Neurologist supply and neurological mortality in 3,144 US counties: analysis code and derived data

Analysis code and derived county-level datasets supporting the study "Neurologist supply and neurological mortality in 3,144 US counties: a national cross-sectional study with negative controls." We identified all Medicare-enrolled neurologists in the United States and mapped them to counties, then examined whether neurologist absence is associated with neurological mortality. Of 3,144 US counties, 1,725 (54.9%) containing 30.7 million residents had no neurologist practising within them, rising to 83.6% among the most rural counties. Counties without a neurologist had 15.3% higher stroke mortality, but also 14.7% higher pancreatic cancer mortality — a prespecified negative control outcome that shares the socioeconomic and rural gradient but lies outside neurological practice. The stroke-specific excess relative to the control was 1.006 (95% CI 0.971 to 1.046), indicating that the apparent association is not specific to stroke. An AI assistant performed data processing, statistical computation, figure generation and preparation of the first manuscript draft under author direction, as disclosed in the manuscript. This deposit exists so that every reported figure can be independently verified rather than taken on trust. CONTENTS- code/ : 14 scripts (Python and shell) covering exposure construction, suppression analysis, negative control comparison, bootstrap inference, distance calculation, and all five figures- data/ : 12 derived county-level datasets, including neurologist supply under three exposure definitions, distance to nearest neurologist, and per-county CDC WONDER suppression status- figures/ : the five manuscript figures at 300 dpi- README.md : full reproducibility documentation, mapping every headline number to the script that produces it- WONDER-pulls.md : exact CDC WONDER query settings for every mortality export SOURCE DATAAll inputs are public and free. Raw files are not redistributed here and should be obtained from source: CMS Doctors and Clinicians National Downloadable File; NPPES monthly dissemination file; CDC WONDER Underlying Cause of Death 2018-2024; HUD USPS ZIP-County crosswalk; NCHS Urban-Rural Classification Scheme for Counties 2023; US Census Gazetteer and cartographic boundary files. CDC WONDER county-level data cannot be retrieved programmatically because the API does not permit sub-national queries. Exports were produced manually using the settings recorded in WONDER-pulls.md. Bootstrap analyses are seeded and reproduce exactly. Requires Python 3.10 or later with pandas, numpy, matplotlib, pyarrow and openpyxl. No GIS libraries needed.