Table 2.

Key themes identified by HCPs^a with regard to the use and integration of AI^b for pain assessment in the NICU^c.

Theme	Description	Representative quote
Technology is used regularly in the NICU	HCPs shared that despite having limited experience with AI specifically, they use technology to inform their clinical decision-making and they feel comfortable using the technology that is currently available.	“It informs everything. I think that’s one of the things that working in intensive care is that we use technology and monitoring to inform a lot of our decisions.”
Concerns of AI integration for pain assessment in the NICU	HCPs identified concerns related to the integration of AI in the NICU. It specifically increased the workload for HCPs and increased distress, knowing they were potentially inflicting pain on an infant. They also reported that constant pain monitoring could increase stress for parents and that added machinery could inhibit parent-child bonding. Concerns were also identified with regard to the overreliance on what the algorithm reported and the overuse of pain pharmaceuticals to manage pain.	Increased HCP distress: “I’m not sure cause you imagine like how upsetting it would be like you know I’m doing a diaper change and this thing is telling me the baby is in pain.” Increased workload: “I think there would be some negative feedback towards having extra work to be done.” Fear of overreliance on the AI: “The disadvantages would be that we become over reliant on it. And just because the machine says the baby’s not in pain, then it could be dismissed as the baby isn’t in pain, when actually if you look at the baby, you can tell they’re in pain.” Increased parent stress: “It can cause stress ... Unnecessary stress.” Impeding parent-child bonding: “I can see it taking away from looking at babies...you see parents, particularly looking at their monitor alarms, for whatever reason, they look more at the monitor than actually what their baby’s doing.”
AI has the potential to improve pain assessment and management	HCPs indicated there are several ways in which integrating constant pain monitoring in the NICU could improve clinical care, including the development of new therapies, early diagnosis of difficulties, detection of changes in clinical presentation, increased awareness of infant pain, increased efficiency of pain assessment, increased standardization of pain assessment, and increased collaboration between HCPs and parents.	“I think it’s good that um there is a form of technology that can give us more information about pain in this population because I think there’s a lot of unknown and I think well I know for myself like I said I can’t honestly say that I’m always thinking about if this baby is in pain or what kind of pain this baby is in when doing a procedure.” “I think it would give them more time to obviously focus on other aspects of their work instead of having to score every half an hour or so to proceed and enter the data as it is at the moment.”
Requirements for implementation of AI in NICU	HCPs described structural (ie, machine size and invasiveness of machinery) requirements for implementing AI in the NICU. Specifically, machinery would need to be small and noninvasive. HCPs indicated that training staff to understand and interpret the output provided by the technology is important. They also indicated that the algorithm would need to be properly validated and sensitive for detecting pain in diverse patient groups and situations.	Structural requirements: “It depends how invasive the technology is. When you have a 450 gram baby in front of you. Even putting on things like more monitors actually occludes your that visual assessment of the child. So I think there can be barriers.” Importance of training: “I think obviously, it’s all about training ... everybody understands how it works and the benefits.”
AI is a tool to inform clinical pain assessment and management	HCPs indicated that AI in the NICU should be viewed as a tool to inform clinical decision-making but not as a replacement. They also indicated that the integration of this technology would have implications for the training of new HCPs to ensure they have the ability to understand how this tool could inform their own clinical assessment.	“I like using technology but as long as it doesn’t replace my ability to provide comfort and care” “If I’m gonna make it’s just detection of pain, I think it’d be fairly comfortable with that. Because then I can react to that. Whereas if it’s making medical decision on the treatment, a baby’s receiving, I think that will be a completely different scenario.”
Ethical concerns with constant pain monitoring may occur	HCP indicated the need to be aware of ethical concerns like the potential bias in AI algorithms, disagreements between HCPs and the AI’s output, and the implications of constant pain monitoring without intervening. HCPs also indicated that algorithms would need to be audited and monitored over time.	“And then you have to decide, what you want to do about it. And then you have to decide, in a medical-legal issue whether to believe A.I. or the clinician and that will be interesting.”

^aHCP: health care professional.

^bAI: artificial intelligence.

^cNICU: neonatal intensive care unit.