Bridging the Reproducibility Divide: Open Source Software's Role in Standardizing Healthcare AI

John Wu; Zhenbang Wu; Jimeng Sun

cs.CYMar 2, 2026

Bridging the Reproducibility Divide: Open Source Software's Role in Standardizing Healthcare AI

Authors: John Wu, Zhenbang Wu, Jimeng Sun

Organizations: University of Illinois, Urbana-Champaign, USA

Abstract

Our analysis of recent AI4H publications reveals that, despite a trend toward utilizing open datasets and sharing modeling code, 74% of AI4H papers still rely on private datasets or do not share their code. This is especially concerning in healthcare applications, where trust is essential. Furthermore, inconsistent and poorly documented data preprocessing pipelines result in variable model performance reports, even for identical tasks and datasets, making it challenging to evaluate the true effectiveness of AI models. Despite the challenges posed by the reproducibility crisis, addressing these issues through open practices offers substantial benefits. For instance, while the reproducibility mandate adds extra effort to research and publication, it significantly enhances the impact of the work. Our analysis shows that papers that used both public datasets and shared code received, on average, 110% more citations than those that do neither--more than doubling the citation count. Given the clear benefits of enhancing reproducibility, it is imperative for the AI4H community to take concrete steps to overcome existing barriers. The community should promote open science practices, establish standardized guidelines for data preprocessing, and develop robust benchmarks. Tackling these challenges through open-source development can improve reproducibility, which is essential for ensuring that AI models are safe, effective, and beneficial for patient care. This approach will help build more trustworthy AI systems that can be integrated into healthcare settings, ultimately contributing to better patient outcomes and advancing the field of medicine.

Figures & tables

Figure 1: Paper Data Collection and Analysis: (a) Paper scraping details: We scrape PDF files directly from conference webpages, extracting and cleaning important paper details like title, authors, abstracts, and emails. We query BioC to scrape paper information from PubMed’s Open Access Database, related to research papers that contain the terms AI, deep learning, machine learning, healthcare, electronic health records, and electronic medical records into XML formats. (b) Stacked bar chart of total number of papers scraped each year, showing a steady increase in paper counts over time. (c) Key analyses performed: Mapping citation counts to each paper using scholarly services like SerpAPI [ 63 ] , Semantic Scholar [ 38 ] , and PMIDcite [ 24 ] ; scraping affiliation details by classifying emails or affiliations sourced from PubMed’s MEDLINE API [ 24 ] ; classifying each paper’s topic by its abstract and title with a medically fine-tuned large language model (OpenBioLLM-70b); checking for code sharing by identifying specific code keywords (e.g., GitHub, Zenodo, Colab, GitLab) in each paper’s main text (excluding the references); assessing public dataset usage by checking for famous dataset keywords in the main text and cross-referencing each paper with the PapersWithCode API. We go into further detail in Appendix A.1 .

Figure 2: Trends in Public Dataset Usage: (a) Distribution of public dataset usage over time and across venues: MIMIC is the most commonly used or mentioned public dataset; private datasets dominate overall each year; conferences use more public datasets than journal papers. (b) Rate of public dataset usage across topics: Biosignal papers use the most public datasets. (c) Rate of public dataset usage across affiliations: Industry surprisingly uses the most public datasets. (d) Distribution of public dataset usage over time: Papers using public datasets have higher citation counts on average every year. The maximum number of citations on the y-axis is capped at 50 to focus on non-outlier behavior. (e) Cumulative distribution of using public datasets: Papers that use public datasets tend to have greater citation counts, regardless of outlier status.

Figure 3: Trends in Code Sharing: (a) Code-sharing percentage across different venues over time. Conference papers share code significantly more often than other PubMed papers. (b) Rate of code sharing by topic: Biomedicine papers share code slightly more often. (c) Rate of code sharing by affiliation: Industry shares slightly less than others. (d) Distribution of citation counts for papers with and without code over time: Papers that share code generally have at least the same number of citations or more each year. Citation counts are limited to 50 or below to better visualize typical citation numbers beyond outlier papers. (e) Cumulative distribution of code sharing vs. no code sharing: Papers that share code overall have more citations, regardless of outlier status.

Category	Accuracy	Precision	Recall	F1 Score	Estimated Rate	True Rate
Public Dataset Usage	0.77	0.90	0.60	0.72	0.33	0.50
Code Sharing	0.87	0.71	0.71	0.71	0.23	0.23
Topic Classification	0.93	0.94	0.93	0.93	-	-

Table I: Manual validation of our automated approach on 30 randomly sampled papers, examining code sharing, public dataset usage, and topic classification. Comparing true rates from manual review against estimated rates from keyword searches, we found exact agreement between methods for code sharing (23% both estimated and true), while public dataset usage was significantly underestimated (33% estimated vs 50% true). Topic classification performed exceptionally well, while code sharing detection was reliable but conservative, and public dataset detection showed high precision but missed some instances due to varied datasets not covered in the PapersWithCode API or our initial pool of popular datasets.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Panel of Figure 4 — Figure 4: Prompt for standardizing noisily extracted conference paper titles.

Panel of Figure 5 — Figure 5: Prompt template for extracting and cleaning email addresses from author details using a large language model. The {text} is the extracted author detail text from the noisy pdf files.

Category	Email Domain Suffixes	Affiliation Keywords
Industry	• .com • .co • .org • .io	• google, ibm, iqvia, microsoft, apple, amazon • facebook, intel, nvidia, deargen, riken • corporation, inc, llc, ltd, gmbh, company • biotech, analytics, corp, tno • leo innovation lab, widex, axispoint • optum, tencent, analytix
Academic	• .edu • .ac • .eth.ch • .cispa.de • dkfz-heidelberg.de	• university, school, college, institute of technology • polytechnic, instituto superior tecnico • universit, hospital, medical center • health science, academia, faculty • department of, school of medicine • research institute, national institute • cancer center, clinic • basque center for applied mathematics • association, fau erlangen-nurnberg • national, isi foundation, institute for health • amsterdam umc, iust, mayo clinic, NIH • .ac.uk, universidad

Appendix

Table II: Keyword Classification Criteria for Industry and Academic Affiliations. Papers were classified as academic, industry, or mixed based on author affiliations. Paper author details with only academic keywords were labeled academic, only industry keywords as industry, and those containing both as mixed affiliations.

Dataset	Search Terms
MIMIC	Medical Information Mart for Intensive Care
eICU	eICU Collaborative Research Database
UK Biobank	UK Biobank
Chest X-Ray14	Chest X-Ray14, NIH Chest X-ray
ADNI	ADNI, Alzheimer’s Disease Neuroimaging Initiative
PhysioNet	PhysioNet

Appendix

Table III: Dataset names and their corresponding search terms used in the analysis.

Figure 9: Papers with higher citation counts show increased usage of public datasets, as also demonstrated by the cumulative distribution plot in Figure 2 e. This trend indicates a positive correlation between paper visibility and researchers’ tendency to use public datasets. However, the majority of papers have less than 10 citations and the vast majority of such papers use private datasets, suggesting that the benefits of reproducible practices are not noticeable for the vast majority of papers.

Figure 10: Figure shows the relationship between paper citations and code availability across different citation thresholds. The proportion of papers with available code increases with citation count, from 24% for papers with any citations to 53% for highly-cited papers ( $\geq$ 300 citations). While most papers in the dataset (n=2,610) have relatively few citations, those with higher citation counts demonstrate improved code sharing practices. This follows the observations in the empirical distribution plot in Figure 3 e.

Open-Source Software Development

See all 14 papers

The Shift Toward Open and Reproducible AI Research
Jun 15, 2026Kevin L Coakley, Thijs Snelleman, Holger Hoos +1ML Reproducibility Scientific Reproducibility
Department of Computer Science, Norwegian University of Science and Technology, Sem Sælands Vei 7-9, Trondheim, 7031, Trønderlag, Norway. · San Diego Supercomputer Center, University of California San Diego, 9500 Gilman Dr, La Jolla, 92093, California, USA. · Chair of AI Methodology, RWTH Aachen University, Theaterstrasse 35-39, Aachen, 52062, North Rhine-Westphalia, Germany. +1
Code Sharing In Prediction Model Research: A Scoping Review
Mar 16, 2026Thomas Sounack, Raffaele Giancotti, Catherine A. Gao +8Scientific Reproducibility ML Reproducibility
Dana-Farber Cancer Institute · University of Calabria · Northwestern Feinberg School of Medicine +9
Perspective on Bias in Biomedical AI: Preventing Downstream Healthcare Disparities
Apr 16, 2026Michal Rosen-Zvi, Yoav Kan-Tor, Michael Danziger +6Algorithmic Fairness Algorithmic Bias
IBM Research - Israel · Faculty of Medicine, The Hebrew University · Israel +6

Jun 15, 2026J/K move · Enter open · S save

The Shift Toward Open and Reproducible AI Research

Kevin L Coakley, Thijs Snelleman, Holger Hoos, Odd Erik Gundersen

Department of Computer Science, Norwegian University of Science and Technology, Sem Sælands Vei 7-9, Trondheim, 7031, Trønderlag, Norway. · San Diego Supercomputer Center, University of California San Diego, 9500 Gilman Dr, La Jolla, 92093, California, USA. · Chair of AI Methodology, RWTH Aachen University, Theaterstrasse 35-39, Aachen, 52062, North Rhine-Westphalia, Germany.+1

ML Reproducibility Scientific Reproducibility

Open paper AskPDF arXiv

The reproducibility crisis has directed the AI research community toward improving documentation practices. Several studies have identified methodological issues, and in response, the most impactful venues in the field have introduced reproducibility checklists. We seek to understand whether documentation practices have changed over time by assessing all published papers at five leading AI conferences over the past decade. Seven reproducibility variables were identified, quality-assured and used to analyse 56 800 publications. Our analysis reveals that in the period 2014 to 2024, documentation practices have improved; papers sharing both code and data increased nearly sixfold, from 11% to 64% Building on empirical reproducibility rates from a prior study, we estimate - inferred from documentation practices, not direct testing - that reproducibility increased from 28% in 2014 to 64% in 2024. Improvements in documentation practices predate the introduction of reproducibility checklists, suggesting these changes reflect a broader movement toward open science rather than a direct response to formal requirements.

Abstract

Open-Source Software Development

Explore similar work

The Shift Toward Open and Reproducible AI Research

Code Sharing In Prediction Model Research: A Scoping Review

Perspective on Bias in Biomedical AI: Preventing Downstream Healthcare Disparities