rdoe_persona_userprofile
rdoe_persona_select_up
Article

Unlocking new possibilities for better patient care with artificial intelligence

Contributing lab leaderPeter McCaffrey

Artificial intelligence (AI) technology has progressed rapidly over the past few years. It has shown value in its ability to process complex algorithms and complete routine tasks quickly and accurately, with performance levels comparable to experts in the relevant fields.1

As test utilization grows increasingly important in providing answers to healthcare professionals, the application of AI to laboratory medicine is gaining attention as a way to manage the huge amount of complex test results produced in labs every day. 

At the recent Association for Diagnostics and Laboratory Medicine (ADLM) conference, Peter McCaffrey, MD, MS, FCAP, Chief AI Officer at the University of Texas Medical Branch, discussed how data-rich laboratory environments can utilize AI to enable efficient processing, and ultimately bring the lab to the front and center of the patient care workflow.

Article highlights:

  • Artificial intelligence technology has progressed rapidly in recent years and has many practical applications in the healthcare space.
  • Laboratories could benefit from artificial intelligence, which has the potential to change the way they operate.
  • Regulatory challenges to the adoption of artificial intelligence must be overcome to enable more widespread use and benefits.
animated speaking bubble

Join our community and stay up to date with the latest laboratory innovations and insights.

Subscribe now

Reducing inefficiency and increasing patient interaction

Many day-to-day, routinely structured tasks or functions can be made more efficient and less labor intensive through the introduction of AI, and this is where Dr. McCaffrey sees an initial use in the healthcare space.

He points out that a lot of doctors’ time is taken up with tasks that are not focused on human interaction with a patient. He believes AI can help with unburdening people from documentation, and by removing some of the more inefficient administrative tasks that can interrupt patient care workflows. “We, at the University of Texas Medical Branch (UTMB), see opportunity in offloading the things that are not the doctor talking to the patient about their health, (but) learning what they need, following them, counseling them, the sort of humanistic touch,” he says.

To enable doctors to provide the best possible care, Dr. McCaffrey sees the lab playing an “enormous role.” “If you think about what doctors want and what people want from healthcare, they want clinical inference all the time on all kinds of things,” he points out, and this is where he says the test results from the lab come in. “The humanistic things like enriching the relationship are really driven in large part by what we provide to the physicians — the guidance we give and the results that we give.”  

Currently, the interpretation of results takes a lot of time and requires physician expertise, which is costly in terms of resources, but increasingly AI is able to improve performance in this area and reduce the time required for the task. “I think we're going to see a trend where there's a cheapening, if you will, of the value of an interpretation, but an expectation to do it 10 times more as a lab,” Dr. McCaffrey says. 

Potential to change the way labs operate

There are many applications for AI in the form of predictive tools, generative tools, and operational tools that can help process and analyze individual data, and even give recommendations for the next steps. Dr. McCaffrey believes these AI tools mean the clinical laboratory will become critical to the patient care workflow.

Predictive tools

AI in the form of computational software tools can now help to predict individuals at risk for certain conditions. When his lab receives a complete blood count, Dr. McCaffrey asks what information can be taken from it. “Can we predict who's going to become anemic? Can we use that to triage workflows for colonoscopies or follow-ups? We do this across radiology as well…if you get a chest x-ray, or a CT at UTMB those will be scrutinized for aortic calcification, coronary artery calcification,” he says.

He believes that adopting “opportunistic screening” as a routine would bring better outcomes for patients. This is because patients are creating data feeds no matter why they’re at the hospital, whether it's an image, a tissue specimen, or a blood tube, there is valuable diagnostic information in that. He says, “Our perspective is it shouldn't require an individual to decide it's worth scrutinizing for it to be scrutinized, it should be scrutinized basically for free all the time so that the things that are anomalous about it can be used to move the care workflow forward. And I think this brings better outcomes to patients.”

Generative tools

A more complex application of AI in healthcare is through generative AI. Dr. McCaffrey has seen the introduction of this technology in the areas of interpretation and imperative guidance, resulting in AI-guided workflows. For example, at UTMB, the team has put together a process for toxicology interpretation with a draft written by chat GPT-4, with some prompt engineering and using Retrieval-Automated Generation (RAG) technology which can tap into a huge amount of validated medical knowledge to provide additional information.3

The workflow now follows the steps below:

  • A urine mass spectrometry drug screen is ordered.
  • Completed liquid chromatography–mass spectrometry results are produced.
  • Results are picked up by an AI agent (a software program to perform specific tasks).
  • Using RAG technology, the AI agent pulls additional information from electronic health records such as active meds, patient age, patient sex, patient ethnicity, and other parameters.

  • The knowledge base of interpretive guidance is used to write the report, which is pushed back into electronic health records automatically.
  • Physician looks at their outstanding list and can sign out the patient report, or edit it if required.

This process allows more context to be provided to physicians and this is where Dr. McCaffrey hopes to see AI driving things forward to enrich healthcare provider knowledge and improve patient care.

Operational tools

Dr. McCaffrey also sees many uses for AI in a more clerical or administrative capacity. Currently, UTMB is using AI to help with tasks such as prior authorization, clerical documentation, and collating the variety of faxes and emails that come from different stakeholders. 

 

UTMB staff have also found value in utilizing AI to answer Standard Operating Procedure (SOP) guidance questions and for document review. When previously Dr. McCaffrey might have been asked for input, instead staff can now ask an AI agent some questions, such as, ‘What do you do with this tube?’ They have also used AI to perform tasks like examining SOPs for contradictions between policies, which has flagged some inaccuracies. “That says 5 [millitres], that says 4 mills, that says a green top, that says an orange top.” he says “You know, it's like 10,000 pages of PDF, no one's going to look through that, but AI can look through that. There's another area where it's very helpful.”

 

  • A urine mass spectrometry drug screen is ordered.
  • Completed liquid chromatography–mass spectrometry results are produced.
  • Results are picked up by an AI agent (a software program to perform specific tasks).
  • Using RAG technology, the AI agent pulls additional information from electronic health records such as active meds, patient age, patient sex, patient ethnicity, and other parameters.

  • The knowledge base of interpretive guidance is used to write the report, which is pushed back into electronic health records automatically.
  • Physician looks at their outstanding list and can sign out the patient report, or edit it if required.

This process allows more context to be provided to physicians and this is where Dr. McCaffrey hopes to see AI driving things forward to enrich healthcare provider knowledge and improve patient care.

Challenges to the adoption of AI

Although there is clear value for AI in the laboratory space, there are some challenges to more widespread adoption such as attitudes and uncertainty over regulations.4

“We don't have regulatory clarity, we don't have technical clarity, we don't have political clarity,” Dr. McCaffrey says. “We don't have logistical clarity on how to deploy something, or where, or what to do about monitoring it, and what the extent of our behavior is there. So, this needs to be crystallized.”

He believes clarity is possible and that market pressure, along with a collective will for change will bring about solutions to these challenges. He notes, “We’ve brought forth lots of complicated stuff, like mass spec, in the past. I don't think this is more complex really, just a slightly different use case.”

His hope for the future, once a regulatory framework is in place, is that labs will be given more ‘providence’ over care pathways due to the large amount of information they can offer to healthcare providers. In practice, this could mean AI-automated detections and nominations into patient care pathways for certain conditions. “We should be doing that more standardly and I think AI, finally, can let us do that and I have good faith that it will show the value in doing that,” he says.

For more examples of the use of AI from Dr. McCaffrey, watch the full presentation, “Digital Solutions Unlock New Possibilities for Better Care.

Want to be the first to receive the latest insights from industry leaders? Sign up for our newsletter.

  1. Hou H et al. (2024). Clinica Chimica Acta, 559, 119724. Paper available from https://www.sciencedirect.com/science/article/abs/pii/S0009898124019752 [Accessed October 2024]
  2. Forbes. (2024). Article available from https://www.forbes.com/sites/bernardmarr/2024/06/17/what-jobs-will-ai-replace-first/ [Accessed October 2024]
  3. KeyReply. (2024). Article available from https://www.keyreply.com/blog/leveraging-rag-to-rethink-healthcare-a-new-era-of-ai-integration [Accessed October 2024]
  4. Mennella C et al. (2024). Heliyon 10, e26297. Paper available from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10879008/ [Accessed October 2024]