TRIMEDX Chief Information Officer Brad Jobe was recently interviewed for a 24x7 Magazine article about how AI can be used in the HTM field. Jobe shared how AI has the potential to empower BMETs and health system staff. The full article, as it appeared July 23, 2024, is below.
It seems like artificial intelligence (AI) is everywhere, creating surreal videos, helping college students cheat on essays, and improving efficiency within certain industries. HTM is no different, and software vendors are already exploring AI to help healthcare facilities maintain their equipment and reduce the workload on biomeds.
“On the show floors of [the Healthcare Information and Management Systems Society] and AAMI this year, every vendor seems to incorporate AI into their products in one way or another, from improving images to optimizing workloads to sniffing network chatter for device analytics,” says Erin Sparnon, MEng, CSSBB, AAMIF, FACCE, artificial intelligence business integration strategist for ICA, Inc.
AI is still a new technology, and it may be a while before the dust settles on this latest tech boom. But healthcare technology experts already see exciting potential applications.
“One of the most exciting potential applications of AI is fully capitalizing on the wealth of data collected in the healthcare industry,” says Brad Jobe, TRIMEDX’s chief information officer. “Empowering critical health system staff to perform their best work with the support of AI can reframe this complex technology as an enhancement for clinical operations instead of a disruptive force.”
What is AI?
One reason AI is discussed so often is that it is a somewhat broad term. Terms like large language models (LLMs), machine learning (ML), and deep learning (DL) are often used to describe how an AI works. But in a fundamental sense, AI is complex computer software and hardware designed to simulate human intelligence. An AI is presented with a data set to analyze, which it interprets to offer new insights in return.
Depending on the veracity of the data and complexity of the AI being used, it can result in something as important as learning to spot cancer in X-ray images or something as novel as generating a five-paragraph book report.
“Most people, when talking about AI, are referring to generative AI, which has seen immense growth and brought AI to the public eye in a significant way due to the release of ChatGPT 3 and several other similar implementations,” says Aaron Hanna, chief technology officer at NVRT Labs in San Antonio. “Generative AI refers to algorithms and models designed to generate new data samples that resemble a given dataset. This is giving AI the ability to parse data and reason over it and create new data, unlocking the incredible potential for all businesses, including HTM.”
According to ICA’s Sparnon, AI is most used in HTM as a ML algorithm to extract insight from large datasets. For example, she says, AI can be used to crunch CMMS data to predict failure modes and support alternative equipment maintenance (AEM) decisions or to track uptime and utilization analytics.
“Machine learning could provide transformational benchmarking insights from maintenance and repair data, in terms of predicting needed maintenance, reducing downtime, and giving clarity to the real service burden and useful life of any particular model,” she adds.
AI in Predictive Maintenance
One area in which AI could make an impact in HTM is predictive maintenance. An AI trained on repair and service data could provide insights into what equipment could be headed for failure, guiding biomeds to where their efforts are most needed.
If an AI detects an imminent failure or malfunctioning equipment, it could go a step further by proactively sending an automated work order to clinical engineering. According to Jobe, the AI could help biomeds optimize preventative maintenance and maximize uptime while minimizing interference with patient care. By streamlining PMs and reducing unpredictability in maintenance schedules, AI could help HTM staff be more efficient with their resources.
“For example, if a certain part needs to be replaced on a machine, AI-powered systems could also warn the BMET that there is a high probability that another part will need to be replaced within three weeks,” says Jobe. “This allows the BMET to order and replace both parts at the same time, instead of working on the machine twice in a matter of weeks.”
AI could also help HTM shops validate supplier recommendations with actual data from the field or track lifecycle management by comparing repair history against industry benchmarks to identify devices that are being used more intensively.
Sparnon cautions that to reach these kinds of capabilities, AI needs to be properly trained on a big enough data pool, and users will need a solid data science skill set to complement their HTM experience for relevant and practical insights. But if the data already exists and is being tracked, the potential of AI to optimize operations is compelling.
“Think of all the analyses you’d love to do if you had access to really good cross-industry, vendor-neutral benchmarked data and a team of data scientists to run analytics for you,” says Sparnon. “When properly deployed on a big enough data pool, AI and ML can help get you most of the way there.”
How AI Can Affect the HTM Shortage
Another compelling use of AI is training new HTM professionals. NVRT Labs, a software development and extended reality training company, offers HTM training, and Aaron Hanna says that AI’s ability to personalize training could impact the workforce.
“Personalized training that adjusts to the user is already a reality in other verticals, and I believe it will present more effective and efficient ways to train students and fill the knowledge gaps of seasoned workers,” says Hanna. “AI could also provide a way to generate and run real-world scenarios during the education of new techs and provide practical training even before getting into the field.”
Hanna suggests personalized training could be combined with immersive experiences through virtual or augmented reality, providing biomeds with more realistic training. This enhanced training approach can also help address another challenge for HTM: the shortage of new workers.
TRIMEDX’s Brad Jobe has observed firsthand that the aging workforce and declining number of new technicians threaten the health system’s ability to maintain equipment. Additionally, as experienced BMETs retire, their valuable knowledge and expertise risk being lost.
AI can address this problem in two ways: by expediting and enhancing quality training and supplementing the shrinking labor pool. According to Jobe, AI has the potential to increase a BMET’s efficiency in daily tasks and allow them to focus on their skill set and responsibilities.
“The majority of all data is private, and in our industry, an immense amount of vital information is stored in the brains of experienced BMETs,” Hanna adds. “When they retire, this information is lost. At NVRT, we are developing AI with the intent to retain as much of this information as possible and make it accessible to technicians everywhere in a way that is simple to interact with and provides significant value to their day-to-day.”
Challenges for AI
For all its promise as an efficiency booster, implementing AI into HTM is likely to present new challenges. That’s why Jobe says it’s crucial to recognize and address the risks associated with new digital technologies.
Sparnon concurs, saying AI is only as good as the data you give and the questions you ask.
One reason for this is that raw data often contains statistical noise that AI might not recognize as easily as a human with real-world experience. As miraculous as AI seems on the surface, it’s still important to remember that AI is only simulating intelligence at its current level, Sparnon says. Ask ChatGPT a question, and it will quickly give you a thorough answer, but if you push it further, you may find the facade of intelligence slip.
Sparnon says that creating useful and relevant insights from medical device data will require both data science and HTM skillsets working together, as not all findings are insights. Take infusion pump repairs, for example.
“If a data scientist finds that there’s a peak of infusion pump repairs every October, we need the HTM to remember that [October] is pump PM month, so of course, we’re going to find things when we touch 700 devices at once,” says Sparnon.
That leads to the last challenge, which is training people to use AI effectively. Aside from the costs of implementing new technology, the job of an HTM professional has only become more complex, requiring new skill sets and knowledge to adapt, and AI will be no different.
For now, Sparnon thinks that the most successful AI applications in HTM will be site-specific, narrow applications that provide insights into current work by comparing it against prior work within a single facility. By keeping the focus narrow, the likelihood of errors and misunderstandings is reduced. However, to unlock AI’s full potential, she notes that investment in data science may be necessary.
“Data scientists have common sense and can explain how they got to their conclusions, which is why we need them more than ever to evaluate HTM-focused AI applications for safety, effectiveness in local conditions, and efficiency,” says Sparnon. “I’d love to see data science and analytics grow into just another tool in the HTM box over time, as did network management and cybersecurity.”
After all, she says, growing and understanding the system ensures the safety of information, devices, and, most importantly, patients.