What question did this study set out to answer?

To evaluate whether a multimodal large language model can predict peer-reviewed publication from conference poster content alone and to determine the institutional and demographic factors influencing publication success.

September 17, 2026The American Surgeon

Can AI Predict Publication? Multimodal Large Language Models and the Structural Determinants of Surgical Scholarship

View Full Paper

Ask AI

Bookmark

Authors

SKSohail Khan GMGavin McAfee AOAlex Chiodo Ortiz

Discussion

Member takes

Overview

Retrospective study reveals that AI predicts surgical abstract publication with modest domain-specific accuracy, indicating that unmeasured structural factors drive peer-reviewed acceptance.

Key Points

To evaluate whether a multimodal large language model can predict peer-reviewed publication from conference poster content alone and to determine the institutional and demographic factors influencing publication success.
Retrospectively analyzed N=260 poster abstracts presented at the 2021–2022 American Association for the Surgery of Trauma (AAST) Annual Meetings, confirming publication through bibliographic searches and inferring author demographics from public data.
Scored poster images using GPT-4.1 across a six-domain rubric averaged over 30 iterations.
Assessed publication predictors using t-tests, Pearson chi-square or Fisher exact tests, and multivariable logistic regression.
GPT-4.1 predicted publication with an overall accuracy of 58.5% (P = .009), with domain-specific accuracy reaching 74.2% in Violence, Societal, and Behavioral topics and 62.2% in Hemorrhage, Resuscitation, and Vascular Control, but declining to chance in Critical Care and Systems Optimization.
Overall, 142 of 260 abstracts (54.6%) were published at a mean of 13.4 months, with multicenter origin identified as the sole independent predictor of publication (P = .02).
Hispanic investigators were significantly underrepresented among first authors (P = .03) and senior authors (P = .01), and were entirely absent from the published hemorrhage and vascular control cohort.

Cite This Study

Khan et al. (2026) studied this question.

synapsesocial.com/papers/6aabb7a65f706d05830e6e24 https://doi.org/10.1177/00031348261487667

View Full Paper

Ask AI

Bookmark

September 17, 2026The American Surgeon

Can AI Predict Publication? Multimodal Large Language Models and the Structural Determinants of Surgical Scholarship

View Full Paper

Ask AI

Bookmark

Authors

SKSohail Khan GMGavin McAfee AOAlex Chiodo Ortiz

Discussion

Member takes

Overview

Retrospective study reveals that AI predicts surgical abstract publication with modest domain-specific accuracy, indicating that unmeasured structural factors drive peer-reviewed acceptance.

Key Points

To evaluate whether a multimodal large language model can predict peer-reviewed publication from conference poster content alone and to determine the institutional and demographic factors influencing publication success.
Retrospectively analyzed N=260 poster abstracts presented at the 2021–2022 American Association for the Surgery of Trauma (AAST) Annual Meetings, confirming publication through bibliographic searches and inferring author demographics from public data.
Scored poster images using GPT-4.1 across a six-domain rubric averaged over 30 iterations.
Assessed publication predictors using t-tests, Pearson chi-square or Fisher exact tests, and multivariable logistic regression.
GPT-4.1 predicted publication with an overall accuracy of 58.5% (P = .009), with domain-specific accuracy reaching 74.2% in Violence, Societal, and Behavioral topics and 62.2% in Hemorrhage, Resuscitation, and Vascular Control, but declining to chance in Critical Care and Systems Optimization.
Overall, 142 of 260 abstracts (54.6%) were published at a mean of 13.4 months, with multicenter origin identified as the sole independent predictor of publication (P = .02).
Hispanic investigators were significantly underrepresented among first authors (P = .03) and senior authors (P = .01), and were entirely absent from the published hemorrhage and vascular control cohort.

Cite This Study

Khan et al. (2026) studied this question.

synapsesocial.com/papers/6aabb7a65f706d05830e6e24 https://doi.org/10.1177/00031348261487667

View Full Paper

Ask AI

Bookmark