Healthleap Raises $38 Million to Push AI Screening Deeper Into Inpatient Care
Healthleap has raised $38 million in combined seed and Series A financing to expand an AI screening platform that sits inside the electronic health record and looks for hospitalized patients whose conditions may be missed during routine care. The company’s first clinical target is malnutrition, a complication that often hides in plain sight until recovery stalls, length of stay grows, or discharge plans unravel.
The funding round included Sequoia Capital, First Round Capital, and Hummingbird Ventures. Healthleap now wants to extend the same workflow to congestive heart failure, aspiration pneumonia, delirium, pressure ulcers, and certain hospital-acquired infections. In practical terms, the company is trying to move risk detection earlier in the hospital course, before the damage from an overlooked diagnosis becomes harder and more expensive to reverse.
Malnutrition Sits at the Center of the First Use Case
Malnutrition is one of the clearest examples of how a condition can be clinically consequential without drawing enough attention at the bedside. Hospitalized patients who are confused, intubated, or critically ill may be unable to answer screening questions, and those patients are often the very ones who need the most careful assessment. Traditional screening depends heavily on manual workflows performed by an admitting nurse or an in-hospital dietitian, which creates an obvious vulnerability when the patient cannot communicate or when staffing is stretched thin.
Healthleap’s current model uses weight, body mass index, laboratory results, diagnoses, medications, intake and output, nutrition orders, and vital signs. Those inputs are combined into a priority score from 0 to 10 for each patient, with a second column indicating whether risk has changed since the last assessment. Dietitians see the result in an existing EHR worklist each morning and sort the list to begin with the highest-risk patients.
The appeal is operational as much as clinical. Instead of interrupting care teams with a flood of alerts, the platform places the score where clinicians already work. In a hospital, friction often determines whether a signal gets acted on at all.
What the Screening Model Is Trying to Fix
The problem Healthleap is addressing is familiar in acute care: diagnoses that are common, consequential, and still easy to miss when attention is focused on the most urgent problem. Malnutrition can worsen wounds, weaken recovery, complicate discharge, and amplify vulnerability to downstream complications. Earlier identification allows nutrition support to begin when it may still change the trajectory of the admission rather than merely documenting decline after it has already taken hold.
The company screens every adult inpatient daily. That cadence matters because nutritional and functional status can deteriorate quickly in the hospital, especially when oral intake falls, inflammation rises, or the admission itself triggers a cascade of immobility and deconditioning. A daily review also gives the care team repeated opportunities to catch patients whose risk was not obvious on admission but becomes clearer over time.
One sentence captures the clinical logic: missed inpatient risk becomes expensive later.
Evidence, Workflow, and the Push for Validation
Healthleap says its model demonstrated 88% higher sensitivity than nursing screening, citing a peer-reviewed study linked from its website. The article did not provide the study’s sample size, full methodology, or comparative specificity, so the available evidence should be read as an early signal rather than a complete validation package. Even so, the performance claim reflects a central tension in hospital screening: a manual process may be simpler to explain, but an AI layer can surface patients whose risk is buried in chart data that no individual clinician has time to synthesize repeatedly.
The platform is already in use at more than 50 hospitals, including Penn Medicine, Houston Methodist, Cedars-Sinai, Emory, and UMass Memorial. Broad adoption across major health systems suggests the workflow fits enough real-world environments to survive more than one pilot phase, though broader deployment does not replace the need for condition-specific validation each time the model expands.
Healthleap’s co-founder and clinician Jemima Meyer said the technology does not send interruptive alerts. Physicians and nurses do not see pop-ups. The same nightly chart review and EHR connection drive each condition, which reduces workflow fragmentation and keeps the burden on existing worklists rather than on bedside interruptions.
The Business Case Follows the Clinical One
Healthleap framed the stakes in economic terms as well as clinical terms. The company says each severe malnutrition diagnosis and treatment can support improved patient outcomes and generate about $11,000 in reimbursement for hospitals. Meyer also said each patient flagged had a 1.6-day shorter average hospital stay. Those figures are important because hospitals operate in a financial environment where avoiding a complication and securing the right diagnosis can affect both care quality and revenue.
The reimbursement argument is not a replacement for patient benefit, but it explains why hospital leaders pay attention. If an AI screening layer identifies patients who would otherwise leave with unresolved disease, the downstream benefit reaches well beyond the nutrition consult. It can alter discharge timing, reduce preventable deterioration, and redirect scarce clinician time toward patients most likely to benefit from intervention.
Expansion Will Depend on Condition-by-Condition Proof
Healthleap’s next stage is more difficult than scaling a software contract. Each new screening target will need further clinical validation before it can be trusted in the same way as the malnutrition workflow. Congestive heart failure, aspiration pneumonia, delirium, pressure ulcers, and hospital-acquired infections are not interchangeable use cases; each carries different signals, different pathways to injury, and different standards for what counts as meaningful early detection.
A model that works in one domain may fail in another if the underlying data patterns are less stable or the interventions are less tightly linked to a change in outcome. Delirium, for example, may surface through subtle chart signals that are easy to miss but harder to convert into a single actionable score. Pressure ulcers and hospital-acquired infections bring yet another challenge, because prevention often depends on rapid multidisciplinary response, not just detection.
The company’s strategy depends on the idea that chart-derived risk can be organized into a bedside workflow fast enough to change outcomes. Health systems are not short on data; they are short on reliable methods for separating urgent signal from background noise. Healthleap is betting that an EHR-native screening layer can do that work better than manual review alone, and that the same framework can be extended without breaking clinical trust.
For now, the most concrete measure of progress is not the size of the financing round. It is whether a quiet score embedded in the worklist can keep malnutrition, and eventually other overlooked inpatient conditions, from escaping detection until the patient is already headed toward worse outcomes.