Automated CPET Interpretation: International Validation Study

Change in agreement between key KOLs and the algorithm-generated CPET interpretation reports on the survey, comparing baseline (prior to implementation) to the 6-month study period. Agreement will be assessed per interpretation domain using percent agreement and Cohen's kappa, with comparisons made both at the individual KOL level and across all KOLs.

Baseline agreement on the survey between the KOLs and the CPET interpretation algorithm will be assessed. The survey assesses CPET interpretation both without and with access to the automated report. Agreement will be assessed per interpretation domain using percent agreement and Cohen's kappa, with comparisons made both at the individual KOL level and across all KOLs. This establishes the initial level of alignment before implementation and allows evaluation of the immediate effect of having access to the report.

Agreement at 6 months on the survey between the KOLs and the CPET interpretation algorithm will be assessed. Again the survey assesses CPET interpretation both without and with access to the automated report. Agreement will be assessed per interpretation domain using percent agreement and Cohen's kappa, with comparisons made both at the individual KOL level and across all KOLs.

Change in self-reported confidence scores before and after using the automated report. Confidence will be rated on a 5-point Likert scale across interpretation domains. Within-timepoint comparisons (without vs. with algorithm) and across-time comparisons (baseline vs. 6 months) will be made.

Feedback on automated protocols generated via the platform during the 6-month implementation period. Changes will be categorised as: accept, accept with minor change, accept with major change or reject. This helps to quantify how the automated interpretation is accepted or refined in practice.

Descriptive evaluation of center-level CPET interpretation reports practices at three timepoints: at baseline, 3 months and 6 months on a random sample (n=10) of included participants of CPET interpretation reports per center, using structured qualitative coding of report content across interpretation domains to assess potential shifts in reporting style, terminology, and completeness.