From Association to Action: Clarifying the Incremental Value of Machine‐Reported RAD/RVH

1.

To the Editor:

We read with great interest the study by Fuchigami, which evaluated machine‐reported right axis deviation or right ventricular hypertrophy (RAD/RVH) against echocardiographic pulmonary hypertension phenotypes in 68 905 ECG–echocardiography pairs [1]. The large linked cohort, patient‐level cluster‐robust analysis, and multiple sensitivity analyses provide valuable real‐world evidence. We agree with the author that RAD/RVH should be regarded as a supportive clue rather than an exclusionary screening test.

However, the clinical interpretation of “high specificity” deserves further qualification. For the specific Echo‐PH phenotype, specificity was 96.7%, but sensitivity was only 7.5%, the positive likelihood ratio was 2.30, and the positive predictive value was 35.5% [1]. Thus, a positive machine report produces only a modest probability shift, whereas a negative report provides almost no reduction in probability, with a negative likelihood ratio of 0.96. Moreover, the adjusted probability of Echo‐PH specific remained 30.0% among RAD/RVH‐positive pairs, indicating that most positive reports were not accompanied by the primary phenotype. These findings support contextual interpretation but do not establish RAD/RVH as a stand‐alone rule‐in trigger.

A useful next step would be to quantify its incremental clinical value beyond information already available at the bedside. A baseline model containing demographics, comorbidities, ICU status, and readily available quantitative ECG measurements could be compared with the same model plus machine‐reported RAD/RVH. Changes in discrimination, calibration, Brier score, and decision‐curve net benefit would clarify whether the machine label meaningfully improves decisions such as expedited echocardiography or structured right heart review [2]. Subgroup analyses according to heart failure, chronic lung disease, pulmonary embolism, and ICU status may also identify settings in which the signal is most informative.

Finally, because the reference outcome was echocardiographic rather than hemodynamic, its proposed use should remain explicitly limited to identifying right heart phenotypes; current guidelines retain right heart catheterization for confirmation of pulmonary hypertension [3]. External validation across ECG vendors, coupled with an analysis excluding uncertain RVH phrases, would further strengthen transportability.

In summary, this study establishes machine‐reported RAD/RVH as a pragmatic marker of right heart disease burden. Demonstrating incremental net benefit, rather than association alone, would provide the clearest bridge from routinely generated ECG text to an actionable clinical workflow.