Artificial intelligence (AI) is changing the way that hospitals are operating. Recently, AI in hospital management has attracted the attention of many healthcare leaders. With the research currently underway, we know that AI can prevent maternal mortality, death from sepsis, hospital falls, open heart surgery complications, and much more.
As the healthcare system moves towards value-based care, AI can play a key role in reducing utilization and promoting positive outcomes.
In this blog, we’ll take a look at a few AI use cases that can help prevent complications in hospitals.
Sepsis Watch
If a patient develops sepsis in a hospital, they have a 28 percent likelihood of death due to organ failure from widespread inflammation. The Duke Institute for Health Innovation developed a deep learning model that acts as an earlier warning detection within the Duke Health system.
The Duke Approach goes beyond simple detection. Their system has a three-stage process for handling patients and risks. Based on healthcare records and intake information, the AI model assesses a risk (low, medium, or high) for each patient and triages them for the next step. If they are evaluated as a low to medium risk, they are further screened for additional factors.
All high-risk patients are put under monitoring by nurses, which ensures that they are constantly observed during their hospital stay. These nurses are not the standard nurses, but well-trained specialists in a centralized team called Rapid Response Team (RRT). Lastly, these nurses have a pre-created bundle of treatment which ensures that rapid response is rapid.
It’s one thing to have model predictions, it’s another to use those predictions to make systems that prevent hospital deaths. Leveraging AI systems has required that the staff be open to a new way to administer care, and everyone has benefited from it.
Maternal Mortality
The United States is a world leader in so many fields. With more access to healthcare than most of the world, we should not be leading the world in maternal mortality. The US Centers for Disease Control and Prevention (CDC) suggests that about 60 percent of these deaths are preventable, and that figure is exacerbated among underrpresented groups
AI researchers have taken this cause on to help identify high-risk pregnancies and monitoring capabilities to improve outcomes. These models largely depend on available historical data for the patient and patients with similar profiles.
There are two methods that we could find being used to build these models. The first is just classifying for risk similar to the septic risk model. The second is trying to forecast when an adverse event is likely to occur which can allow for intervention before it happens. Both look at current complications, complications found in exams, and socioeconomic factors that can contribute to the issue. Researchers are in the early stages of testing to determine if AI can be used to do general triage and monitoring.
Hospital Falls
A fall may not seem life-threatening, but for some patients, it could lead to a long stay in the hospital and additional complications. Increased extended hospital stays can impact all staffing, especially night shift staffing. El Camino Hospital, a non-profit health system in Silicon Valley, partnered with a healthcare technology startup called Qventus to effectively reduce fall rates by 29 percent.
Traditional prevention plans involved sending volunteers to any patient room. Qventus took a step back and looked at data from electronic health records, nurse call-light, bed-alarm data, medications, and vitals to have an AI system predict the patients with the highest likelihood of falling.
This data was then used to increase patient monitoring and assign direct volunteers to the patients who need more frequent checks. This model was not a generalized ‘if the button is pressed too often then add staff’ model. Factors such as if the button was pressed more than the average of people with a similar condition on that particular floor.
Artificial Intelligence was used to inform decisions around operation practices. It was not left to simple dashboard outputs but implemented into the operational structure of the hospital. Real-time alerts were sent out to the nurse’s stations if a risk factor changed. This allowed teams to prioritize who was seen by the available staff.
Open Heart Surgery Complications
Going into surgery can be a frightening idea. Unfortunately, sometimes people die after surgery from the simplest issues. AI researchers are trying to reduce those risks now.
Houston Methodist is using an AI brain ultrasound model to prevent perioperative embolism and lower the risk of stroke from open-heart surgery. There was research conducted to detect emboli in 1982, but it didn’t get very far as it was extremely expensive to implement.
Houston Methodist was able to modernize the previously conducted research with AI. This research looked at Transcranial Doppler (TCD). According to Houston Methodist, “TCD can detect not only the presence of blood flow but also its depth, direction, and resistance (DDR). It provides great sensitivity in detecting foreign particles”. Essentially, it produces a signal output that is read by sonographers.
Now, an additional algorithm is built to assist in the reading of the results. Nurses can be trained to use the system in a couple of hours.
Conclusion
In each of these cases, AI solutions were developed to assist in the treatment plan implementation or level of risk of the patient outside of human biases. Each experiment tried to train professionals in working with AI to do their jobs even better. No one wants to be treated by a robot but with the help of computers, we may be able to help improve their quality of life.
Looking to harness the power of AI at your organization? Be sure to check out our free guide on How to Implement a Successful AI Strategy for Your Company.
