With an increasing number of people living with chronic respiratory failure, the use of home mechanical ventilation (HMV) has escalated in recent years. Understanding the complexity of care, from navigating device selection and qualification to titration and compliance, can be overwhelming for clinicians. This process is further convoluted by the vast amount of electronic data available and a lack of shared terminology among device manufacturers. The use of artificial intelligence (AI) in modern medicine is rapidly evolving to process this information. It is important to highlight its use in HMV by both clinicians and insurance payors.
Use by clinicians
Clinical decision-support AI platforms, such as OpenEvidence and UpToDate Expert AI, differ from general large language models (LLMs), such as ChatGPT and Google Gemini, by using retrieval-augmented generation (RAG) to ground results in reputable sources.1 These platforms reference current guidelines and provide citations. It is estimated that 81% of physicians in the United States use AI, and OpenEvidence estimates that 40% of physicians use the platform daily.2,3
For general purposes, such as the broad principles of device qualification and differentiating respiratory assist devices (RADs) from HMV, these programs can be useful. Platforms like OpenEvidence now use programs such as “EvidenceGrade” to help clinicians review the strength of citations. While this provides some transparency, the inherent “black box” problem with AI algorithms remains, where the internal logic is unclear. Limitations include data bias and a lag between changes in current practice and updated guidelines.
In addition, the material presented is limited by user queries and still requires data synthesis. For example, depending on how the question is phrased, OpenEvidence still cites the need for sleep oximetry based on the 1999 US Centers for Medicare and Medicaid Services (CMS) policy for noninvasive ventilation (NIV) qualification in COPD, despite this no longer being required with the 2025 national coverage determination (NCD) update. The HMV-specific oxygen bleed-in requirement of 4 liters is also not referenced.4 Often citations come from organizations like CHEST and the American Thoracic Society, and the data are usually “insurance agnostic.”
Furthermore, due to limited data regarding disease-specific HMV modes and settings, which often relies on provider experience, AI mixes inputs from pediatric and adult sources to provide recommendations on settings such as inspiratory time, trigger, and cycle. While the general principles are often correct, direct application should be used with caution. The lack of universal terminology between device manufacturers is a significant limitation of AI when attempting to transition between devices. For example, NIV can refer to RADs, HMV, or BiPAP, which can confuse programs.
Finally, the role of AI in remote patient monitoring is an important area for growth. With the availability of large amounts of ventilator data via cloud-based data monitoring through modems, there is an opportunity to create proactive, rather than reactive, patient care.5 AI has shown potential in critical care ventilation, though there is currently no single data platform.6 The availability of this information is double-edged, however, as patients can already run reports from platforms like Open Source CPAP Analysis Reporter (OSCAR) and SleepHQ through LLMs to provide data interpretations.
Use by insurance payors
While AI is used by clinicians to navigate clinical care, there is an increase in the use of clinical decision-support tools by insurers to determine medical necessity and coverage. These tools have been accused of producing high rates of care denial—up to 16 times higher, in fact—which is particularly concerning in the context of durable medical equipment like RADs and ventilators.7,8 Several insurers use Optum’s (under UnitedHealth Group) InterQual criteria for device approval. The criteria are developed using sources like CMS, literature review, and independent clinician panels. InterQual has tools such as “AutoReview” and “Auth Accelerator” that use AI to pull medical record data to match with criteria, similar to Tennr, and issues reviews, though InterQual states that there are no automated coverage denials and that final reviews are manual.9
Conclusion
Ultimately, it is important to recognize the numerous facets of modern medicine that interface with AI. It remains a clinical tool and not a subspeciality team member or clinical expert. As it applies to the landscape of chronic respiratory failure, we must be aware of data biases given the heterogeneity of the patient population and the limited evidence available.
Input matters: Pulmonary medicine has not developed a shared language for HMV, which limits the application of AI. It is important for clinical experts to continue to address practice gaps to optimize the role of AI in patient care.
References
1. Amugongo LM, Mascheroni P, Brooks S, Doering S, Seidel J. Retrieval augmented generation for large language models in healthcare: a systematic review. PLOS Digit Health. 2025;4(6):e0000877. doi:10.1371/journal.pdig.0000877
2. Center for Digital Health and AI. 2026 physician survey on augmented intelligence. American Medical Association. Published March 2026.
3. American College of Emergency Physicians. OpenEvidence. ACEP and OpenEvidence expand access to emergency medicine resources. Published December 15, 2025.
4. Centers for Medicare & Medicaid Services. Noninvasive positive pressure ventilation (NIPPV) in the home for the treatment of chronic respiratory failure consequent to COPD—decision memo. Published June 9, 2025.
5. Crimi C, Lujan M, Duiverman M. Artificial intelligence and the future of telemonitoring in home mechanical ventilation. ERJ Open Res. 2026;12(3):01717-2025. doi:10.1183/23120541.01717-2025
6. Gallifant J, Zhang J, Del Pilar Arias Lopez M, et al. Artificial intelligence for mechanical ventilation: systematic review of design, reporting standards, and bias. Br J Anaesth. 2022;128(2):343-351. doi:10.1016/j.bja.2021.09.025
7. US Senate, Committee on Homeland Security and Governmental Affairs. Refusal of Recovery: how Medicare Advantage Insurers Have Denied Patients Access to Post-Acute Care: Majority Staff Report. US Government Publishing Office; 2024.
8. Raza S, Gerke S, Silcox C, Hendricks-Sturrup R, Shachar C. Medicare advantage becoming a disadvantage with use of artificial intelligence in prior authorization review. NPJ Digit Med. 2026;9(1):208. doi:10.1038/s41746-026-02387-x
9. Optum Inc. InterQual AutoReview. 2024.
