\documentclass[10pt]{article}\usepackage[a4paper,margin=1.8cm]{geometry}\usepackage[T1]{fontenc}\usepackage{lmodern}\usepackage{longtable,booktabs}\begin{document}
\noindent\textbf{TRIPOD+AI checklist} (Collins et al., \emph{BMJ} 2024;385:e078378), applicable items for the evaluation of an existing model. Item wording is abbreviated; numbering follows the TRIPOD+AI main checklist.\par\smallskip\noindent Manuscript: \emph{Green means go: an ECG foundation model over-calls normal athletes, and athlete-criteria labels correct it}\par\medskip
\begin{longtable}{lp{2.8cm}p{5.4cm}p{6.2cm}}\toprule Item & Section & Content & Where reported\\\midrule\endhead
1 & Title & Identifies evaluation of an AI prediction model, population and outcome & Title\\
2 & Abstract & Structured summary & Abstract\\
3a-c & Background & Context, rationale, existing models and gaps & Introduction\\
4 & Objectives & Study objectives, evaluation of an existing model & Introduction (Questions)\\
5a-b & Data sources & Sources, dates/settings of data & Methods: Model; Non-athletes; Athletes\\
6a-c & Participants & Eligibility, setting, treatments & Methods: Non-athletes; Athletes (green-only rules, exclusions)\\
7 & Data preparation & Preprocessing, quality checks & Methods: Model (published preprocessing unchanged)\\
8a-c & Outcome & Definition of outcome, blinding of assessment & Methods (dataset and cardiologist labels, never model-derived; mapping fixed before outputs)\\
9a-c & Predictors & Input data definition & Methods: Model (12-lead, 10 s, 500 Hz)\\
10 & Sample size & Justification & Methods (all available eligible ECGs); Limitations (small athlete groups)\\
11 & Missing data & Handling & Methods (162 of 163 PF12RED recordings readable)\\
12a-g & Analytical methods & Thresholds, performance measures, uncertainty & Methods: Operating points (supplemental table S1); Statistics\\
13 & Class imbalance & Handling & NA (fixed operating points; no training except readout)\\
14 & Fairness & Approaches to assess fairness & Results (ancestry, descriptive); EDI statement; Limitations\\
15 & Model output & Output and thresholds & Methods: Model; Operating points\\
16 & Training vs evaluation & Differences between development and evaluation data & Methods: Model (no test data in training); Limitations (devices)\\
17 & Ethical approval & Ethics & Declarations (public de-identified data; not human-subjects research)\\
18a-f & Open science & Funding, conflicts, protocol, registration, data and code availability & Declarations; pre-registration of athlete analysis and fixes\\
19 & Patient and public involvement & PPI & Declarations\\
20a-c & Participants (results) & Flow, characteristics & Results; Methods (counts and exclusions)\\
21 & Model development & Number of participants and outcome events & Methods: Fix 1 (training folds)\\
22 & Model specification & Specification of evaluated model & Methods: Model (released weights, unchanged)\\
23a-b & Model performance & Performance with CIs, subgroups & Results; Tables 1--2; Figures 1--3\\
24 & Model updating & Results of updating & Results: Fix 1 and Fix 2\\
25 & Interpretation & Overall interpretation & Discussion\\
26 & Limitations & Limitations & Limitations\\
27a-c & Usability & Handling of poor-quality input, user interaction, implications & Discussion: clinical meaning; what would work\\
\bottomrule\end{longtable}\end{document}