Comment on “The impact of underdiagnosis and delay in care for patients with moderate and severe aortic stenosis: A real-world database cohort study”

Dear Editor,

We read with great interest the study by Peterson et al. [1], quantifying the disparities in formal aortic stenosis diagnosis following echocardiographic detection.

1. Sponsor identification of the cohort through EMR-specific severity fields introduces a structural selection step not fully accounted for in the disparity analysis

The study population was restricted to a small subset of echocardiograms with structured severity fields recorded directly in the EMR portion of the database, reducing an initial pool of over three million eligible patients to 39,185 with AS-specific readings and further to 3693 meeting moderate or severe criteria [2]. This funnel depends entirely on which sites and providers used structured reporting fields rather than free-text or PDF reports, a documentation practice itself likely correlated with institutional resources, academic affiliation, and possibly patient demographics [3]. Since the entire disparity analysis rests on this pre-selected structured-data subset, the reported racial and sex gaps in subsequent diagnosis could be confounded by which institutions and provider types generate structured severity data in the first place, a possibility the discussion does not address despite its direct bearing on whether these disparities generalize to the broader undiagnosed AS population invisible to this dataset.

2. The post hoc landmark analysis addressing immortal time bias is presented as confirmatory rather than as a partial correction with its own limitations

The authors appropriately recognized that AVR is essentially uncodeable without a preceding AS diagnosis, producing implausible hazard ratios exceeding 100 in the primary model, and responded with a 30-day landmark reanalysis [4]. However, this correction only addresses immortal time within the diagnosis-to-AVR pathway; it does not resolve the same structural dependency operating in the primary time-to-second-echocardiogram outcome, where formal diagnosis plausibly triggers the very documentation and billing pathways that generate a recorded follow-up study. A patient whose AS is never formally coded may still receive clinically identical surveillance imaging that goes uncaptured as an AS-related encounter in claims data, meaning the 1.7–2.3-fold associations for follow-up imaging likely retain some of the same coding-dependency artifact that necessitated the landmark correction for AVR, even after that correction was applied.

3. Comorbidity burden was measured only at baseline, leaving open whether undiagnosed patients had less severe or differently distributed valve disease that reduced clinical urgency

The Elixhauser scores were similar between the diagnosed and undiagnosed groups, which the discussion uses to argue that disparities are not explained by differing health status. However, this index captures the overall comorbidity burden but not AS-specific clinical context, such as symptom documentation, ejection fraction, or the specific echocardiographic parameter driving severity classification, information the authors note was substantially missing across the cohort. Two patients with identical Elixhauser scores could differ meaningfully in symptomatic burden or borderline-versus-clear-cut severity grading, and without this granularity, similar comorbidity scores cannot fully rule out clinical justification for differential diagnostic follow-through, even as the demographic patterns remain concerning on their own terms.

4. Conclusion

This study makes a valuable and clinically urgent contribution by quantifying a diagnosis-documentation gap with direct consequences for guideline-concordant care, and its finding that formal diagnosis predicts subsequent management supports the paper's call for systematic interventions, such as electronic provider notifications. Characterizing how structured-field availability itself varies by site and patient demographics, and incorporating symptom or severity-parameter granularity into the comorbidity comparison, would strengthen confidence that the reported disparities reflect genuine inequities in clinical recognition rather than artifacts of which patients' echocardiograms produced analyzable structured data in the first place.

CRediT authorship contribution statement

Ram Krishna Rathore: Conceptualization, Data curation, Formal analysis. Tanvi Pankaj Vyas: Writing – review & editing, Writing – original draft, Visualization. Muradova Emma Vladimirovna: Writing – review & editing, Writing – original draft, Visualization.

Informed consent

Not applicable.

Ethical approval

Not applicable (Commentary on a previously published article).

Declaration of Generative AI and AI-assisted technologies in the writing process

Generative AI tools were used only for language refinement and formatting assistance. All scientific interpretation, critical appraisal, and manuscript preparation were performed independently by the authors.

Funding

None.

Declaration of competing interest

The authors declare that there are no conflicts of interest.

Data availability

No new data were generated or analyzed in this study.