ABSTRACT
Background
More than one in four older adults experience a fall each year. While exercise programs are effective in reducing fall‐related injuries (FRI), participation remains low due to access barriers. The primary aim of this study was to evaluate whether older adults who registered for Nymbl, a self‐guided, asynchronous, balance application, were associated with fewer FRIs as compared to age‐similar individuals non‐registrants.
Methods
This retrospective cohort study used data from Kaiser Permanente Colorado, linked to Nymbl registration and usage records based on patient name and demographic information between February 2018 and September 2024. The cohort included individuals aged 60 and older with continuous health plan enrollment for 12 months before and after Nymbl registration (or a randomly assigned index date). Logistic regression models with inverse probability of treatment weighting estimated the association between Nymbl registration and FRIs during the 12‐month follow‐up, stratified by history of FRIs. Marginal effects reported the absolute risk difference associated with Nymbl registration. Secondary analyses examined dose–response associations of Nymbl usage and whether the association of Nymbl was additive to participation in other exercise programs.
Results
We identified 3735 individuals who registered for Nymbl and 114,219 age‐eligible non‐registrants. Among individuals with a prior FRI, Nymbl registration was associated with a 3.83 percentage point reduction in acute FRIs; however, no significant association was estimated for individuals without a baseline FRI. Secondary analysis indicated that at least five sessions were required to achieve a meaningful reduction in FRIs, and associations were limited to those not already participating in other exercise programs.
Conclusion
Findings from this study suggest that asynchronous, self‐guided balance applications may reduce FRIs among older adults with a history of falls who are not otherwise engaged in structured exercise programs. Remotely delivered fall prevention programs may help overcome access barriers and can be used to supplement in‐person and guided exercise programs.
Keywords: asynchronous application, balance exercise, digital health, fall prevention
Summary
- Key Points
-
○Registering for Nymbl, a self‐guided balance application for fall prevention, was associated with a 3.83 percentage point reduction in fall‐related injuries among older community dwelling adults with a history of falls.
-
○No reduction in fall‐related injuries was associated with Nymbl registration among individuals who were already participating in the Silver Sneakers Fitness program.
-
○Self‐directed digital balance programs may reduce recurrent fall‐related injuries and should be considered as an option for preventive care, especially for individuals with a history of falls who are not already engaged in other exercise interventions.
-
○
- Why Does This Matter?
-
○The U.S. Preventive Services Task Force recommends exercise‐based interventions for adults aged 65 and older at increased risk of falling; however, participation in fall prevention programs is low and access is a significant barrier. Findings from this study suggest that self‐directed, asynchronous, digital balance programs may help reduce recurrent fall‐related injuries, supporting their potential value in expanding access and improving health outcomes.
-
○
1. Introduction
More than one in four adults aged 65 and older experience a fall each year, resulting in approximately 3 million emergency department visits and 800,000 hospitalizations annually [1]. In response to this significant healthcare burden, the US Preventive Services Task Force (USPSTF) recommended exercise interventions to prevent falls in community‐dwelling adults 65 years or older who are at increased risk with a Grade “B” recommendation, indicating moderate certainty of a net benefit [2]. Despite the high prevalence of fall‐related injuries (FRIs), participation in fall prevention programs remains low [3, 4]. Limited access remains a significant barrier, as many older adults struggle to attend in‐person programs due to transportation challenges or scheduling constraints [5, 6]. To address access barriers, the USPSTF's most recent recommendation emphasized the need for interventions that improve the availability and accessibility of effective fall prevention strategies, including asynchronous, remotely delivered exercise programs [2].
Prior evaluations of asynchronous, remotely delivered fall prevention interventions are limited. While findings have been mixed, access to a self‐directed digital exercise program has been shown to reduce FRIs by up to 18%, particularly among individuals at low to intermediate fall risk [7, 8]. However, these studies typically have relied upon patient‐reported fall outcomes, which often have very high reported rates of fall outcomes, with limited ability to identify severe fall injuries that require medical attention vs. minor falls, and may be affected by recall bias [7, 9]. Additionally, it remains unclear whether remotely delivered interventions offer benefits that are additive to those of traditional, in‐person exercise programs. Moreover, individuals who participate in clinical trials tend to be healthier and more motivated than the broader population, which may limit the generalizability of trial findings [10, 11, 12].
To better understand the association of asynchronous exercise programs in reducing FRIs using real‐world evidence, the primary aim of this study was to evaluate whether registration for Nymbl, a self‐guided balance training application, was associated with reduced FRIs among community‐dwelling older adults compared with age‐similar adults who did not register. We examined the association of Nymbl on both primary fall prevention (preventing initial FRIs) and secondary prevention (preventing recurrent FRIs). This research addresses a key knowledge gap identified by the USPSTF by examining the association between a digitally delivered fall prevention program and fall‐related outcomes, which may support efforts to expand access for individuals who face barriers to traditional in‐person or guided exercise interventions and inform the design of future randomized trials [2].
2. Methods
2.1. Study Setting and Data
This study was conducted within Kaiser Permanente Colorado (KPCO), an integrated healthcare system that provided both clinical care and insurance to approximately 550,000 members, of which 110,000 members were enrolled in Medicare, which KPCO offered exclusively through Medicare Advantage plans. KPCO's coordinated care model delivers comprehensive care management for patients with FRIs, encompassing initial treatment, recovery‐focused care, and outpatient follow‐up [13, 14]. To improve physical well‐being and help prevent FRIs, KPCO offers various exercise programs, such as Silver Sneakers, a fitness program for older adults that provides free access to thousands of fitness centers and guided classes.
Through a partnership between Nymbl Science and the Denver Regional Council of Governments (DRCOG), all adults aged 60 and older living in the Denver metro area were eligible to download the Nymbl app at no cost starting February 1, 2019. DRCOG promoted Nymbl through community‐based outreach and local media to individuals meeting age and residency criteria; however, KPCO did not advertise or encourage Nymbl use among members during the study period. Nymbl is an asynchronous mobile application, accessible via smartphone or tablet, designed to improve balance and walking stability to help prevent falls [15, 16]. The Nymbl application uses a dual‐tasking approach, combining physical activities such as balance or strength exercises with simultaneous cognitive challenges, like image recall. Evidence suggests that dual‐task training can improve both static and dynamic balance, as well as executive function, particularly among older adults with a history of falls [17]. A more detailed description of the Nymbl balance application can be found in Figure S1.
Nymbl Science provided KPCO with a complete list of individuals who downloaded and registered for the Nymbl application during its partnership with DRCOG in accordance with a data transfer agreement. The list of Nymbl users was linked to KPCO members at the individual level using name, birthdate, and demographic information. Data provided by Nymbl Science included the date of registration and timestamps of when the application was used. Electronic health record (EHR) and administrative data were obtained from the KPCO Virtual data warehouse (VDW), which captured care delivered within KPCO as well as claims data for services provided outside the KPCO system. The VDW is a standardized data repository containing information on diagnostic and procedural codes, patient‐reported responses to health questionnaires, demographics, enrollment, and census‐based socioeconomic status measures [18]. The cohort was limited to KPCO members aged 60+ who were insured between February 1, 2019 and September 30, 2024. Study protocol and human subjects considerations were reviewed and approved by the Kaiser Permanente Interregional Institutional Review Board (#2163506).
2.2. Measures
The primary outcome was a FRI episode during the 365‐day period after the index date (described below). To quantify FRI episodes, we applied a previously published algorithm that identified FRIs using International Classification of Diseases version 10 diagnostic codes [19]. This algorithm categorized FRI episodes into three levels of acuity based on injury type and care setting are as follows: (1) the acute care algorithm, which captured the most severe and clinically valid injuries, including those to the head, face, limbs, neck, and trunk treated in hospital or emergency department settings; (2) the balanced algorithm, which built upon the acute care definition by including severe injuries treated in non‐emergency outpatient settings, offering a middle ground between specificity and comprehensiveness; and (3) the inclusive algorithm, which maximized FRI detection by incorporating nearly all injury types regardless of setting, allowing for broad identification of FRIs.
The primary exposure was registration for the Nymbl application. Participants were classified into two groups: individuals who registered for Nymbl and an age‐eligible comparison group who did not register for Nymbl. The date of registration was used as the index date among those who registered for Nymbl. Because the comparison group lacked a natural index date for evaluating FRI outcomes, each individual in the comparison group was assigned a randomly selected index date based on the distribution of registration dates among Nymbl users (Figure S2) [20, 21]. This approach ensured equivalent follow‐up intervals across both groups to assess FRI outcomes. All individuals were required to have continuous enrollment within KPCO for 12 months before and 12 months after their index date.
Baseline covariates were measured during the 12‐month period prior to the index date and included data on an individual's demographics, comorbidity profile, healthcare utilization, and socioeconomic status. Age in years was included as a continuous measure. Race was self‐reported and categorized into mutually exclusive groups: Asian, Black, Unknown, White, and, due to small sample sizes, a combined single category labeled “Other Minority Races,” which included Native Hawaiian/Pacific Islander, Native American/Alaskan Native, and Multiracial individuals. Ethnicity was captured as a separate binary variable indicating Hispanic or Non‐Hispanic/Unknown. Frailty was included as a binary indicator based on diagnostic and procedural code value sets defined by the National Committee for Quality Assurance [22]. Comorbidity burden was measured using the Deyo adaptation of the Charlson comorbidity index, based on diagnoses codes from the year prior to the index date [23]. The number of diagnosed chronic conditions was grouped into four categories are as follows: 0 conditions, 1 condition, 2 conditions, and 3+ conditions. Utilization of acute medical services was captured using two binary indicators: one indicator for a hospital admission and a separate indicator for an emergency department visit. Socioeconomic status was assessed using the neighborhood deprivation index (NDI), a composite measure incorporating housing, poverty, employment, occupation, and education at the census tract level. Higher NDI values indicated greater levels of deprivation within the census tract [24].
2.3. Statistical Analysis
Baseline characteristics of Nymbl registrants and age‐eligible non‐registrants were compared using chi‐square tests for categorical variables. Unadjusted differences in FRI outcomes between groups were calculated to help illustrate absolute differences in FRI outcomes. Inverse probability of treatment weighting (IPTW) was used to balance baseline characteristics by Nymbl registration [25]. The predicted probability (propensity score) of Nymbl registration was estimated using a logistic regression model. Overlap of propensity scores between Nymbl registrants and age‐eligible non‐registrants was depicted using density plots. Standardized mean differences were used to examine the balance of covariates after applying IPTW.
The association of Nymbl registration on FRI outcomes during the 12 months after the index date was estimated using a logistic regression weighted by the IPTW [26]. To better understand Nymbl's association with both primary fall prevention (preventing initial FRI) and secondary prevention (preventing recurrent FRI), our primary analysis stratified individuals by whether they experienced any FRI in the 12‐month period prior to the index date. Separate models were estimated for each of the three FRI acuity outcomes: acute, balanced, and inclusive. Marginal effects (ME) and 95% confidence intervals (CI) were reported holding all other covariates at their mean values. In this study, the ME is interpreted as the percentage point (pp) difference in the probability of an FRI outcome between Nymbl registrants and the age‐eligible comparison group.
2.4. Pre‐Specified Secondary Analysis
To assess whether the amount of Nymbl utilization was associated with differential associations on FRI outcomes, we conducted a dose–response analysis limited to individuals who registered for Nymbl. We summed the number of Nymbl sessions for each individual during the first 90 days after registration and categorized them into three groups are as follows: 0 sessions (registered but did not use Nymbl), 1–4 sessions, and > = 5 sessions. FRI outcomes were then assessed during days 91–365 post‐registration. Because we allowed for three levels of exposure, multinomial regression models were used to estimate the dose–response association using registrants with 0 sessions as the referent group. Due to a sharp reduction in sample size when limiting our cohort to Nymbl registrants, it was necessary to collapse race into White versus Non‐White and diagnosed chronic conditions were regrouped as 0, 1–2, or 3+ to maintain model stability.
2.5. Post Hoc Secondary Analysis
In a complementary analysis, we stratified our models by self‐reported fall risk among a subset of our cohort who completed a health risk assessment (HRA) for Medicare beneficiaries. The HRA collects self‐reported medical, functional, and social data, including information on recent falls and balance concerns (Figure S3) [27]. Among individuals who completed the HRA, we re‐estimated logistic regression models stratified by a binary variable of self‐reported fall history or balance impairment.
To evaluate whether Nymbl was additive to other exercise interventions, we re‐estimated our primary regression models further stratifying by whether individuals participated in the Silver Sneakers fitness program at any point during the 12‐month pre‐index period. All analyses were performed using SAS Software version 9.4 (SAS Institute Inc., Cary, North Carolina) and STATA Statistical Software version 18.0 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC).
3. Results
After applying exclusion criteria, our cohort included 117,954 individuals (Figure S4), 3735 (3.2%) who registered for Nymbl, and 114,219 (96.8%) who were age‐eligible but did not register for Nymbl (Table 1). Compared to age‐eligible non‐registrants, Nymbl registrants were more likely to have a pre‐index FRI (acute: 3.1% vs. 2.0%, balanced: 3.4% vs. 2.2%, Inclusive: 5.2% vs. 3.1%), be 76 or older (35.6% vs. 28.1%), female (72.8% vs. 54.9%), White (90.1% vs. 77.1%), and reside in the least deprived census tracts (48.8% vs. 42.0%). No statistically significant differences were observed between groups in the number of diagnosed chronic conditions or in utilization of acute medical services during the pre‐index period. When stratified by both Nymbl registration and Silver Sneakers participation, elevated pre‐index FRIs were only present among Nymbl registrants, and not among Silver Sneakers users (Table S1).
TABLE 1.
Baseline characteristics by Nymbl registration.
| Nymbl registrant | p | |||
|---|---|---|---|---|
| No | Yes | Total | ||
| Unique patients, N (%) | 114,219 (96.8) | 3735 (3.2) | 117,954 | |
| Pre‐index FRI, N (%) | ||||
| Acute algorithm | 2302 (2.0) | 117 (3.1) | 2419 (2.1) | < 0.001 |
| Balance algorithm | 2541 (2.2) | 128 (3.4) | 2669 (2.3) | < 0.001 |
| Inclusive algorithm | 3594 (3.1) | 194 (5.2) | 3788 (3.2) | < 0.001 |
| Self reported fall risk, N (%)* | 0.719 | |||
| Yes | 6333 (5.5) | 242 (6.5) | 6575 (5.6) | |
| No | 7402 (6.5) | 292 (7.8) | 7694 (6.5) | |
| Pre‐index silver sneakers use, N (%) | < 0.001 | |||
| Yes | 16,439 (14.4) | 1136 (30.4) | 17,575 (14.9) | |
| No | 97,780 (85.6) | 2599 (69.6) | 100,379 (85.1) | |
| Age at index date, mean (Std) | 71.4 (7.9) | 73.3 (7.0) | 71.5 (7.9) | < 0.001 |
| Gender, N (%) | < 0.001 | |||
| Female | 62,728 (54.9) | 2718 (72.8) | 65,446 (55.5) | |
| Race, N (%) | < 0.001 | |||
| Asian | 3375 (3.0) | 31 (0.8) | 3406 (2.9) | |
| Black | 4704 (4.1) | 72 (1.9) | 4776 (4.0) | |
| White | 88,069 (77.1) | 3367 (90.1) | 91,436 (77.5) | |
| Native hawaiian/pacific islander, American indian/alaska native, Multi‐race, or other race | 14,907 (13.1) | 217 (5.8) | 15,124 (12.8) | |
| Unknown | 3164 (2.8) | 48 (1.3) | 3212 (2.7) | |
| Hispanic, N (%) | < 0.001 | |||
| Yes | 11,742 (10.3) | 133 (3.6) | 11,875 (10.1) | |
| No/missing | 102,477 (89.7) | 3602 (96.4) | 106,079 (89.9) | |
| Charlson comorbidity index, N (%) | 0.882 | |||
| 0 | 81,324 (71.2) | 2672 (71.5) | 83,996 (71.2) | |
| 1 | 11,148 (9.8) | 369 (9.9) | 11,517 (9.8) | |
| 2 | 8873 (7.8) | 278 (7.4) | 9151 (7.8) | |
| 3 or more | 12,874 (11.3) | 416 (11.1) | 13,290 (11.3) | |
| Frailty indicator, N (%) | 1981 (1.7) | 61 (1.6) | 2042 (1.7) | 0.641 |
| Utilization (binary indicator), N (%) | ||||
| Emergency department | 17,237 (15.1) | 6.04 (16.2) | 17,841 (15.1) | 0.070 |
| Inpatient admission | 8656 (7.6) | 316 (8.5) | 8972 (7.6) | 0.045 |
| Neighborhood deprivation index quintile, N (%) | < 0.001 | |||
| 1 (least deprived) | 47,981 (42.0) | 1821 (48.8) | 49,802 (42.2) | |
| 2 | 30,352 (26.6) | 1031 (27.6) | 31,383 (26.6) | |
| 3 | 17,560 (15.4) | 498 (13.3) | 18,058 (15.3) | |
| 4 | 14,449 (12.7) | 312 (8.4) | 14,761 (12.5) | |
| 5 (most deprived) | 3877 (3.4) | 73 (2.0) | 3950 (3.3) | |
Note: Neighborhood Deprivation Index measures 13 dimensions of socioeconomic status at the census‐tract level including: wealth and income, education, occupation, and housing conditions.
Self‐reported fall risk limited to individuals who completed a health risk assessment (HRA) for Medicare beneficiaries.
Unadjusted analyses are shown in Table S2. Among individuals with a pre‐index FRI, Nymbl registration was associated with a 4.90 pp. lower probability of an acute FRI (3.61% vs. 8.51%), a 4.43 pp. lower probability of a balanced FRI (4.64% vs. 9.07%), and a 3.91 pp. lower probability of an inclusive FRI (7.22% vs. 11.13%) compared to non‐registrants.
Propensity scores showed a high degree of overlap between Nymbl registrants and age‐eligible non‐registrants, indicating there was adequate common support to satisfy the positivity assumption (Figure S5) [28]. Mean standardized differences were < 10% for most covariates after applying IPTWs, suggesting that observable characteristics were well balanced between groups (Table S3) [29].
Primary results from IPTW weighted logistic regression models stratified by pre‐index FRI status are presented in Figure 1. Among individuals with a pre‐index FRI episode, Nymbl registration was associated with a 3.83 pp. reduction in acute FRIs (ME = −3.83; 95% CI: −7.40 to −0.25), a 3.80 pp. reduction in balanced FRIs (ME = −3.80; 95% CI: −7.46 to −0.13), and a non‐statistically significant 3.83 pp. reduction in inclusive FRIs (ME = −3.83; 95% CI: −7.85 to 0.19). In contrast, among individuals who did not have a pre‐index FRI, no significant association was estimated between Nymbl registration and FRI outcomes.
FIGURE 1.
Adjusted marginal effect (ME) of Nymbl registration on occurrence of a fall‐related injury (FRI) episode by acuity level stratified by Pre‐index FRI episode.
Estimates from dose–response analyses among Nymbl users are shown in Figure 2 (corresponding distribution of Nymbl sessions shown in Figure S6). Compared to individuals who registered but did not use Nymbl, logging five or more sessions was associated with a 1.19 pp. reduction in acute FRIs (ME = −1.19; 95% CI: −2.25 to −0.12), a 1.38 pp. reduction in balanced FRIs (ME = −1.38; 95% CI: −2.52 to −0.23), and a 1.81 pp. reduction in inclusive FRIs (ME = −1.81; 95% CI: −3.25 to −0.37). No significant associations were estimated among individuals who used Nymbl 1–4 times compared to those who registered but did not use Nymbl.
FIGURE 2.
Among nymbl registrants: Adjusted marginal effect (ME) of Nymbl use during 90 days post‐registration on fall‐related injury (FRI) episode during days 91–365 post‐registration (dose–response).
Results using self‐reported fall or balance impairment as the stratification variable are shown in Figure 3. Regardless of whether individuals did or did not report prior fall or balance concerns, we estimated no significant associations between Nymbl registration and FRI outcomes at any acuity level.
FIGURE 3.
Adjusted marginal effect (ME) of nymbl registration on occurrence of a fall‐related injury (FRI) episode by acuity level stratified by self‐reported risk of fall or balance difficulty.
Estimates stratified by both pre‐index FRI episode and participation in the Silver Sneakers Fitness program are shown in Figure 4. Among individuals who had a pre‐index FRI but did not participate in Silver Sneakers, we estimated that Nymbl registration was associated with a non‐statistically significant 4.47 pp. reduction in acute FRIs (ME = −4.47; 95% CI: −9.07 to 0.12), a 5.44 pp. reduction in inclusive FRIs (ME = −5.44; 95% CI: −10.38 to −0.50), and a non‐statistically significant 4.52 pp. reduction in balanced FRIs (ME = −4.52; 95% CI: −9.21 to 0.16). No significant associations were estimated for individuals with a pre‐index FRI episode who participated in Silver Sneakers, or for those without a pre‐index FRI episode, regardless of Silver Sneakers participation.
FIGURE 4.
Adjusted marginal effect (ME) of nymbl registration on occurrence of a fall‐related injury (FRI) episode by acuity level stratified by pre‐index FRI episode and silver sneakers participation.
4. Discussion
In this study, the use of an asynchronous, digital application designed to improve balance was associated with a 3.83 pp. reduction in acute FRI episodes among individuals with a prior FRI. These findings suggest that self‐directed digital fall prevention programs may provide an effective and more accessible alternative to in‐person or instructor‐led exercise programs for secondary fall prevention [7, 8, 19, 30]. Prior randomized trials of unsupervised digital exercise interventions, widely regarded as the gold standard for evaluating efficacy, have reported fall injury reductions between 8% and 18% among at risk populations [7, 8]. Differences in outcome ascertainment and study inclusion criteria may partially explain the larger estimated reduction in FRIs associated with the digital exercise program evaluated in this study. The FRI algorithm used in this study was specifically designed to capture medically attended fall injuries using utilization and diagnosis codes (prioritizing specificity over sensitivity). In contrast, participant selection and outcomes in randomized trials have often relied on patient‐reported falls, which are subject to high reporting rates (> 50% of patients reported a fall) and may not distinguish between severe and minor falls.
The cohort in the present study and participants in prior randomized trials of unsupervised digital exercise interventions were predominantly White females, underscoring the need for targeted efforts to increase participation among minority and male populations. While fall risk may vary by sex, race, and ethnicity, fall‐related mortality rises with age across all groups and has been increasing over the last 20 years [31]. Personalizing digital fall‐reduction interventions to align with individual preferences and cultural contexts may improve engagement and increase accessibility across diverse populations [5].
Findings from our dose–response analysis among Nymbl registrants indicated that completing at least five sessions was necessary for an associated reduction in FRIs. However, of the 3735 individuals who registered for Nymbl, only 789 (21%) completed five or more sessions. This is consistent with previous studies showing that most participants in fall prevention interventions do not achieve the recommended dose of exercise activity [7, 8]. Given the abundance of digital exercise applications, sustaining user engagement over longer time horizons is critical to realizing the full benefits of these programs. As digital fall prevention strategies become increasingly embedded into routine care, efforts should focus on promoting sustained engagement through approaches such as integration with primary care providers, personalized goal setting, and continuous refinement of the user experience [32, 33, 34].
Analysis stratified by self‐reported fall risk factors found that Nymbl was not associated with reduction in acute FRIs using self‐reported data, although the direction of the association suggests Nymbl could be associated with FRI reductions among those reporting fall risk factors. This may indicate that reliance on patient reported fall risk factors may be insufficient to adequately identify individuals at high risk. For example, one study found that fewer than 40% of healthcare providers routinely assess fall history among older adults, highlighting missed opportunities for prevention [3]. Moreover, randomized trials of fall prevention interventions that used self‐reported fall risk to identify participants, such as STRIDE, reported no reductions in medically attended FRIs despite employing interventions more intensive than Nymbl [35, 36]. Additionally, specific components of fall risk such as gait impairment are more nuanced, with the relationship between gait speed and fall risk being non‐linear [37]. Individuals with faster gait speeds may be more agile and experience less severe injuries, while slower gait speeds may reflect underlying frailty, increasing the likelihood of more serious FRIs, complicating efforts to predict FRIs [38, 39]. Moreover, long‐term risk factors such as gait speed and impaired balance can interact with short‐term triggers, including acute illness or environmental hazards, adding additional complexity to FRI prediction [40, 41, 42]. Although many fall risk assessment tools are available, current evidence suggests that no single measure offers sufficient accuracy to predict FRIs in community‐dwelling older adults [43, 44]. As such, a patient's history of prior falls remains one of the most reliable indicators of future FRI risk [43].
Sensitivity analyses evaluating the association of Nymbl registration among individuals who were already participating in an exercise program indicated that the associated reductions in FRI were limited to those who were not taking part in the Silver Sneakers program. This finding suggests that engaging in exercise, regardless of its primary intent or mode of delivery, may be effective in reducing FRIs. Prior evidence has suggested that brisk walking can be more effective than structured balance training in preventing fall injuries [45]. Offering individuals a choice in program format (e.g., in‐person vs. virtual, synchronous vs. asynchronous, guided vs. self‐directed) may enhance long‐term engagement, contributing to larger and more sustained reductions in FRI episodes.
5. Limitations
This study has several limitations. Participation in Nymbl was voluntary and registration among age‐eligible individuals was low (3.2%), which may reflect selective participation. If those who chose to register for Nymbl were more aware of their fall risk, motivated, or simultaneously engaged in other fall prevention strategies, the estimated association of Nymbl may be overstated. Because the study cohort was predominantly White, non‐Hispanic females, our findings may not extend to more diverse populations. While our EHR data likely captured hospitalizations and emergency department visits related to severe fall injuries, less severe FRIs may have gone unreported for individuals who did not seek medical care, nor could we determine if there were injuries attributable to Nymbl use [46, 47]. If individuals who registered for Nymbl were more engaged with the healthcare system, they may have had more opportunities to report FRIs and been less susceptible to recall bias than non‐registrants, potentially resulting in an underestimation of Nymbl's true impact [40]. Additionally, the 12‐month follow‐up period limited our ability to assess the long‐term associations of Nymbl or to evaluate multiple FRI episodes within the same individual.
6. Conclusion
Findings from this study suggest that asynchronous, self‐directed digital balance programs are associated with FRI reductions, strengthening the evidence for integrating remote digital interventions into USPSTF guidelines for fall prevention [2]. However, these associations appear to be limited to individuals with a history of prior FRIs and those not already engaged in other exercise programs, thus implementation should target high‐risk, sedentary populations. Alternatively, the mechanism explaining these findings may involve other unobserved factors, such as an individual's motivation to prevent secondary FRIs being associated with Nymbl registration. With the increasing number of fall prevention applications readily available, sustained patient engagement will be essential to realizing the full impact of digitally delivered interventions. Future research should assess the effectiveness of asynchronous exercise programs through randomized trials and across multiple different fall‐prevention programs, evaluate whether effects vary by participation in other fitness programs such as Silver Sneakers, and quantify long‐term durability and potential cost savings.
Author Contributions
All authors met the criteria for authoriship as follows: Study concept and design: Kris Wain, Wendolyn Gozansky, John Litten, Deanna McQuillan, Claudia Steiner. Acquisition of data: Andrea Daddato, Andrew Jessen, Chris Jentz, Kris Wain. Analysis and interpretation of data: Kris Wain, Wendolyn Gozansky. Preparation of manuscript: Andrea Daddato, Wendolyn Gozansky, Andrew Jessen, John Litten, Deanna McQuillan, Claudia Steiner, Chris Jentz, Kris Wain.
Funding
This work was funded by internal resources from the Kaiser Foundation Health Plan of Colorado, the Colorado Permanente Medical Group, and the Institute for Health Research.
Disclosure
This work was funded by internal resources from Kaiser Permanente.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figure S1: Description of Nymbl balance application.
Figure S2: Distribution of index dates among nymbl participants as compared to distribution of randomly assigned index for age‐eligible non‐participants.
Figure S3: Patient self‐reported fall‐related injury risk factors from the health risk assessment (HRA) tool.
Figure S4: Consort diagram.
Figure S5: Propensity score distribution after IPTW by Nymbl registration, stratified by Pre‐index fall‐related injury.
Figure S6: Distribution of Nymbl sessions in the first 90 days following registration among registrants.
Table S1: Baseline characteristics by Nymbl registration and silver sneakers use.
Table S2: Calculation of unadjusted marginal effects (ME) of Nymbl registration on fall related injury (FRI) episode by acuity level stratified by pre‐index FRI episode.
Table S3: Standardized mean differences by Nymbl registration, stratified by pre‐index fall‐related injury.
Acknowledgments
The authors have nothing to report.
References
- 1. Kakara R., Bergen G., Burns E., and Stevens M., “Nonfatal and Fatal Falls Among Adults Aged ≥65 Years ‐ United States, 2020‐2021,” MMWR. Morbidity and Mortality Weekly Report 72, no. 35 (2023): 938–943, 10.15585/mmwr.mm7235a1. [DOI] [PubMed] [Google Scholar]
- 2. Nicholson W. K., Silverstein M., Wong J. B., et al., “Interventions to Prevent Falls in Community‐Dwelling Older Adults: US Preventive Services Task Force Recommendation Statement,” JAMA 332, no. 1 (2024): 51–57, 10.1001/jama.2024.8481. [DOI] [PubMed] [Google Scholar]
- 3. Smith M. L., Stevens J. A., Ehrenreich H., et al., “Healthcare Providers' Perceptions and Self‐Reported Fall Prevention Practices: Findings From a Large New York Health System,” Frontiers in Public Health 3 (2015): 17, 10.3389/fpubh.2015.00017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Wiseman J. M., Quatman C. E., and Quatman‐Yates C. C., “Examining Factors Influencing Older Adult Engagement in Fall Prevention: A Comparative Analysis Among Stakeholders,” Journal of the American Geriatrics Society 73, no. 4 (2025): 1135–1143, 10.1111/jgs.19330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Fernandes J. B., Fernandes S. B., Almeida A. S., Vareta D. A., and Miller C. A., “Older Adults' Perceived Barriers to Participation in a Falls Prevention Strategy,” Journal of Personalized Medicine 11, no. 6 (2021): 450, 10.3390/jpm11060450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Tzeng H. M., Okpalauwaekwe U., and Lyons E. J., “Barriers and Facilitators to Older Adults Participating in Fall‐Prevention Strategies After Transitioning Home From Acute Hospitalization: A Scoping Review,” Clinical Interventions in Aging 15 (2020): 971–989, 10.2147/cia.S256599. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Pettersson B., Lundin‐Olsson L., Skelton D. A., et al., “Effectiveness of the Safe Step Digital Exercise Program to Prevent Falls in Older Community‐Dwelling Adults: Randomized Controlled Trial,” Journal of Medical Internet Research 27 (2025): e67539, 10.2196/67539. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Delbaere K., Valenzuela T., Lord S. R., et al., “E‐Health StandingTall Balance Exercise for Fall Prevention in Older People: Results of a Two Year Randomised Controlled Trial,” BMJ 373 (2021): n740, 10.1136/bmj.n740. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Hoffman G. J., Ha J., Alexander N. B., Langa K. M., Tinetti M., and Min L. C., “Underreporting of Fall Injuries of Older Adults: Implications for Wellness Visit Fall Risk Screening,” Journal of the American Geriatrics Society 66, no. 6 (2018): 1195–1200, 10.1111/jgs.15360. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Tan Y. Y., Papez V., Chang W. H., Mueller S. H., Denaxas S., and Lai A. G., “Comparing Clinical Trial Population Representativeness to Real‐World Populations: An External Validity Analysis Encompassing 43 895 Trials and 5 685 738 Individuals Across 989 Unique Drugs and 286 Conditions in England,” Lancet Healthy Longevity 3, no. 10 (2022): e674–e689, 10.1016/s2666-7568(22)00186-6. [DOI] [PubMed] [Google Scholar]
- 11. Barsha R. A. A., Assari S., Byiringiro S., et al., “Motivation to Clinical Trial Participation: Health Information Distrust and Healthcare Access as Explanatory Variables and Gender as Moderator,” Journal of Clinical Medicine 14, no. 2 (2025): 485, 10.3390/jcm14020485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Van Spall H. G., Toren A., Kiss A., and Fowler R. A., “Eligibility Criteria of Randomized Controlled Trials Published in High‐Impact General Medical Journals: A Systematic Sampling Review,” JAMA 297, no. 11 (2007): 1233–1240, 10.1001/jama.297.11.1233. [DOI] [PubMed] [Google Scholar]
- 13. McHugh M. D., Aiken L. H., Eckenhoff M. E., and Burns L. R., “Achieving Kaiser Permanente Quality,” Health Care Management Review 41, no. 3 (2016): 178–188, 10.1097/hmr.0000000000000070. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Kanter M. H., Lindsay G., Bellows J., and Chase A., “Complete Care at Kaiser Permanente: Transforming Chronic and Preventive Care,” Joint Commission Journal on Quality and Patient Safety 39, no. 11 (2013): 484–494, 10.1016/s1553-7250(13)39064-3. [DOI] [PubMed] [Google Scholar]
- 15. “Nymbl_Science,” (2024), https://nymblscience.com/balance.
- 16. Papi E., Chiou S. Y., and McGregor A. H., “Feasibility and Acceptability Study on the Use of a Smartphone Application to Facilitate Balance Training in the Ageing Population,” BMJ Open 10, no. 12 (2020): e039054, 10.1136/bmjopen-2020-039054. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Park J. H., “Is Dual‐Task Training Clinically Beneficial to Improve Balance and Executive Function in Community‐Dwelling Older Adults With a History of Falls?,” International Journal of Environmental Research and Public Health 19, no. 16 (2022): 10198, 10.3390/ijerph191610198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Ross T. R., Ng D., Brown J. S., et al., “The HMO Research Network Virtual Data Warehouse: A Public Data Model to Support Collaboration,” EGEMS (Washington, DC) 2, no. 1 (2014): 1049, 10.13063/2327-9214.1049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Min L., Tinetti M., Langa K. M., Ha J., Alexander N., and Hoffman G. J., “Measurement of Fall Injury With Health Care System Data and Assessment of Inclusiveness and Validity of Measurement Models,” JAMA Network Open 2, no. 8 (2019): e199679, 10.1001/jamanetworkopen.2019.9679. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Jacob J., Schmedt N., Hickstein L., Galetzka W., Walker J., and Enders D., “Comparison of Approaches to Select a Propensity Score Matched Control Group in the Absence of an Obvious Start of Follow up for This Group: An Example Study on the Economic Impact of the DMP Bronchial Asthma,” Das Gesundheitswesen 82, no. S 02 (2020): S151–s157, 10.1055/a-0948-5356. [DOI] [PubMed] [Google Scholar]
- 21. Wain K., Maiyani M., Carroll N. M., et al., “Patterns of Medical Care Cost by Service Type Associated With Lung Cancer Screening,” Medical Care 63 (2025): 656–665, 10.1097/mlr.0000000000002169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Shield BCB , “HEDIS Advanced Illness and Frailty Guide,” (2026), https://www.bcbsm.com/amslibs/content/dam/public/providers/documents/advanced‐illness‐frailty‐guide.pdf.
- 23. Kallogjeri D., Gaynor S. M., Piccirillo M. L., Jean R. A., E. L. Spitznagel, Jr. , and Piccirillo J. F., “Comparison of Comorbidity Collection Methods,” Journal of the American College of Surgeons 219, no. 2 (2014): 245–255, 10.1016/j.jamcollsurg.2014.01.059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Messer L. C., Laraia B. A., Kaufman J. S., et al., “The Development of a Standardized Neighborhood Deprivation Index,” Journal of Urban Health 83, no. 6 (2006): 1041–1062, 10.1007/s11524-006-9094-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Chesnaye N. C., Stel V. S., Tripepi G., et al., “An Introduction to Inverse Probability of Treatment Weighting in Observational Research,” Clinical Kidney Journal 15, no. 1 (2022): 14–20, 10.1093/ckj/sfab158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Gabriel E. E., Sachs M. C., Martinussen T., et al., “Inverse Probability of Treatment Weighting With Generalized Linear Outcome Models for Doubly Robust Estimation,” Statistics in Medicine 43, no. 3 (2023): 534–547. [DOI] [PubMed] [Google Scholar]
- 27. Permanente K , “KP Medicare Total Health Assessment (MTHA) Expert Clinical Panel. Medicare Total Health Assessment Questionnaire,” (2025), https://mydoctor.kaiserpermanente.org/ncal/Images/Medicare%20Total%20Health%20Assessment%20Questionnaire_tcm75‐487922.pdf.
- 28. Petersen M. L., Porter K. E., Gruber S., Wang Y., and van der Laan M. J., “Diagnosing and Responding to Violations in the Positivity Assumption,” Statistical Methods in Medical Research 21, no. 1 (2012): 31–54, 10.1177/0962280210386207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Austin P. C., “Balance Diagnostics for Comparing the Distribution of Baseline Covariates Between Treatment Groups in Propensity‐Score Matched Samples,” Statistics in Medicine 28, no. 25 (2009): 3083–3107, 10.1002/sim.3697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Kiami S. R., Sky R., and Goodgold S., “Facilitators and Barriers to Enrolling in Falls Prevention Programming Among Community Dwelling Older Adults,” Archives of Gerontology and Geriatrics 82 (2019): 106–113, 10.1016/j.archger.2019.01.006. [DOI] [PubMed] [Google Scholar]
- 31. Garnett M. F., Weeks J. D., and Zehner A. M., “Unintentional Fall Deaths in Adults Age 65 and Older: United States, 2023,” NCHS Data Brief 532 (2025): 1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Alkhuzaimi F., Rainey D., Wilson C. B., and Bloomfield J., “The Impact of Mobile Health Interventions on Service Users' Health Outcomes and the Role of Health Professions: A Systematic Review of Systematic Reviews‐Protocol,” Systematic Reviews 13, no. 1 (2024): 199, 10.1186/s13643-024-02624-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Xu L., Shi H., Shen M., et al., “The Effects of mHealth‐Based Gamification Interventions on Participation in Physical Activity: Systematic Review,” JMIR mHealth and uHealth 10, no. 2 (2022): e27794, 10.2196/27794. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Mönninghoff A., Kramer J. N., Hess A. J., et al., “Long‐Term Effectiveness of mHealth Physical Activity Interventions: Systematic Review and Meta‐Analysis of Randomized Controlled Trials,” Journal of Medical Internet Research 23, no. 4 (2021): e26699, 10.2196/26699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Ganz D. A., Yuan A. H., Greene E. J., et al., “Effect of the STRIDE Fall Injury Prevention Intervention on Falls, Fall Injuries, and Health‐Related Quality of Life,” Journal of the American Geriatrics Society 70, no. 11 (2022): 3221–3229, 10.1111/jgs.17964. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Bhasin S., Gill T. M., Reuben D. B., et al., “A Randomized Trial of a Multifactorial Strategy to Prevent Serious Fall Injuries,” New England Journal of Medicine 383, no. 2 (2020): 129–140, 10.1056/NEJMoa2002183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Tinetti M. E., Doucette J., Claus E., and Marottoli R., “Risk Factors for Serious Injury During Falls by Older Persons in the Community,” Journal of the American Geriatrics Society 43, no. 11 (1995): 1214–1221, 10.1111/j.1532-5415.1995.tb07396.x. [DOI] [PubMed] [Google Scholar]
- 38. Middleton A., Fritz S. L., and Lusardi M., “Walking Speed: The Functional Vital Sign,” Journal of Aging and Physical Activity 23, no. 2 (2015): 314–322, 10.1123/japa.2013-0236. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Quarch L., Galica A., and Jones R., “The Non‐Linear Relationship Between Gait Speed and Falls: The MOBILIZE Boston Study,” Journal of the American Geriatrics Society 59, no. 6 (2011): 1069–1073. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Sanders K. M., Stuart A. L., Scott D., Kotowicz M. A., and Nicholson G. C., “Validity of 12‐Month Falls Recall in Community‐Dwelling Older Women Participating in a Clinical Trial,” International Journal of Endocrinology 2015 (2015): 210527, 10.1155/2015/210527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Dormosh N., Schut M. C., Heymans M. W., et al., “Predicting Future Falls in Older People Using Natural Language Processing of General Practitioners' Clinical Notes,” Age and Ageing 52, no. 4 (2023): afad046, 10.1093/ageing/afad046. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Appeadu M. K. and Bordoni B., “Falls and Fall Prevention in Older Adults,” in StatPearls [Internet] (StatPearls Publishing, 2023). [PubMed] [Google Scholar]
- 43. Meekes W. M., Korevaar J. C., Leemrijse C. J., and van de Goor I. A., “Practical and Validated Tool to Assess Falls Risk in the Primary Care Setting: A Systematic Review,” BMJ Open 11, no. 9 (2021): e045431, 10.1136/bmjopen-2020-045431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Beck Jepsen D., Robinson K., Ogliari G., et al., “Predicting Falls in Older Adults: An Umbrella Review of Instruments Assessing Gait, Balance, and Functional Mobility,” BMC Geriatrics 22, no. 1 (2022): 615, 10.1186/s12877-022-03271-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Okubo Y., Osuka Y., Jung S., et al., “Walking Can Be More Effective Than Balance Training in Fall Prevention Among Community‐Dwelling Older Adults,” Geriatrics & Gerontology International 16, no. 1 (2016): 118–125, 10.1111/ggi.12444. [DOI] [PubMed] [Google Scholar]
- 46. Faul M., Stevens J. A., Sasser S. M., et al., “Older Adult Falls Seen by Emergency Medical Service Providers: A Prevention Opportunity,” American Journal of Preventive Medicine 50, no. 6 (2016): 719–726, 10.1016/j.amepre.2015.12.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Yosef T., Pasco J. A., Tembo M. C., Williams L. J., and Holloway‐Kew K. L., “Falls and Fall‐Related Injuries: Prevalence, Characteristics, and Treatment Among Participants of the Geelong Osteoporosis Study,” Frontiers in Public Health 12 (2024): 1454117, 10.3389/fpubh.2024.1454117. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Description of Nymbl balance application.
Figure S2: Distribution of index dates among nymbl participants as compared to distribution of randomly assigned index for age‐eligible non‐participants.
Figure S3: Patient self‐reported fall‐related injury risk factors from the health risk assessment (HRA) tool.
Figure S4: Consort diagram.
Figure S5: Propensity score distribution after IPTW by Nymbl registration, stratified by Pre‐index fall‐related injury.
Figure S6: Distribution of Nymbl sessions in the first 90 days following registration among registrants.
Table S1: Baseline characteristics by Nymbl registration and silver sneakers use.
Table S2: Calculation of unadjusted marginal effects (ME) of Nymbl registration on fall related injury (FRI) episode by acuity level stratified by pre‐index FRI episode.
Table S3: Standardized mean differences by Nymbl registration, stratified by pre‐index fall‐related injury.