Abstract
-
Purpose
Stroke is a leading cause of morbidity and mortality worldwide, and rapid early detection is essential. Although screening tools such as FAST (face, arm, speech, and time) are widely used, they may miss posterior strokes, whereas BE-FAST (balance-eyes, face, arm, speech, and time) demonstrates greater sensitivity. However, its implementation in digital formats remains limited. This study aimed to develop, validate, and implement BE-ALERT (balance-eyes-arm weakness-language difficulties-extreme headache-reaction slowed or confusion-time to response) as a community-based early stroke detection application.
-
Methods
This research and development study consisted of development, validation, diagnostic accuracy testing, and community implementation. The development and validation phase included 160 family caregivers of patients with stroke who were aged ≥18 years. The diagnostic accuracy phase included 500 family caregivers who accompanied consecutive patients with suspected stroke in the emergency department. Family caregivers completed the BE-ALERT assessment while accompanying the patients, and their assessment results were compared with neurologist-confirmed diagnoses.
-
Results
Validation showed a content validity index of 0.95 and a Cronbach α reliability coefficient of .82. Community implementation among 160 participants was associated with higher stroke knowledge scores (82.1) and stronger intention to seek immediate treatment (4.5). Receiver operating characteristic curve analysis yielded an area under the curve of 0.836, indicating good diagnostic accuracy. BE-ALERT showed a sensitivity of 85.0%, specificity of 82.1%, negative predictive value of 94.5% and System Usability Scale score of 74.
-
Conclusion
BE-ALERT is a practical, accurate, and well-accepted tool for community-based early stroke detection. It may support stroke screening, public education, and timely treatment-seeking behavior.
-
Key Words: Stroke; Early diagnosis; Mobile applications; Mass screening; Sensitivity and specificity
INTRODUCTION
Stroke occurs when cerebral blood flow is interrupted, resulting in a serious medical condition that requires immediate treatment [
1]. It is a leading cause of death and disability in adults and is one of the most common reasons for emergency department visits [
2]. Prompt treatment is essential to reduce the risk of severe brain injury and improve functional recovery [
3]. Stroke can result in long-term sequelae and increase the risk of recurrence, making early detection critically important. Family members play an essential role in recognizing symptoms and detecting recurrence [
4]. A recent multicenter study reported a 1-year stroke recurrence rate of 9.96%, with a combined death or recurrence rate of 21.83% during the same period, underscoring the importance of early detection of post-stroke complications [
5]. Earlier medical intervention is associated with reduced brain injury; however, many families continue to struggle to recognize stroke symptoms and respond appropriately [
4]. Family-based educational interventions have been associated with improved ability to detect recurrence symptoms and provide recovery support [
6].
Early stroke detection can be performed using various screening tools, including the Cincinnati Prehospital Stroke Scale, FAST (face, arm, speech, and time), Los Angeles Prehospital Stroke Screen, Melbourne Ambulance Stroke Screen, Medic Prehospital Assessment for Code Stroke, and Recognition of Stroke in the Emergency Room. These instruments are structured and designed for ease of use [
7]. The BE-FAST (balance-eyes, face, arm, speech, and time) method allows individuals to self-screen for stroke symptoms. However, effective early detection also depends on healthcare system responsiveness and family caregiving capacity [
8]. Limited public knowledge regarding early detection tools and technologies remains a significant barrier to optimal stroke prevention and treatment [
9].
Globally, stroke is the second leading cause of death and the third leading cause of disability [
10]. According to the World Health Organization, more than 12 million new stroke cases occur annually [
11]. In Indonesia, the prevalence of stroke is approximately 8.3 per 1,000 population (0.83%) [
12]. Delays in recognizing early stroke symptoms frequently result in patients missing the therapeutic “golden period” for reperfusion therapy. FAST and BE-FAST have reported sensitivities of approximately 77% [
13]; however, stroke assessment instruments vary in complexity and length, making it challenging to determine the most appropriate tool during emergencies [
2].
Commonly used screening tools for stroke detection include FAST and BE-FAST. FAST can support early stroke identification at the emergency medical services (EMS) call-taker level, with an acceptable sensitivity of 87.5% for stroke cases subsequently diagnosed by EMS. However, its specificity (17.4%), positive predictive value (PPV) (34.0%), and overall accuracy (40.4%) are low, indicating that FAST is more appropriate as a screening tool to trigger early suspicion and timely resource allocation than as a stand-alone diagnostic tool [
14]. FAST may expedite triage and follow-up when used by EMS call-takers, but definitive diagnosis requires additional examinations by trained medical personnel. Prior studies have recommended improving public and call-taker knowledge of stroke risk factors and symptoms and further evaluating the influence of social and educational variables on FAST screening performance [
15].
BE-FAST is a highly sensitive screening tool for detecting acute ischemic stroke in hospitalized patients, including those meeting criteria for reperfusion therapy; however, specificity is moderate, and false positives among non-cerebrovascular conditions may occur [
16]. Although BE-FAST sensitivity among hospitalized patients is slightly lower than among patients presenting from the community (85% vs. 94%), it retains clinical value, particularly when the patient has an intact level of consciousness [
17].
BE-ALERT (balance-eyes-arm weakness-language difficulties-extreme headache-reaction slowed or confusion-time to response) was developed from existing screening tools (FAST and BE-FAST) for use in the community via a web-based application. In Indonesia, prehospital delays are common; contributing factors include limited awareness of stroke symptoms, cultural beliefs, low socioeconomic status, and poor transportation access, which may result in arrival after the 3.0- to 4.5-hour therapeutic window from symptom onset. The application uses plain language and was tested for validity and reliability. In addition, it provides guidance on accessing healthcare facilities during an event and instructions for home management. Because the success of stroke therapy depends strongly on rapid hospital arrival—and delays in symptom recognition remain common—the design, validation, and implementation of BE-ALERT as an early stroke detection tool are expected to support community-based stroke screening and education.
METHODS
1. Study Design
This study used a research and development (R&D) design with a descriptive quantitative approach to evaluate whether the developed application was valid, reliable, practical, and suitable for use. The study is reported in accordance with the Standards for Reporting Diagnostic Accuracy Studies (STARD) guidelines.
2. Setting and Samples
Participants in phase 1 were family caregivers of patients with stroke who provided care in the home, inpatient ward, or emergency department of a government hospital in Central Java, Indonesia, between February and March 2025. Caregivers were eligible if they were aged ≥18 years, lived in the same household as the patient, owned an Android or iOS smartphone, and were willing to participate. Those with cognitive or communication impairments that could interfere with study participation were excluded. Consecutive sampling was used to recruit 160 family caregivers. This phase assessed caregivers’ knowledge of early stroke symptom recognition and examined whether knowledge differed by age, sex, and educational level. The findings were used to identify educational needs and guide development of the BE-ALERT application. The minimum required sample size for phase 1 was estimated using G*Power version 3.1.9 for an F-test (linear multiple regression, fixed model, R² deviation from zero) with one predictor (total BE-ALERT score). Assuming a medium effect size (f²=0.15), α=.05, and statistical power (1−β) of .95, the minimum required sample size was approximately 90 participants. To improve statistical precision and account for potential missing data, 160 participants were included.
The expert panel comprised six purposively selected specialists: two neurologists, two nursing experts with experience in stroke management, and two health information technology experts. The panel conducted content validation of the BE-ALERT application during the development phase.
Phase 2 consisted of diagnostic accuracy testing of the completed BE-ALERT application and was conducted from April to July 2025. During this predefined recruitment period, 500 family caregivers accompanying 500 consecutive patients with suspected stroke who presented to the Emergency Department of K.R.M.T. Wongsonegoro Regional Hospital, Semarang, Indonesia, were recruited using consecutive sampling. All eligible caregiver-patient pairs presenting during the study period were invited to participate, yielding a final sample of 500 pairs. The phase 2 sample size was determined by the number of eligible consecutive caregiver–patient pairs recruited during the predefined study period. The sample included both stroke-positive and stroke-negative cases, allowing sensitivity, specificity, and other diagnostic accuracy measures to be estimated with 95% confidence intervals.
Eligible participants were family caregivers aged ≥18 years who lived in the same household as the patient and accompanied the patient to the emergency department. Caregivers completed the BE-ALERT assessment on their smartphones while awaiting the patient’s clinical evaluation. The neurologist-confirmed diagnosis, established through neurological examination and appropriate diagnostic investigations, served as the reference standard for diagnostic performance. If caregivers had difficulty accessing or completing the application, trained emergency department nurses, the principal investigator, and research assistants who had received standardized training provided technical assistance; however, caregivers entered all responses directly. The content, scoring algorithm, and assessment results of BE-ALERT were not disclosed to the treating neurologists and were not incorporated into routine clinical assessment or decision-making. Emergency department nurses recorded the neurologist-confirmed diagnosis and supporting examination results on study record forms for research purposes. The research team then used these records as the reference standard for diagnostic accuracy analysis.
3. Measurements/Instruments
The BE-ALERT application was developed based on the BE-FAST framework and incorporates a rapid-action messaging feature. The reference standard for stroke diagnosis was the neurologist-confirmed diagnosis. Application usability was assessed with the System Usability Scale (SUS). BE-ALERT is a family-based observational screening tool designed to help caregivers recognize early stroke symptoms by directly observing patients with suspected stroke. The application includes seven assessment components: balance, eyes, arm weakness, language difficulties, extreme headache, reaction slowed or confusion, and time to response. Detailed scoring criteria for each BE-ALERT component are presented in
Supplementary Data 1. Six components are scored on a four-point scale (1–4), whereas the time to response component is scored on a two-point scale (1–2). Higher scores indicate greater impairment or delayed response, yielding a total possible BE-ALERT score ranging from 7 to 26. For the diagnostic accuracy analysis, participants with a total BE-ALERT score >7 were classified as BE-ALERT positive, indicating possible stroke risk, whereas those with a total score ≤7 were classified as BE-ALERT negative. When the total score exceeded 7, the application displayed the message, “High risk, go to the hospital immediately,” advising users to seek immediate medical evaluation. When the total score was ≤7, the application displayed the message, “No signs of stroke found at this time,” while recommending further medical evaluation if symptoms persisted or worsened. This predefined threshold was used to compare the BE-ALERT assessment as the index test with the neurologist-confirmed diagnosis as the reference standard.
This study employed a modified Borg and Gall R&D framework consisting of seven sequential stages: (1) needs identification, (2) application design and development, (3) content validation, (4) reliability and construct testing, (5) implementation testing, (6) diagnostic accuracy testing, and (7) revision and finalization.
The first stage involved identifying users' needs through a literature review, focus group discussions, and surveys of family caregivers of patients with stroke. These findings were used to develop assessment domains and items based on the BE-FAST framework and to design the initial BE-ALERT prototype. Content validity was then evaluated by an expert panel using the content validity index (CVI), including the item-level content validity index (I-CVI ≥0.78) and scale-level content validity index/average (S-CVI/Ave ≥0.90). Reliability and construct testing were conducted among 160 family caregivers using Cronbach α and the Kuder-Richardson Formula 20 (KR-20), with values ≥.70 considered acceptable.
After reliability testing, implementation testing was conducted through community education sessions using BE-ALERT. Changes in participants’ knowledge and intention to respond to stroke symptoms were evaluated with the Wilcoxon signed-rank test (p<.05), and application usability was assessed with the SUS, with a score ≥68 indicating acceptable usability. Findings from these stages were used to refine the application before diagnostic accuracy evaluation.
A separate diagnostic accuracy study was then conducted to evaluate the finalized BE-ALERT application against a clinical reference standard. Family caregivers accompanying consecutive patients with suspected stroke completed the BE-ALERT web-based assessment on their smartphones while awaiting the patient’s clinical evaluation in the emergency department. If caregivers required assistance accessing or completing the application, trained emergency department nurses, the principal investigator, and research assistants who had received standardized training provided technical support; however, caregivers entered all responses directly.
Before data collection began, the research team informed the physician responsible for the emergency department that a diagnostic accuracy study would be conducted as part of the institutional research approval process. However, the content, scoring algorithm, and assessment results of BE-ALERT were not disclosed to the treating neurologists and were not incorporated into routine clinical assessment or decision-making. The reference-standard diagnosis was established independently according to routine hospital practice through neurological examination and supporting diagnostic investigations performed by the treating neurologist. Emergency department nurses then documented the neurologist-confirmed diagnosis and supporting examination results in the study data collection forms, which the research team used as the reference standard for diagnostic accuracy analysis. The BE-ALERT assessment was completed before the final clinical diagnosis was established, and both the index test and reference-standard diagnosis were completed during the same emergency department visit.
The diagnostic performance of the BE-ALERT application was evaluated using receiver operating characteristic (ROC) curve analysis by comparing the BE-ALERT assessment as the index test with the neurologist-confirmed diagnosis as the reference standard. Sensitivity, specificity, PPV, negative predictive value (NPV), and the area under the ROC curve (AUC) were calculated with corresponding 95% confidence intervals. The final stage involved revising BE-ALERT on the basis of findings from content validation, reliability testing, diagnostic accuracy evaluation, and implementation testing. Revisions focused on the application interface, assessment workflow, wording, navigation, and user guidance, in accordance with expert recommendations and user feedback. After these revisions were incorporated, the final version of BE-ALERT and its accompanying user guide were finalized for community and clinical use.
5. Ethical Considerations
Written informed consent was obtained from all participants. The study approval was obtained from K.R.M.T. Wongsonegoro Regional Hospital, Semarang City (No. 033/Kom.EtikRSWN/II/2025).
6. Data Analysis
Validation analyses included content validity assessment with the CVI and reliability testing with KR-20. Internal consistency was evaluated using Cronbach α, with values ≥.70 considered acceptable. Diagnostic accuracy was evaluated by comparing the BE-ALERT assessment as the index test with the neurologist-confirmed diagnosis as the reference standard, which was established through routine neurological examination and supporting diagnostic investigations according to hospital practice. Diagnostic performance was assessed by calculating sensitivity, specificity, PPV, NPV, ROC curve analysis, and AUC. Sensitivity, specificity, PPV, NPV, and AUC were reported with corresponding 95% confidence intervals (CIs), with CIs for diagnostic accuracy measures estimated using the Wilson method. Implementation testing compared preintervention and postintervention knowledge and intention-to-act scores using the Wilcoxon signed-rank test (p<.05). Application usability was summarized with the SUS. All statistical analyses were performed using IBM SPSS Statistics for Windows, version 27.0 (IBM Corp., Armonk, NY, USA).
RESULTS
1. Participants
A total of 160 family caregivers of patients with stroke participated in the needs assessment phase. For knowledge of stroke symptoms, 45 participants (28.1%) had poor knowledge, 59 (36.9%) had adequate knowledge, and 56 (35.0%) had good knowledge. The most frequently requested BE-ALERT feature was risk logging (53 participants, 33.1%), followed by educational videos (50 participants, 31.3%), easy access (30 participants, 18.8%), and emergency alerts (27 participants, 16.9%).
Knowledge scores differed significantly according to age group. Young adults had significantly lower knowledge scores than adults (59.46±12.35 vs. 74.83±14.27, t=−3.76,
p<.001). No significant difference in knowledge scores was observed according to sex, with mean scores of 74.75±15.60 among males and 72.49±13.83 among females (t=0.97,
p=.333). Knowledge scores also differed significantly according to educational level (F=5.89,
p<.001). Participants with primary education had the highest mean knowledge score (79.83±13.66), followed by those with further education (71.98±13.20), no schooling (71.81±15.13), and higher education (67.13±14.55) (
Table 1).
The needs assessment informed the development of BE-ALERT. Participants most frequently requested risk logging, educational videos, easy access, and emergency alert features, and these preferences were incorporated into the application design (
Table 1).
1) Development domain
The development of the BE-ALERT domain components is presented in
Table 2.
2) BE-ALERT application prototype
Based on the domain results described above, a prototype BE-ALERT application was developed with the following main features.
(1) Main page
The “Early Detection” button directs users to the BE-ALERT form. The “Stroke Education” button provides short videos, infographics, and articles. The “History & Risk Factors” button allows users to record blood pressure, blood glucose, history of hypertension/diabetes, and smoking habits. The “Emergency” button allows users to call the emergency service number (119) with one click.
(2) The early detection form (BE-ALERT)
A structured questionnaire with component-specific response options includes visual icons to help users navigate the questions. For example, an arm icon represents the “arm weakness” component, which assesses weakness or paralysis of the arms. If the total BE-ALERT score is >7, the application automatically advises users to take the patient to the hospital immediately.
(3) Results and recommendations
After the form is completed, the application displays a status message. If the total BE-ALERT score is greater than 7, the application displays the message, "High risk, go to the hospital immediately." If no stroke signs are identified (score ≤7), the application displays the message, "No signs of stroke found at this time," and advises further medical evaluation if symptoms persist or worsen.
The design and development stage was completed by compiling BE-ALERT domains and items and integrating them with multimedia-based stroke education features. Initial testing involved experts and potential users to ensure content suitability and ease of use. The resulting BE-ALERT application has the following main functions: (1) identifying stroke symptoms through simple BE-ALERT-based questions; (2) providing a risk indication score and recommendations to seek immediate medical attention; and (3) delivering interactive educational content on risk factors, symptoms, and the importance of the golden period.
The application was reviewed by experts and tested in the community to evaluate content performance, reliability, diagnostic accuracy, and usability.
4. Content Validation
The content validity assessment yielded I-CVI values ranging from 0.83 to 1.00 and an S-CVI/Ave of 0.95 (
Table 2). Reliability testing yielded a Cronbach’s α coefficient of .82 and a KR-20 coefficient of .80.
Diagnostic accuracy was evaluated using BE-ALERT assessments completed by 500 family caregivers accompanying patients with suspected stroke. According to the neurologist-confirmed reference standard, 120 patients had stroke and 380 were classified as nonstroke. Using the predefined BE-ALERT positivity threshold of a total score >7, 102 accompanying patients were classified as true positives, 312 as true negatives, 68 as false positives, and 18 as false negatives (
Table 3).
BE-ALERT demonstrated a sensitivity of 85.0% (95% CI, 77.0%–91.5%), specificity of 82.1% (95% CI, 77.9%–85.7%), PPV of 60.0% (95% CI, 52.5%–67.2%), and NPV of 94.5% (95% CI, 92.1%–97.0%). ROC analysis yielded an AUC of 0.836 (95% CI, 0.792–0.879) (
Table 3).
Community implementation of the BE-ALERT application was evaluated among 160 family caregivers. The median stroke knowledge score increased from 65.4 (interquartile range [IQR], 60.0–70.0) before the intervention to 82.1 (IQR, 78.0–88.0) after the intervention (Wilcoxon signed-rank test,
p<.001). The median intention-to-act score increased from 3.1 (IQR, 2.0–3.5) before the intervention to 4.5 (IQR, 4.0–5.0) after the intervention (
p<.001). The mean SUS score was 74±11 (
Table 4).
DISCUSSION
BE-ALERT demonstrated good diagnostic performance, with an AUC of 0.836, sensitivity of 85.0%, specificity of 82.1%, and NPV of 94.5%. These findings suggest that the application can discriminate between patients with and without stroke in community-based screening while maintaining a relatively low likelihood of missed stroke cases. The moderate PPV should be interpreted in relation to disease prevalence and the intended role of BE-ALERT as a screening tool rather than a diagnostic tool. The mean SUS score of 74 also indicates acceptable user experience, consistent with previous studies showing that usable mobile health applications may improve public awareness and timely healthcare seeking.
The high I-CVI values and S-CVI/Ave indicate that the assessment items adequately represented the intended constructs and were judged relevant by the expert panel. Similarly, the Cronbach α and KR-20 coefficients indicate acceptable internal consistency and reliability. Together, these findings support the psychometric quality of BE-ALERT before diagnostic accuracy evaluation. This is consistent with previous studies showing that expert-reviewed digital health assessment tools with high content validity and acceptable internal consistency provide a stronger basis for subsequent diagnostic accuracy testing.
Sensitivity and specificity are measures of diagnostic test accuracy. Sensitivity refers to a test’s ability to correctly identify individuals with disease, whereas specificity refers to its ability to correctly identify individuals without disease [
2]. A model with high sensitivity but low specificity may detect most true-positive cases but generate many false alarms. Conversely, a model with low sensitivity but high specificity may miss clinically important cases. Therefore, diagnostic tools should balance sensitivity and specificity according to clinical needs and intended use [
3]. In this study, BE-ALERT demonstrated good overall performance, with high sensitivity and specificity and a moderate PPV, supporting its ability to identify a large proportion of true-positive stroke cases while classifying negative cases with relatively low error.
Stroke is not limited to older adults and can occur in younger populations; however, incidence is most common between ages 45 and 80 years. In Indonesia, public messaging emphasizes early recognition and rapid hospital presentation, including the slogan “SeGeRa Ke RS” [
4]. The government also promotes early action through screening approaches such as FAST. Nevertheless, misconceptions persist, including beliefs that stroke can be treated with massage, bloodletting from the ear, or inserting needles into affected body parts, which may delay hospital presentation [
5]. In response to these contextual barriers and prior evidence, the researchers developed a community-oriented early stroke detection application to support symptom recognition.
The primary objective of this study was to evaluate the validity of a community-based stroke screening application in families caring for stroke patients. Expert I-CVI results indicated that BE-ALERT was acceptable for community use. After the intervention, participants demonstrated improved knowledge and reported better usability of the application. These findings suggest that the application may help families screen for signs and symptoms of stroke and potential recurrence and seek prompt medical assistance.
Factors contributing to screening failure, recurrence, and stroke complications are often related to delayed detection, suboptimal risk factor management, and barriers to service delivery and patient behavior [
6]. Stroke prevalence data are more readily available than incidence data, reflecting limited population-based surveillance in many regions. Much of the available evidence is derived from hospital-based case series, which may show substantial heterogeneity across settings [
7]. In addition, prior systematic reviews have noted limited evidence regarding community-based interventions to prevent stroke [
8].
In this study, public knowledge of stroke symptoms remained limited, with a substantial proportion of respondents categorized as poor or adequate, underscoring the need for digital, application-based stroke education such as BE-ALERT. This finding aligns with prior studies reporting that public knowledge of stroke warning signs and emergency response remains generally low, despite the implementation of FAST/BE-FAST campaigns in many countries. Video-based e-health interventions and mobile applications have been associated with improved patient and family knowledge of stroke risk factors and symptoms, although they may not consistently outperform conventional education. Therefore, local context and user characteristics should be considered when developing BE-ALERT educational content [
9-
11].
Respondents’ preferences for easily accessible features, audiovisual education, risk recording, and emergency alerts are consistent with literature indicating that age, sex, and education level influence adoption and sustained use of e-health applications [
12]. Although smartphone and internet use are widespread, older adults living alone with chronic conditions may have low digital health literacy, which has been associated with sex, age, and education level [
13]. Public awareness of stroke symptoms also varies across countries and is not solely determined by national development status. Educational initiatives in schools and mass media may help improve stroke awareness [
12]. However, prior reports indicate that only about half of participants demonstrate good stroke awareness, highlighting the need to strengthen stroke education efforts in Thailand [
6]. In addition, stroke literacy in a Vietnamese community improved following a single culturally tailored educational session, underscoring the importance of tailored interventions and the need to address barriers related to basic literacy and aging [
14].
Limited internet access highlights the need for applications that support offline functionality to expand educational reach. Overall, user-driven digital platforms may help bridge gaps in stroke awareness, improve healthcare-seeking behavior, and enhance community preparedness for stroke emergencies. However, differences between rural and urban areas may reflect not only rurality but also sociodemographic factors such as education, income, and healthcare access. These considerations suggest that digital application-based education may support more effective understanding than conventional methods in some contexts. Increasing knowledge is important because better understanding of stroke symptoms and risk factors has been associated with earlier detection and reduced treatment delays [
15]. Digital health studies often have design limitations, such as small sample sizes within single regions, and therefore may require multicenter validation to assess reliability and generalizability across populations [
16]. In addition, onset-to-door time and onset-to-needle time are critical outcomes, yet many digital health studies have not evaluated their effects comprehensively [
17]. Future research should consider multicenter designs, long-term clinical impact evaluation, and assessment of cultural factors and digital literacy to improve validity and acceptability.
This study has several limitations. First, the implementation sample was limited to urban areas; therefore, generalizability to rural settings requires further testing. Second, the study did not evaluate the application’s impact on onset-to-door time or onset-to-treatment time, although these outcomes are important for stroke-related clinical outcomes. Third, sociocultural factors and community digital literacy were not examined in depth, although they may affect application acceptance. Overall, limitations include a relatively small sample from a single region, the absence of long-term evaluation of onset-to-door outcomes, and incomplete assessment of users’ sociocultural and digital literacy factors.
Technology-based health interventions tailored to user needs, together with improved digital health literacy, may support optimal self-care for stroke patients in the digital era [
18]. This study is innovative; however, additional evidence regarding long-term clinical outcomes and cross-population validation is needed before BE-ALERT can be recommended as a standard community-based screening approach in Indonesia. Prior work has emphasized the importance of caregiver-oriented, evidence-based mHealth solutions and cross-sector collaboration among healthcare professionals, technology developers, and policymakers to create applications that are accessible, effective, and sustainable [
19]. An additional limitation is that BE-ALERT is currently available only in Indonesian and has been implemented only in Indonesia.
The findings suggest that BE-ALERT has promising psychometric and diagnostic properties for community-based stroke screening. The high CVI and acceptable reliability coefficients indicate that the application content is relevant and internally consistent. The observed sensitivity, specificity, NPV, and AUC also suggest that BE-ALERT may help identify individuals with suspected stroke while minimizing missed cases. Community implementation was associated with improved stroke knowledge and stronger intention to seek prompt medical attention, indicating potential educational benefits. The SUS score exceeded the commonly accepted usability threshold, suggesting that the application is generally acceptable to users. Nevertheless, larger multicenter studies are needed before widespread implementation can be recommended. The iterative revision process allowed the application to be refined before finalization by incorporating expert recommendations, diagnostic evaluation findings, and user feedback from implementation testing.
CONCLUSION
This study showed that BE-ALERT is a family-based, community-oriented application for early stroke screening. The application demonstrated good content validity, acceptable reliability, high diagnostic accuracy, and good usability, suggesting that it may support early recognition of stroke symptoms and reduce prehospital delays, particularly for posterior circulation stroke, which may not be identified with the conventional FAST approach. Community implementation improved caregivers’ knowledge of stroke symptoms and intention to seek immediate medical care. These findings support the potential integration of BE-ALERT into community stroke education and early detection programs to help family caregivers recognize suspected stroke and facilitate timely access to emergency care.
-
CONFLICTS OF INTEREST
The authors declared no conflict of interest.
-
AUTHORSHIP
Study conception and design - LTH, N, TLY, and TWL; analysis - LTH and N; interpretation of the data - LTH; and drafting or critical revision of the manuscript for important intellectual content - LTH, SDMR, M, P, SR, and CA.
-
FUNDING
None.
-
ACKNOWLEDGEMENT
The researchers would like to thank various parties who have helped and participated in the research process.
-
DATA AVAILABILITY STATEMENT
The data can be obtained from the corresponding author.
SUPPLEMENTARY MATERIAL
Table 1.General Characteristics and Needs Assessment of Participants (N=160)
|
Indicators |
Categories |
n (%) |
Knowledge M±SD |
t or F (p) |
|
Aspects assessed |
Public knowledge about stroke symptoms |
|
|
|
|
Poor |
45 (28.1) |
- |
- |
|
Adequate |
59 (36.9) |
- |
- |
|
Good |
56 (35.0) |
- |
- |
|
Requirements for digital application features |
|
|
|
|
Easy to access |
30 (18.8) |
- |
- |
|
Emergency alerts |
27 (16.9) |
- |
- |
|
Educational videos |
50 (31.3) |
- |
- |
|
Risk logging |
53 (33.1) |
- |
- |
|
Characteristics |
Age |
|
|
–3.76 (<.001) |
|
Young adults |
13 (8.1) |
59.46±12.35 |
|
|
Adults |
147 (91.9) |
74.83±14.27 |
|
|
Sex |
|
|
0.97 (.333) |
|
Male |
77 (48.1) |
74.75±15.60 |
|
|
Female |
83 (51.9) |
72.49±13.83 |
|
|
Education |
|
|
5.89 (<.001) |
|
No schooling |
37 (23.1) |
71.81±15.13 |
|
|
Primary education |
52 (32.5) |
79.83±13.66 |
|
|
Further education |
41 (25.6) |
71.98±13.20 |
|
|
Higher education |
30 (18.8) |
67.13±14.55 |
|
Table 2.BE-ALERT Content Validity Test Results
|
BE-ALERT components |
No. of experts value 3–4 |
I-CVI |
Description |
|
Balance |
6/6 |
1.00 |
Very valid |
|
Eyes |
5/6 |
0.83 |
Valid |
|
Arm weakness |
6/6 |
1.00 |
Very valid |
|
Language difficulties |
6/6 |
1.00 |
Very valid |
|
Extreme headache |
5/6 |
0.83 |
Valid |
|
Reaction slowed or confusion |
6/6 |
1.00 |
Very valid |
|
Time to response |
6/6 |
1.00 |
Very valid |
Table 3.Diagnostic Accuracy Tests (N=500)
|
BE-ALERT |
Stroke (+) |
Stroke (–) |
|
BE-ALERT (+) |
102 (TP) |
68 (FP) |
|
BE-ALERT (–) |
18 (FN) |
312 (TN) |
Table 4.Changes in Stroke Knowledge, Intention to Act, and System Usability after Community Implementation (N=160)
|
Pretest |
Posttest |
Δ Median |
p
|
|
Respondent’s knowledge in using the application (n=160) |
65.4 (60.0–70.0) |
82.1 (78.0–88.0) |
+16.7 |
<.001 |
|
Intention to act (n=160) |
3.1 (2.0–3.5) |
4.5 (4.0–5.0) |
+1.4 |
<.001 |
|
SUS score (n=160) |
- |
74±11 |
- |
- |
REFERENCES
- 1. Denti L, Marcomini B, Riva S, Schulz PJ, Caminiti C; for EROI (Educazione e Ritardo di Ospedalizzazione per Ictus) study group. Cross-cultural adaptation of the stroke action test for Italian-speaking people. BMC Neurol. 2015;15:76. https://doi.org/10.1186/s12883-015-0335-z
- 2. Trevethan R. Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice. Front Public Health. 2017;5:307. https://doi.org/10.3389/fpubh.2017.00307
- 3. Monaghan TF, Rahman SN, Agudelo CW, Wein AJ, Lazar JM, Everaert K, et al. Foundational statistical principles in medical research: sensitivity, specificity, positive predictive value, and negative predictive value. Medicina (Kaunas). 2021;57(5):503. https://doi.org/10.3390/medicina57050503
- 4. Ministry of Health Republic of Indonesia. Cegah Stroke dengan Aktivitas Fisik [Prevent stroke through physical activity] [Internet]. Jakarta: Ministry of Health Republic of Indonesia; 2024 [cited 2025 November 29]. Available from: https://kemkes.go.id/eng/cegah-stroke-dengan-aktivitas-fisik
- 5. Ministry of Health Republic of Indonesia. Kenali Gejala Stroke dengan Metode FAST [Recognize the symptoms of stroke using the FAST method] [Internet]. Jakarta: Ministry of Health Republic of Indonesia; 2021 [cited 2025 November 29]. Available from: https://kemkes.go.id/eng/%20kenali-gejala-stroke-dengan-metode-fast
- 6. Dharmasaroja P, Uransilp N. Stroke awareness and knowledge in the at-risk population: a community-based study. Cureus. 2024;16(4):e57756. https://doi.org/10.7759/cureus.57756
- 7. Pandian JD, Padma Srivastava MV, Aaron S, Ranawaka UK, Venketasubramanian N, Sebastian IA, et al. The burden, risk factors and unique etiologies of stroke in South-East Asia Region (SEAR). Lancet Reg Health Southeast Asia. 2023;17:100290. https://doi.org/10.1016/j.lansea.2023.100290
- 8. Nowrin I, Bhattacharyya DS, Saif-Ur-Rahman KM. Community-based interventions to prevent stroke in low-income and middle-income countries: a protocol for a systematic review and meta-analysis. BMJ Open. 2022;12(8):e063181. https://doi.org/10.1136/bmjopen-2022-063181
- 9. Mehta S, Agarwal S, Patel V, Shah Y, Doshi V, Porwal AK, et al. Assessing public awareness of stroke: knowledge of warning signs, risk factors, and treatment responses. Eur J Cardiovasc Med. 2024;14(6):7-12. https://doi.org/10.61336/ejcm/24-06-2
- 10. Park SJ, Lim YS. Development and effectiveness of a mobile application-based health management program for middle-aged men with andropause: a non-equivalent control group pretest-posttest study. Korean J Adult Nurs. 2025;37(3):231-44. https://doi.org/10.7475/kjan.2025.0204
- 11. Favilla CG, Reehal N, Cummings SR, Burdett R, Stein LA, Shakibajahromi B, et al. Personalized video-based educational platform to improve stroke knowledge: a randomized clinical trial. J Am Heart Assoc. 2024;13(15):e035176. https://doi.org/10.1161/JAHA.124.035176
- 12. van Elburg FR, Klaver NS, Nieboer AP, Askari M. Gender differences regarding intention to use mHealth applications in the Dutch elderly population: a cross-sectional study. BMC Geriatr. 2022;22(1):449. https://doi.org/10.1186/s12877-022-03130-3
- 13. Hwang M, Kim G, Lee S, Park YH. Digital health literacy and associated factors among older adults living alone in South Korea: a cross-sectional study. Res Community Public Health Nurs. 2024;35(4):389-400. https://doi.org/10.12799/rcphn.2024.00766
- 14. Ly J, Blair C, Badge H, Camit M, Do K, Pham T, et al. A culturally-specific education strategy to improve stroke health literacy in Vietnamese communities in South Western Sydney. Dialogues Health. 2025;6:100211. https://doi.org/10.1016/j.dialog.2025.100211
- 15. Tsao CW, Aday AW, Almarzooq ZI, Alonso A, Beaton AZ, Bittencourt MS, et al. Heart disease and stroke statistics-2022 update: a report from the American Heart Association. Circulation. 2022;145(8):e153-639. https://doi.org/10.1161/CIR.0000000000001052
- 16. Merino M, Del Barrio J, Nuno R, Errea M. Value-based digital health: a systematic literature review of the value elements of digital health care. Digit Health. 2024;10:20552076241277438. https://doi.org/10.1177/20552076241277438
- 17. Estrela M, Semedo G, Roque F, Ferreira PL, Herdeiro MT. Sociodemographic determinants of digital health literacy: a systematic review and meta-analysis. Int J Med Inform. 2023;177:105124. https://doi.org/10.1016/j.ijmedinf.2023.105124
- 18. Cho MK, Han A, Lee H, Choi J, Lee H, Kim H. Current status of information and communication technologies utilization, education needs, mobile health literacy, and self-care education needs of a population of stroke patients. Healthcare (Basel). 2025;13(10):1183. https://doi.org/10.3390/healthcare13101183
- 19. Long Tuan Kechik TS, Musa KI, Abdullah JM, Kamalakannan S, Sidek NN, Hamzah N, et al. A narrative review on mobile health (mHealth) app for stroke care and rehabilitation intervention for Malaysia. Malays J Med Sci. 2025;32(3):49-72.
, Nursalam3
E-SUBMISSION
