Just try it! That was the advice on integrating AI into occupational healthcare from Enterprise Health’s president, Jeff Donnell at this year’s Association of Occupational Health Professionals in Healthcare (AOHP) annual conference.
Despite the momentum around this revolutionary technology, most attendees at Donnell’s session had yet to integrate AI into clinic practices and workflows, although they acknowledged that AI will soon play a role in their workplace. With their curiosity piqued, Donnell not only made the case for integrating AI into the practice of occupational healthcare but also addressed some common challenges and concerns. From the outset, he clearly established that in the occupational health space, AI is not going to replace clinicians or take away their livelihoods. In fact, in a profession that struggles to find enough occupational healthcare clinicians, AI will improve working conditions and play a key role in preventing clinician burnout.
A clear example of this is in how AI can improve electronic health records (EHRs) and patient encounters. “When EHRs were launched years ago there were all these great promises,” explained Donnell. “Not only would we get rid of paper, but we were going to have more accurate, easily accessible records. And using EHRs would save all this time, but then reality set in.” The reality that Donnell referenced is the endless clicking through templates to find the right sections to record the patient encounter, time spent staring at a keyboard entering notes instead of connecting with the patient, and still requiring additional hours of data entry after work.
In using AI as a scribe during a patient encounter, clinicians can keep their focus on the patient. Moreover, AI can create a host of supporting materials: SOAP notes, documents translated into a patient’s working language, referral letters, and work status letters to identify just a few items. “It’s really about going back to those things that normally get done during pajama time (after work),” Donnell shared. “You can do them in real time, or at the end of the day. What would have taken 10, 15, or 20 minutes to complete per patient would just take a couple of minutes with an assist from AI.”
However, it’s not just as a scribe where AI demonstrates potential in the occupational health setting. Donnell also shared a unique application that he sees – data chasing. That is, being able to search, apply, and put data to work to identify trends within a workplace, such as vaccine uptake or recurrent injuries. For example, he shared, “we have all these employees who need flu shots, or respirator tests. AI can analyze data from all the company’s OEHRs and identify those employees who need them. An AI agent can then call them, fill out questionnaires, make appointments, and send reminders. The entire process requires little additional input from a human clinician or clinic administrator after they enter the prompt.”
With all these evident benefits, what’s holding occupational health clinics back from stepping boldly into the AI era? Donnell identified two core issues – understanding and governance. Because AI is often presented as a black box and people see the widely available AIs, like ChatGPT, that scrape the internet to feed their Large Language Models (LLM) and algorithms there are well-founded concerns about poor quality information resulting in wrong and potentially harmful information, especially in a clinical setting. Donnell had two responses to this concern. First, he shared that occupational health clinics should be using AI solutions that are trained on specific data sets relevant to medical practice, occupational healthcare, and each organization’s unique data. Even then, he shared, AI has its issues. “It’s OK to use AI, but it is not a human replacement. AI will hallucinate, and AI will make stuff up. Independent medical judgement reigns supreme.”
However, concerns over governance are the biggest reason that occupational health clinics are hesitant to deploy AI. With Google searches for guidance on AI governance growing 7,000 percent over the last five years, it’s clear that everyone is still trying to work this out. Donnell noted not only the obvious need for governance over data privacy and security, but also how to write the policy documents for actual use, reporting breaches, disciplinary policy, managing shadow AI tools, and coordinating all teams within an organization – at a minimum IT, occupational health clinicians, and legal – to agree on policies and procedures. This already complex work is further complicated by how quickly AI is evolving and the requirements of state agencies and international organizations.
Even after highlighting the complications and potential pitfalls of AI in occupational healthcare, Donnell was quick to return to the positive. “Today there are emerging standards, like the FAVES framework, healthcare AI certification programs like the Drummond pDSI-risk assessment that will help guide the adoption of AI in occupational healthcare,” he explained. “While there’s currently only one body certifying AI for healthcare, more guidance will be coming.
And just as more guidance and structure are coming to the occupational healthcare community on how to implement AI, more AI applications will also be coming. And for Donnell, this means it’s time for occupational healthcare workers to get familiar with AI for all the reasons he shared throughout his presentation. However, he saved perhaps the most important for last.
“Your job won’t be replaced by AI, but you will likely be replaced by someone who knows how to use AI,”he shared in his closing remarks. In other words, now is the best time for occupational health care workers to just try it!