Firth Penalised Logistic Sensitivity Model: Clinical Predictors for Two Patient Cohorts

A 5.4 KB dataset published on figshare by Alejandro Hernández-Arango on 2026-04-30. It contains results from a penalised logistic regression (Firth method) fitted with three predictors each for two patient cohorts. Cohort A (n=121) used CXR total compromise ratio, NLP-IDSA score, and age, achieving an AUC of 0.756 and Brier score of 0.116.

Use Cases

• Comparing model performance (AUC, Brier score) between two distinct patient cohorts.
• Analyzing the predictive power of CXR total compromise ratio and NLP-IDSA score across different clinical settings.
• Studying the application of Firth's penalised logistic regression method to mitigate small-sample bias.
• Investigating the impact of different predictor sets (age vs. heart rate) on model sensitivity.

Strengths

• Model performance metrics (AUC, Brier score) are provided for both cohorts, allowing for direct comparison.
• Cohort sizes are explicitly stated (n=121 and n=27), clarifying the sample scale.
• The specific predictors used for each model are clearly listed in the description.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (5.4 KB), indicating limited scope, likely containing only model results or summary statistics.

Provenance

Source

figshare

Collection Method

Likely contains results from a clinical prediction modeling study using penalised logistic regression.

Freshness

Last updated 2026-04-30 17:49:50; freshness should be verified.