Pneumonia and pulmonary thromboembolism (PTE) are both respiratory diseases; their diagnosis is difficult due to their similarity in symptoms, medical subjectivity, and the large amount of information from different sources necessary for a correct diagnosis. Analysis of such clinical data using computational tools could help medical staff reduce time, increase diagnostic certainty, and improve patient care during hospitalization. In addition, no studies have been found that analyze all clinical information on the Mexican population in the Spanish language. Therefore, this work performs automatic diagnosis of pneumonia and pulmonary thromboembolism using machine-learning tools along with clinical laboratory information (structured data) and clinical text (unstructured data) obtained from electronic health records. A cohort of 173 clinical records was obtained from the Mexican Social Security Institute. The data were preprocessed, transformed, and adjusted to be analyzed using several machine-learning algorithms. For structured data, naïve Bayes, support vector machine, decision trees, AdaBoost, random forest, and multilayer perceptron were used; for unstructured data, a BiLSTM was used. K-fold cross-validation and leave-one-out were used for evaluation of structured data, and hold-out was used for unstructured data; additionally, 1-vs.-1 and 1-vs.-rest approaches were used. Structured data results show that the highest AUC-ROC was achieved by the naïve Bayes algorithm classifying PTE vs. pneumonia (87.0%), PTE vs. control (75.1%), and pneumonia vs. control (85.2%) with the 1-vs.-1 approach; for the 1-vs.-rest approach, the best performance was reported in pneumonia vs. rest (86.3%) and PTE vs. rest (79.7%) using naïve Bayes, and control vs. diseases (79.8%) using decision trees. Regarding unstructured data, the results do not present a good AUC-ROC; however, the best F1-score were scored for control vs. disease (72.7%) in the 1-vs.-rest approach and control vs. pneumonia (63.6%) in the 1-to-1 approach. Additionally, several decision trees were obtained to identify important attributes for automatic diagnosis for structured data, particularly for PTE vs. pneumonia. Based on the experiments, the structured datasets present the highest values. Results suggest using naïve Bayes and structured data to automatically diagnose PTE vs. pneumonia. Moreover, using decision trees allows the observation of some decision criteria that the medical staff could consider for diagnosis.

- Importance: Many COVID-19 prognostic factors for disease severity have been identified and many scores have already been proposed to predict death and other outcomes. However, hospitals in developing countries often cannot measure some of the variables that have been reported as useful. - Objective: To assess the sensitivity, specificity, and predictive values of the novel LOW-HARM score ( L ymphopenia, O xygen saturation, W hite blood cells, H ypertension, A ge, R enal injury, and M yocardial injury). - Design: The score was designed using data from already published cohorts of patients diagnosed with COVID-19. Afterwards, it was calculated it in 438 consecutive hospital admissions at twelve different institutions in ten different cities in Mexico. - Setting: Twelve hospitals in ten different cities in Mexico. - Participants: Data from 438 patients was collected. Data from 400 patients (200 deaths and 200 survivors) was included in the analysis. - Exposure: All patients had an infection with SARS-CoV-2 confirmed by PCR. - Main Outcome: The sensitivity, specificity, and predictive values of different cut-offs of the LOW-HARM score to predict death. - Results: Mean scores at admission and their distributions were significantly lower in patients who were discharged compared to those who died during their hospitalization 10 (SD: 17) vs 70 (SD: 28). The overall AUC of the model was 95%. A cut-off > 65 points had a specificity of 98% and a positive predictive value of 96%. More than a third of the cases (36%) in the sample had a LOW-HARM score > 65 points. - Conclusions and relevance: The LOW-HARM score measured at admission is highly specific and useful for predicting mortality. It is easy to calculate and can be updated with individual clinical progression. KEY POINTS: Question: Is it possible to predict mortality in patients diagnosed with COVID-19 using easy-to-access and easy-to-measure variables? Findings: The LOW-HARM score ( L ymphopenia, O xygen saturation, W hite blood cells, H ypertension, A ge, R enal injury, and M yocardial injury) is a one-hundred-point score that, when measured at admission, had an overall AUC of 95% for predicting mortality. A cut-off of ≥ 65 points had a specificity of 98% and a positive predictive value of 96%. Meaning: The LOW-HARM score measured at admission is highly specific and useful for predicting mortality in patients diagnosed with COVID-19. In our sample, more than a third of patients met the proposed cut-off.

Resumen La educación médica tiene como eje fundamental la generación de habilidades para la atención de precisión a los pacientes; sin embargo, una de las herramientas esenciales para la reflexión médica en cada caso radica, también, en el desarrollo de habilidades científicas que permitan al médico generar un pensamiento crítico y ofrecer nuevas alternativas basadas en la evidencia. No obstante, existen barreras importantes que dificultan que el proceso de formación en investigación permee las habilidades del médico. Los hospitales escuela requieren reflexionar acerca de las barreras modificables, con el fin de ofrecer a los médicos en formación un ambiente propicio y motivador hacia la ciencia. El presente artículo se fundamenta en la reflexión de estudios científicos que versan sobre las dificultades para la incorporación de habilidades en investigación.