to non-pandemic settings (Thwaites et al., 2023). The enrolled cohort had a lower-than-expected baseline fall rate, potentially limiting the ability to detect intervention effects and reducing statistical power (Johnston et al., 2019; USPSTF, 2024). The researchers assumed a 50% fall rate among individuals identified as high-risk in our modeling, based on findings from prior studies. However, this figure reflects the outcome incidence within a high-risk subgroup rather than the baseline fall risk in the general older adult population, which is typically 25%–30%. As such, the researchers’ assumptions may overestimate the absolute risk of falls when applied more broadly. This should be considered when interpreting the generalizability and real-world applicability of the results. Second, PCP fidelity in reviewing and recommending interventions, patient adherence to recommended interventions and the fidelity of telemedicine delivery were not systematically measured, introducing uncertainty regarding the actual exposure to fall prevention strategies (Johnston et al., 2023; Rein et al., 2024; United States Preventive Services Task Force, 2024). Third, patients were not excluded if they received external fall prevention interventions or received care for a fall outside of the study health system, which could introduce unmeasured confounding and limit attribution of outcomes to the study intervention, as is common in pragmatic fall prevention trials. Fourth, the pragmatic design allowed primary care providers discretion in implementing recommendations, which may have led to variability in intervention uptake and diluted potential effects. A further limitation is the potential for spillover between study arms due to the randomization of PCP to intervention and SOC on alternating weeks. Providers exposed to the intervention may have incorporated elements of the STEADI-based approach into their standard care practice, either consciously or unconsciously, during SOC weeks. This contamination could attenuate observed differences between groups and bias results toward the null, a recognized challenge in pragmatic and cluster-randomized trials. Fifth, study protocols called for intervention arm participants to be assessed for fall risk via videoconferencing; however, many (over half) were unable to accommodate video and had the assessments conducted over the phone as a workaround to inexperience with videoconferencing tools. This resulted in their not receiving the full STEADI assessment (e.g., mobility assessments) as designed. Finally, self-reported falls are subject to recall bias, and the study may have been underpowered to detect differences in secondary outcomes such as medically treated falls or service utilization. Future research should address these limitations by ensuring adequate sample size, monitoring intervention fidelity and adherence, employing objective outcome measures, and considering randomization strategies that minimize provider-level contamination.

Conclusion

This randomized trial found that a STEADI-based fall prevention approach delivered via telemedicine in primary care did not significantly reduce falls compared to standard care. While telemedicine interventions can be feasible and may improve access, their effectiveness depends on intervention fidelity and patient engagement. Future research can optimize implementation and asses long-term outcomes strategies and exploring new methods to improve and measure patient engagement and adherence.

Supplementary Material

Supplementary data are available at The Gerontologist online.

Data availability

The data are not publicly available due to participant privacy concerns and institutional data use agreements. Deidentified data can be made available to qualified researchers for replication purposes upon reasonable request, subject to review and approval by the corresponding institutional review board and in accordance with applicable data sharing policies. Requests for data access should be directed at the corresponding author. The analytic code used for primary analyses is available upon reasonable request.

The studies reported in this manuscript were preregistered at ClinicalTrials.gov (NCT05390736).