RESUMEN
OBJECTIVE: To review and critically appraise published and preprint reports of prognostic models of in-hospital mortality of patients in the intensive-care unit (ICU) based on neural representations (embeddings) of clinical notes. METHODS: PubMed and arXiv were searched up to August 1, 2022. At least two reviewers independently selected the studies that developed a prognostic model of in-hospital mortality of intensive-care patients using free-text represented as embeddings and extracted data using the CHARMS checklist. Risk of bias was assessed using PROBAST. Reporting on the model was assessed with the TRIPOD guideline. To assess the machine learning components that were used in the models, we present a new descriptive framework based on different techniques to represent text and provide predictions from text. The study protocol was registered in the PROSPERO database (CRD42022354602). RESULTS: Eighteen studies out of 2,825 were included. All studies used the publicly-available MIMIC dataset. Context-independent word embeddings are widely used. Model discrimination was provided by all studies (AUROC 0.75-0.96), but measures of calibration were scarce. Seven studies used both structural clinical variables and notes. Model discrimination improved when adding clinical notes to variables. None of the models was externally validated and often a simple train/test split was used for internal validation. Our critical appraisal demonstrated a high risk of bias in all studies and concerns regarding their applicability in clinical practice. CONCLUSION: All studies used a neural architecture for prediction and were based on one publicly available dataset. Clinical notes were reported to improve predictive performance when used in addition to only clinical variables. Most studies had methodological, reporting, and applicability issues. We recommend reporting both model discrimination and calibration, using additional data sources, and using more robust evaluation strategies, including prospective and external validation. Finally, sharing data and code is encouraged to improve study reproducibility.
RESUMEN
Online forums play an important role in connecting people who have crossed paths with cancer. These communities create networks of mutual support that cover different cancer-related topics, containing an extensive amount of heterogeneous information that can be mined to get useful insights. This work presents a case study where users' posts from an Italian cancer patient community have been classified combining both count-based and prediction-based representations to identify discussion topics, with the aim of improving message reviewing and filtering. We demonstrate that pairing simple bag-of-words representations based on keywords matching with pre-trained contextual embeddings significantly improves the overall quality of the predictions and allows the model to handle ambiguities and misspellings. By using non-English real-world data, we also investigated the reusability of pretrained multilingual models like BERT in lower data regimes like many local medical institutions.
