How healthcare informatics are aiding personalized and proactive monitoring of heart failure

• Healthcare professionals are looking for new ways to manage heart failure as they seek to reduce hospitalizations
• Frequent NT-proBNP testing has the potential to give doctors a better understanding of a patient’s condition and could support earlier intervention (McDonagh et al, 2023; Mebazaa et al, 2022)
• Passively collecting data through wearables can create large data patterns that can be analyzed by algorithms, providing crucial insights for physicians

Advances in technology are reshaping patient care by offering new possibilities for proactive disease management. Cutting-edge innovations like wearable devices and data analytics promise to revolutionize how chronic conditions are monitored and treated. By leveraging these tools, healthcare providers may be able to shift from reactive interventions to proactive strategies, providing early warnings and tailored care.

Integrating artificial intelligence (AI) and machine learning (ML) enhances accuracy, enabling predictive analytics for identifying potential health risks.¹ As technology evolves, its role in patient care extends beyond diagnostics, fostering holistic disease management that emphasizes prevention, early intervention, and improved quality of life.

Healthcare Transformers had the pleasure of sitting down with Javier Echenique, CEO and co-founder of GPx, to learn more about transformative patient care and proactive, remote disease management.

HT: Can you describe to us the unmet need that you’re trying to address when it comes to remote disease management that you’re trying to solve?

Javier Echenique: Chronic diseases, which generally require regular monitoring for effective management,² are the number one cause of death and disability globally^3,4 Our goal is to help people live their best lives despite their conditions. Diabetes is a great example of a disease that’s been revolutionized by advanced diagnostics and therapies. We want to build on this model of continuous glucose monitoring in diabetes and apply it to other diseases.

We’ve decided to focus initially on heart failure because these patients often have limited access to remote monitoring options. Although patients have regular follow-up appointments with their doctors, it is not always possible to avoid hospitalization.

To give patients a much better chance, we want to provide doctors and patients with early warning signs that the patient’s condition is changing for the worse. This gives people more time to adjust their medication and stabilize their condition so that they can stay out of hospital. There’s a huge economic aspect to this. Globally, heart failure (HF) is the leading cause of hospitalization in patients over the age of 65, which drives up the economic burden on health systems.^5,6 For example, in the US, hospitalizations were shown to be the largest component of direct medical costs for HF, ranging from 49% to 73% of total costs.⁷ Therefore, if we can improve the management of the disease we can potentially reduce the economic burden of this condition. Revolutionizing disease management using technology.

HT: On your website you mention that doctors can monitor their patient’s blood biomarkers remotely without needles or implants. Could you please elaborate on this a little more and explain to our audience what is meant by that?

Javier Echenique: Just as measuring glucose is the gold standard for understanding the status of a patient’s diabetes, measuring a patient’s NT-proBNP has become the gold standard for understanding the status of a person’s heart failure as supported by its Class 1 recommendation in heart failure guidelines.^{8- 10}

As things stand at the moment, NT-proBNP home testing is not readily available and currently point of care tests must be conducted by a qualified healthcare professional. There’s no patient self-test version currently available.¹¹

Data from our survey have shown that doctors want to do this kind of testing weekly.¹² Yet right now, because of all these barriers, all data indicate that this test is done on average only twice a year per patient.¹² Clearly, this means there is a huge gap between what is needed and what is currently possible.

To overcome this, we have looked at how you can get the data doctors need without having to do regular blood tests. We believe the answer lies in algorithms that correlate information that can be picked up from a smartwatch to clinically show deviations in NT-proBNP. This approach would allow doctors to quickly spot if there were any clinically significant changes in a person’s status, and would allow them to intervene more quickly.

HT: When it comes to the post-acute phase in heart failure, how will the “bloodless” blood test help to facilitate early discharge and prevent patients from being readmitted to hospitals in the critical three week to six month window post an acute event?

Javier Echenique: Algorithms have the potential to help doctors identify changes in the patient and provide important indicators about their status following discharge. As a result of this information, physicians may be able to make more timely adjustments to patients’ treatment plans to reflect any changes in their status.

HT: Why is it so important to continually monitor blood biomarkers in that critical window, and also beyond it, and how will it help patients and doctors to more proactively manage the condition?

Javier Echenique: The system could alert the doctor to schedule an appointment with the patient before their health deteriorates further. At that point, they could go through the whole series of bringing the patient to the clinic, doing the physical examination, running a blood panel that includes NT-proBNP, adjusting their medication, and so on. With that much time in advance, the patient’s condition could be stabilized.¹³

HT: How do you think using informatics alongside blood-based diagnostics will make it easier to continuously monitor patients with heart failure?

Javier Echenique: I’ve spent more than half my career working in heart-failure devices and something we’re finding out very quickly is that the amount of information that a smartwatch can collect from you passively while you sleep, while you work, while you eat is enormous. And it’s not about each individual data point, but it’s these large patterns that are created minute to minute, hour to hour, day-to-day, week to week that algorithms can collect and look into so that we can really understand a person’s condition.

And let’s be clear, AI has made this a completely different world. You could try to be very artistic using our knowledge of medical device algorithms, but there’s really a limit to what you can do with all this information.

When you apply ML to it, it opens up a new world. We have collected volumes of data every minute, looking at heart rate, activity levels, and sleep patterns, and computers can draw so much knowledge from that. If you do a test every month, or every three months, or every six months, you’re missing tons of information about what happened in between those points. And so we end up with an incomplete picture of a patient’s health.

HT: What is the role of humans in algorithm development? And how important are data scientists in the development of your “bloodless” blood tests?

Javier Echenique: The data scientists at GPx combine deep physiological knowledge with the ability to develop algorithms. Rule-based algorithms in the medical world require that you have good knowledge of human physiology, how heart failure works, and the mechanisms of heart failure.

If you were just a pure ML data scientist, you might just throw all the data into the computer and let it come to a conclusion but then you have to ask, does this actually make physiological sense? With rule-based algorithms, you start by applying physiological knowledge before providing the data. This is the only way to ensure that the results that come out make physiological sense. We have found that machine learning can be used to enhance rule-based algorithms. If you have a well-performing rule-based algorithm and you plug that into an ML model along with other data and tell the computer we know this from a rule-based world, the results that are produced are extraordinary.

HT: Finally, you made the decision to take part in the Startup Creasphere program. What made you do that? What were the deciding factors for you in the decision-making process?

Javier Echenique: Roche has a proven track record of being extremely serious when they select you to be their partner. This is evident in the Startup Creasphere program, which supports the co-creation of solutions and services that have the potential to transform healthcare.¹⁴ It was therefore my obvious first choice for GPx when we started looking for an innovation program to support our development efforts.

We have a whole team that has spent many hours with us trying to figure out our technology, understanding it at a deep level, and asking really tough questions. And then once we get past that sort of trust, really trying to see, well, what would make the most business sense, it’s clear that they’re bought into this. They’re spending their time so every moment that we spend with them is highly valuable for our company.

GPx successfully participated in Startup Creasphere, a leading digital health accelerator that strives to transform healthcare together with startups.