Introduction: Bipolar disorder (BD) is a chronically progressive mental condition, associated with a reduced quality of life and greater disability. Patient admissions are preventable events with a considerable impact on global functioning and social adjustment. While machine learning (ML) approaches have proven prediction ability in other diseases, little is known about their utility to predict patient admissions in this pathology.Aim: To develop prediction models for hospital admission/readmission within 5 years of diagnosis in patients with BD using ML techniques.Methods: The study utilized data from patients diagnosed with BD in a major healthcare organization in Colombia. Candidate predictors were selected from Electronic Health Records (EHRs) and included sociodemographic and clinical variables. ML algorithms, including Decision Trees, Random Forests, Logistic Regressions, and Support Vector Machines, were used to predict patient admission or readmission. Survival models, including a penalized Cox Model and Random Survival Forest, were used to predict time to admission and first readmission. Model performance was evaluated using accuracy, precision, recall, F1 score, area under the receiver operating characteristic curve (AUC) and concordance index.Results: The admission dataset included 2,726 BD patients, with 354 admissions, while the readmission dataset included 352 patients, with almost half being readmitted. The best-performing model for predicting admission was the Random Forest, with an accuracy score of 0.951 and an AUC of 0.98. The variables with the greatest predictive power in the Recursive Feature Elimination (RFE) importance analysis were the number of psychiatric emergency visits, the number of outpatient follow-up appointments and age. Survival models showed similar results, with the Random Survival Forest performing best, achieving an AUC of 0.95. However, the prediction models for patient readmission had poorer performance, with the Random Forest model being again the best performer but with an AUC below 0.70.Conclusion: ML models, particularly the Random Forest model, outperformed traditional statistical techniques for admission prediction. However, readmission prediction models had poorer performance. This study demonstrates the potential of ML techniques in improving prediction accuracy for BD patient admissions.

Objective: To estimate Diabetes mellitus (DM) progression at one and two years in terms of glycemic targets and development of complications. Research design and methods: We analyzed a retrospective cohort of adult DM patients treated in a Health Maintenance Organization in Colombia, including those with at least one glycosylated hemoglobin (HbA1c) measurement in 2018, 2019, and 2020. We defined four disease transition stages based on metabolic goals according to HbA1c levels and complications: 1. Within HbA1c goals and without complications; 2. Outside goals and without complications, 3. Within goals, but with complications, and 4. Outside goals and with complications. We applied Natural Language Processing (NLP) techniques to extract relevant clinical information from Electronic Health Records. Machine learning (ML) models were used to predict patient progression. Results: A total of 23,802 patients were included. Despite achieving initial glycemic control, more than 60% of patients who started within HbA1c targets and without complications developed chronic complications within two years. Our models, which achieved up to 80% accuracy and F1 scores above 74%, identified key predictors of disease progression. Adherence to dyslipidemia treatment guidelines significantly reduced the likelihood of HbA1c deterioration and complications, whereas non-adherence to pharmacological treatments increased the risk of complications. These findings suggest that HbA1c control alone is insufficient to prevent disease progression and that a more comprehensive management approach—including lipid control, kidney function monitoring, and improved adherence to clinical guidelines—is necessary. Conclusions: Patient compliance with pharmacological treatments, professional adherence to clinical practice guidelines, and lifestyle interventions play a crucial role in diabetes progression. While our models provide strong predictive capabilities, improving data quality and integration remains essential for better forecasting and intervention strategies.

Background: Hemoglobin (Hb) and hematocrit (Hct) reference intervals (RIs) are critical for diagnosing hematological disorders. However, existing reference values often do not account for demographic and environmental variability. Particularly in countries with altitude gradients, such as Colombia, the absence of locally adjusted intervals may lead to the misclassification of anemia and polycythemia. Therefore, this study aims to establish sex-, age-, and altitude-specific reference intervals for Hb and Hct within the Colombian adult population via an indirect, big-data-based methodology. Methods: This retrospective cross-sectional study used 3.1 million Hb and Hct test results nationwide between 2022 and 2024. After applying the exclusion criteria, Hb data from 667,857 individuals and Hct data from 662,024 individuals were included. The population was stratified by sex, age, and altitude into <1100 m above sea level (m.a.s.l.), 1100–2000 m.a.s.l., and 2000–3000 m.a.s.l. Reference intervals (RIs) were estimated via the refineR algorithm, and the results were compared across altitude categories and against World Health Organization (WHO) anemia and polycythemia thresholds. Results: Hb and Hct concentrations increased with altitude in all sexes and age groups. Compared with women, men presented higher mean values and narrower RIs, whereas older adults presented greater variability. Compared with WHO thresholds, a significant proportion of individuals living above 2000 m exceeded polycythemia cutoffs without clinical evidence of disease, suggesting the need for altitude-adjusted diagnostic criteria. Conclusions: This study provides the first large-scale, data-driven reference intervals for Hb and Hct in Colombia, adjusted for altitude, sex, and age. The implementation of locally derived RIs may improve diagnostic accuracy and prevent the over- or underdiagnosis of hematological disorders, with direct implications for clinical decision-making and public health policy.