edithatogo/uci-diabetes-130-us-hospitals

UCI Diabetes 130-US Hospitals, 1999–2008 — source mirror

Real publicly released hospital encounter data, not synthetic data. This is an unofficial provenance-bearing mirror for statistical analysis and RareBurden public-data node engineering. It is not a monogenic-diabetes cohort, a country node deployment, clinical validation or evidence of hospital participation in RareBurden. The train split is a hosting convention, not a study-design split.

Attribution and rights

Clore, J., Cios, K., DeShazo, J., & Strack, B. (2014). Diabetes 130-US Hospitals for Years 1999–2008 [Dataset]. UCI Machine Learning Repository. DOI: 10.24432/C5230J.

The UCI source page lists CC BY 4.0. Rights were checked on 2026-09-06. Retain attribution, licence and change notices. No endorsement by the source authors, UCI or hospitals is implied.

Files and transformations

• —raw/uci-296.zip: unchanged upstream ZIP, including source CSV and ID mapping.
• —data/encounters.parquet: all source CSV columns and rows, stored as strings to preserve identifiers and original categorical/missing-value codes.
• —manifest.json: source URL, byte hashes, transformation versions and checks.
• —prepare.py: conversion and round-trip verification script; run from the parent directory containing both dataset folders. No network access is performed.

The source ? marker is retained, not silently interpreted as zero or a diagnosis. Encounter and patient identifiers are different; repeated admissions must not be counted as distinct people. No linkage to outside datasets or re-identification has been performed. Public availability does not make individual clinical use appropriate. No raw rows are placed in the RareBurden code repository.

Limits

Selected inpatient encounters from 1999–2008 do not represent contemporary or national diabetes prevalence. These records do not establish genetically confirmed monogenic case status. Readmission categories and ascertainment require the upstream documentation. No privacy attack, clinical recommendation or causal treatment-effect inference is supported by this mirror.