30 August 2024 in Research Square
Aims This study aims to assess the feasibility and impact of the implementation of the ChatGLM for real-world data (RWD) extraction in hospital settings. The primary focus of this research is on the effectiveness of ChatGLM-driven data extraction compared with that of manual processes associated with the electronic source data repository (ESDR) system. Methods and results The researchers developed the ESDR system, which integrates ChatGLM, electronic case report forms (eCRFs) and electronic health records (EHRs). The LLaMA (Large Language Model Meta AI) model was also deployed to compare the extraction accuracy of ChatGLM in free-text forms. A single-center retrospective cohort study served as a pilot case. Five eCRF forms of 63 subjects, including free-text forms and discharge medication, were evaluated. Results Data collection involved electronic medical and prescription records collected from 13 departments. The ChatGLM-assisted process was associated with an estimated efficiency improvement of 80.7% in the eCRF data transcription time. The initial manual input accuracy for free-text forms was 99.59%, the ChatGLM data extraction accuracy was 77.13%, and the LLaMA data extraction accuracy was 43.86%. The challenges associated with the use of ChatGLM focus on prompt design, prompt output consistency, prompt output verification, and integration with hospital information systems. Conclusion The main contribution of this study is to validate the use of ESDR tools to address the interoperability and transparency challenges of using ChatGLM for RWD extraction in Chinese hospital settings.
10 January 2024 in medRxiv
Background There is a lack of tools specifically designed to assess mortality risk in patients with atrial fibrillation (AF). The aim of this study was to utilize machine learning methods for identifying pertinent variables and developing an easily applicable prognostic score to predict 1-year mortality in AF patients. Methods This single-center retrospective cohort study based on the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database focused on patients aged 18 years and older with AF. The study thoroughly scrutinized patient data to identify and analyze variables, encompassing demographic variables, comorbidities, scores, vital signs, laboratory test results, and medication usage. The variable importance from XGBoost guided the development of a logistic model, forming the basis for an AF scoring model. Decision curve analysis was used to compare the AF score with other scores. Python and R software were used for data analysis. Results A cohort of 59,595 AF patients was obtained from the MIMIC-IV database; these patients were predominantly elderly (median age 77.3 years) and male (55.6%). The XGBoost model effectively predicted 1-year mortality (Area under the curve (AUC): 0.833; 95% confidence intervals: 0.826-0.839), underscoring the significance of the Charlson Comorbidity Index (CCI) and the presence of metastatic solid tumors. The CRAMB score (Charlson comorbidity index, readmission, age, metastatic solid tumor, and blood urea nitrogen maximum) outperformed the CCI and CHA2DS2-VASc scores, demonstrating superior predictive value for 1-year mortality. In the test set, the area under the ROC curve (AUC) for the CRAMB score was 0.756 (95% confidence intervals: 0.748-0.764), surpassing the CCI score of 0.720 (95% confidence intervals: 0.712-0.728) and the CHA2DS2-VASc score of 0.609 (95% confidence intervals: 0.600-0.618). Decision curve analysis revealed that the CRAMB score had a consistently positive effect and greater net benefit across the entire threshold range than did the default strategies and other scoring systems. The calibration plot for the test set indicated that the CRAMB score was well calibrated. Conclusions This study’s primary contribution is the establishment of a benchmark for utilizing machine learning models in construction of a score for mortality prediction in AF. The CRAMB score was developed by leveraging a large-sample population dataset and employing XGBoost models for predictor screening. The simplicity of the CRAMB score makes it user friendly, allowing for coverage of a broader and more heterogeneous AF population.