The outpatient POCUS and AI litmus test: why AAA screening can scale and others won't (Part 2)

In Part 1, I argued that POCUS has not achieved outpatient scale despite billions in investment, not because the technology failed, but because commercial architecture was never engineered around payment realities. TAM models assumed frictionless reimbursement. The system is not frictionless.

The pattern becomes clearer when examining specific applications. Not all clinical utility translates to commercial viability. The difference is payment architecture.

Where AI investment has misaligned

Lung ultrasound (LUS) to support heart failure management represents strong signal detection with weak reimbursement persistence. U.S. cardiology practice remains anchored to volume echocardiography and biomarker-driven workflows. Lung ultrasound is episodically useful for symptomatic evaluation, but payer coverage is not structured for longitudinal monitoring. Very few cardiologists even leverage ultrasound or are trained in LUS for this purpose, so it is not a common tool of consideration. Medicare covers lung ultrasound (CPT 76604) for symptomatic evaluation but does not recognize it as a chronic disease monitoring service with recurring reimbursement. There is no endorsed screening or chronic management pathway and arguably little HEOR evidence, their currency, to support it as standard of care. The clinical utility is real. The payment architecture for routine deployment does not exist.

Gestational age estimation offers high social value in access deserts and improves dating accuracy when early prenatal care is available. But reimbursement is anchored to comprehensive obstetric ultrasounds. Most commercial payers reimburse comprehensive OB ultrasounds (CPT 76805) at $150 to $300, while isolated biometry has no distinct reimbursement pathway outside of that comprehensive study, so it becomes a nice-to-have feature but likely with added cost. Gestational age alone is a component finding, not a standalone payable service. Without integration into a bundled OB care model with aligned incentives, the economic value remains theoretical.

AI-guided limited echocardiography addresses sonographer scarcity and can greatly expand user bases for image acquisition, critical when sonographers are in great shortage internationally and even in the U.S. there are only about 14,000 cardiac-trained sonographers. However, available products are limited by the lack of AI guidance for at least color Doppler. The limited scans often fail to answer the clinical question. Quantitative hemodynamic assessment, valve evaluation, and longitudinal surveillance typically require comprehensive studies. Limited echoes that appear abnormal frequently convert to full echocardiograms. Color Doppler remains non-negotiable for ruling out the presence or progression of valvular disease, murmurs, and heart failure assessment. While the technology is fascinating for reducing initial friction, it risks not reducing total cost and in some cases may increase downstream utilization.

Carotid POCUS for stroke risk screening illustrates what happens when clinical logic exists without payment alignment. Screening asymptomatic individuals for carotid stenosis is generally not reimbursed and is considered investigational by most payers. Unlike AAA screening, there is no USPSTF endorsement for population-level carotid screening. The anatomy is accessible. The technology works. But to date, evidence of asymptomatic stenosis has been shown to lead to potentially unnecessary additional testing and interventions that ultimately have not improved outcomes. What may be thought of as a great idea has incomplete HEOR evidence, lacks guideline support, and lacks coverage pathways, leaving no clear path to scale.

Venous compression ultrasound for DVT is covered, but only with adequate view quality and documentation. Poor documentation leads to frequent denials if studies are abbreviated or incomplete. This is a case where the payment pathway exists but product design often fails to enforce the alignment to view requirements and documentation standards needed to capture it. The reimbursement is available. The workflow to reliably access it is not.

These are not failures of engineering. They are mismatches between clinical capability and payment logic.

The emerging exception: AI-assisted interpretation

CPT pathways for AI-assisted interpretation are evolving, and progress is occurring. Some payers now cover AI-assisted interpretation when the applicable technology has been used by the provider, for example Ultromics or US2.ai. This represents an evolving payment architecture, and diligent market access and HEOR evidence are required to apply for a CPT or HCPCS billing code. If successful, subsequent deliberate payer engagement for medical and coverage policy would still be needed.

The lesson is not that payment architecture is immutable. It is that changing it requires intentional effort, typically years before commercialization.

Why AAA screening is structurally different

Abdominal aortic aneurysm (AAA) screening represents the structural inverse.

AAA screening is USPSTF Grade B recommended for men aged 65 to 75 who have ever smoked, and of undetermined evidence in women of similar age. The clinical pathway is established. The disease progression is well-characterized, so testing is justified for preventable harm. Defined surveillance intervals exist based on aneurysm diameter. The payer logic is straightforward. Prevent catastrophic rupture, avoid high-cost emergency surgery, reduce mortality. Medical necessity is unambiguous.

Yet screening rates remain suboptimal. Estimates suggest that only one-third of eligible adults have been screened, largely due to sonographer shortages, imaging center capacity constraints, and lack of practical primary care deployment models. The anatomy is relatively simple. The finding is binary, aneurysm present or absent, with diameter measurement as an anchor to guide surveillance or intervention. Interpretation variance is low. The workflow is naturally suited to task-shifting with appropriate supervision and quality oversight. Beyond surveillance, even post-operative care often calls for AAA ultrasound to assess for endoleaks, stent position, and sac diameter, along with flow patterns in the sac and branch vessels with the addition of color Doppler.

AAA screening has coverage, guideline support, a defined population, and a broken delivery model. Quite the opportunity that few, if any, are actively targeting. AI-enabled ultrasound does not need to create the market. It needs to operationalize an existing one.

The litmus test

The pattern is consistent.

AAA checks every box. Lung ultrasound, gestational age estimation, carotid screening, and limited echo guidance do not, unless tightly capitated delivery models align incentives across the care continuum as a path to gain real-world HEOR evidence.

Just because reimbursement mechanisms for the others do not exist, it does not mean they may not be valued in care. But it does mean that a market access strategy, HEOR evidence, and a payer engagement plan will need to be solidified beyond just selling devices and features.

What this means for product strategy

Bolster your MRD approach and business case. The traditional medical device development sequence is to identify clinical need, build technology, validate clinically, then figure out reimbursement. That sequence is why billions have been invested without achieving outpatient scale.

In such a path, the Market Requirements Document (MRD) often has less rigor before developing and proceeding with the Product Requirements Document (PRD). Commercial evaluation, validation, and payment architecture analysis should precede and constrain product design decisions.

• Target population size and prevalence. What is the annual incidence or prevalence of the clinical condition? What percentage meets inclusion criteria for the intended use? This determines addressable market size and payer interest. A large prevalent population with clear diagnostic criteria is fundamentally different from a niche indication requiring complex patient identification.
• Care setting and continuum placement. Is the product intended for primary care, specialty clinics, inpatient, outpatient, or home settings? Is it for early detection, diagnosis, triage, monitoring, or follow-up? Site of service dictates the incumbent revenue cycle constraints affecting reimbursement rates, supervision requirements, and who controls adoption decisions.
• Procedure utilization volume. How often is the associated procedure performed per patient per year? Can this solution be deployed at scale across provider types? A one-time screening exam has different economics than longitudinal monitoring. High-frequency utilization supports recurring revenue but may trigger utilization management scrutiny.
• Existing CPT or HCPCS codes. Are there existing procedure codes that can support billing? Are these codes limited by provider type, setting, imaging completeness, or supervision requirements? Building around existing codes accelerates reimbursement feasibility. Requiring new codes can add years and uncertainty.
• Payer mix of target population. What percentage of the clinical population is covered by Medicare, Medicaid, Commercial, or cash-pay? Are they in value-based plans such as Medicare Advantage, ACOs, or MCOs, or in traditional fee-for-service? Payer mix shapes coverage strategy and pricing. Medicare Advantage populations may value cost avoidance differently than fee-for-service Medicare.
• Payer line of business alignment. Is this condition managed heavily in Medicaid MCOs, Medicare Advantage, or Commercial plans? Do payers in these segments have incentives aligned with the innovation? A technology that reduces downstream utilization creates value for risk-bearing entities but may represent revenue loss for fee-for-service providers.
• Clinical workflow and delegation potential. Can image acquisition or data input be delegated to non-licensed staff? How is supervision, documentation, and interpretation structured? Delegation potential affects labor economics and scalability. Products that require physician-performed acquisition face different constraints than those enabling MA or RN acquisition with remote interpretation, all of which may have implications on technical fee coverage and reimbursement, independent of state scope-of-practice allowances.
• HEOR evidence requirements. What clinical outcomes, cost offsets, or access improvements must be demonstrated? What value levers matter to payers, such as utilization reduction, earlier diagnosis, network adequacy and expansion, or medical cost reduction and avoidance? Understanding payer evidence requirements before clinical trial design prevents building a clinical evidence dossier that does not support coverage decisions.

These questions should be answered with data and not assumptions before product requirements are finalized. The answers constrain what should be built, not just how it should be marketed.

Connecting product to commercialization strategy

• Start with payment architecture, not clinical capability. Before building, map the CPT landscape, medical policy language, prior authorization requirements, and audit exposure. If the payment pathway does not exist or requires policy evolution, that is a commercialization variable, not an afterthought.
• Engage payers before commercialization, not after. Medical policy language can evolve, but it requires deliberate engagement. The companies building AI interpretation pathways today started payer conversations years ago. Companies that treat payer relations as a post-launch function will face the same structural barriers that have constrained the industry for a decade.
• Design documentation workflows around medical policy language. If payer policies require specific views, medical necessity documentation, or defined clinical indications, product design should enforce compliance and not assume clinicians will adapt. DVT ultrasound is covered, but poor documentation leads to denials. The product should make compliant documentation the path of least resistance, or make an informed decision to pursue a new code.
• Model denial rates and prior authorization exposure as core variables. A 20% to 25% effective revenue haircut from denials, downcoding, or bundling changes unit economics fundamentally. TAM models that ignore this are theoretical.
• Build for service delivery, not device sales. Ultrasound outside hospital walls is a payment-governed diagnostic service. The companies that recognize this will design business models around adoption, utilization, outcomes, workflows, and payer alignment, not hardware margins around a feature-laden commodity.

The path forward

The technology is improving. The clinical evidence is evolving. The question is no longer whether AI features and POCUS can work in outpatient settings, but whether the evidence supports payment architecture that drives scale.

AAA screening is the proof of a clear untapped opportunity. It demonstrates that when clinical utility aligns with coverage, guidelines, defined populations, and delivery gaps, the potential exists for AI-enabled POCUS to operationalize markets that traditional imaging infrastructure cannot serve.

The companies that recognize this pattern will build differently. They will start with payment architecture. They will understand the business relationship of their providers with payers, including the revenue cycle. They will engage payers as partners, not obstacles. They will design products around audit defensibility and documentation compliance. They will model economics honestly.

Handheld POCUS has the potential to reshape ambulatory medicine. However, that potential will only be realized when commercial architecture matches clinical capability.

This is the second of a two-part series. If you missed it, start with Part 1: The ultrasound TAM mirage, why billions have not unlocked outpatient POCUS scale.

All views, analyses, and services presented here reflect my independent professional perspective, informed by more than two decades of real-world experience, and do not represent the views or positions of any current or former employer or affiliated organization.