Validity, Judgment, and Authority: The Hypotension Prediction Index as a Test Case for Clinical AI Alignment
The Hypotension Prediction Index (HPI) estimates from the arterial waveform the likelihood of impending hypotension and, when the index is high, prompts review of a secondary screen that displays pulse pressure or stroke volume variation (PPV/SVV), peripheral arterial dP/dt, and dynamic arterial elastance (Eadyn) under the headings preload, contractility, and afterload. Its predictive performance is supported by meta-analyses, but the physiological meaning of the secondary-screen variables depends on conditions that the monitor does not verify, so that a numerically correct value can be physiologically uninterpretable. This narrative review summarizes the evidence on the prediction and on the validity conditions of each variable, using a companion conceptual framework that distinguishes the capability to evaluate from the authority to act. We separate the documented functions of the current software from extensions that we propose: a runtime validity gate that would verify, flag, or withhold interpretations whose prerequisites cannot be established, and a bedside sequence for clinicians. The framework, the proposed PPV checklist, and the taxonomy tables are expert-derived, have not been tested against clinical outcomes or usability data, and concern a single manufacturer's product. Within these limits, HPI illustrates a principle that we propose for clinical AI: prediction accuracy does not guarantee interpretation validity, and interpretation validity does not confer therapeutic authority.
Authors
- Christian de Tymowski (ORCID: https://orcid.org/0000-0001-6899-9912)
- Dan Longrois (ORCID: https://orcid.org/0000-0001-5864-8975)
- Stefan Andrei (ORCID: https://orcid.org/0000-0002-0216-1380)
- David Holcman
- Pierre-Grégoire Guinot
Institutions
- Université Paris Cité (FR)
- University of Cambridge (GB)
- Hôpital Louis-Mourier (FR)
- Université Sorbonne Paris Nord (FR)
- CHU Dijon Bourgogne (FR)
- Université Bourgogne Franche-Comté (FR)
- Institut de Biologie de l'École Normale Supérieure (FR)
- Laboratoire de Recherche Vasculaire Translationnelle (FR)
- Hôpital Bichat-Claude-Bernard (FR)
Publication Details
- Journal
- The Journal of Critical Care Medicine
- Published
- 2026-11-16
- DOI
- https://doi.org/10.62838/jccm-2026-0068
- Primary Topic
- Hemodynamic Monitoring and Therapy
- Type
- article
- Field-Weighted Citation Impact
- 0.00