A retrospective study on machine learning-assisted stroke recognition for medical helpline calls.
Ontology highlight
ABSTRACT:
SUBMITTER: Wenstrup J
PROVIDER: S-EPMC10730829 | biostudies-literature | 2023 Dec
REPOSITORIES: biostudies-literature
Publications
A retrospective study on machine learning-assisted stroke recognition for medical helpline calls.
Wenstrup Jonathan J Havtorn Jakob Drachmann JD Borgholt Lasse L Blomberg Stig Nikolaj SN Maaloe Lars L Sayre Michael R MR Christensen Hanne H Kruuse Christina C Christensen Helle Collatz HC
NPJ digital medicine 20231219 1
Advanced stroke treatment is time-dependent and, therefore, relies on recognition by call-takers at prehospital telehealth services to ensure fast hospitalisation. This study aims to develop and assess the potential of machine learning in improving prehospital stroke recognition during medical helpline calls. We used calls from 1 January 2015 to 31 December 2020 in Copenhagen to develop a machine learning-based classification pipeline. Calls from 2021 are used for testing. Calls are first transc ...[more]
Similar Datasets
Project description:BackgroundTelephone triage is used globally in out-of-hours primary care, to prioritize who needs urgent assessment. Even though children rarely are severely ill, calls about sick children are among the most prevalent, mainly due to parental worry. Pediatric calls are considered challenging, as the call-handler must rely on parents' second-hand information. We aimed to investigate if parents' worry can be used as a predictor of severe illness, and if the content of the calls varies between different grades of worry.MethodsIn a convergent mixed methods study design we asked patients to rate their degree-of-worry before talking to a call-handler. We used quantitative data of degree-of-worry, triage- and patient outcome in pediatric calls (n = 2857), and the qualitative content from 54 calls with subsequent hospitalization ≥24 h.ResultsHigh degree-of-worry was associated with hospitalization ≥24 h (OR 3.33, 95% CI 1.53-7.21). Qualitative findings both confirmed and expanded knowledge of degree-of-worry. Worry was the predominant cause for contact overall, and was mainly triggered by loss-of-control. In calls with high degree-of-worry, the prevalence of loss-of-control was especially high, and the parents had additionally often contacted healthcare services recently. Parents with a foreign accent often rated their worry as high, and these callers were often ignored or interrupted. Calls with low degree-of-worry seemed to occur early during the disease.ConclusionHigh degree of parental worry was associated with severe illness. At the end of calls, call-handlers should ensure that the parent has regained control of the situation to reach increased reassurance and to prevent renewed unnecessary contact. Safety-netting is crucial, as many parents made contact early during the illness and deterioration may develop later. The scoring of parental degree-of-worry may be used as an indicator of potentially severe illness and can easily be implemented at out-of-hours call-centers globally.Trial registrationOriginal study registered at clinicaltrials.gov ( NCT02979457 ).
| S-EPMC9012025 | biostudies-literature
Project description:OBJECTIVES:To identify sociodemographic and health-related characteristics of callers' making repeated calls within 48 hours to a medical helpline, compared with those who only call once. SETTING:In the Capital Region of Denmark people with acute, non-life-threatening illnesses or injuries are triaged through a single-tier medical helpline for acute, healthcare services. PARTICIPANTS:People who called the medical helpline between 18 January and 9 February 2017 were invited to participate in the survey. During the period, 38 787 calls were handled and 12 902 agreed to participate. Calls were excluded because of the temporary civil registration number (n=78), the call was not made by the patient or a close relative (n=699), or survey responses were incomplete (n=19). Hence, the analysis included 12 106 calls, representing 11.131 callers' making single calls and 464 callers' making two or more calls within 48 hours. Callers' data (age, sex and caller identification) were collected from the medical helpline's electronic records. Data were enriched using the callers' self-rated health, self-evaluated degree of worry, and registry data on income, ethnicity and comorbidities. The OR for making repeated calls was calculated in a crude, sex-adjusted and age-adjusted analysis and in a mutually adjusted analysis. RESULTS:The crude logistic regression analysis showed that age, self-rated health, self-evaluated degree of worry, income, ethnicity and comorbidities were significantly associated with making repeated calls. In the mutually adjusted analysis associations decreased, however, odds ratios remained significantly decreased for callers with a household income in the middle (OR=0.71;95% CI 0.54 to 0.92) or highest (OR=0.68;95% CI 0.48 to 0.96) quartiles, whereas immigrants had borderline significantly increased OR (OR=1.34;95% CI 0.96 to 1.86) for making repeated calls. CONCLUSIONS:Findings suggest that income and ethnicity are potential determinants of callers' need to make additional calls within 48 hours to a medical helpline with triage function.
| S-EPMC6720138 | biostudies-literature
Project description:BackgroundParents often contact out-of-hours services due to worry concerning febrile children, despite the children rarely being severely ill. As telephone triage of children is challenging, many children are referred to hospital assessment. This study investigated if video triage resulted in more children staying at home. Secondary aims included safety, acceptability and feasibility of this new triage tool.MethodsIn this prospective quality improvement study, nurse call-handlers enrolled febrile children aged 3 months-5 years to video or telephone triage (1:1), with follow-up within 48 h after call. The setting was an out-of-hours call-center for non-urgent illness in Copenhagen, Denmark, receiving over 1 million calls annually and predominately staffed by registered nurses. Main outcome measure was difference in number of children assessed at hospital within 8 h after call between video-and telephone triage group. Rates of feasibility, acceptability and safety (death, lasting means, transfer to intensive care unit) were compared between the triage groups.ResultsThere was no difference in triage outcome (home care vs. hospital referral) or number of patients assessed at hospital between triage groups. However, more video triaged patients received in-hospital treatment, testing and hospitalization.ConclusionVideo triage was feasible to conduct, acceptable to parents and as safe as telephone triage. The study did not show that more children stayed at home after video triage, possibly because the allocation strategy was not upheld, as video triage sometimes was chosen in cases of complex and severe symptoms, and this likely has changed study outcome.Trial registrationClinicaltrials.gov.: Id NCT04074239. Registered 2019-08-30. https://clinicaltrials.gov/ct2/show/study/NCT04074239.
| S-EPMC10464404 | biostudies-literature
Project description:ImportanceEmergency medical dispatchers fail to identify approximately 25% of cases of out-of-hospital cardiac arrest (OHCA), resulting in lost opportunities to save lives by initiating cardiopulmonary resuscitation.ObjectiveTo examine how a machine learning model trained to identify OHCA and alert dispatchers during emergency calls affected OHCA recognition and response.Design, setting, and participantsThis double-masked, 2-group, randomized clinical trial analyzed all calls to emergency number 112 (equivalent to 911) in Denmark. Calls were processed by a machine learning model using speech recognition software. The machine learning model assessed ongoing calls, and calls in which the model identified OHCA were randomized. The trial was performed at Copenhagen Emergency Medical Services, Denmark, between September 1, 2018, and December 31, 2019.InterventionDispatchers in the intervention group were alerted when the machine learning model identified out-of-hospital cardiac arrest, and those in the control group followed normal protocols without alert.Main outcomes and measuresThe primary end point was the rate of dispatcher recognition of subsequently confirmed OHCA.ResultsA total of 169 049 emergency calls were examined, of which the machine learning model identified 5242 as suspected OHCA. Calls were randomized to control (2661 [50.8%]) or intervention (2581 [49.2%]) groups. Of these, 336 (12.6%) and 318 (12.3%), respectively, had confirmed OHCA. The mean (SD) age among of these 654 patients was 70 (16.1) years, and 419 of 627 patients (67.8%) with known gender were men. Dispatchers in the intervention group recognized 296 confirmed OHCA cases (93.1%) with machine learning assistance compared with 304 confirmed OHCA cases (90.5%) using standard protocols without machine learning assistance (P = .15). Machine learning alerts alone had a significantly higher sensitivity than dispatchers without alerts for confirmed OHCA (85.0% vs 77.5%; P < .001) but lower specificity (97.4% vs 99.6%; P < .001) and positive predictive value (17.8% vs 55.8%; P < .001).Conclusions and relevanceThis randomized clinical trial did not find any significant improvement in dispatchers' ability to recognize cardiac arrest when supported by machine learning even though artificial intelligence did surpass human recognition.Trial registrationClinicalTrials.gov Identifier: NCT04219306.
| S-EPMC7788469 | biostudies-literature
Project description:The human voice stands out for its rich information transmission capabilities. However, voice communication is susceptible to interference from noisy environments and obstacles. Here, we propose a wearable wireless flexible skin-attached acoustic sensor (SAAS) capable of capturing the vibrations of vocal organs and skin movements, thereby enabling voice recognition and human-machine interaction (HMI) in harsh acoustic environments. This system utilizes a piezoelectric micromachined ultrasonic transducers (PMUT), which feature high sensitivity (-198 dB), wide bandwidth (10 Hz-20 kHz), and excellent flatness (±0.5 dB). Flexible packaging enhances comfort and adaptability during wear, while integration with the Residual Network (ResNet) architecture significantly improves the classification of laryngeal speech features, achieving an accuracy exceeding 96%. Furthermore, we also demonstrated SAAS's data collection and intelligent classification capabilities in multiple HMI scenarios. Finally, the speech recognition system was able to recognize everyday sentences spoken by participants with an accuracy of 99.8% through a deep learning model. With advantages including a simple fabrication process, stable performance, easy integration, and low cost, SAAS presents a compelling solution for applications in voice control, HMI, and wearable electronics.
| S-EPMC11894117 | biostudies-literature
Project description:Quantifying sequence-specific protein-ligand interactions is critical for understanding and exploiting numerous cellular processes, including gene expression regulation and signal transduction. Given their importance, next-generation sequencing (NGS) based assays that characterize such recognition with high-throughput are increasingly being used to profile a range of protein classes and interactions. However, these methods do not measure the biophysical parameters that have long been used to uncover the quantitative rules underlying sequence recognition. We developed a highly flexible machine learning framework, called ProBound, to quantify sequence recognition in terms of biophysical parameters based on NGS data. ProBound quantifies transcription factor (TF) behavior with models that accurately predict binding affinity over a range exceeding that of previous resources, captures the impact of DNA modifications and conformational flexibility of multi-TF complexes, and infers specificity directly from \textit{in vivo} data such as ChIP-seq without peak calling. When coupled with a new assay called Kd-seq, it quantifies the absolute affinity of protein-ligand interactions. Its applicability extends beyond thermodynamic equilibrium binding, to the kinetics of kinase-substrate interactions. Altogether, ProBound provides a versatile algorithmic framework for understanding sequence recognition in a wide variety of biological contexts.
2021-06-02 | GSE175942 | GEO
Project description:BACKGROUND:Little is known regarding paediatric medical emergency calls to Danish Emergency Medical Dispatch Centres (EMDC). This study aimed to investigate these calls, specifically the medical issues leading to them and the pre-hospital units dispatched to the paediatric emergencies. METHODS:We performed a retrospective, observational study on paediatric medical emergency calls managed by the EMDC in the Region of Southern Denmark in February 2016. We reviewed audio recordings of emergency calls and ambulance records to identify calls concerning patients ≤ 15 years. We examined EMDC dispatch records to establish how the medical issues leading to these calls were classified and which pre-hospital units were dispatched to the paediatric emergencies. We analysed the data using descriptive statistics. RESULTS:Of a total of 7052 emergency calls in February 2016, 485 (6.9%) concerned patients ≤ 15 years. We excluded 19 and analysed the remaining 466. The reported medical issues were commonly classified as: "seizures" (22.1%), "sick child" (18.9%) and "unclear problem" (12.9%). The overall most common pre-hospital response was immediate dispatch of an ambulance with sirens and lights with a supporting physician-manned mobile emergency care unit (56.4%). The classification of medical issues and the dispatched pre-hospital units varied with patient age. DISCUSSION:We believe our results might help focus the paediatric training received by emergency medical dispatch staff on commonly encountered medical issues, such as the symptoms and conditions pertaining to the symptom categories "seizures" and "sick child". Furthermore, the results could prove useful in hypothesis generation for future studies examining paediatric medical emergency calls. CONCLUSION:Almost 7% of all calls concerned patients ≤ 15 years. Medical issues pertaining to the symptom categories "seizures", "sick child" and "unclear problem" were common and the calls commonly resulted in urgent pre-hospital responses.
| S-EPMC5756442 | biostudies-literature
Project description:ObjectivesTo keep healthcare systems sustainable for future demands, many countries are developing a centralised telephone line for out-of-hours primary care services. To increase the quality of such services, more information is needed on factors that influence caller satisfaction. The aim of this study was to identify demographic and call-related characteristics that are associated with the patient satisfaction of callers to a medical helpline in Denmark.DesignRetrospective cohort study on patient registry data and questionnaire results.SettingNon-emergency medical helpline in the Capital Region of Denmark.ParticipantsA random sample of 30 402 callers to the medical helpline between May 2016 and May 2018.Primary and secondary outcome measuresResponses of a satisfaction questionnaire were linked to demographic and call-related dispatch data. Associations between the characteristics were analysed with multivariable logistic regression analysis with satisfaction as the dependent variable. A subgroup analysis was performed on callers for children aged between 0 and 4 years.ResultsOf the 30 402 analysed callers, 73.0% were satisfied with the medical helpline. Satisfaction was associated with calling for a somatic injury (OR: 1.96, 95% CI: 1.72 to 2.23), receiving a face-to-face consultation (OR: 2.27, 95% CI: 2.04 to 2.50) and a waiting time less than 10 min (OR: 1.82, 95% CI: 1.56 to 2.08). Callers for a 0-year to 4-year-old patient were more likely to be satisfied when they called for a somatic illness or received a telephone consultation, compared with the rest of the population (p<0.0001).ConclusionCallers were in general satisfied with the medical helpline. Satisfaction was associated with reason for encounter, triage response and waiting time. People calling for 0-year to 4-year-old patients were, compared with the rest of the population, more frequently satisfied when they called for a somatic illness or received a telephone consultation.
| S-EPMC6797474 | biostudies-literature
Project description:ObjectivesEmergency Medical Communication Centers (EMCC) have a key role in the prehospital chain-of-stroke-survival by recognizing stroke patients and reducing prehospital delay. However, studies on EMCC stroke recognition report both substantial undertriage and overtriage. Since mis-triage at the EMCC challenges the whole chain-of-stroke-survival, by occupying limited resources for non-stroke patients or failing to recognize the true stroke patients, there is a need to achieve a more comprehensive understanding of the dispatchers' routines and experiences. The aim of this study was to explore factors affecting EMCC dispatcher's decision-making in stroke calls.Materials and methodsA qualitative exploratory study, based on individual semi-structured interviews of 15 medical dispatchers from EMCC Oslo, Norway. Interviews were conducted during August and October 2022 and analyzed using the principles of thematic analysis.ResultsWe identified four themes: [1] Pronounced stroke symptoms are easy to identify [2]. Non-specific neurological symptoms raise suspicion of acute stroke but are difficult to differentiate from other medical conditions [3]. Consistent use of the Criteria Based Dispatch (CBD) protocol may increase EMCC overtriage [4]. Contextual conditions at EMCC can affect dispatchers' decision-making process and the ability for experiential learning.ConclusionsMedical dispatchers at the EMCC perceive vague and non-specific stroke symptoms, such as dizziness, confusion or altered behaviour, challenging to differentiate from symptoms of other less time-critical medical conditions. They also perceive the current CBD protocol in use as less supportive in assessing such symptoms. High workload and strict EMCC response time interval requirements hinder the gathering of essential patient information and the ability to seek guidance in cases of doubt, potentially exacerbating both EMCC undertriage and overtriage. The absence of feedback loops and other strategies for experiential learning in the EMCC hampers the medical dispatcher's ability to evaluate their own assessments and improve dispatch accuracy.
| S-EPMC11566283 | biostudies-literature
Project description:BackgroundMany machine learning heuristics integrate well with Electronic Medical Record (EMR) systems yet often fail to surpass traditional statistical models for biomedical applications.ObjectiveWe sought to compare predictive performances of 12 machine learning and traditional statistical techniques to predict the occurrence of Hospital Acquired Pressure Injuries (HAPI).MethodsEMR information was collected from 57,227 hospitalizations acquired from Dartmouth Hitchcock Medical Center (April 2011 to December 2016). Twelve classification algorithms, chosen based upon classic regression and recent machine learning techniques, were trained to predict HAPI incidence and performance was assessed using the Area Under the Receiver Operating Characteristic Curve (AUC).ResultsLogistic regression achieved a performance (AUC = 0.91 ± 0.034) comparable to the other machine learning approaches. We report discordance between machine learning derived predictors compared to the traditional statistical model. We visually assessed important patient-specific factors through Shapley Additive Explanations.ConclusionsMachine learning models will continue to inform clinical decision-making processes but should be compared to traditional modeling approaches to ensure proper utilization. Disagreements between important predictors found by traditional and machine learning modeling approaches can potentially confuse clinicians and need to be reconciled. These developments represent important steps forward in developing real-time predictive models that can be integrated into EMR systems to reduce unnecessary harm.
| S-EPMC9243224 | biostudies-literature