Artificial intelligence is transforming the role of advanced practice providers (APPs) by streamlining documentation, simplifying clinical workflows, and helping identify patients who might otherwise fall through the cracks. Beyond improving efficiency, AI is enabling providers to spend more time with patients, strengthen care coordination, and support earlier disease detection. In this interview with BioPharma BoardRoom, Dr. Javier Zulueta, Chief Medical Officer of Pulmonology at Qure.ai, explains how AI-powered tools are reshaping clinical practice, improving outcomes through earlier intervention, addressing workforce shortages, and highlighting the safeguards needed to ensure responsible adoption in healthcare.
Where are AI tools having the greatest impact on APP workflow and efficiency? / 2. Are you able to devote more time to patient education, care coordination, chronic disease management, or preventive care?
AI is giving APPs and physicians the possibility of dedicating more time to the real mission of their practice, spending quality time with patients, while significantly reducing the time spent on administrative work.
Clear examples include AI that listens to the conversation between provider and patient and instantly writes a structured report, which can also incorporate relevant information already in the EMR. Today, one of the major challenges when we see patients in our offices is that while we are talking to the patient, we have to simultaneously type the conversation into our computers. Many patients feel providers are distant because of this lack of eye contact during visits.
Another example is the ability of LLM-based AI models to extract information from the electronic medical record, allowing the provider to spend more time intellectually reviewing data rather than searching for it.
AI has also proven very effective at flagging patients for medical problems unrelated to the one that brought them into the healthcare system, a clear real-world example being incidental lung nodules. Research shows that more than 60% of individuals with a lung nodule detected on a radiological imaging test will be lost to follow-up, despite a statement by the radiologist on the report recommending follow-up. LLM-based AI can search for patients with abnormalities on their diagnostic testing that have gone unnoticed because the test was ordered for unrelated reasons. A navigator-based program can then have easy access to all patients with incidental findings and use AI to seamlessly facilitate their journey through a diagnostic workflow.
At Mount Sinai Morningside Hospital, we have shown that this approach can reduce the proportion of patients lost to follow-up from 60% down to 10%. The most important aspect is that up to 5% of patients lost to follow-up may have lung cancer in early stages. In fact, at Mount Sinai Morningside, among patients with nodules who were at risk of being lost to follow-up and were identified by AI, 75% of those eventually diagnosed with cancer were found at stage I.
Another benefit we experienced in our ILN program at Mount Sinai Morningside is the drastic reduction in the time navigators had to spend diving into the electronic medical record to extract each patient’s history. That information can now be automatically extracted by LLM-based AI, reducing navigator time from anywhere between 2 and 15 minutes per patient to virtually none.
Has AI introduced any new risks or workflow challenges?
As with any new technology in healthcare, special care must be taken to prevent associated risks. Any diagnostic test that is highly sensitive can potentially increase the risk of unnecessary and potentially harmful procedures. In the use case mentioned before, opportunistic detection of incidental lung nodules may lead to more biopsies of benign lesions. Even if biopsy complication rates are low, any complication in a patient who does not have a significant or potentially deadly disease is unacceptable, so extreme care must be taken in the workflow protocols. On the other hand, AI may also create a scenario where better characterization of nodules actually reduces the need for biopsies of benign lesions, in other words, increasing the specificity of the diagnostic process and reducing false positives. As we implement AI in real-world scenarios, we will have to monitor outcomes very closely to weigh the benefits and harms.
Can AI help address workforce shortages or improve access to care?
Yes, most definitely. One clear example is the effectiveness of AI applied to conventional chest X-rays to detect tuberculosis in rural areas of low- and middle-income countries with severe shortages of radiologists. qXR, the first AI tool developed by Qure, is approved for use in areas where there are no radiologists, for the detection of tuberculosis on chest X-rays.
Another use case currently being developed in the United States is the implementation of programs for incidental abnormalities. Without AI, the finding of an incidental abnormality on a diagnostic test requires significant human effort to navigate the patient through the healthcare system. With AI, this process can be done much more efficiently, with a significantly reduced need for human resources.
What unintended consequences or risks have emerged as APPs adopt AI tools? Looking ahead, how do you see AI changing the role of APPs over the next several years?
As mentioned before, the use of AI tools may result in an increase in unnecessary procedures if strict protocols are not followed. However, as experience is gained during the implementation process, I believe AI itself is going to help with adherence to guidelines and the avoidance of unnecessary procedures. To guarantee this, strict and close monitoring should be maintained throughout the implementation process.
