This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.
Recent falls prevention guidelines recommend early routine fall risk assessment among older persons.
The purpose of this study was to develop a Falls Screening Mobile App (FallSA), determine its acceptance, concurrent validity, test-retest reliability, discriminative ability, and predictive validity as a self-screening tool to identify fall risk among Malaysian older persons.
FallSA acceptance was tested among 15 participants (mean age 65.93 [SD 7.42] years); its validity and reliability among 91 participants (mean age 67.34 [SD 5.97] years); discriminative ability and predictive validity among 610 participants (mean age 71.78 [SD 4.70] years). Acceptance of FallSA was assessed using a questionnaire, and it was validated against a comprehensive fall risk assessment tool, the Physiological Profile Assessment (PPA). Participants used FallSA to test their fall risk repeatedly twice within an hour. Its discriminative ability and predictive validity were determined by comparing participant fall risk scores between fallers and nonfallers and prospectively through a 6-month follow-up, respectively.
The findings of our study showed that FallSA had a high acceptance level with 80% (12/15) of older persons agreeing on its suitability as a falls self-screening tool. Concurrent validity test demonstrated a significant moderate correlation (
These results suggest that FallSA is a valid and reliable fall risk self-screening tool. Further studies are required to empower and engage older persons or care givers in the use of FallSA to self-screen for falls and thereafter to seek early prevention intervention.
Falls among older persons are a major health and socioeconomic concern globally [
Falls comprise multifactorial etiology, and they occur as a result of interactions between multiple intrinsic and extrinsic factors. Current evidence-based practice guidelines recommend early falls screening using a multifactorial fall risk assessment tool [
Currently, there are several app-based fall risk assessment tools, including Lindera, Steady, and Aachen fall prevention apps. Lindera is a smartphone-based app designed to facilitate the health care professional (nursing staff) to perform a structured multifactorial fall risk assessment among older adults [
Although multifactorial app-based fall risk assessment tools are currently available, they lack comprehensive information regarding their reliability and validity as self-assessment tools. Prior to the use of fall risk apps, their diagnostic accuracy in discriminating between fallers and nonfallers and predicting actual falls among at-risk older persons must be demonstrated. Existing fall risk assessment tools in clinical settings were found to have only moderate diagnostic accuracy [
This study regarding FallSA was divided into 4 phases comprising (1) development, (2) acceptance among older persons, (3) concurrent validity and test-retest reliability, and (4) discriminative ability and predictive validity.
Ethical approval was obtained from the secretariat for research of ethics of Universiti Kebangsaan Malaysia (UKM 1.5.3.5/244/NN-060-2013 and UKM PPI/111/8/JEP-2018-559). Prior to all studies, participants were given information about the study and were required to provide informed verbal consent.
Prior to the development of FallSA, a literature review was conducted to identify current gaps in the literature regarding fall risk mobile screening apps. Next, several group discussions and meetings were conducted to identify the intended features, interface, and design to meet the functional needs of older persons. After which, preparation of the proposed product features and design using international guidelines for ease of use and usability of graphic user interface for older persons (ISO/IEC 2001) was done.
FallSA was developed based on a model established from our team’s large-scale study report (Neuroprotective Model for Healthy Longevity), which was designed to evaluate the magnitude of cognitive decline and its risk factors through comprehensive multifactorial assessment [
The input button, navigation, and arrangements were developed following the same guidelines. The navigation buttons of FallSA were structured in a rectangle with some simple instruction text at the bottom of the interface; it is displayed as an icon shape on the current user section. In addition, colored navigation buttons were employed in questions sections with a selection of yes in green and no in red. In order to increase ease of use among older persons, a simple navigation system flow was used by having only 2 main selections: new user and current user. The app features for wording, questions, and instructions were bilingual in English and Malay languages, which are commonly used among Malaysian population.
Prior to the use of FallSA, participants were shown an informed consent on data protection policy at the beginning of the app interface. Personal information or data were only saved in the user’s device and could only be assessed by the user and researcher. It is also noteworthy that the information stored in FallSA will be handled in a similar way to hard copies of medical records, as declared in the Data Protection Disclaimer and Laws of Malaysia Act 709 and Personal Data Protection Act 2010. Since FallSA is a self-screening mobile app for identifying fall risk and it does not provide any diagnosis or suggestion to change current medical dosages, it is not categorized as a medical device, which would require an approval from health authorities.
FallSA was developed by a freelance mobile software developer with more than 3 years’ experience in a related field, based on the proposed product features and design of the researcher team. The development of FallSA was conducted according to the waterfall software development process using agile principles for better quality, time, and cost-effective mobile app development.
FallSA user acceptance testing was tested in the real world by 15 older persons.
A convenience sampling method was used to recruit the participants from 2 senior citizen clubs via letters of invitation. Participants were community-dwelling older persons aged 60 years and over and were from the main ethnicities in Malaysia (Malay, Chinese, and Indian). Inclusion criteria included being 60 years and above, able to comprehend the Malay or English language, able to use a smartphone, and able to ambulate with or without assistive devices with minimal supervision. Older persons with acute illnesses (unstable heart diseases and vestibular disorders) were excluded.
Participants were given information regarding the procedures and consent to participate in this study. Participant level of smartphone proficiency was determined verbally before conducting the study. Participants were presented with FallSA on a researcher’s mobile, and they were asked to open the app and follow its instructions with minimal guidance from the researcher. They then needed to complete both levels of navigation (new user and return user) in order to finish a real-life environment response of the software system. The FallSA test took approximately 15 minutes to complete. Upon completion, participants were required to provide feedback using a series of closed- and open-ended questions regarding their understanding of FallSA in relation to its features and design: (1) color contrast, (2) graphics or illustration, (3) font size, (4) presentation of instructional video, and (5) overall FallSA suitability. A modified version of a technology acceptance model survey was also used. The technology acceptance model is rated on 7-point Likert scale and has a high internal reliability (Cronbach 𝛼=.96) and positive correlation between each determinant (perceived usefulness, perceived ease of use, intention and attitude toward use of mobile technology) [
Phase 3 was a cross-sectional study to determine the concurrent validity and test-retest reliability of FallSA among community-dwelling older persons.
Participants were recruited among community-dwelling older persons at two other senior citizen clubs. A total of 91 community-dwelling older persons participated in this study. The sampling method and its inclusion and exclusion criteria were similar to the phase 2 study.
In order to identify the concurrent validity of FallSA, the researcher validated FallSA with an existing fall risk assessment tool. To date, there is no gold standard assessment for fall risk [
Upon screening based on inclusion and exclusion criteria, participants were briefed on the assessments and informed consent was given. Next, participants proceeded with anthropometry measurements (height and weight), followed by collecting their sociodemographic data, fall history, and associated characteristics. Participants were then asked to use the latest version of the FallSA with minimal guidance from the trained enumerator. Participants selected their language preference (Malay/English) and then could proceed to the first section, registration of sociodemographic information. Upon completion of the registration section, participants were directed to the page on physical performance test instruction. Participants were asked to use all methods of instruction: wording, audio, and video. Following this, the interface changed to the physical performance test, with the start and stop timer being a big red button for ease of use among older persons. Two trials were attempted, and the mean was calculated. Next, participants were directed to 4 closed-ended questions regarding fall risk. Report of the overall result was presented upon completion with the whole assessment taking approximately 15 minutes to accomplish.
Participants were given a break of at least 15 minutes before proceeding to the short-form PPA. Trained physiotherapists with more than 1 year of experience were employed to administer the PPA. Participants were asked to perform the 5 tests included in the PPA: (1) edge contrast sensitivity (vision), (2) peripheral sensation (proprioception), (3) finger press (reaction time), (4) standing on the medium-density foam rubber mat (body sway), and (5) knee extension (lower limb strength). On average, the PPA was completed within 30 minutes.
After a 15-minute break upon completion of the PPA, participants repeated the FallSA test to determine its reliability. After verifying their identity card number and date of birth, participants directly proceeded to the physical performance test. Each participant took 10 minutes to complete this FallSA test.
Phase 4 was a cross-sectional study to identify the discriminative ability of FallSA followed by a 6-month prospective follow-up to examine its predictive validity in identifying risk of falls among the older persons in Peninsular Malaysia.
A total of 610 community-dwelling older persons from Peninsular Malaysia (Johor, Selangor, Perak, and Kelantan) participated in this cross-sectional validation study. Participants were selected using a multistage random sampling method. Participants aged 60 years and over able to ambulate independently with or without assistive devices were included in this study, whereas those unable to comprehend and follow instructions, having severe medical conditions, having severe vision or hearing impairments, or having cognitive impairments (dementia or depression) were excluded from this study.
Participant sociodemographic data, fall history, and medical condition were obtained. FallSA assessment was performed after a demo and trial session to assess their risk of fall. In the discriminative study, FallSA scores were compared among fallers and nonfallers according to their past 1-year fall history data. A faller was defined as someone who had one or more falls in the past year and nonfaller as someone without any falls [
Next, the validation study was conducted whereby participants were provided with a fall diary to document fall incidence monthly for a period of 6 months. In addition, participants were contacted via phone monthly to obtain their fall incidence and any other feedback and to remind participants about documenting their falls if any occur. The diaries were collected after 6 months to determine the predictive validity of FallSA. At the end of the 6-month follow-up, participants who had a risk of falls were advised verbally to seek further fall risk assessment and management at their primary health care settings.
The data were analyzed using SPSS (version 22.0, IBM Corp). Researchers performed the normality test for continuous variables in advance by using a Shapiro-Wilk test, Kolmogorov-Smirnov test, kurtosis, skewness ratio, histogram, stem and leaf, or box plot.
Descriptive analysis was performed on the sociodemographic profile. In phase 2, a Pearson correlation analysis was conducted to determine the acceptance level of FallSA among Malaysian community-dwelling older persons. In phase 3, the concurrent validity of FallSA was analyzed using Spearman correlation, kappa for agreement, sensitivity, and specificity. The test-retest reliability was conducted to identify internal consistency (Cronbach alpha), and the intraclass correlation (ICC) and Bland-Altman agreement were determined between 2 trials of the physical performance test. In the final phase, the discriminative ability of FallSA was analyzed using an independent
The figures below depict the screen shots of FallSA (final version): selection of the Malay or English language, which are commonly used in the Malaysian population (
User initial navigation and language selection.
Physical Performance Test (TUG) results.
Sociodemographic and self-rated multifactorial questionnaire.
Fall risk reports.
Participants listed 4 aspects in the FallSA app that were found to be not suitable. First was the physical performance test as the users were required to learn and understand how to perform the TUG test independently. However, most of the participants agreed that the TUG test was easy to use after minimal guidance. Second, participants found the font, especially in the disclaimer interface section, to be small for them. Participants stated that they had difficulty reading the instructions and information when using the app. Third, the system hanged occasionally, probably due to an unstable server and nonsynchronization between the apps and online database. Last, 15% (3/15) of participants requested more interesting graphics or illustrations. Technology Acceptance Model (TAM) survey results showed the presence of a high correlation (0.70 and above) between all determinants (perceived usefulness, perceived ease of use, intention and attitude towards usage of mobile technology).
Sociodemographic information and user evaluation of different aspects of FallSA based on gender.
Characteristic | Males (n=6), n (%) | Females (n=9), n (%) | Total (n=15), n (%) | ||||
|
|||||||
|
|
||||||
|
|
60-69 | 4 (36) | 7 (64) | 11 (73) | ||
|
|
≥70 | 2 (50) | 2 (50) | 4 (27) | ||
|
|
||||||
|
|
Malay | 2 (29) | 5 (71) | 7 (47) | ||
|
|
Chinese | 3 (60) | 2 (40) | 5 (33) | ||
|
|
Indian | 1 (33) | 2 (67) | 3 (20) | ||
|
|
||||||
|
|
None | 3 (75) | 1 (25) | 4 (27) | ||
|
|
Primary | 2 (33) | 4 (67) | 6 (40) | ||
|
|
Secondary | —a | 3 (100) | 3 (20) | ||
|
|
Tertiary | 1 (50) | 1 (50) | 2 (13) | ||
|
|
||||||
|
|
Married | 2 (25) | 6 (75) | 8 (53) | ||
|
|
Widowed | 2 (50) | 2 (50) | 4 (27) | ||
|
|
Divorced | 1 (50) | 1 (50) | 2 (13) | ||
|
|
Single | 1 (100) | — | 1 (7) | ||
|
|||||||
|
|
||||||
|
|
Understand | 5 (36) | 9 (64) | 14 (93) | ||
|
|
Did not understand | 1 (100) | — | 1 (7) | ||
|
|
||||||
|
|
Suitable | 5 (36) | 9 (64) | 14 (93) | ||
|
|
Not suitable | 1 (100) | — | 1 (7) | ||
|
|
||||||
|
|
Suitable | 6 (43) | 8 (57) | 14 (93) | ||
|
|
Not suitable | — | 1 (100) | 1 (7) | ||
|
|
||||||
|
|
Easy to read | 4 (40) | 6 (60) | 10 (67) | ||
|
|
Hard to read | 2 (40) | 3 (60) | 5 (33) | ||
|
|
||||||
|
|
Easy to understand | 6 (40) | 9 (60) | 15 (100) | ||
|
|
Hard to understand | — | — | — | ||
|
|
||||||
|
|
Suitable | 5 (42) | 7 (58) | 12 (80) | ||
|
|
Not suitable | 1 (33) | 2 (67) | 3 (20) |
aNot applicable.
Participant characteristics are shown in
The concurrent validity results between FallSA and PPA are shown in
As for test-retest reliability, there was a significant high reliability between repeated FallSA tests (
Sociodemographic data of the participants based on gender.
Characteristics | Males (n=40), n (%) | Females (n=51), n (%) | Total (n=91), n (%) | |
|
||||
|
60-69 | 19 (32) | 41 (68) | 60 (66) |
|
≥70 | 21 (68) | 10 (32) | 31 (34) |
|
||||
|
Malay | 29 (47) | 33 (53) | 62 (68) |
|
Chinese | 9 (36) | 16 (64) | 25 (28) |
|
Indian | 2 (50) | 2 (50) | 4 (4) |
|
||||
|
None | —a | 2 (100) | 2 (2) |
|
Primary | 8 (53) | 7 (47) | 15 (17) |
|
Secondary | 19 (41) | 27 (59) | 46 (51) |
|
Tertiary | 13 (46) | 15 (54) | 28 (31) |
|
||||
|
Married | 37 (56) | 29 (44) | 66 (73) |
|
Widowed | 1 (6) | 17 (94) | 18 (20) |
|
Divorced | — | 2 (100) | 2 (2) |
|
Single | 2 (40) | 3 (60) | 5 (6) |
aNot applicable.
Concurrent validity of FallSA against Physiological Profile Assessment (PPA).
Variables | Kappa | Sensitivity (%) | Specificity (%) | ROCa | |||||||
|
|||||||||||
|
Males | .538 | .494 | 88.9 | 74.2 | 0.749 | |||||
|
Females | .502 | .429 | 70.7 | 90.0 | 0.790 | |||||
|
|||||||||||
|
Low | .511 | .628 | 81.8 | 83.3 | 0.902 | |||||
|
High | .427 | .489 | 80.0 | 69.2 | 0.776 | |||||
Total score | .518 | .516 | 80.4 | 71.1 | 0.794 |
aROC: receiver operating characteristic.
Reliability of the FallSA total score.
Characteristic | Cronbach alpha | ICCa | 95% CI | SEMb | |||||
Total score FallSA | .948 | .948 | .921-.966 | 1.11 | |||||
|
|||||||||
|
Male | .920 | .920 | .849-.958 | 0.98 | ||||
|
Female | .929 | .929 | .875-.959 | 1.21 | ||||
|
|||||||||
|
Low | .974 | .974 | .929-.991 | 0.88 | ||||
|
High | .942 | .942 | .908-.963 | 1.15 |
aICC: intraclass correlation.
bSEM: standard error of the mean.
Limit of agreement.
A total of 1005 community-dwelling older persons aged 60 years and above within Peninsular Malaysia (Selangor, Perak, Johor, and Kelantan states) were screened for the inclusion criteria; 395 were excluded due to having a score ≥5 on the Geriatric Depression Scale, a score ≤21 on the Mini Mental State Examination, or failing to complete the FallSA test. The sociodemographic data of the participants is shown in
After 6 months, 74.4% (454/610) of participants were successfully followed up via fall diary and monthly phone calls. The 26.6% (156/610) who were dropouts were excluded from the follow-up analysis. About 14.5% (66/454) of the community-dwelling older persons had a fall after the 6-month follow-up. FallSA scores were compared with actual falls reported after the 6-month follow-up; 3% (3/66) of participants categorized as low risk and 18% (63/66) of participants categorized as at-risk experienced a fall.
The cutoff, sensitivity and specificity values, and positive and negative likelihood ratios of FallSA are presented in
The ROC of FallSA is shown in
Baseline sociodemographic data classified based on fallers and nonfallers.
Variables | Total (n=610) | Fallers (n=111; 18.2%) | Nonfallers (n=499; 81.8%) | |||||
Age, mean (SD) | 71.78 (4.7) | 72.01 (4.6) | 71.73 (4.8) | .58 | ||||
MMSEa, mean (SD) | 26.48 (2.4) | 26.43 (2.4) | 26.49 (2.4 | .82 | ||||
GDSb, mean (SD) | 1.85 (1.3) | 1.96 (1.3) | 1.83 (1.3) | .32 | ||||
PASEc, mean (SD) | 125.76 (54.4) | 125.49 (50.8) | 125.82 (55.3) | .95 | ||||
|
—d | — | — | .001 | ||||
|
Male | 332 (54.4) | 45 (13.6) | 287 (86.4) | — | |||
|
Female | 278 (45.6) | 66 (23.7) | 212 (76.3) | — | |||
|
— | — | — | .85 | ||||
|
Malay | 361 (59.2) | 65 (18.0) | 296 (82.0) | — | |||
|
Chinese | 218 (35.7) | 40 (18.3) | 178 (81.7) | — | |||
|
Indian | 31 (5.1) | 6 (19.4) | 25 (80.6) | — | |||
|
— | — | — | .79 | ||||
|
None | 63 (10.3) | 10 (15.9) | 53 (84.1) | — | |||
|
Primary | 306 (50.2) | 60 (19.6) | 246 (80.4) | — | |||
|
Secondary | 199 (32.6) | 34 (17.1) | 165 (82.9) | — | |||
|
Tertiary | 31 (5.1) | 7 (22.6) | 24 (77.4) | — | |||
|
Other | 11 (1.8) | 0 | 11 (100) | — | |||
|
— | — | — | — | ||||
|
Hypertension | 83 (13.6) | 18 (21.7) | 65 (78.3) | .83 | |||
|
Diabetes | 163 (26.7) | 34 (20.9) | 129 (79.1) | .84 | |||
|
Heart disease | 62 (10.2) | 15 (24.2) | 47 (75.8) | .55 | |||
|
Arthritis | 209 (34.3) | 44 (21.1) | 165 (78.9) | .15 | |||
|
— | — | — | .001 | ||||
|
No falls | 499 (81.8) | 0 | 499 (100) | — | |||
|
1 fall | 68 (11.1) | 68 (100) | 0 | — | |||
|
≥2 falls | 43 (7.1) | 43 (100) | 0 | — | |||
|
— | — | — | .47 | ||||
|
< meds | 440 (72.1) | 75 (17.0) | 365 (83.0) | — | |||
|
≥ meds | 170 (27.9) | 36 (21.2) | 134 (78.8) | — | |||
|
— | — | — | .56 | ||||
|
Yes | 105 (17.2) | 17 (16.2) | 88 (83.8) | — | |||
|
No | 505 (82.8) | 94 (18.6) | 411 (81.4) | — | |||
|
— | — | — | .05 | ||||
|
Yes | 31 (5.1) | 11 (35.5) | 20 (64.5) | — | |||
|
No | 579 (94.9) | 100 (17.3) | 479 (82.7) | — | |||
FallSA score, mean (SD) | 5.05 (2.15) | 7.33 (1.77) | 4.55 (1.86) | .001 | ||||
|
— | — | — | .001 | ||||
|
Low risk | 132 (21.6) | 1 (0.8) | 131 (99.2) | — | |||
|
At risk | 478 (78.4) | 110 (23.0) | 368 (77.0) | — |
aMMSA: Mini Mental State Examination.
bGDS: Geriatric Depression Scale.
cPASE: Physical Activity Scale for Elderly.
dNot applicable.
Criterion values and coordinates of the receiver operating characteristic curve in the FallSA fall risk score.
Criterion | Sna | Spb | +LRc | –LRd | PPVe | NPVf |
≥2 | 100 | 0 | 1 | —g | 14.5 | — |
>2 | 95.45 | 20.36 | 1.20 | 0.22 | 16.9 | 96.3 |
>3 | 95.45 | 25.26 | 1.28 | 0.18 | 17.8 | 97.0 |
>4 | 82.35 | 54.90 | 1.91 | 0.25 | 24.6 | 95.9 |
>5h | 81.82 | 63.92 | 2.27 | 0.28 | 27.8 | 95.4 |
>6 | 65.15 | 81.70 | 3.56 | 0.43 | 37.7 | 93.2 |
>7 | 48.48 | 91.75 | 5.88 | 0.56 | 50.0 | 91.3 |
>8 | 33.33 | 95.62 | 7.61 | 0.70 | 56.4 | 89.4 |
>9 | 6.06 | 99.23 | 7.84 | 0.95 | 57.1 | 86.1 |
>10 | 4.55 | 99.48 | 8.82 | 0.96 | 60.0 | 86.0 |
>11 | 0 | 100 | — | 1 | — | 85.5 |
aSn: sensitivity.
bSp: specificity.
cLR+: positive likelihood ratio.
dLR–: negative likelihood ratio.
ePPV: positive predictive value.
fNPV: negative predictive value.
gNot applicable.
hCutoff score.
Receiver operating characteristic of FallSA based on the 6-month follow-up.
In our study, we successfully developed an accepted and validated mobile app, FallSA, that has the potential to identify a risk of falls among a multiethnic Malaysian older population. The FallSA score manifested a moderate to high discriminative ability and predictive validity in classifying fallers and nonfallers and predicting falls among older persons. A cutoff score of >5.0 is recommended in our FallSA predictive results. With this cutoff, FallSA had a positive likelihood ratio of 2.27, pooled sensitivity of 82%, and specificity of 64% with an average AUC of 0.802. However, only 4.5% of older persons screened to have a low fall risk using FallSA had experienced falls in the previous 6 months in our study. This suggests that the present sensitivity and specificity of the FallSA test would be more useful for nonfallers.
Our findings showed that 80% of the participants agreed that FallSA is suitable for use as a self-screening tool among Malaysian community-dwelling older persons. Similarly, the Aachen fall prevention app [
There were several rounds of adjustments in the development stages (requirement design, implementation, and verification) of the FallSA app that included changes in concept, design, graphics, contents, navigation flow, and technical and language corrections. In the user acceptance stage, the design of FallSA was well accepted and older persons had a positive attitude toward FallSA’s adoption with some minor use issues. We addressed all the highlighted issues in the acceptance test before proceeding to conduct reliability and validity tests. For example, font size in FallSA was increased within the limitations of smartphone display size. In addition, we provided bilingual and simplified video instructions for the TUG test. Increasing the number of icons used in the graphic user interface and rectifying problems in the coding system that caused the system to hang during use were also addressed.
Upon addressing participant feedback regarding FallSA, we examined the concurrent validity of FallSA against PPA among 91 community-dwelling older persons. To date, there is no gold standard assessment tool to measure fall risk. However, PPA was identified as the most comprehensive evidence-based practice fall risk assessment tool available. Our study results showed that there is a significant moderate correlation between FallSA and PPA fall risk measurements (
Compared to the concurrent validity of other fall screening tools such as the Fall Risk Questionnaire [
Test-retest analysis of first and second FallSA score results showed that all parameters had an excellent test-retest reliability with Cronbach 𝛼=.948 (ICC .948, 95% CI .921-.966; SEM 1.11), suggesting FallSA is consistent in assessing fall risk among older persons. In addition, a good agreement was demonstrated with Bland-Altman analysis, having a small mean difference and narrow limits of agreement between the first and second FallSA assessments.
The fall prevalence demonstrated in our phase 4 study was 18.2%, which corresponded to earlier local studies conducted among community-dwelling older persons in Malaysia (15% to 18%) [
There was a moderate to high discriminative ability and predictive validity in discriminating fallers and nonfallers and predicting falls among older persons using FallSA. The pooled predictive sensitivity (82%) and specificity (64%) of FallSA are comparable to the Aachen fall prevention app (sensitivity 57.0%; specificity 76.7%) in discriminating between fallers and nonfallers. It is noteworthy that the FallSA predictive validity was conducted prospectively based on actual falls, while the Aachen fall prevention app used its primary outcome in a cross-sectional manner. Using a FallSA cutoff score of >5.0, 95.5% of falls after the 6-month follow-up among community-dwelling older persons were predicted. The number of false negative results can be reduced along with a lower negative likelihood ratio and further reinforced with a higher sensitivity value [
Comparably, the AUC (0.84) and sensitivity value (93%) of Lindera were much higher than FallSA. This discrepancy is possibly due to the variation in study methodology whereby the discriminative ability of Lindera was based on retrospective fall information, whereas for FallSA it was conducted prospectively. Moreover, FallSA is developed as a self-screening fall risk assessment tool to support the older persons as an early fall screening tool, while Lindera is specifically designed to assist health care professionals in clinical settings to identify fall risk among the older people. FallSA having a higher average sensitivity compared to specificity value is more suitable for early fall screening and prevention. This will allow a greater proportion of older persons to be screened and participate in fall prevention programs. This is in line to support the call for routine fall risk assessment and early management in the updated National Institute for Health and Care Excellence (NICE) fall prevention statements [
FallSA has a higher sensitivity value when compared to traditional fall risk assessment using the TUG test. TUG test sensitivity as a fall risk assessment was reported to range from 30.5% to 67.5% [
The model used in FallSA was derived from both urban and rural community-dwelling older populations. Hence FallSA’s use among older persons living in institutions and with frailty will require adaptations to the present model used. Another limitation of FallSA is that it does not provide information for specific impairment areas for tailored interventions as PPA does. However, FallSA is meant for preliminary fall risk self-screening. Last, the duration of follow-up for FallSA may be relatively short in this study, and the changes of health status among older persons may not have been accounted for.
The main strength of FallSA is that it was developed systematically and followed by testing its acceptance concurrent validity, reliability, and discriminative and predictive validity. In addition, we used prospective fall monitoring to identify the predictive validity of FallSA. Note that our study comprised older persons from Peninsular Malaysia and all 3 main ethnicities (Malays, Chinese and Indians) and the results can be generalized to the entire community-dwelling older population.
Clinically, FallSA has the potential to be used as a self-screening, caregiver administered, or at primary health care settings as an early fall risk screening tool. This will assist in the annual fall risk screening among older adults as outlined in the NICE updated fall prevention guidelines [
In this study, we successfully developed a mobile app (FallSA) to identify fall risk among community-dwelling older persons that was accepted, valid, and reliable. FallSA is a short multifactorial assessment tool as it integrates sociodemographic, clinical, and physical fall risk factors. Although NICE updated fall prevention guidelines have recommended an annual fall risk screening, community-dwelling older persons in our local setting tend to visit primary health care settings more often. Since this test is self-administered, we suggest at least a biannual fall risk screening among older persons. To the best of our knowledge, FallSA is one of the most comprehensively tested fall risk self-assessment tools. FallSA has the potential to be used as one of the fall prevention strategies with the ultimate aim of maintaining independence and improving quality of life as long as possible among older persons. Future studies are required to empower and engage older persons or care givers in the use of FallSA to self-screen for falls and thereafter seek early prevention intervention.
area under the curve
Falls Risk Screening App
intraclass correlation
Long-Term Research Grant Scheme—Toward Useful Ageing
National Institute for Health and Care Excellence
Physiological Profile Assessment
receiver operating characteristic
Timed Up and Go
The authors would like to thank the participants for their support. The work presented in this paper was supported by grant LRGS/BU/2012UKM/L/01 (LRGS TUA) awarded by the Ministry of Education and grant Inovasi-2019-001 from Universiti Kebangsaan Malaysia.
None declared.