For HTM, the potential of AI is still ahead

February 19, 2020
by Sean Ruck, Contributing Editor
The downtown campus of the University of Maryland Medical System alone has close to 30,000 pieces of medical equipment. With those types of numbers, ill-advised purchasing and servicing decisions can quickly become major financial missteps. That’s why Sean Connolly, imaging services manager, is committed to finding ways to streamline preventative maintenance to make it more predictable and efficient.

Alternative equipment maintenance (AEM) programs have emerged as a promising method for reducing equipment service expenses, maximizing equipment uptime and creating a more efficient HTM program in general. Once validated, an AEM program can work outside of suggested manufacturer maintenance recommendations in a safe, effective way. It doesn't take a lot of imagination to see how AI could bring those programs to the next level.

By allowing some smart automation, the efficacy of the AEM programs can ramp up, further reducing downtime and costs. But there’s some translating that needs to be done in order for HTM to hit that point. A Columbus, Ohio-based company called Mass Technologies is busy zeroing in on equipment that can safely be purchased in higher quantities without having to worry about all the eggs going into one basket, which is a benefit. “We’re looking at standardizing technologies,” says Elizer Kotapuri, the company's chief clinical technology officer. “If I have one car and the service training costs so much, the question becomes ‘is there a cost benefit in investment to support one piece of equipment?’ However, if we have 20 cars, same make and model, it’s an economy of scale. You standardize on technology because there is cost efficiency.”

Eliezer Kotapuri
That economy of scale also means AI will be working with a more robust data set. Pulling information from 10,000 items and plugging in their maintenance details will weed out anomalies where devices far outlast their suggested maintenance, or break down earlier than they’re supposed to. But getting back to Kotapuri’s point, the economy of scale is only effective if the data is standardized. Taking three devices of the same make and model and entering the first as a “sports car” the second as a “red car” and the third as “standard transmission” muddies the data until either someone manually corrects or the AI evolves enough to do it on its own.

For HTM, AI is still a new frontier. Connolly says that, to his knowledge, there are not currently any specific tools available to help facilities determine if AI use is right for their organization. “That question came up during AAMI when we gave our talk, that it’s the perfect opportunity for the HTM community to come together and establish standard code to use throughout the industry so our databases can learn.”

Sean Connolly
That standardization, according to Connolly, would address things like preventative maintenance codes, corrective maintenance, operator error and more. “The machine would be learning from the same database,” he says. “That’s something we haven’t done well as an HTM community. Some people have said that maybe we need to look at an AAMI or the Joint Commission to tell us what these standards are. I would have to disagree and say that we should be telling the Joint Commission and AAMI what these standards should be based on our experience of working every day with this equipment.”

Ultimately, Connolly believes a well-implemented AI program catering to the HTM department could provide a return on investment in just a year. After that year, it would be almost all savings with very little, if any, investment needed.

The two believe that AI could be integrated into existing systems so use would be seamless, but training would still be needed to keep everyone entering data on the same page. Making sure everyone’s using the same codes, speaking the same language to keep the AI healthy and effective and even creating effective search queries.

The caveat to this conversation is that there’s not currently a lot out there for AI in HTM right now, so organizations probably shouldn’t start writing checks just yet. “I’d hold off until something more standard is out there, but it really has the potential to explode in the HTM field and bring a lot of value,” says Connolly.