Cardiac Surgery Patient EHR Data for Predicting Acute Kidney Injury

130 prospectively recruited patients from a multicenter observational trial, with data used to predict acute kidney injury after cardiac surgery. The dataset, created by Janis Fliegenschmidt and last updated in March 2026, contains electronic health record data analyzed with machine learning to generate risk scores. 33.1% of patients developed acute kidney injury within 72 hours postoperatively, and 18.5% developed acute kidney disease by day 30.

Use Cases

• Training machine learning models for acute kidney injury risk prediction based on EHR data.
• Validating AI-based risk scoring tools for postoperative complications in cardiac surgery.
• Analyzing perioperative risk factor progression based on the described increase in predicted risk scores over time.
• Benchmarking EHR-only predictive models against traditional clinical risk assessment methods.

Strengths

• Data is from a prospectively recruited cohort of 130 patients within a multicenter trial.
• Model performance reported with an AUROC of 0.79 for CSA-AKI and 0.83 for AKD.
• Released under a permissive CC-BY-4.0 license for reuse.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is from a single center within a multicenter study, which may limit generalizability.

Provenance

Source

Secondary analysis of data from a multicenter observational trial.

Collection Method

Electronic health record data from prospectively recruited patients.

Time Range

Perioperative period and 30-day follow-up.

Freshness

Last updated 2026-03 18 07:43:51; freshness should be verified.

Geography

Data from one center; specific location unknown.