Design: retrospective two-campus cohort study with paired comparison of classification discrimination
Population: 2,251 patients hospitalised with acute pulmonary embolism, 2013-2025
Primary outcome: 30-day all-cause mortality discrimination by area under the curve
Effect: AHA/ACC 0.885 (95% CI 0.858-0.912) vs ESC 0.763 (0.721-0.806), difference 0.122

A Chinese tertiary hospital across two campuses studied 2,251 patients hospitalised with acute pulmonary embolism between 2013 and 2025, comparing the 2026 AHA/ACC clinical categories against the 2019 ESC scheme for 30-day mortality. The classifications differ in structure: the ESC scheme defines high risk by haemodynamic instability, while the AHA/ACC categories run from subclinical embolism through cardiopulmonary failure.

The AHA/ACC categories discriminated better - area under the curve 0.885 (95% CI 0.858-0.912) against 0.763 (0.721-0.806), a difference of 0.122. For embolism-related death, 0.942 against 0.826. Sensitivity for embolism-related death at the analysis-defined high-risk thresholds was 83.8% for AHA/ACC categories D and E against 58.8% for ESC high risk, at similar specificity.

The number worth carrying is a small one: 37 patients fell into AHA/ACC categories D or E but only ESC intermediate risk, and 75.7% of them were dead within 30 days. That is the discordant group, and it is where a classification choice has consequences. Two cautions. Removing the AHA/ACC extremes narrowed the advantage, so part of it comes from the broader span rather than better separation in the middle. And this is one Chinese centre, retrospective, with outcomes under the care actually given - a classification cannot be credited for deaths that happened despite treatment decisions it did not drive.

A patient with normal blood pressure but severe cardiopulmonary compromise may be ESC intermediate risk and still very high risk.
Of 37 patients discordant in that direction, 75.7% died within 30 days.
Much of the AHA/ACC advantage came from spanning a wider severity range, not from better separation in the middle.
Single-centre retrospective data from one country; external validity is untested.
Do not let a normal blood pressure alone downgrade concern in a patient who looks unwell.

Why it matters

It names a group that one widely used classification calls intermediate risk and that died at a rate no one would call intermediate.

The statistics, in plain English

An area under the curve of 0.885 against 0.763 is a substantial difference in ranking ability, but the attenuation when the extreme categories were excluded shows where it comes from: a scheme that spans a wider range of severity will separate better simply by including milder and sicker patients. The discordant group of 37 is small, so 75.7% is an imprecise estimate of a striking pattern rather than a reliable figure.