Baseline Characteristics for ICU AKI Prediction Using Anomaly Detection

Medical Information Mart for Intensive Care (MIMIC-III/IV) and eICU Collaborative Research Database data from 81,876 and 140,237 ICU admissions, respectively. The dataset contains baseline characteristics for a study testing a deep-learning anomaly signal on creatinine-eGFR time series to predict near-term kidney replacement therapy and mortality. Author Yoonjin Kang published the data under a CC-BY-4.0 license in May 2026.

Use Cases

• Training risk stratification models based on creatinine-eGFR time series anomaly scores.
• Benchmarking new AKI prediction methods against established KDIGO criteria.
• Analyzing associations between anomaly signals and clinical outcomes like kidney replacement therapy.
• Validating time-series anomaly detection algorithms on clinical data from multiple ICU databases.

Strengths

• Derived from two large, established clinical databases: MIMIC-III/IV (81,876 admissions) and eICU (140,237 admissions).
• Includes a substantial number of time-series instances (381,700 internally, 494,684 externally).
• External validation was performed using a separate database (eICU), suggesting a test for generalizability.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the baseline characteristics file is unknown, which may limit suitability assessment.
• The dataset is very small (5.5 KB), indicating it likely contains summary statistics rather than raw patient-level time-series data.

Provenance

Source

Medical Information Mart for Intensive Care (MIMIC-III/IV) and the eICU Collaborative Research Database.

Collection Method

Seven-step daily creatinine–eGFR instances were built from ICU admissions, starting within 48 hours before admission.

Freshness

Last updated 2026-05-13 17:34:41; freshness should be verified.