Endovascular thrombectomy activity as a candidate marker of organised acute stroke-service capacity: a GBD-based European panel study — reproduction code

Published: 8 October 2026| Version 1 | DOI: 10.17632/r9p47jkz8w.1
Contributors:
,
,
,
,
,
,

Description

This deposit provides the analysis code for a repeated cross-sectional ecological panel study examining whether national endovascular thrombectomy (EVT) activity is a cross-national marker of organised acute stroke-service capacity in 39 European and adjacent countries. The primary balanced three-wave Aguiar panel (2016, 2019, 2020; 117 country-year observations) is complemented by a four-wave extension adding 2023 European Stroke Organisation (ESO) Stroke Service Tracker data (27 of 39 countries reporting wave 4; 144 country-year observations). The deposit contains code only. The underlying country-year datasets are not redistributed: every input is derived from existing aggregate sources obtainable from their original publishers — GBD 2023 (IHME GBD Results Tool), IAEA IMAGINE, OECD Health Statistics, WHO GHO, World Bank WDI and the Aguiar de Sousa et al. surveys are openly available; the 2023 ESO Stroke Service Tracker data are not public and are available from the ESO on request. data_dictionary.md documents the provenance of every column, so the analytic panel can be reassembled from the originals. Contents: data_dictionary.md (column-level documentation + provenance); statistical_results.json (cached regression output — coefficients, CIs, p-values reported in the manuscript); statistical_analysis.py (Stage 1, every regression); reproduce_all.py (four-stage orchestrator); requirements.txt (Python 3.13 pin set); scripts/ (panel_tables.py Tables 1–3, panel_figures.py Figures 1–3 + Supp Figure 2, panel_supplement.py 14 supplementary tables + Supp Figure 1, plus helpers tex_utils.py, interpolate_ct.py, data_prep.py); LICENSE-MIT.txt. Reuse: code MIT. Running the pipeline requires reassembling the country-year panel from the sources above. Please cite the linked manuscript when using this code or any derived outputs.

Files

Steps to reproduce

This deposit is code only; the underlying country-year panel is NOT included. First reassemble panel_frozen.csv and master_table.csv from their original sources (GBD 2023 via the IHME GBD Results Tool; IAEA IMAGINE, OECD Health Statistics, WHO GHO, World Bank WDI and the Aguiar de Sousa et al. surveys, all openly available; the 2023 ESO Stroke Service Tracker wave on request from the European Stroke Organisation). Per-column provenance is in data_dictionary.md. Then: 1. Set up the pinned Python 3.13 environment: python -m venv .venv && source .venv/bin/activate (Windows: .venv\Scripts\activate); pip install -r requirements.txt 2. python reproduce_all.py --check-only (preflight: required scripts + linearmodels importable; does not need the data) 3. python reproduce_all.py (runs all 4 stages once the panel is in place) Stage 1 (statistical_analysis.py) reads panel_frozen.csv and master_table.csv and writes statistical_results.json: primary Mundlak within–between model (country-clustered robust SE), supporting specifications (pooled OLS, two-way FE, first differences, FE excluding 2020, cross-section), Mundlak sensitivities S1–S5, the 4-wave extension + balanced-27 sub-restriction, comparator exposures, adjusted EVT models, the alternative endpoint (GBD ischaemic stroke deaths rate), and influence diagnostics. Stages 2–4 (scripts/panel_tables.py, panel_figures.py, panel_supplement.py) produce Tables 1–3, Figures 1–3, Supplementary Figures 1–2 and the 14 supplementary tables. The shipped statistical_results.json records the exact reported numbers for verification. Headline (primary 3-wave Aguiar Mundlak, ischaemic stroke MIR, fully standardised): between-country β = −0.42 SD MIR per SD EVT (95% CI −0.79 to −0.05; p = 0.027; 39 countries, 117 country-year observations); four-wave extension MIR between β = −0.38 (p = 0.036; 144 observations).

Institutions

Categories

Epidemiology, Public Health, Stroke, Health Services Research, Cerebrovascular Disease

Licence