Can AI help close a persistent gap in cardiovascular care?

This week, we’re running a Strategy Bootcamp on cardiovascular (CV) care because no matter where Union members sit in the ecosystem—health systems, payers, employers, digital health, devices, or investing—you can’t make smart decisions in healthcare without a working grasp of cardiovascular services. It’s one of the biggest spend categories, one of the most operationally complex service lines, and one of the fastest-moving markets in terms of policy, innovation, and site-of-care shifts. 

As I worked on this project, two related dynamics stood out to me: A longstanding clinical standardization problem, and, a new potential AI solution set. This pair is well-known to CV insiders, but might be flying under the radar for those who don’t work day-to-day with this clinical area.  

• The persistent clinical standardization problem: Women are more likely to present “atypically,” later, and be underdiagnosed for common and serious cardiovascular conditions. People who live in this world know the pattern and have been trying to correct it forever using clinical standardization and public communication levers. The challenge is that the pattern stubbornly remains – we have yet to see equal detection, more effective pathways, or better outcomes. 

• A new AI solution?: There’s a real push underway to use AI-enabled diagnostics to help close that gap, through better pattern recognition, earlier risk identification, and more consistent interpretation of signals that humans can miss under time pressure. This is striking to me because it's so common to talk about AI as a potential source of bias. This case helps illustrate the flip side: a growing share of innovation in clinical AI is explicitly aimed at reducing bias, by standardizing detection and catching risk earlier across populations that have historically been missed. 

  
  

In this post, I’ll break down why the underdiagnosis problem has persisted for so long, and what new tools, workflows, and clinical resources are starting to show up to narrow the gap. 

Drivers of the underdiagnosis problem in women 

The underdiagnosis of cardiovascular disease in women isn’t the result of a single failure. It’s the cumulative effect of how cardiovascular care has been designed, studied, and operationalized over decades. Much of modern cardiology was built around male-pattern disease: classic chest pain, obvious ECG changes, and obstructive coronary disease. Women, by contrast, are more likely to present with subtler or “atypical” symptoms (e.g., fatigue, shortness of breath, nausea, or diffuse discomfort) that don’t always trigger the same diagnostic pathways. When time-pressured clinicians are working within rigid protocols, these differences can lead to delayed testing, delayed referral, or missed diagnoses altogether. 

The problem is reinforced by structural factors beyond the exam room. Risk calculators, diagnostic thresholds, and clinical trial data have historically underrepresented women, meaning the tools meant to guide care often perform less reliably for them. Add in fragmented care, shorter visits, and payer-driven utilization controls, and the system quietly selects against early detection in women, especially younger women and those without “textbook” risk profiles. The result is a persistent gap: women are more likely to be diagnosed later, when disease is more advanced and treatment options are narrower, more expensive, and higher risk. 

From a strategy perspective, this isn’t just a clinical equity issue, it’s also a downstream cost, quality, and reputation issue. Late diagnosis drives higher-acuity admissions, worse outcomes, and greater penalty exposure. Understanding why this gap exists is the first step toward redesigning detection, workflows, and technology in ways that actually close it. 

Union's Strategy Bootcamp for member, 'How to Speak CV Care', goes through major patterns to know in this area -- including but not limited to demographic gaps in care 

Solutions that have been tried—and failed to completely solve

Over the past two decades, there have been multiple high-profile efforts to raise awareness and change practice. Campaigns like the American Heart Association’s Go Red for Women succeeded in elevating public and professional awareness, reframing heart disease as a women’s health issue rather than a “male disease.” These efforts meaningfully increased knowledge and visibility, particularly around symptom recognition and prevention. 

At the clinical level, guidelines have evolved to better acknowledge sex-based differences in presentation, risk factors, and outcomes. Professional societies have emphasized earlier risk assessment, greater attention to non-obstructive coronary disease, and improved representation of women in cardiovascular trials. Health systems have also launched women’s heart clinics and specialty programs aimed at improving access and coordination.

And yet, the impact on everyday diagnosis has been uneven. Awareness campaigns change mindsets, but they don’t automatically change workflows; these efforts have run up against the problem drivers outlined in the section above and short of solving the problem completely. Guidelines help, but they still rely on clinician recognition and judgment in busy, resource-constrained settings. Specialty programs often reach a subset of patients, but not the millions moving through primary care, emergency departments, and outpatient cardiology pathways where early detection decisions are made.

From a systems perspective, these efforts often struggled because they were layered onto existing processes rather than embedded into them. They depended heavily on human vigilance, education, and advocacy, rather than redesigning the underlying diagnostic machinery.  

Enter the current push toward AI-enabled diagnostics: Its approach is is different in that it doesn't rely on awareness or vigilance, it aims to change the default. The idea is to hard-wire more equitable detection into the tools and pathways clinicians already use—at scale.  

What’s different about AI-driven approaches to closing cardiovascular care gaps 

Instead of depending on a clinician to recognize an atypical pattern in a high-pressure moment, innovators are proposing that AI tools continuously analyze ECGs, imaging, vitals, and longitudinal data to flag risk earlier and more consistently. When designed and validated thoughtfully, these systems can surface patterns that have historically been discounted or missed, helping standardize detection across sex, age, and presentation type. In that sense, AI’s real promise in cardiovascular care isn’t automation just for efficiency's sake (the use case industry sources most often start with). Instead it’s (also) about embedding equity into routine workflows, so earlier and more accurate diagnosis becomes the norm rather than the exception. 

Three keys to changing the default approach in cardiovascular care 

Changing default behavior and processes in CV diagnoses requires directly targeting the same mechanisms that have historically led to the underdiagnosis problem. These switches focus on three main areas. 

• Signal reinterpretation. AI-enabled ECG and echocardiography tools are trained on large, diverse datasets to detect subtle patterns (e.g., changes in rhythm, ventricular function, or valve dynamics) that may not meet traditional thresholds but still signal elevated risk. These tools can flag abnormalities earlier and more consistently, even when symptoms don’t look “classic,” helping counteract reliance on male-pattern presentations. 

• Risk stratification beyond obstructive disease. Women are more likely to have non-obstructive coronary disease, microvascular dysfunction, or heart failure with preserved ejection fraction, conditions that are harder to diagnose with standard pathways. AI-driven imaging analysis and predictive models are increasingly being used to identify plaque characteristics, myocardial strain, or early functional changes that would otherwise be dismissed as low-risk. Instead of asking, “Is there a blockage?” these tools help reframe the question to, “Is this patient on a high-risk trajectory?” 

• Standardizing decision-making across settings, not just specialty centers. Embedding AI into routine workflows (i.e., ECGs in primary care, echoes in community hospitals, remote monitoring in outpatient settings) reduces variability driven by time pressure, experience level, or unconscious bias. When risk flags and diagnostic prompts are triggered consistently, fewer patients depend on being the “right” kind of patient to get the “right” workup. 

  
  

From a strategy standpoint, this is what makes AI different from prior efforts: it’s not an awareness campaign or a specialty program. It’s an attempt to rewire the front end of cardiovascular detection so that earlier, more equitable diagnosis happens by default, at scale, and where most women actually receive care. 

Spotlight on the AHA’s investment in Ultromics  

Recently, the American Heart Association (AHA) invested in Ultromics, an AI company focused on improving cardiovascular diagnosis through advanced echocardiography analysis. That investment matters because it reflects a strategic bet by one of the most influential cardiovascular institutions on AI as a tool to improve detection.  

Echocardiograms are performed millions of times each year in the U.S. and are often the first-line cardiac test, particularly for women, who are more likely to present with nonspecific symptoms such as fatigue, dyspnea, or exercise intolerance rather than classic chest pain. Despite this, women are significantly more likely than men to have their cardiac symptoms initially labeled as non-cardiac, and studies show women experience diagnostic delays of years for conditions like coronary artery disease and heart failure. 

Ultromics applies AI to echocardiography to extract signals beyond traditional measurements like ejection fraction. This matters because women are disproportionately affected by conditions that are harder to detect with standard echo reads, including non-obstructive coronary artery disease and heart failure with preserved ejection fraction (HFpEF). HFpEF now accounts for over 50% of all heart failure cases, and women represent nearly 60% of that population. Yet HFpEF remains underdiagnosed and undertreated, in part because routine imaging often appears “normal” using conventional thresholds. 

Ultromics’ AI models analyze thousands of data points within standard echo studies to identify subtle functional and structural patterns associated with coronary disease and heart failure risk. In validation studies, AI-enhanced echo analysis has demonstrated the ability to improve detection of coronary artery disease even in patients without significant visible blockages, an especially important advance given that women are more likely to have ischemia without obstructive coronary disease. By enhancing diagnostic sensitivity without requiring new tests, radiation, or invasive procedures, this approach directly addresses one of the root causes of underdiagnosis. 

From a systems and equity perspective, the AHA’s involvement underscores a strategic shift. Again, full credit to previous efforts, such as the awareness campaign Go Red for Women, which successfully increased recognition that heart disease is the leading cause of death in women. This type of awareness was valuable but, in itself not enough to close outcome gaps. The bet here is that AI-enabled echocardiography represents a different kind of intervention: one that embeds more equitable detection into everyday workflows at scale. If broadly adopted, it’s possible that tools like Ultromics could have meaningful impact on reducing diagnostic variability across sex, geography, and care settings, turning a routine test into a more powerful early-warning system for women’s cardiovascular disease. 

Summary 

This problem/solution set is only example of the many interesting and important dynamics at work in today’s cardiovascular care space. Cardiovascular strategy is full of factors that we should all be tracking because they materially affect cost, access, outcomes, and risk across the healthcare system. In the strategy bootcamp, we’ll go deeper into additional topics like these, breaking down how cardiovascular care actually operates, where the pressure points are, and what leaders across healthcare, industry, and investment need to understand to make better-informed decisions. 

Union members: Join us live tomorrow 

How to Speak Cardiovascular Care 

Thursday, January 22, 2026 1-2 PM ET 

Registration details available here