Class imbalance on medical image classification: towards better evaluation practices for discrimination and calibration performance.

PURPOSE: This work aims to assess standard evaluation practices used by the research community for evaluating medical imaging classifiers, with a specific focus on the implications of class imbalance. The analysis is performed on chest X-rays as a case study and encompasses a comprehensive model performance definition, considering both discriminative capabilities and model calibration. MATERIALS AND METHODS: We conduct a concise literature review to examine prevailing scientific practices used when evaluating X-ray classifiers. Then, we perform a systematic experiment on two major chest X-ray datasets to showcase a didactic example of the behavior of several performance metrics under different class ratios and highlight how widely adopted metrics can conceal performance in the minority class. RESULTS: Our literature study confirms that: (1) even when dealing with highly imbalanced datasets, the community tends to use metrics that are dominated by the majority class; and (2) it is still uncommon to include calibration studies for chest X-ray classifiers, albeit its importance in the context of healthcare. Moreover, our systematic experiments confirm that current evaluation practices may not reflect model performance in real clinical scenarios and suggest complementary metrics to better reflect the performance of the system in such scenarios. CONCLUSION: Our analysis underscores the need for enhanced evaluation practices, particularly in the context of class-imbalanced chest X-ray classifiers. We recommend the inclusion of complementary metrics such as the area under the precision-recall curve (AUC-PR), adjusted AUC-PR, and balanced Brier score, to offer a more accurate depiction of system performance in real clinical scenarios, considering metrics that reflect both, discrimination and calibration performance. CLINICAL RELEVANCE STATEMENT: This study underscores the critical need for refined evaluation metrics in medical imaging classifiers, emphasizing that prevalent metrics may mask poor performance in minority classes, potentially impacting clinical diagnoses and healthcare outcomes. KEY POINTS: Common scientific practices in papers dealing with X-ray computer-assisted diagnosis (CAD) systems may be misleading. We highlight limitations in reporting of evaluation metrics for X-ray CAD systems in highly imbalanced scenarios. We propose adopting alternative metrics based on experimental evaluation on large-scale datasets.

Development and validation of nonattendance predictive models for scheduled adult outpatient appointments in different medical specialties.

INTRODUCTION: Nonattendance is a critical problem that affects health care worldwide. Our aim was to build and validate predictive models of nonattendance in all outpatients appointments, general practitioners, and clinical and surgical specialties. METHODS: A cohort study of adult patients, who had scheduled outpatient appointments for General Practitioners, Clinical and Surgical specialties, was conducted between January 2015 and December 2016, at the Italian Hospital of Buenos Aires. We evaluated potential predictors grouped in baseline patient characteristics, characteristics of the appointment scheduling process, patient history, characteristics of the appointment, and comorbidities. Patients were divided between those who attended their appointments, and those who did not. We generated predictive models for nonattendance for all appointments and the three subgroups. RESULTS: Of 2,526,549 appointments included, 703,449 were missed (27.8%). The predictive model for all appointments contains 30 variables, with an area under the ROC (AUROC) curve of 0.71, calibration-in-the-large (CITL) of 0.046, and calibration slope of 1.03 in the validation cohort. For General Practitioners the model has 28 variables (AUROC of 0.72, CITL of 0.053, and calibration slope of 1.01). For clinical subspecialties, the model has 23 variables (AUROC of 0.71, CITL of 0.039, and calibration slope of 1), and for surgical specialties, the model has 22 variables (AUROC of 0.70, CITL of 0.023, and calibration slope of 1.01). CONCLUSION: We build robust predictive models of nonattendance with adequate precision and calibration for each of the subgroups.

[Telemedicine: validation of a questionnaire to evaluate the experience of health professionalsTelemedicina: validação de um questionário para avaliar a experiência dos profissionais de saúde]. / Telemedicina: validación de un cuestionario para evaluar la experiencia de los profesionales de la salud.

Objective: This objective of this work is to develop and validate a questionnaire to evaluate health professionals' experience with telemedicine systems. Methods: Based on an abbreviated, locally validated Spanish-language version of the patient questionnaire developed by Parmanto et al., a group of experts developed a version to evaluate the experience of health professionals who provide telemedicine services. The psychometric behavior of the items was tested in an initial sample of 129 professionals, using exploratory factor analysis. The comprehensibility of the items was then assessed through cognitive interviews. Finally, in a new sample of 329 professionals, the construct validity of the questionnaire was evaluated by means of confirmatory factor analysis (CFA); its criteria of external validity were assessed by comparing the score with that of a summary question. Results: A 12-item questionnaire was obtained, with a two-factor structure and acceptable adjustment indicators documented through CFA. Reliability, convergent validity, and discriminant validity were appropriate. The criteria of external validity showed optimal results. Conclusions: The instrument obtained has adequate psychometric properties and will contribute to the objective evaluation of the experience of health professionals who perform telemedicine.

Objetivo: Desenvolver e validar um questionário para avaliar a experiência dos profissionais de saúde com os sistemas de telemedicina. Métodos: Com base na versão abreviada em espanhol ­ e validada localmente ­ do questionário para pacientes desenvolvido por Parmanto et al., um grupo de especialistas gerou uma versão de consenso para avaliar a experiência de profissionais de saúde que prestam serviços de telemedicina. O comportamento psicométrico dos itens foi testado em uma primeira amostra de 129 profissionais, por meio de análise fatorial exploratória. Em seguida, sua compreensibilidade foi avaliada por meio de entrevistas cognitivas. Por fim, em uma nova amostra de 329 profissionais, avaliou-se a validade de construto do questionário por meio de uma análise fatorial confirmatória (AFC), e sua validade de critério externo, mediante a avaliação de sua pontuação com a de uma pergunta resumo. Resultados: Obteve-se um questionário de 12 itens com estrutura de dois fatores, com indicadores de ajuste aceitáveis, documentados pela AFC. A confiabilidade, a validade convergente e a validade discriminante foram adequadas. A validade de critério externo apresentou ótimos resultados. Conclusões: O instrumento obtido possui propriedades psicométricas adequadas e contribuirá para a avaliação objetiva da experiência dos profissionais que realizam telemedicina.

Prevalence and outcomes of acute-on-chronic liver failure among cirrhotic patients admitted for an acute decompensation.

BACKGROUND: Acute-on-chronic liver failure (ACLF) is a common syndrome that occurs in patients with advanced chronic liver disease. It consists of the rapid failure of various organs and is associated with high short-term mortality. We aim to describe the main features and outcomes of inpatients who developed ACLF and to identify the factors associated with in-hospital and 28-day mortality. PATIENTS AND METHODS: All patients meeting ACLF criteria with advanced chronic liver disease admitted for decompensation from January 2014 to December 2016 were identified. Clinical and biological data were collected at the time of ACLF diagnosis and at 3-7 days thereafter, as well as in-hospital and 28-day mortality. RESULTS: Eighty nine out of 354 admission episodes (28%) developed ACLF, which was present at the time of admission in 72% of cases. A precipitating factor was identified in 83% of cases, the most frequent being infection (53%) and gastrointestinal bleeding (19%). In the multivariate regression analysis, the ACLF grade at 3-7 days after diagnosis was predictive of in-hospital mortality and 28-day mortality, and lower creatinine and bilirubin levels at the time of ACLF diagnosis and a precipitating factor other than bacterial infection were associated with ACLF reversion at 3-7 days. CONCLUSIONS: ACLF is a frequent complication among patients with chronic liver disease admitted for acute decompensations and is associated with a high mortality rate and is related to the number of organs involved. Bacterial infection is the most frequent precipitating factor of ACLF and probably entails a worse prognosis.

Defective Influenza A Virus RNA Products Mediate MAVS-Dependent Upregulation of Human Leukocyte Antigen Class I Proteins.

Influenza A virus (IAV) increases the presentation of class I human leukocyte antigen (HLA) proteins that limit antiviral responses mediated by natural killer (NK) cells, but molecular mechanisms for these processes have not yet been fully elucidated. We observed that infection with A/Fort Monmouth/1/1947(H1N1) IAV significantly increased the presentation of HLA-B, -C, and -E on lung epithelial cells. Virus entry was not sufficient to induce HLA upregulation because UV-inactivated virus had no effect. Aberrant internally deleted viral RNAs (vRNAs) known as mini viral RNAs (mvRNAs) and defective interfering RNAs (DI RNAs) expressed from an IAV minireplicon were sufficient for inducing HLA upregulation. These defective RNAs bind to retinoic acid-inducible gene I (RIG-I) and initiate mitochondrial antiviral signaling (MAVS) protein-dependent antiviral interferon (IFN) responses. Indeed, MAVS was required for HLA upregulation in response to IAV infection or ectopic mvRNA/DI RNA expression. The effect was partially due to paracrine signaling, as we observed that IAV infection or mvRNA/DI RNA-expression stimulated production of IFN-ß and IFN-λ1 and conditioned media from these cells elicited a modest increase in HLA surface levels in naive epithelial cells. HLA upregulation in response to aberrant viral RNAs could be prevented by the Janus kinase (JAK) inhibitor ruxolitinib. While HLA upregulation would seem to be advantageous to the virus, it is kept in check by the viral nonstructural 1 (NS1) protein; we determined that NS1 limits cell-intrinsic and paracrine mechanisms of HLA upregulation. Taken together, our findings indicate that aberrant IAV RNAs stimulate HLA presentation, which may aid viral evasion of innate immunity.IMPORTANCE Human leukocyte antigens (HLAs) are cell surface proteins that regulate innate and adaptive immune responses to viral infection by engaging with receptors on immune cells. Many viruses have evolved ways to evade host immune responses by modulating HLA expression and/or processing. Here, we provide evidence that aberrant RNA products of influenza virus genome replication can trigger retinoic acid-inducible gene I (RIG-I)/mitochondrial antiviral signaling (MAVS)-dependent remodeling of the cell surface, increasing surface presentation of HLA proteins known to inhibit the activation of an immune cell known as a natural killer (NK) cell. While this HLA upregulation would seem to be advantageous to the virus, it is kept in check by the viral nonstructural 1 (NS1) protein, which limits RIG-I activation and interferon production by the infected cell.

Toward a holistic definition for Information Systems for Health in the age of digital interdependence.

The article's main objective is to propose a new definition for Information Systems for Health, which is characterized by the identification and involvement of all the parts of a complex and interconnected process for data collection and decision-making in public health in the information society. The development of the concept was through a seven-step process including document analysis, on-site and virtual sessions for experts, and an online survey of broader health professionals. This new definition seeks to provide a holistic view, process, and approach for managing interoperable applications and databases that ethically considers open and free access to structured and unstructured data from different sectors, strategic information, and information and communication technology (ICT) tools for decision-making for the benefit of public health. It also supports the monitoring of the Sustainable Development Goals and the implementation of universal access to health and universal health coverage as well as Health in All Policies as an approach to promote health-related policies across sectors. Information Systems for Health evolves from preconceptions of health information systems to an integrated and multistakeholder effort that ensures better care and better policy-making and decision-making.

El objetivo principal de este artículo es proponer una nueva definición de los sistemas de información para la salud, que se caracterizan por la identificación y la participación de todas las partes involucradas en un complejo proceso interconectado de recopilación de datos y toma de decisiones en el ámbito de la salud pública en la sociedad de la información. El concepto se desarrolló en un proceso de siete pasos que incluyó el análisis de documentos, sesiones presenciales y virtuales con expertos y una encuesta en línea a profesionales de la salud en general. Esta nueva definición procura ofrecer un criterio holístico, un proceso y un enfoque para la gestión de bases de datos y aplicaciones interoperables que considere desde un punto de vista ético el acceso abierto y gratuito a datos estructurados y no estructurados de diferentes sectores, información estratégica y herramientas de tecnologías de la información y de la comunicación (TIC) para la toma de decisiones en beneficio de la salud pública. También brinda apoyo al seguimiento de los Objetivos de Desarrollo Sostenible y la ejecución del acceso universal a la salud y la cobertura universal de salud, así como la salud en todas las políticas como iniciativa para promover políticas relacionadas con la salud en todos los sectores. El concepto de sistemas de información para la salud implica una evolución desde lo que se consideraba anteriormente sistemas de información de salud hacia un esfuerzo integrado por parte de varios interesados directos que garantiza una mejora en la atención, la formulación de políticas y la toma de decisiones.

O principal objetivo deste artigo é propor uma nova definição para Sistemas de Informação em Saúde, que são caracterizados pela identificação e participação de todas as partes de um processo complexo e interconectado para a coleta de dados e tomada de decisão em saúde pública na sociedade da informação. O conceito foi desenvolvido por um processo de sete passos incluindo análise de documentos, sessões presenciais e virtuais com especialistas e uma pesquisa on-line com profissionais de saúde generalistas. A nova definição busca oferecer uma visão, um processo e uma abordagem holística para gerenciar aplicativos e bases de dados interoperáveis que consideram eticamente o acesso aberto e gratuito a dados estruturados e não estruturados de diferentes setores, informações estratégicas e ferramentas de tecnologia da informação e comunicação (TIC) para tomadas de decisão em prol da saúde pública. Também sustenta o monitoramento dos Objetivos de Desenvolvimento Sustentável e a implementação do acesso universal à saúde e da cobertura universal de saúde, assim como a Saúde em Todas as Políticas como uma abordagem para promover políticas relacionadas à saúde em vários setores. Os Sistemas de Informação em Saúde evoluíram de pré-conceitos dos sistemas de informação de saúde para um esforço integrado e com muitas partes interessadas, assegurando melhor cuidado, formulação de políticas e tomada de decisão.

[Implementation of telemedicine in the Americas: Barriers and facilitatorsBarreiras e facilitadores para a implementação da telemedicina nas Américas]. / Barreras y facilitadores a la implementación de la telemedicina en las Américas.

With millions of people in the world in situations of physical distancing because of COVID-19, information and communication technology (ICT) has become as one of the principal means of interaction and collaboration. The following advantages of ICT have been cited since the start of the new millennium: increased access to information and service delivery, educational strengthening, quality control of screening programs, and reduction of health care costs. In the case of telemedicine, however, a number of barriers-especially technological, human and social, psychosocial, anthropological, economic, and governance-related-have stood in the way of its adoption. The past 20 years have seen an increase in the availability of resources and technical capacity, improvements in digital education, empowerment of patients regarding their treatment, and increased public interest in this area. Successes have included the use of interdisciplinary teams, academic and professional networking, and virtual medical consultations. After reviewing the state of telemedicine in the Region of the Americas, the authors recommend the urgent adoption of measures aimed at implementing national telemedicine policies and programs, including a regulatory framework and adequate funding. Implementation of the measures should be integrated and interoperable and include the support of academic networks and the collaboration of specialized institutions. The policies should generate an enabling context that ensures sustainability of the progress achieved, bearing in mind the possible barriers mentioned.

Com milhões de pessoas no mundo em situação de distanciamento físico devido à COVID-19, as tecnologias da informação e comunicação (TICs) se enquadraram como um dos meios principais de interação e colaboração. Já no início deste milênio, começaram a ser mencionadas as seguintes vantagens: maior acesso à informação e à prestação de serviços; fortalecimento da educação; controle de qualidade dos programas de detecção e redução dos custos na atenção à saúde. No entanto, entre as principais barreiras de adoção da telemedicina se encontram as de caráter: tecnológico; humano e social; psicossocial e antropológico; de governança e econômico. Nestes 20 anos, houve um aumento nos recursos e na capacidade técnica, uma melhora na educação digital, um empoderamento do paciente em seu tratamento e um maior interesse público nessa área. Em especial, são consideradas bem-sucedidas a constituição de equipes interdisciplinares e as redes acadêmicas e profissionais, e as consultas médicas virtuais. Após revisar o estado da telemedicina na Região das Américas, os autores recomendam a adoção de medidas urgentes para implementar políticas e programas nacionais de telemedicina, incluindo o marco normativo e o orçamento necessário. Essa implementação deve ser realizada de maneira integral e interoperável e sustentada por redes acadêmicas, de parceria e instituições especializadas. Tais políticas devem gerar um contexto favorável, dando sustentabilidade ao avanço obtido e considerando os aspectos mencionados nas possíveis barreiras.

Association between non-attendance to outpatient clinics and emergency department consultations, hospitalizations and mortality in a Health Maintenance Organization.

BACKGROUND: Non-attendance to scheduled medical appointments in outpatient clinics is a problem that affects patient health and health-care systems. OBJECTIVE: Evaluate association of non-attendance to scheduled appointments in outpatient clinics and Emergency Department (ED) visits, hospitalizations and mortality. METHODS: Retrospective cohort study of outpatients enrolled in 2015 to 2016 in the Hospital Italiano de Buenos Aires HMO with over five scheduled appointments. Individual non-attendance proportion was obtained by dividing missed over scheduled appointment numbers in the 365 days prior to index date. Outcomes were evaluated with a Cox proportional-hazards or Fine and Gray model for competing risks. We adjusted by several variables. RESULTS: Sixty-five thousand two hundred sixty-five adults were included. Mean age was 63.6 years (SD 18.16), 29.9% male. Outpatients had average 10.18 (SD 5.59) appointments. Non-attendance the year before the index appointment had a median of 20%. A 10% increase in non-attendance was significantly associated with ED visits (asHR 1.19; 95%CI 1.08-1.32, P < .001) and all-cause mortality (aHR 7.57; 95%CI 4.88-11.73, P < .001). In the matched subcohort analysis we observed a crude significant association of non-attendance with ED visits (P < .001) and all-cause mortality (P < .001). DISCUSSION: Our findings show non-attendance could be a marker of health events that lead to emergency department evaluations and/or death.

Causes of appointment attendance, nonattendance, and cancellation in outpatient consultations at a university hospital.

BACKGROUND: Nonattendance to scheduled appointments in outpatient clinics is a frequent problem in ambulatory medicine with an impact on health systems and patients' health. The characterization of nonattendance is fundamental for the design of appropriate strategies for its management. AIMS: To identify causes of nonattendance of scheduled ambulatory medical appointments by adult patients. METHODS: Case and two controls study nested in a prospective cohort. A telephone-administered questionnaire was applied within the first 72 hours to identify the causes of attendance, nonattendance, or cancellation in patients who had a scheduled appointment to which they had been present, absent, or cancelled. RESULTS: A total of 150 absences (cases), 176 attendances, and 147 cancellations (controls) in a prospective cohort of 160 146 scheduled appointments (2012/2013) were included. According to self-reports in telephone interviews, the most frequent causes of nonattendance were forgetting 44% (66), unexpected competing events 15.3% (23), illness or unwellness 12% (18), work-related inconvenience 5.3% (8), transport-related difficulties 4.7% (4), and cause that motivated appointment scheduling already resolved 4.7% (4). DISCUSSION: The main cause of nonattendance is forgetting the scheduled appointment, but there is a proportion of different causes that do not respond to reminders but could respond to different strategies.

Ferroelectret-based Hydrophone Employed in Oil Identification-A Machine Learning Approach.

This work focuses on acoustic analysis as a way of discriminating mineral oil, providing a robust technique, immune to electromagnetic noise, and in some cases, depending on the applied sensor, a low-cost technique. Thus, we propose a new method for the diagnosis of the quality of mineral oil used in electrical transformers, integrating a ferroelectric-based hydrophone and an acoustic transducer. Our classification solution is based on a supervised machine learning technique applied to the signals generated by an in-home built hydrophone. A total of three statistical datasets entries were collected during the acoustic experiments on four types of oils. The first, the second, and third datasets contain 180, 240, and 420 entries, respectively. Eighty-four features were considered from each dataset to apply to two classification approaches. The first classification approach is able to distinguish the oils from the four possible classes with a classification error less than 2%, while the second approach is able to successfully classify the oils without errors (e.g., with a score of 100%).

Venous thrombosis following pancreaticoduodenectomy with venous resection.

BACKGROUND: Addition of en bloc segmental venous reconstruction (VR) to pancreaticoduodenectomy (PD) for venous involvement of pancreatic tumors increases the complexity of the operation and may increase complications. The long-term mesenteric venous patency rate and oncologic outcome has not been well defined. METHODS: Our prospective database was reviewed to assess 90-day postoperative outcomes for patients who underwent PD or PD + VR (September 2004-June 2016). Two independent observers reviewed CT scans to determine long-term vein patency. In patients with pancreatic ductal adenocarcinoma, the impact of VR on 5-year overall survival was assessed using multivariate Cox proportional hazards regression. Student's t-test was used to evaluate continuous variables and the chi-square test for categorical variables. RESULTS: Three hundred ninety-three patients underwent PD (51 PD + VR). Patients undergoing PD + VR had longer operations (561 ± 119 versus 433 ± 89 min, P < 0.00001) and greater blood loss (768 ± 812 versus 327 ± 423 cc, P < 0.00001). There was no difference in 90-day mortality, overall postoperative complication rates, complication severity grades, reoperation, readmission, or length of stay. 26.7% experienced venous thrombosis. Most thromboses occurred in the first year after surgery, but we also observed late thrombosis in 1 patient after 89-month follow-up. Among 135 patients with pancreatic ductal adenocarcinoma, survival was significantly longer in the PD-alone group (31.3 months [95% confidence interval: 22.9-40.0] versus 17.0 [95% confidence interval: 13.0-19.1], plog-rank = 0.013). CONCLUSIONS: PD + VR does not increase short-term morbidity, but venous thrombosis is frequent and can occur long after surgery. Survival is inferior when VR is required especially in the absence of neoadjuvant chemotherapy.

Fast Morphological Gallbladder Changes Triggered by a Hypercholesterolemic Diet.

INTRODUCTION AND AIM: Obesity is a worldwide epidemic problem, described as a risk factor for hepatic diseases, such as non-alcoholic fatty liver disease and other pathologies related to development of cholesterol crystals and cholesterol gallbladder stones. It has been reported that cholesterol overload may cause hepatic damage; however, little is known about the effects of an acute hypercholesterolemic diet on the gallbladder. The aim of this manuscript was to evaluate the impact of a cholesterol-rich diet on the gallbladder. MATERIAL AND METHODS: The study included ten eight-week-old C57BL6 male mice, which were divided into two study groups and fed different diets for 48 h: a hypercholesterolemic diet and a balanced Chow diet. After 48 h, the mice were analyzed by US with a Siemens Acuson Antares equipment. Mice were subsequently sacrificed to carry out a cholesterol analysis with a Refloton System (Roche), a crystal analysis with a Carl Zeiss microscope with polarized light, and a histological analysis with Hematoxylin-eosin staining. RESULTS: The hypercholesterolemic diet induced an increase in gallbladder size and total cholesterol content in the bile, along with important histological changes. CONCLUSION: Cholesterol overloads not only trigger hepatic damage, but also affect the gallbladder significantly.

User-centered design improves the usability of drug-drug interaction alerts: Experimental comparison of interfaces.

Clinical Decision Support Systems can alert health professionals about drug interactions when they prescribe medications. The Hospital Italiano de Buenos Aires in Argentina developed an electronic health record with drug-drug interaction alerts, using traditional software engineering techniques and requirements. Despite enhancing the drug-drug interaction knowledge database, the alert override rate of this system was very high. We redesigned the alert system using user-centered design (UCD) and participatory design techniques to enhance the drug-drug interaction alert interface. This paper describes the methodology of our UCD. We used crossover method with realistic, clinical vignettes to compare usability of the standard and new software versions in terms of efficiency, effectiveness, and user satisfaction. Our study showed that, compared to the traditional alert system, the UCD alert system was more efficient (alerts faster resolution), more effective (tasks completed with fewer errors), and more satisfying. These results indicate that UCD techniques that follow ISO 9241-210 can generate more usable alerts than traditional design.

Hyperlipidemic microenvironment conditionates damage mechanisms in human chondrocytes by oxidative stress.

BACKGROUND: Currently, two pathogenic pathways describe the role of obesity in osteoarthritis (OA); one through biomechanical stress, and the other by the contribution of systemic inflammation. The aim of this study was to evaluate the effect of free fatty acids (FFA) in human chondrocytes (HC) expression of proinflammatory factors and reactive oxygen species (ROS). METHODS: HC were exposed to two different concentrations of FFA in order to evaluate the secretion of adipokines through cytokines immunoassays panel, quantify the protein secretion of FFA-treated chondrocytes, and fluorescent cytometry assays were performed to evaluate the reactive oxygen species (ROS) production. RESULTS: HC injury was observed at 48 h of treatment with FFA. In the FFA-treated HC the production of reactive oxygen species such as superoxide radical, hydrogen peroxide, and the reactive nitrogen species increased significantly in a at the two-dose tested (250 and 500 µM). In addition, we found an increase in the cytokine secretion of IL-6 and chemokine IL-8 in FFA-treated HC in comparison to the untreated HC. CONCLUSION: In our in vitro model of HC, a hyperlipidemia microenvironment induces an oxidative stress state that enhances the inflammatory process mediated by adipokines secretion in HC.

Development of an ASR System for Medical Conversations.

In this work we document the development of an ASR system for the transcription of conversations between patient and doctor and we will point out the critical aspects of the domain. The system was trained with an acoustic base of spontaneous speech that has a domain language model and a supervised phonetic dictionary. Its performance was compared with two systems: a) NeMo End-to-End Conformers in Spanish and b) Google API ASR (Automatic Speech Recognition) Cloud. The evaluation was carried out on a set of 208 teleconsultations recorded during the year 2020. The WER (Word Error Rate) was evaluated in ASR, and Recall and F1 for recognized medical entities. In conclusion, the developed system performed better, reaching 72.5% accuracy in the domain of teleconsultations and an F1 for entity recognition of 0.80.

Automatic Speech Recognition System to Record Progress Notes in a Mobile EHR: A Pilot Study.

Creating notes in the EHR is one of the most problematic aspects for health professionals. The main challenges are the time spent on this task and the quality of the records. Automatic speech recognition technologies aim to facilitate clinical documentation for users, optimizing their workflow. In our hospital, we internally developed an automatic speech recognition system (ASR) to record progress notes in a mobile EHR. The objective of this article is to describe the pilot study carried out to evaluate the implementation of ASR to record progress notes in a mobile EHR application. As a result, the specialty that used ASR the most was Home Medicine. The lack of access to a computer at the time of care and the need to perform short and fast evolutions were the main reasons for users to use the system.

Development and Usage Patterns of a Home-Grown Drug Information Tool.

Drug information tools help avoid medication errors, a common cause of avoidable harm in health care systems. We sought to describe the design, development process and architecture of an electronic drug information tool, as well as its overall use by health professionals. We developed a tool that can be accessed by all health professionals in a tertiary level university hospital. The functionalities of eDrugs are organized into two main parts: Drug Summary sheet, and Prescription Simulator. Most users accessed eDrugs to use the Drug summary sheet. Clinical information and antimicrobial drugs were the most accessed drug information and drug group. The analysis of log data provides insights into the information priorities of health professionals.

Improving Waiting Time for Chemotherapy with Ahead-of-Time Drug Preparation.

Waiting time for chemotherapy infusion is a fundamental factor to measure quality of care. It has been shown that a prolonged waiting time is related to a higher incidence of anticipatory nausea and poor patient adherence to scheduled appointments and recommended oncology treatment programs. Some chemotherapy regimens can be prepared hours ahead-of-time, due to long stability. We aimed to study the effect of an informatic-led workflow redesign intervention, facilitating workflow changes in the Oncology Pharmacy, on patient waiting time. This intervention included changes on EHR processes and the chemotherapy CPOE. Their main effect was allowing ahead-of-time preparation of selected chemotherapy regimes. We conducted a cross sectional study, comparing waiting times pre and post intervention periods. A total of 4600 programmed chemotherapy episodes were included. We found a 26.5 % decrease in the mean wait time in the post intervention period (p > 0.02). We were able to show a decrease in waiting time and a measurable impact of the intervention. This evaluation produced valuable and actionable data for Oncology units and adds a valuable, Latin American experience to the literature.