18 Open Healthcare Datasets-2025 Update

"The healthcare industry continues its digital transformation, driven by the availability of open-source datasets. These datasets provide data scientists, researchers, and medical professionals with valuable insights to improve patient outcomes, streamline operations, and foster innovative treatments. Here are 18 top open-source healthcare datasets that are making a significant impact in healthcare research and can be helpful for those working in AI and data science. This is an updated version of our 2024 blog on open healthcare datasets, with new additions and revised information for 2026."

"MIMIC-IV (Medical Information Mart for Intensive Care) MIMIC-IV is a comprehensive critical care dataset containing de-identified health records of ICU patients. An update to the widely-used MIMIC-III database, MIMIC-IV includes data from 2008-2019, such as vital signs, laboratory tests, medications, procedures, diagnoses, and even deidentified clinical notes. It is freely accessible (with a data use agreement) and is widely used for developing predictive models and analyzing ICU practices, enabling research in patient monitoring, outcome prediction, and more."

"eICU Collaborative Research Database The eICU Collaborative Research Database is a large multi-center critical care dataset of clinical data collected from ICU patients across multiple hospitals. Established in 2014 through a collaboration between MIT and Philips Healthcare, it encompasses data from over 200,000 ICU stays across more than 200 hospitals. This database enables researchers to conduct robust studies on a wide range of critical care topics, including ICU practices, patient deterioration, treatment outcomes, and comparative effectiveness of interventions."

Open-source healthcare datasets enable development of predictive models, diagnostics, and operational improvements across clinical settings. An updated 2026 compilation includes eighteen leading datasets spanning critical care EHRs, multi-center ICU data, and large labeled medical imaging collections. Datasets provide de-identified ICU records with vital signs, labs, medications, procedures, diagnoses, and clinical notes for outcome prediction and patient monitoring. Multi-center ICU datasets cover hundreds of hospitals and hundreds of thousands of stays, supporting comparative effectiveness and deterioration studies. Large chest X-ray collections with labeled thoracic disease categories support imaging-based AI and diagnostic algorithm development.

Read at Medium