Final optimized model with full benchmark

Optimize an existing Logistic Regression baseline for `Outcome`.
Your final report must include a before-vs-after benchmark (Accuracy, Precision, Recall, F1, ROC-AUC, PR-AUC, Confusion Matrix, and Runtime).
A subgroup error analysis by Age and BMI buckets to track False Positives/Negatives.
Conclude with a summary explaining how your selected settings specifically addressed class imbalance or feature variance to improve performance.

Use the existing Logistic Regression baseline as the starting point; the payload does not name a file or notebook entry point. Compare the baseline and optimized model on all requested metrics, include the Age and BMI subgroup error analysis, and finish with the requested explanation of class imbalance or feature variance.

Written by the indexing model from the issue text.