Code for "Reassessing the management of uncomplicated urinary tract infection: A retrospective analysis
Preface: Important Note on Replication The data used in this study is not publicly available.
Repository Overview Setup
Data pre-processing and analysis/Cohort Generation
- Feature Generation 1-Final.ipynb: Generates Treatment History Confounders
- Feature Generation 2-Final.ipynb: Generates Provider Features using NPI Database
- Feature Generation 3-Final.ipynb: Generates Condition 2 year history Features
- Feature Generation 4-Final.ipynb: Generates additional condition history covariates using concept ancestors for following condition categories
- Feature Generation 5-Final.ipynb: Generates lab, lab existence features, censor and outcome variables
- Cohort_Generation.ipynb: Generates the cohort based on defined inclusion/exclusion criteria
Data pre-processing and analysis/Covariate Tables
- Antibiotic prevalence analysis.ipynb: Creates supplementary table 5
- Table 1.ipynb: Creates table 1
Data pre-processing and analysis/ATE Tables and Plots The causal analysis notebooks utilize either the omop or domain-knowledge derived features and second line or alternatives treatment to do cross-validated grid search for propensity/censorship models. These notebooks also contain the code to use the model to generate boostrap ATEs, shapley value plots and the calibration plots.
- Grid Search and Causal Analysis (final) – alternatives o The notebook used to perform the causal analysis on first line vs alternatives
- Grid Search and Causal Analysis (final) – second line o The notebook used to perform the causal analysis on first line vs second-line