Who needs opioids?

Harvard School of Dental Medicine faculty, residents, and students published a paper in Current Surgery Reports examining how AI and machine learning could help clinicians predict postoperative opioid needs and identify patients at greater risk for misuse, prolonged use, or inadequate pain control. The paper notes that oral and maxillofacial surgery procedures account for more than 60 percent of opioids prescribed by dentists in the U.S., and that in a prospective study of third-molar surgery patients only 7 percent required opioids when recovery followed an uneventful course, while more than half of opioids prescribed after dental surgery remain unused. Senior author Tim Wang, chief resident of the Oral and Maxillofacial Surgery program at the Dental School and Massachusetts General Hospital, said "the fundamental problem is that opioid prescribing is still largely standardized, while patients' actual pain medication needs are highly individualized.

Health https://news.harvard.edu/gazette/section/health/ Who needs opioids? As concerns about misuse persist, paper suggests role for AI in making pain treatment more precise For a patient having their wisdom teeth removed, the days after surgery can bring an important question: How much pain medication will I actually need? Over-the-counter medications such as acetaminophen and ibuprofen provide adequate pain relief for most, but some patients experience breakthrough pain that does not respond to these medications and may benefit from an opioid. For clinicians, the challenge is knowing before surgery which patients will need additional pain relief. Harvard School of Dental Medicine faculty, residents, and students are exploring whether artificial intelligence could help make those decisions more precise. In a new paper published in Current Surgery Reports https://link.springer.com/article/10.1007/s40137-026-00528-5 , the authors examine how AI and machine learning could help clinicians predict postoperative opioid needs and identify patients who may be at greater risk for misuse, prolonged use, or inadequate pain control. “The fundamental problem is that opioid prescribing is still largely standardized, while patients’ actual pain medication needs are highly individualized. Two patients can undergo essentially the same operation, performed by the same surgeon, and have very different experiences with pain and swelling afterward due to biology and pain tolerance,” said Tim Wang, senior author and chief resident of the Oral and Maxillofacial Surgery https://www.hsdm.harvard.edu/oral-and-maxillofacial-surgery-program OMFS program at the Dental School and Massachusetts General Hospital. Rethinking ‘just-in-case’ prescribing Oral and maxillofacial surgery procedures lead to the most dental opioid prescriptions, accounting for more than 60 percent of opioids prescribed by dentists in the U.S. Procedures such as third-molar extraction can also represent a patient’s first exposure to prescription opioids. But the amount prescribed does not always reflect how much a patient ultimately needs. The authors highlight a prospective study of patients undergoing third-molar surgery in which only 7 percent required opioids when recovery followed an uneventful course. Other research has found that more than half of opioids prescribed after dental surgery remain unused, creating opportunities for misuse or diversion. “The challenge for the surgeon is that we have to make the prescribing decision before we know how that individual patient will respond in terms of postoperative pain,” said Wang. The authors describe this practice as “just-in-case” prescribing, providing medication in anticipation of potential pain rather than basing the prescription on a patient’s individual risk and expected needs. At the same time, efforts to reduce unnecessary prescribing must not leave patients who do need additional pain relief without adequate treatment. For Samat Borbiev, D.M.D. ’28, and Clark Morgan, D.M.D. ’28, student authors who contributed to the paper, the challenge is finding that balance. “With more breakthroughs in clinical biomarkers and AI, we can help merge the two to create more tools for physicians to help them make informed decisions on pain control, management, and also nudge physicians when a patient may need more counseling and time to help manage their pain,” said Borbiev. That balance can be particularly difficult because the factors that influence pain and opioid response extend beyond the procedure itself. “Ultimately, the amount of clinical, genetic, and social data that may influence an ideal pain-control regimen is greater than what one provider can fully collect and consider, opening the door for AI to aid in clinical decision-making,” Morgan noted. Using AI to account for individual differences The authors point to machine learning as one way clinicians could account for these differences. AI models can analyze large amounts of clinical, behavioral, and biological data to identify patterns associated with opioid use, misuse, and prolonged use. They highlight research involving more than 560,000 Medicare beneficiaries in which machine learning methods outperformed traditional statistical approaches in predicting opioid overdose risk. Other models have demonstrated the ability to predict opioid use disorder and prolonged postoperative opioid use with varying degrees of accuracy. For clinicians, the potential value of these tools lies in translating those predictions into more individualized care. AI could potentially be incorporated into electronic health records to help determine the appropriate type, quantity, or duration of an opioid prescription, flag patients who may benefit from additional monitoring, and enhance existing prescription drug monitoring programs. The promise of artificial intelligence is that it may allow us to move away from a one-size-fits-all approach toward more individualized prescribing.” Tim Wang “The promise of artificial intelligence is that it may allow us to move away from a one-size-fits-all approach toward more individualized prescribing,” said Wang. AI could also help clinicians quickly review prescription drug monitoring data and identify relevant patterns in a patient’s history, he added. The technology could also continue to support care after a patient leaves the clinic. The authors describe systems that could incorporate patient-reported pain, medication use, and adherence to refine recommendations over time, while also helping clinicians provide targeted education about pain management and medication use. “It’s true that an inpatient setting provides relative ease for pain monitoring and management per the physician perspective, but the moment a patient is discharged we lose the benefit of direct supervision,” said Matthew Watt, D.M.D. ’26, OMFS resident and a contributor to the paper. “AI can be leveraged with greater return in the at-home setting, where we can utilize these models for more tailored pain regimens.” The authors emphasize that AI is not a replacement for clinical judgment. Much of the existing research is retrospective, and many models have undergone limited external validation. Their performance can also vary depending on the populations and data used to train them. Despite these challenges, Wang said the goal is not simply to reduce opioid prescribing, but to make it more precise without compromising patient outcomes. “For patients, that could mean a pain-management plan shaped not only by the procedure they underwent, but also by the individual recovering from it,” he said.