Seeking new ways to improve more lives in more ways

As the world has become more digital, the amount of data created on a daily basis is astronomical. Access to this data can facilitate better decision-making across a variety of industries, including healthcare.

In 2021, HCA Healthcare collected data from more than 35 million patient encounters – from vital signs and device readouts to lab results and health history. We use this data to help our more than 93,000 nurses and 45,000 active and affiliated physicians quickly spot patterns that point to certain health conditions, and make clinical decisions based on that knowledge. Because time is often of the essence, being able to speed up any part of that decision-making process can be the difference between life and death.

That’s where technology comes in. Advanced technology like machine learning and artificial intelligence has the power to analyze vast amounts of data quickly, allowing caregivers to recognize patterns, identify problems – or potential future problems – and make decisions faster.

We know this technology saves lives because we’ve seen it firsthand, and we’re excited about the implications that AI has to transform even more aspects of patient care so that we can continue to provide the best possible outcomes for everyone who comes to an HCA Healthcare facility.

In the Wall Street Journal, read more about how HCA Healthcare prioritizes innovation to address the needs of those we serve: How Hospitals Are Using AI to Save Lives.

Where we started: SPOT

AI has dramatically transformed how we identify potential sepsis cases. Using AI, we are able to identify sepsis six hours earlier than clinicians. Early recognition and treatment have reduced sepsis mortality across 160 hospitals by almost 30%.  

Sepsis is a life-threatening emergency. It is the body’s extreme response to an infection, occurring when an infection you already have triggers a chain reaction throughout your body. Infections that lead to sepsis most often start in the lung, urinary tract, skin or gastrointestinal tract. Without timely treatment, sepsis can rapidly lead to tissue damage, organ failure, and death. Anyone can get an infection, and almost any infection, including COVID-19, can lead to sepsis. In a typical year, the Centers for Disease Control and Prevention states that at least 1.7 million adults in America develop sepsis and nearly 270,000 die as a result of sepsis. Further – one in three patients who die in a hospital has sepsis.

Like most hospitals, previously, our nurses manually monitored our patients for sepsis. Patients can become septic at any time, but since it’s not realistic to monitor every patient 24/7, nurses were checking for sepsis during shift change. They used a checklist to look for certain criteria, such as lactate value and blood pressure. But often, it was different clinicians doing the checks and noticing different issues at different times, so it was difficult to see the full picture of symptom changes indicating the patient was becoming septic. This was costing precious time.

Since AI is effective at spotting patterns in data, we saw an opportunity to use it here. HCA Healthcare created an algorithm driven, real time system called SPOT (Sepsis Prediction and Optimization of Therapy) to better identify the signs of sepsis.

Our clinical and data science experts developed SPOT using data from millions of hospitalizations. SPOT continuously monitors vital signs, lab results, nursing reports and other data that can inform treatment, and recognizes critical data points in patients’ electronic health records to quickly alert care teams to important, often subtle, changes in a patient’s condition so they can take appropriate action.

When it sees data consistent with sepsis risk, SPOT will signal with an alert, which clinicians can then respond to by examining the patient in person and making a clinical judgment. This is key because clinical judgment is still necessary. While not every SPOT alert is an actual sepsis patient, the system helps us see which patients are most likely to have contracted sepsis.

The beauty of SPOT is two-fold. First, continuous monitoring saves time and lives. Second, it integrates intelligence into workflows. SPOT doesn’t create new workflows or disrupt existing processes – it merely saves time in the current workflow. When the algorithm identifies sepsis risk for a patient, the same next steps occur. It just happens that the algorithm is better and faster at spotting those risks than a human is – which leads to better patient outcomes.

Expanding AI’s use

Once we got SPOT up and running in our hospitals, we were able to see how well it worked – and started thinking about how we could leverage the same type of tool for other patient care challenges. That led to the development of NATE.

NATE, which stands for Next-gen Analytics for Treatment and Efficiency, delivers data-rich views of a facility using real-time analytics and machine learning solutions. It uses the same foundation of real-time data intake as SPOT, but its algorithms help us do things like deciding when and who to evacuate from hospitals in a natural disaster to who is at higher risk for readmission.

For example, we have a huge number of hospitals in hurricane-prone Florida. When there is a risk of evacuation, you need to know exactly who is in the hospital and what their condition is to make contingency plans. Before NATE, we were manually making evacuation plans in command centers. With NATE, we are able to plug in all the relevant data elements about patients and use a dashboard to pull the necessary information about everyone in the hospital at any given moment.

So, when our teams needed new tools to respond to COVID-19, NATE was able to deliver. With an urgent need to treat acute respiratory distress syndrome (ARDS), we developed an overlay for NATE named C-ARDS (COVID ARDS). C-ARDS brought real-time analytic insights to assist clinicians in applying the appropriate clinical protocol for mechanically ventilated patients, specifically COVID-19 patients.

The impact of C-ARDS was tremendous on several fronts. For physicians and care teams, C-ARDS helped illustrate the effectiveness of the treatment for ARDS, which had been known for years, and these insights helped decrease anxiety within the physician and nursing communities. Additionally, C-ARDS enabled communication and collaboration between respiratory therapists, intensivists, and nurses to support high quality patient care. The ability to monitor closely key data points for COVID-19 patients resulted in decreasing their lengths of stay and helping to increase survival for COVID-19 in our facilities by 28%.

The secret: make AI work for caregivers – not the other way around

The key to success for SPOT, NATE and the advanced technology and innovation work we’re currently doing is workflow integration. Whenever we think about a new application of AI, our data scientists are collaborating directly with clinical staff to determine where using algorithms and predictive models would be most useful and how they can most easily fit into the current patient care workflow.

A common challenge we have seen in an industry eager to use data to use machine learning to advance care, generating adoption among caregivers. We believe our success with the adoption of SPOT was due to the way the technology fit naturally within the existing workflow.

This key learning is shaping new projects, and we’re excited about the many promising possibilities to use AI and machine learning to improve hospitals and patient care.

Building trust with technology

To effectively use innovation to transform patient care, it is essential to build caregivers’ trust in the technology. As much positive impact as AI has the potential to make, it can only go as far as the caregivers are willing to trust it and the signals it sends. This is why we engage directly with care teams as we design and deploy algorithms so they are deeply involved and understand and trust the signals. We ask them if they were to design a solution, what would it look like? What would you want it to tell you? And then, we build the solutions collaboratively so that the final product gives them the information they’re looking for in the way that they’re looking for it in their workflow.

We’re also very clear about what the algorithms are doing. For example, with SPOT, we were very transparent about how it worked, explaining what data we were looking at and why it alerted when it did. That transparency helped us build trust with our caregivers.

What’s next?

Using AI to improve patient care is as much about the integration into workflows and trust-building with the care team as it is the technology. We started small, with one use case – SPOT. SPOT met our nurses where they were, integrated into their current workflow, and helped reduce the time it took to monitor for sepsis, while still leaving the ultimate decision-making in the hands of our clinicians. We will continue to build, adjust and apply lessons learned from using AI in SPOT and other applications to deliver healthier tomorrows for our patients.

Through the clinically led integration of technology into care delivery, we’re excited to leverage bold innovations to address current and anticipated challenges facing caregivers. The expertise of our clinical care, engineering and technology groups – and the voices of our frontline clinicians – will be key in future endeavors as we work together to test, measure and scale solutions.