Postoperative UTI Prediction Model for Ureteral Stone Patients, 2020-2024

826 patient records from a retrospective cohort study of ureteral stone surgery patients from January 2020 to January 2024. The dataset was used to develop an AutoML-based predictive model for postoperative urinary tract infection, created by Ding Li and shared under a CC-BY-4.0 license.

Use Cases

• Training binary classification models for postoperative infection risk based on demographic and hematological indicators.
• Conducting feature importance analysis to identify key predictors like preoperative urinary retention and serum albumin levels.
• Developing clinical decision support systems for ureteral stone surgery based on quantified risk thresholds.
• Studying interaction effects between clinical factors, such as stone size and catheter indwelling time.

Strengths

• Model performance metrics are provided, with the AutoML model achieving an ROC-AUC of 0.9251 on the test set.
• The study employed specific techniques like SMOTE-ENN hybrid sampling and SHAP analysis for interpretability.
• Data covers a defined four-year period from January 2020 to January 2024.

Limitations

• The dataset is very small at 14.1 KB, suggesting limited scope or that the primary file is a research document rather than the raw data.
• Column-level documentation is absent; field semantics must be inferred from the research description after download.
• Row count is unknown, which may limit suitability assessment for machine learning tasks.

Provenance

Source

figshare

Collection Method

Retrospective cohort design using structured electronic medical records from ureteral stone surgery patients.

Time Range

January 2020 to January 2024

Freshness

Last updated 2026-05-26 04:39:10; freshness should be verified.