Multidrug-resistant organism surveillance in a Chinese tertiary hospital (2017-2026): de-identified record-level detection data and analysis code

De-identified surveillance dataset and complete analysis code for a retrospective cohort of multidrug-resistant organism (MDRO) detections in a Chinese tertiary hospital (May 2017 - August 2026). Files: 01_cleaned_MDRO_surveillance_dataset_record_level.csv (3,031 record-level MDRO detections, 57 fields: ward, pathogen, resistance phenotype, specimen, timestamps, infection status, risk factors); 02_episode_level_analysis_dataset.csv (1,576 hospitalisation episodes with outcomes and episode-level risk factors); 00_cleaning_log.json (data-cleaning audit trail); code/ (Python scripts that reproduce every reported number); README.md (variable definitions, de-identification statement, run order and key results). Key result: infection (versus colonisation) at first MDRO detection was independently associated with in-hospital mortality (adjusted Fine-Gray subdistribution hazard ratio 2.62, 95% CI 1.45-4.75, P = 0.001; cause-specific HR 2.23; logistic OR 2.72), with attenuation of the crude estimate attributable to ICU adjustment. Ecological analyses cover diversity, year-to-year turnover and seasonality over the continuous surveillance window 2020-05 to 2026-08 (76 months); ward-level co-detection networks are descriptive. Privacy: all identifiers were replaced with sequential study codes (patient_no, episode_id, specimen_no, record_id); the dataset contains no names, national/hospital ID numbers, contact details or free-text identifiers. Chinese-language categorical labels from the source system were translated into English. Software: Python 3.8.1 with numpy, pandas, scipy and matplotlib; the lifelines package is used only for independent cross-validation.