Baseline Characteristics for Intraoperative Hypotension and Kidney Injury Study

319,699 surgical cases from a tertiary hospital in China between 2013 and 2023 were used to develop a Transformer model for predicting transient intraoperative hypotension. The dataset, authored by Shouqiang Zhu and shared on figshare, includes baseline characteristics for a study that externally validated the model using data from South Korea and assessed associations with postoperative kidney injury.

Use Cases

• Analyze associations between intraoperative hypotension burden and postoperative acute kidney injury based on cumulative MAP thresholds mentioned in the description
• Benchmark machine learning models (e.g., Transformer vs. XGBoost) for time-series prediction of clinical events based on vital sign data
• Study baseline patient characteristics for a large surgical cohort spanning a decade
• Assess model generalizability by comparing internal (China) and external (South Korea) validation performance metrics

Strengths

• Dataset underpins a study with a large training cohort of 319,699 surgical cases
• Model performance metrics are reported (e.g., AUCs up to 0.904, recall ≥88.3%)
• Includes external validation from an independent dataset in South Korea, suggesting tested 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 study design is retrospective, a limitation noted by the authors

Provenance

Source

figshare, author Shouqiang Zhu

Collection Method

Retrospective data collection from a tertiary hospital in China, likely from electronic health records.

Time Range

2013–2023

Freshness

Last updated 2026-03-25 17:40:39; freshness should be verified

Geography

Primary data from China; external validation data from South Korea.