Although health systems are struggling to meet increasing demands for healthcare with their fixed amounts of resources, a closer look reveals significant unused capacity throughout those systems. Sophisticated artificial intelligence (AI) solutions already being used in asset-intensive industries can be applied to healthcare to improve asset utilization.
In an August Becker's Hospital Review webinar sponsored by LeanTaaS, Sanjeev Agrawal, president and COO of LeanTaaS, explores the status quo as well as how AI and machine learning can make a significant impact in three key asset-heavy healthcare arenas: operating room scheduling, inpatient bed management and infusion centers.
Four key takeaways were:
1. Changing and improving healthcare requires understanding four realities. "The first reality is that the need for access to healthcare services is rising very fast," Mr. Agrawal said. "Not only is our population increasing, but the proportion of people over age 65 is growing even faster. The need for access to medical care over the next 20 to 30 years is a tsunami."
Other realities, Mr. Agrawal explained, are that the patient experience is inadequate, a lot of existing capacity is not well used and the current cost model is unsustainable. "The only way out of this situation is to do more with less. We have to find a way to make our assets sweat," Mr. Agrawal said.
2. By using AI like other asset-intensive industries do, hospitals can better predict operating room block utilization. Other industries facing cost pressure and pressure to use their assets more efficiently — such as the airline industry — have reinvented themselves using AI and machine learning to unlock unused capacity. "Multiple other industries have successfully used such techniques," Mr. Agrawal said.
Within hospitals, which have important and expensive assets like operating rooms — and where demand is hard to predict — the scarcity mindset discourages surgeons from releasing unused OR time. "Imagine a world where you could predict the likelihood that a specific provider will use a particular block," Mr. Agrawal said. "When the likelihood gets beyond a certain point, imagine being able to send their clinic scheduler a prompt saying, 'I know Dr. Jones has blocked time three weeks from now, and the system is more than 50 percent convinced that he or she is not going to use their time. Please consider releasing it.'" A Colorado hospital showed that with 99 percent certainty that block minutes would not be used, the hospital could release seven OR blocks per month.
3. The same mathematical models that Google Maps employs can be used to manage the ingress and egress of patients by specific unit. By using historic patient admission and discharge data, health systems can accurately predict, for example, activity for a Wednesday in August for a particular unit. "Now, during morning huddles, hospital leaders can focus on the five units likely to have the biggest demand for beds," Mr. Agrawal said. "Imagine being much more confident when you do rounds, prioritizing the patients most likely for discharge."
In the case of a 631-bed specialty hospital, the accuracy of human-predicted patient discharges reached only 64 percent accuracy. But using AI modeling, this hospital increased its predicted discharge accuracy to 95 percent.
4. Just like UPS must predict the volume and shapes of packages to fit into a truck, infusion centers must predict the type and length of treatments to fit into a day. "Each Monday, there's a unique volume and mix of patients that will come in. The nursing roster and shifts, hours of operation and the times when your labs and pharmacy are open are the constraints that result in hundreds of possibilities," Mr. Agrawal said.
"By using constraint-based optimization, magic happens," he continued. "Even if you're able to comply 85 percent of the time, the majority of days will feel much better, more level-loaded where patient wait times are low."
Optimization at the Penn Medicine Infusion Center in Philadelphia resulted in a 25 percent increase in patient volumes, 20 percent increase in patient hours and a 22 percent decrease in wait times.
By using AI, health systems can successfully unlock capacity in areas such as OR scheduling, inpatient bed management and infusion center optimization, resulting in delivering more healthcare services without higher costs or additional staff.
To register for upcoming webinars, click here.