Smartphone videos for infantile epileptic spasms triaging and assessment (VISTA study): Impact of education and standardized clinical history on diagnostic accuracy Authors: Christine L. Shrock, Margaret C. Savage, Lauren Sham, Christopher Cortina, Kathryn P. Gray, I‐Hsiu Lee, Chahnez C. Triki, Jacqueline French, J. Helen Cross, Orrin Devinsky, Jo M. Wilmshurst, Archana A. Patel Published: 2026 ABSTRACT AbstractObjectiveDiagnostic and treatment delays in infantile epileptic spasms syndrome (IESS) increase the risk of poor neurodevelopmental outcomes. Early clinical recognition of IESS is essential, especially in regions lacking expedited access to electroencephalograms (EEG). This study aimed to determine clinicians' accuracy at recognizing infantile epileptic spasms (ES) based on smartphone videos, and the impact of brief IESS education on accuracy, diagnostic confidence, and willingness to treat without EEG.MethodsThis multicenter prospective cohort study took place over seven sessions globally from 2022 to 2023. Smartphone videos of children from the US and South Africa with EEG‐confirmed diagnoses of IESS (6 videos) and non‐epileptic ES‐mimickers (3 videos) were obtained. Staff physicians and trainee participants from multiple subspecialties worldwide viewed videos three times: (1) baseline viewing, (2) after brief IESS training, and (3) with clinical history. Surveys on diagnosis and management were completed after each viewing.ResultsOf 187 participants who attended a session and initiated a survey, 180 (80 trainees [44%]) met the inclusion criteria. Initial diagnostic accuracy averaged 64% (95% confidence interval [CI]: 62–66%) and improved to 72% (69–74%) after IESS training and clinical history (V + T + CHx). Area under the curve for diagnostic performance of smartphone videos was 0.80 (0.78–0.82), and sensitivity was 0.85 (0.83–0.88) after V + T + CHx. The odds of making a correct diagnosis increased by 86% (OR 1.86, CI 1.59–2.18, p < 0.001) after V + T + CHx. Diagnostic confidence and clinician comfort level treating ES without EEG also improved significantly after V + T + CHx (by 0.36 points and 0.45 points, respectively, on 5‐point Likert scales, p < 0.001). Diagnostic accuracy correlated strongly with increased diagnostic confidence and increased clinician comfort level managing patients without an EEG (p < 0.001). Staff physicians had a 24% higher likelihood of making a correct diagnosis than trainees.SignificanceSmartphone videos, especially when enhanced by brief IESS training, can facilitate triage and early identification of infantile ES, reducing diagnostic delays in this time‐sensitive condition.Plain Language SummaryInfantile epileptic spasms syndrome is associated with severe developmental impacts, which can be worsened by delayed treatment. Rapid diagnosis is critical, especially in resource‐limited settings lacking specialists and timely access to diagnostic tests. Our study found that clinician participants identified epileptic spasms, the hallmark seizure type of this condition, based on video alone with moderately high accuracy, and accuracy improved after education and clinical information. Thus, smartphone videos, particularly when enhanced by brief training, may be an effective tool to triage movements concerning for epileptic spasms, potentially improving resource allocation and reducing diagnostic delays in this urgent childhood epilepsy condition. Key points Delayed treatment of infantile epileptic spasms (ES) can impact development; lack of specialists and EEG are barriers to timely diagnosis.This large international study evaluated clinicians' accuracy and comfort level in using smartphone videos to identify ES.Diagnostic accuracy correlated strongly with diagnostic confidence and clinician comfort level treating patients without EEG confirmation.ES training and clinical history improved diagnostic accuracy, confidence, and comfort treating patients without EEG.Smartphone videos can be an effective tool for triaging movements concerning for ES to expedite diagnosis and treatment. Key points Delayed treatment of infantile epileptic spasms (ES) can impact development; lack of specialists and EEG are barriers to timely diagnosis.This large international study evaluated clinicians' accuracy and comfort level in using smartphone videos to identify ES.Diagnostic accuracy correlated strongly with diagnostic confidence and clinician comfort level treating patients without EEG confirmation.ES training and clinical history improved diagnostic accuracy, confidence, and comfort treating patients without EEG.Smartphone videos can be an effective tool for triaging movements concerning for ES to expedite diagnosis and treatment. Delayed treatment of infantile epileptic spasms (ES) can impact development; lack of specialists and EEG are barriers to timely diagnosis. Delayed treatment of infantile epileptic spasms (ES) can impact development; lack of specialists and EEG are barriers to timely diagnosis. This large international study evaluated clinicians' accuracy and comfort level in using smartphone videos to identify ES. This large international study evaluated clinicians' accuracy and comfort level in using smartphone videos to identify ES. Diagnostic accuracy correlated strongly with diagnostic confidence and clinician comfort level treating patients without EEG confirmation. Diagnostic accuracy correlated strongly with diagnostic confidence and clinician comfort level treating patients without EEG confirmation. ES training and clinical history improved diagnostic accuracy, confidence, and comfort treating patients without EEG. ES training and clinical history improved diagnostic accuracy, confidence, and comfort treating patients without EEG. Smartphone videos can be an effective tool for triaging movements concerning for ES to expedite diagnosis and treatment. Smartphone videos can be an effective tool for triaging movements concerning for ES to expedite diagnosis and treatment. INTRODUCTION Infantile epileptic spasms syndrome (IESS) is characterized by onset of epileptic spasms (ES), a specific seizure semiology of 1–2 s tonic contractions of axial muscles and extremities which may be flexor, extensor, or mixed, which tend to cluster, and occur within the first 24 months of life. 1 IESS is often associated with developmental delay or regression and later drug‐resistant epilepsy. 1 , 2 Estimated incidence of IESS is around 1.6–4.5 per 10 000 live births. 1 , 3 Early detection and treatment of IESS can dramatically reduce seizure burden and improve development and quality of life. 4 , 5 , 6 Clinical history and description of abnormal movements may lead to high suspicion for IESS. However, standard of care includes confirmation with a prolonged electroencephalogram (EEG) capturing sleep, 1 , 7 after which standard first‐line treatment with corticosteroids and/or vigabatrin should be started immediately. 8 , 9 Unfortunately, IESS diagnosis and treatment are often delayed due to lack of recognition of ES as a concerning seizure type by caregivers and non‐specialist providers. Spasms can be subtle, may not be reported to a physician, or may be mistaken for other diagnoses such as gastroesophageal reflux (Sandifer syndrome), benign sleep myoclonus, or exaggerated startle reflex. 1 , 10 , 11 , 12 In resource‐constrained settings, limited timely access to video EEG (vEEG) and to reliable pediatric EEG interpretation can further delay diagnosis, 3 , 13 creating an urgent need for strategies to expedite IESS diagnosis. Studies in adults have shown that smartphone videos can help distinguish seizures from psychogenic nonepileptic events, improving diagnostic accuracy compared to history alone, 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 and that videos may be more helpful than EEG when events are rare. 22 However, data in children are more limited. 10 , 23 , 24 , 25 One retrospective study found that videos of clinical events reduced diagnostic lag time in IESS, suggesting that videos can facilitate triage. 10 Although initial IESS evaluation by video is growing globally due to widespread caregiver access to smartphones, clinicians' ability to utilize this modality effectively has not been formally studied to date. While EEG remains foundational for IESS diagnosis, recent recommendations include a qualifier allowing diagnosis without EEG in resource‐limited regions if spasms are witnessed by experienced clinicians and all other clinical criteria are met, 7 since dependence on EEG can significantly delay diagnosis and treatment. Thus, determining how accurately clinicians can triage patients with possible IESS and make a probable diagnosis based on smartphone video alone has potential for high clinical impact. This study aimed to evaluate the accuracy with which clinicians with varying experience in pediatric epilepsy could discriminate ES, the hallmark epileptic event of IESS, from nonepileptic events (NEE) by caregiver smartphone videos. Secondary objectives included assessing (1) confidence, defined as level of clinical confidence in making a diagnosis via video, and comfort level, defined as how comfortable clinicians felt treating suspected IESS without EEG, (2) impact of brief IESS training and clinical history on diagnostic accuracy, confidence, and comfort level, and (3) differences in outcome measures between staff physicians and trainees. METHODS Study design This was an international multicenter prospective cohort study with an educational component. Home smartphone videos of infants with abnormal movements, confirmed by EEG as either ES or NEE, were solicited from caregivers of children evaluated at two tertiary care hospitals for suspected IESS between January 2020 and May 2022. Relevant clinical history was obtained from charts, including age at the time of evaluation, frequency of events, and a structured report of hallmark IESS clinical features (e.g. propensity to cluster, association with behavioral changes, correlation with sleep–wake transitions, and presence of developmental regression). Study participants were shown nine smartphone videos, six of ES and three of NEE. All videos were truncated to include the most relevant 30 s. Participants were blinded to the electroclinical diagnoses of the children in the videos, and watched each video three times during the session: (1) baseline viewing, (2) after watching a brief tutorial video on IESS characteristics, diagnosis, and management, (3) after relevant clinical history was provided. After each viewing, they answered survey questions regarding diagnosis, level of diagnostic confidence, level of comfort managing the patient without EEG, and need for specialist referral (Figure 1). Clinical history was initially withheld in order to isolate the characterization of events based on smartphone videos alone, as this is an increasingly common practice, as well as to optimize the assessment of the educational intervention and to minimize repetition. Participants scored the quality of the videos at the end of the session. The videos were not distributed, and online sessions were not recorded to preserve patient privacy. Pictorial description of study design. The study was run over seven 90‐min standardized sessions, of which two were held in person and five virtually over Zoom, with participants at institutions in the United States, Canada, Chile, Zambia, as well as an International League Against Epilepsy (ILAE) Eastern Mediterranean & Africa‐sponsored webinar with global attendance. The survey, IESS tutorial, and webinar were translated into French and Spanish by native‐speaking child neurologists. Surveys were conducted using REDCap electronic data capture tools hosted at Boston Children's Hospital 26 between July 2022 and May 2023. Institutional Review Board (IRB) approval (from Boston Children's Hospital Institutional Review Board (IRB‐P00035737) and the University of Cape Town, South Africa Human Research Ethics Committee (IRB‐00001938)) and parental consent were obtained for all videos, as well as for retrospective video and clinical history gathering and for the prospective study. Inclusion criteria Inclusion criteria were: (1) being a staff physician or post‐graduate medical trainee (of any specialty), (2) attendance at one of the virtual or in‐person sessions, and (3) completion of >50% of the survey. Video selection criteria Children 3–18 months of age who were evaluated for possible IESS at Boston Children's Hospital (Massachusetts, USA) or Red Cross War Memorial Children's Hospital (Cape Town, South Africa) were identified by retrospective chart review. Smartphone videos of the events in question were obtained from caregivers with consent. Criteria for video inclusion required a vEEG‐confirmed electroclinical diagnosis of the event by a board‐certified epileptologist, caregiver consent for study use, and appropriate quality based on independent review by three clinicians, using a validated quality rating scale. 20 Thirty‐one videos were collected, and nine videos were selected based on sufficient quality and variable ambiguity, ranging from subtle to obvious events. Of the nine videos, six showed ES and three showed NEE. Six videos were from the USA and three were from South Africa. The quality of videos, on a scale from 1 to 10 (10: highest quality), as scored by the three clinician reviewers, ranged from 7 to 10 (mean 8.5). Of the nine video participants included, five were male (mean age 5.7 months, range 4–7 months [Table S1]). A description of each of the videos, with characterization of ES subtype or non‐epileptic events, is provided in Table S2. Educational intervention An 18‐min pre‐recorded IESS tutorial video was created by a board‐certified epileptologist and shown to participants during the session. Core learning objectives were to recognize IESS presentations, including classic flexor and extensor spasms as well as more subtle or atypical spasms, 6 to understand the diagnostic criteria and the urgency of early diagnosis, to realize the utility of home videos as part of the history taking, and to become familiar with the management and treatment of ES. Outcome measures Our primary outcome was diagnostic accuracy, defined as accurate if the participant's diagnosis based on video review matched the true diagnosis (ES or NEE) and not accurate if the participant's diagnosis either did not match the true diagnosis or was “indeterminate”. Secondary outcomes included level of confidence in the diagnosis, level of comfort managing the patient (i.e. treating IESS versus providing reassurance that the event was nonepileptic) without EEG, and likelihood that a non‐specialist provider would refer to a specialist. Each of the secondary outcomes was based on survey responses involving a 5‐point Likert scale. Statistical analyses Statistical analyses were performed using R‐Statistical Software (v4.3.2) for descriptive analyses and modeling for comparative analyses. Data from all sessions were pooled. Mixed‐effects logistic regression models were used to evaluate the diagnostic performance of smartphone videos alone, after training, and with clinical history provided. Ratings were clustered by video, with video (n = 9) modeled as a random effect to account for within‐video correlation. The diagnostic accuracy was summarized using the area under the receiver operating characteristic (ROC) curve (AUC), with 95% confidence intervals (CIs) estimated via nonparametric bootstrap resampling (2000 replicates) 27 and ROC curves generated using the pROC package in R. 28 Diagnostic performance metrics, including diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were estimated using a model‐derived classification threshold selected to maximize sensitivity (i.e. the point closest to the top‐left‐corner of the ROC curve method 29 ). Corresponding 95% CIs were reported. Sensitivity analyses were also conducted, excluding indeterminate responses from the definition of diagnostic accuracy. Separate mixed‐effects linear regression models were fitted to examine the associations between diagnostic accuracy and diagnostic confidence and comfort managing patients (on 5‐point Likert scales), with clustering by videos accounted for as random effects. RESULTS Participants Across the seven sessions, 187 individuals attended at least part of a session and initiated a survey. Seven attendees did not meet inclusion criteria and were excluded: a medical student, an EEG technologist, a parent, and four attendees who completed <50% of the survey (Figure S1). Of the remaining 180 participants, 80 (44%) were trainees. Participants spanned a range of specialties and practice settings and hailed from 78 countries across all six ILAE regions globally (Table 1). Among all participants, 17% identified as epileptologists and 44% as child neurologists. TABLE 1Survey participant demographics.Survey participant demographics n (%)Total number180TrainingStaff Physician98 (54%)Trainee80 (44%)Not specified2 (1%)SpecialtyGeneral Pediatrics49 (27%)Child Neurology79 (44%)Epilepsy31 (17%)Other pediatric subspecialty5 (3%)Adult subspecialty12 (7%)Not specified4 (2%)Type of practiceAcademic health center102 (57%)Non‐academic hospital practice48 (27%)Group practice4 (2%)Solo practice10 (6%)Other9 (5%)Not specified7 (4%)Region (ILAE categorization)Africa58 (32%)Asia and Oceania13 (7%)Eastern Mediterranean11 (6%)Europe18 (10%)Latin America35 (19%)North America33 (18%)Not specified12 (7%) Survey participant demographics. Diagnostic accuracy The average diagnostic accuracy for all nine ES and NEE videos combined, defined as the percent of videos correctly identified by participants and determined using a mixed‐effects logistic regression model, was 64% (95% confidence interval [CI]: 62–66%) on the initial viewing of the videos alone, 65% (CI 63–68%) on the second viewing after brief IESS training (V + T), and 72% (CI 69–74%) on the third viewing with clinical history provided (V + T + CHx) (Table 2). The estimated AUC (95% CI) improved from 0.69 (0.66–0.72) for videos alone to 0.73 (0.70–0.75) for V + T and to 0.80 (0.78–0.82) for V + T + CHx, suggesting good discrimination between ES and NEE cases (Figure 2A). The estimated sensitivity was 0.67 (0.64–0.71) for videos alone and improved to 0.84 (0.81–0.87) for V + T and to 0.85 (0.83–0.88) for V + T + CHx, while the specificity was lower at 0.61 (0.58–0.65) for videos alone, 0.50 (0.47–0.53) for V + T, and 0.54 (0.50–0.58) for V + T + CHx (Table 2). Overall, the odds of participants making a correct diagnosis increased by 86% (OR 1.86, 95% CI 1.59–2.18, p < 0.001) after V + T + CHx, compared to watching the videos alone, after adjusting for video‐level variation (Figure 2B). TABLE 2Measures of diagnostic performance of videos alone, after training, and with clinical history.Diagnostic measuresVideo alone estimate (95% CI)Video + training estimate (95% CI)Video + training + clinical hx estimate (95% CI) Global Measure of Diagnostic Performance AUC0.69 (0.66–0.72)0.73 (0.70–0.75)0.80 (0.78–0.82) Calculated Using Optimized Threshold (cutoff) Diagnostic Accuracy0.64 (0.62–0.66)0.65 (0.63–0.68)0.72 (0.69–0.74)Sensitivity0.67 (0.64–0.71)0.84 (0.81–0.87)0.85 (0.83–0.88)Specificity0.61 (0.58–0.65)0.50 (0.47–0.53)0.54 (0.50–0.58)PPV0.58 (0.54–0.61)0.58 (0.54–0.61)0.70 (0.67–0.73)NPV0.71 (0.67–0.74)0.80 (0.76–0.83)0.75 (0.71–0.78)Abbreviations: AUC, area under the curve; NPV, Negative Predictive Value; PPV, Positive Predictive Value. Measures of diagnostic performance of videos alone, after training, and with clinical history. Abbreviations: AUC, area under the curve; NPV, Negative Predictive Value; PPV, Positive Predictive Value. (A) Receiver operating characteristic (ROC) curve showing diagnostic performance of smartphone video as a diagnostic tool when viewed alone, after ES training, and with clinical history. (B) Univariate mixed‐effects logistic regression model assessing the odds of a correct diagnosis after training and after clinical history compared with video alone, as well as for staff physicians compared with trainees. Interestingly, in comparing ES to NEE videos, diagnostic accuracy improved after training for all six ES videos but declined after training for the three NEE videos; overall trends showed improvement after clinical history for both ES and NEE videos (Figure S2A and Table S6). Diagnostic accuracy varied widely among individual videos, from 43% to 79% on initial viewing and from 41% to 97% after training and clinical history (Figure 3A and Table S4). (A) Distributions of correct diagnoses for each video after each viewing. (B–D) Summary of responses (shown in percentages) for diagnostic confidence (B), comfort level managing patients without EEG (C), and need for specialist referral (D), on a 1–5 Likert scale, after each viewing. Diagnostic confidence and comfort managing patients without EEG When asked how confident participants were in their diagnosis, on a 1–5 Likert scale (1 “not confident”, 5 “extremely confident”), 34% of participants answered 4 “very confident” or 5 “extremely confident” after viewing the videos alone, which increased to 40% after V + T and then to 51% after V + T + CHx (Figure 3B, Table S5). On average, the confidence score increased by 0.15 points (CI 0.08–0.23, p < 0.001) after V + T, compared to video alone, and by 0.36 points (CI 0.28–0.43, p < 0.001) after V + T + CHx (independent of correct diagnosis) (Figure 4A). (A) Mixed‐effects linear regression model assessing mean difference in confidence and comfort level managing patients without EEG after training and clinical history, compared with video alone. (B) Mixed‐effects linear regression model assessing correlation of confidence and comfort level managing patients without EEG with diagnostic accuracy overall, and then broken down by each viewing, shown as mean difference with 95% CI. When asked how comfortable participants felt managing patients without an EEG (i.e. initiating treatment for IESS or providing reassurance for NEE), on a 1–5 Likert scale (1 “not comfortable”, 5 “extremely comfortable”), 23% of participants answered 4 “very comfortable” or 5 “extremely comfortable” after viewing the videos alone, which increased to 30% after V + T and to 40% after V + T + CHx (Figure 3C, Table S5). On average, the comfort score increased by 0.24 points (CI 0.15–0.32, p < 0.001) after V + T, compared to video alone, and by 0.45 points (CI 0.37–0.54, p < 0.001) after V + T + CHx, (independent of correct diagnosis) (Figure 4A). When comparing ES to NEE videos, both confidence and comfort managing patients without EEG improved more notably for ES videos after V + T + CHx than for NEE videos, although sample size limited statistical analyses (Figure S2B,C and Table S6). Participants rated how strongly they felt that patients needed to be referred to a specialist on a 1–5 Likert scale (1 “not at all”, 5 “extremely strongly”), unless they were self‐identified pediatric epilepsy specialists themselves (47 of the 180 participants), in which case they were excluded from this analysis. Among the remaining 133 participants, 39% responded with 4 “very strongly” or 5 “extremely strongly” after viewing the videos alone, which increased to 40% after V + T and to 46% after V + T + CHx. On the other hand, 11% responded with “not at all” initially, 7% after V + T, and 9% after V + T + CHx. Of these responses recommending no referral, when comparing ES to NEE videos, 6% were for ES videos while 21% were for NEE videos initially, 3% were for ES videos while 15% were for NEE videos after V + T, and 4% were for ES videos while 17% were for NEE videos after V + T + CHx (Figure 3D, Table S5). Correlation of diagnostic accuracy with other measures Diagnostic accuracy correlated strongly with confidence in the diagnosis and with comfort level managing patients without EEG. On average, when a participant made a correct diagnosis, the confidence score (on a 1–5 Likert scale) increased by 0.85 points (95% CI 0.78–0.92, p < 0.001) and the comfort level managing patients without EEG (on a 1–5 Likert scale) increased by 0.81 points (CI 0.73–0.88, p < 0.001), after adjusting for video‐level variation (Figure 4B). Staff physicians versus trainees Diagnostic accuracy between staff physicians and medical trainees (residents and fellows) differed significantly, with a 24% higher likelihood of a correct diagnosis by staff physicians than by trainees (OR 1.24, 95% CI 1.09–1.41, p < 0.001) (Figure 2B, Figure S3A, Table S7). Video quality Participants scored the quality of each video on a 1–5 scale (1 “poor”, 5 “excellent”), with an overall average of 3.4 (CI 3.1–3.7) (Table S8). DISCUSSION This international multicenter prospective cohort study demonstrates that smartphone videos can facilitate diagnosing and triaging IESS. It is, to the best of our knowledge, the largest and most regionally diverse study assessing clinician accuracy and comfort level in using smartphone videos to diagnose IESS. We found that clinicians with varying expertise in pediatric seizure disorders were relatively accurate at recognizing ES based on videos alone, and that clinical history and IESS training significantly improved accuracy. Confidence in the diagnosis correlated strongly with diagnostic accuracy, further demonstrating the value of smartphone videos as a triaging tool, as clinicians generally were able to discern if they could make a correct diagnosis or were uncertain. Thus, our findings support the utility of smartphone videos, a relatively sensitive tool, in improving IESS care by allowing effective triaging of patients who need EEG and specialist referral more urgently and improving resource allocation when sparse. This is particularly relevant in regions where EEG is not readily accessible, representing an estimated 40–60% of the world's population. 30 Moreover, our study demonstrated how use of video triaging can be enhanced significantly by brief (<20 min) IESS training and structured brief clinical history focused on diagnostic characteristics of IESS. Our diagnostic accuracy in identifying ES based on smartphone videos alone, after training and with clinical history provided, was 72% using the most conservative estimate where “indeterminate” responses were counted as incorrect responses. Notably, when sensitivity analyses were performed and “indeterminate” responses were removed, the accuracy was 84% after training and clinical history (Table S3), suggesting that the true accuracy may be higher than our conservative estimate if participants had not been given a choice of responding with “indeterminate”. This was comparable to other studies, including one where diagnostic accuracy was 83% for ES and 81% for psychogenic nonepileptic attacks (now called functional/dissociative seizures) among ten epilepsy experts and nine senior neurology residents combined. 16 Another study reported 45% among residents and 94% among expert pediatric neurologists in correctly diagnosing seizures or nonepileptic events in children, 23 and yet another study found a diagnostic accuracy of 88% among epileptologists, 73% among general neurologists, 54% among internal medicine physicians, and 44% among emergency physicians in distinguishing psychogenic nonepileptic seizures from epileptic seizures. 31 In our study, the AUC for smartphone video performance as a diagnostic tool, after training and with clinical history, reached 0.80, the standard level for good clinical utility. 32 Similar to prior studies with higher accuracy among experts than trainees, 16 , 23 we found that staff physicians had a 24% higher likelihood of making an accurate diagnosis than trainees, as would be expected based on additional experience improving diagnostic acumen. This also aligns with prior studies suggesting that determining seizures from videos is a trained skill. 21 , 33 Of note, diagnostic accuracy in our study consistently decreased after IESS training for NEE videos, while it increased for all ES videos, suggesting that once clinicians learned more about the variety of ways that spasms can appear, they reconsidered their initial diagnoses and recognized the need for more clinical information. This drop in specificity for NEE videos suggests that training could be further enhanced to address more subtle distinctions between ES and NEE, but it also fits with the priorities of a triaging tool, as the risk of incorrectly classifying true seizures as benign is the highest concern, and any uncertain cases would preferably be referred for further evaluation. As anticipated, after clinical history was provided, the overall diagnostic accuracy improved for both ES and NEE videos. Furthermore, diagnostic accuracy varied widely among individual videos, from 43% to 79% on initial viewing of the videos alone, reflective of the range of event types and subtlety included among the nine videos. It has been recognized that there is a diagnostic and treatment gap in epilepsy worldwide, with the diagnostic gap in low‐ and middle‐income countries (LMICs) estimated at 38–61%. 34 , 35 , 36 To address this, the ILAE's Intersectoral Global Action Plan (IGAP) on Epilepsy and Other Neurological Disorders put forth strategic objectives to strengthen the public health approach to epilepsy by 2031 and to provide effective, timely, and responsive diagnosis, treatment, and care for neurological disorders. 37 The South Asia allied IESS Research Group published further recommendations, specifically in LMICs, including improving health care providers' knowledge for early recognition, expedited investigation, and prompt treatment of IESS, as well as scaling telehealth services and infrastructure to maximize availability and utilization to mitigate treatment lag. 35 Since approximately 97% of adults ages 18–49 in the USA, and at least 42% of adults in LMICs (as of 2017), own a smartphone, 38 , 39 it is reasonable to recommend that smartphone videos be used to triage the need for EEG in order to expedite IESS diagnosis and treatment, especially in LMICs, where there is, on average, one child neurologist for every 5 million people and very few, if any, epileptologists. 40 , 41 Limitations Our study has several limitations. First, to maximize broad study participation and balance timing (keeping our sessions to 90 min each), we chose to limit our sample size to nine videos, six ES and three NEE from two contrasting resource settings. This limited some of our sub‐analyses and may have introduced repetition bias, since the same nine videos were shown three times. However, in practice, clinicians typically view videos multiple times, and restricting the number of viewings may have made it actually more challenging. Additionally, we could not separately analyze the effect of providing clinical history before IESS training. Second, for the purposes of this study, only ES or NEE were included, while, in practice, there may be other seizure types to consider (e.g. myoclonic seizures, atonic seizures, etc.). Third, the quality of videos was selected to be varied but generally good, while in practice some videos are poor quality. Fourth, the inclusion of “indeterminate” as a survey response option for diagnosis, while it added more information and represented a realistic clinical option, also complicated our interpretation of results, since 12–20% of respondents chose indeterminate rather than a binary diagnosis of ES or NEE. Fifth, the survey question on the need for referral may have been difficult for participants to interpret, as guidelines for referral vary in different regions. Sixth, the study was largely run over Zoom virtual platforms rather than in person, so challenges with internet connections may have influenced response rates, although this platform allowed us to reach a wider and more diverse population globally. Finally, although the training slides and surveys were translated into other languages, language barriers may have affected some participants. CONCLUSIONS Early recognition and treatment of IESS is critical to reduce severe impacts on developmental outcomes, and the limited availability of EEG and of reliable pediatric EEG interpretation in many regions unfortunately is often a critical barrier to timely diagnosis and management. Our study is the first large‐scale study of its kind that provides evidence that smartphone video is a reasonable triaging tool for patients with suspected IESS to optimize resource allocation of EEGs and specialist consultations, particularly in resource‐constrained regions. Moreover, our study is unique in demonstrating that there is added benefit of brief IESS training and structured clinical history in improving diagnostic accuracy based on smartphone video. Thus, we recommend that for assessment of possible IESS in resource‐constrained regions where EEGs and EEG interpretation by pediatric epilepsy specialists are scarce, the combined use of smartphone video, brief IESS training on diagnosis and management, and structured clinical history can provide a mechanism for improving time to diagnosis and treatment of this urgent condition. AUTHOR CONTRIBUTIONS Christine Shrock was involved in study conception and design, designed the data collection instrument, collected data, planned and carried out analyses, and drafted the initial manuscript. Margaret Savage was involved in study conception and design, designed the data collection instrument, assisted with data collection, drafted the initial manuscript, and critically reviewed and revised the manuscript. Lauren Sham was involved in study conception and design, assisted with data collection, and critically reviewed and revised the manuscript. Christopher Cortina planned and carried out analyses and critically reviewed and revised the manuscript. Kathryn Gray planned and carried out analyses and critically reviewed and revised the manuscript. I‐Hsiu Lee planned and carried out analyses and critically reviewed and revised the manuscript. Chahnez Triki assisted with data collection and critically reviewed and revised the manuscript. Jacqueline French, J. Helen Cross, and Orrin Devinsky were involved in study conception and design and critically reviewed and revised the manuscript. Jo Wilmshurst was involved in study conception and design, coordinated and supervised data collection, and critically reviewed and revised the manuscript. Archana Patel was involved in study conception and design, designed the data collection instrument, coordinated and supervised data collection, planned analyses, and critically reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work. FUNDING INFORMATION None. CONFLICT OF INTEREST STATEMENT Lauren Sham has received unrestricted educational funding from Jazz Pharmaceuticals and LivaNova, and has received consulting fees from Jazz Pharmaceuticals. Jacqueline French receives salary support from the Epilepsy Foundation and from Epilepsy Study Consortium for consulting work and/or attending Scientific Advisory Boards for Acadia Pharmaceuticals, Access Industries, Acuta Capital Partners, AFASCI Inc., Agrithera, Inc., Alkermes, Alterity Therapeutics Limited, Angelini Pharma S.p.A, Autifony Therapeutics Limited, Axonis Therapeutics, Bain Capital, Beacon Biosignals, Inc., Biogen, Biohaven Pharmaceuticals, Bloom Science Inc., BMJ Best Practice Group, Bright Minds Biosciences, Inc., Capsida Biotherapeutics, Cerebral Therapeutics, Cerecin Inc., Cerevel, Ceribell, Cognizance Biomarkers, Cowen and Company, LLC, Crossject, EcoR1 Capital, EG 427, Eisai, Encoded Therapeutics, Engrail, EpiMinder, Epitel Inc., Grin Therapeutics, Harmony/Epygenix, Ionis Pharmaceuticals, iQure Pharma Inc., IQVIA RDS Inc., Janssen Pharmaceutica, Jazz Pharmaceuticals, Leal Therapeutics Inc., LivaNova, London Research and Pharmaceuticals, Longboard Pharmaceuticals, Lundbeck, Maplight Therapeutics, Marinus, Medscape/Web MD, Modulight.bio, Montara Therapeutics, Mosaica Therapeutics, Neumarker, Neumirna Therapeutics, Neurelis, Neurocrine, Neurona Therapeutics, NeuroPace, Inc., NeuroPro Therapeutics, Neuroventis, Neurvati, Noema Pharma, Ono Pharmaceutical Co., Otsuka Pharmaceutical Development, Ovid Therapeutics Inc., Praxis, PureTech LTY Inc., QurAlis, Rapport Therapeutics, Inc., Receptor Holdings Inc., Rivervest Venture Partners, Sage Therapeutics, Inc., SK Life Sciences, Stoke, Stream Neuroscience, Supernus, Takeda, Taysha Gene Therapies, UCB Inc., uniQure, Ventus Therapeutics, Vida Ventures Management, Xenon. She has also received research support from the Epilepsy Study Consortium (funded by Eisai and UCB), Epilepsy Study Consortium/Epilepsy Foundation (funded by UCB), GW/FACES/One8Foundation, NINDS and Praxis. J. Helen Cross has acted as an investigator for studies with GW Pharma (Jazz Pharmaceuticals), Ovid Therapeutics, Zogenix (UCB), Vitaflo, Marinus, Stoke Therapeutics, and Ultragenyx. She has been a speaker and on advisory boards for Jazz Pharmaceuticals, Biocodex, Takeda, Biocodex, and UCB; all remuneration has been paid to her department. She holds an endowed chair at UCL Great Ormond Street Institute of Child Health; she holds grants from NIHR, LifeARC, Action Medical Research, and DPFS. All her research is supported by the National Institute of Health Research (NIHR) Biomedical Research Centre at Great Ormond Street Hospital. Orrin Devinsky receives grant support from NINDS, NIMH, MURI, CDC and NSF. He has equity and/or compensation from the following companies: Tilray, Tevard Biosciences, Regel Biosciences, Script Biosciences, Actio Biosciences, Empatica, Ajna Biosciences, Blackrock Neurotech, and California Cannabis Enterprises (CCE). He has received consulting fees or equity options from Emotiv, UCB Pharma, Ultragenyx, Praxis Precision Therapeutics. He holds patents for the use of cannabidiol in treating neurological disorders that are owned by GW and Jazz Pharmaceuticals; he has waived all financial interests. He holds other patents in molecular biology. He is the managing partner of the PhiFund Ventures. Jo M. Wilmshurst is Associate Editor for Epilepsia. Archana A. Patel receives research funding from the NIH, has received consulting fees from the World Health Organization and Global Strategies, and serves on the medical advisory boards of ROW Foundation and Neurotech Foundation. The remaining authors have no conflicts of interest. ETHICAL APPROVAL Appropriate ethical approval for this study was obtained from Boston Children's Hospital (IRB‐P00035737) and University of Cape Town (IRB‐00001938). Informed consent was obtained from caregivers for all included videos of children under these approved protocols. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. Supporting information Figure S1. Figure S2. Figure S3. Table S1. Table S2. Table S3. Table S4. Table S5. Table S6. Table S7. Table S8.