Supplementary file 1_Development and internal validation of a machine learning model for p

1,317 patient records from a two-center retrospective study between 2015 and 2024, used to develop a machine learning model for predicting intracranial infection after spontaneous intracerebral hemorrhage. The dataset includes baseline demographic, clinical, laboratory, and radiological variables collected within 24 hours of admission. It was authored by Yizhao Lin and shared under a CC-BY-4.0 license.

Use Cases

• Developing risk stratification models for intracranial infection based on early clinical and laboratory variables.
• Comparing the performance of machine learning algorithms like LightGBM for binary medical outcomes.
• Analyzing associations between predicted infection risk and long-term functional outcomes at 180 days.
• Conducting internal validation studies for clinical prediction models using retrospective hospital data.

Strengths

• Includes 1,317 patient records from a two-center study, providing a substantial clinical cohort.
• Model development used 12 machine learning algorithms and reported a high test-set AUC of 0.923.
• Data covers a 9-year time range (2015-2024) from hospitals in the same province.

Limitations

• Only internal validation was performed; the description notes further external validation is needed.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the underlying data file is unknown, which may limit suitability assessment.

Provenance

Source

Two hospitals in the same province, as stated in the study.

Collection Method

Retrospective data collection from patient records.

Time Range

2015 to 2024

Freshness

Last updated 2026-05-04 04:16:46; freshness should be verified.

Geography

Hospitals within the same province (specific location not stated).