For chronic pain, AI-assisted therapy may be just as good as standard care

Key takeaways

• Access to cognitive behavioral therapy for chronic pain is limited due to a shortage of therapists.
• Cognitive behavioral therapy delivered via artificial intelligence was noninferior to therapist-delivered cognitive behavioral therapy and required less than half of the therapists’ time.
• At 6 months, twice as many patients who received therapy through the AI intervention had a clinically meaningful improvement in pain-related disability and pain intensity compared with the control group.

Using artificial intelligence to facilitate cognitive behavioral therapy for chronic pain can provide outcomes that are as good as typical interventions while requiring less time with a therapist, according to researchers.

Evidence-based cognitive behavioral therapy for chronic pain (CBT-CP) is a safe and effective alternative to opioid analgesics, John D. Piette, MSc, PhD, a professor at the University of Michigan, and colleagues wrote in the study, published in JAMA Internal Medicine. However, due to a scarcity of therapists and the fact that CBT-CP requires multiple sessions, “many patients have limited access or fail to complete treatment,” the researchers wrote.

To improve access, Piette and colleagues said they developed a CBT-CP intervention that uses AI (AI-CBT-CP) “to automatically adjust the modality of weekly therapist interactions” based on daily patient-reported feedback through interactive voice response calls (IVR).

“Interventions like AI-CBT-CP could allow many more patients to be served effectively by CBT-CP programs using the same number of therapists,” the researchers wrote.

The AI-CBT-CP intervention was evaluated during a randomized, noninferiority, comparative effectiveness trial that included 278 patients with chronic back pain from the Department of Veterans Affairs health system. The researchers said they randomly assigned more patients to the AI-CBT-CP group than the control — 1.4:1 — “to maximize the system’s ability to learn from patient interactions.”

Each participant received a CBT-CP manual that described eight pain coping skills across 10 weekly modules. Participants also chose a behavioral goal during each session and received a walking goal, which was 110% of their average steps the week before.

In the intervention group, patients briefly reported information about mood, pain intensity, sleep, step counts and more in daily IVR calls. According to the researchers, the AI engine used this information to make weekly recommendations for the patients to receive one of three options:

• a recorded IVR-delivered session that included “a voice message with individualized therapist feedback based on the participant’s IVR-reported data;”

• a 15-minute call with a therapist who reinforced and addressed skill practice barriers; or

• a 45-minute call with a therapist “that prioritized problem-solving difficulties with skill practice and progress toward physical activity goals.”

The patient’s weekly status was evaluated based on a score that considered IVR-reported daily step counts and “the patient’s experience of pain-related interference,” Piette and colleagues wrote. The AI-CBT-CP then calculated the score for all three action choices using a multidimensional matrix.

“Like clinicians, AI-CBT-CP can only make effective decisions about treatment course if it has feedback from reliable and valid assessments about patient status over time,” the researchers wrote. “In this study, brief, daily IVR calls were successful in obtaining this feedback; AI-CBT-CP had the data it needed to make a decision 94% of the time.”

Meanwhile, patients in the control were only offered 45-minute, therapist-delivered telephone CBT-CP sessions.

The researchers found that AI-CBT-CP was noninferior to standard care for pain-related outcomes. It also was associated with an improvement in Roland-Morris Disability Questionnaire (RMDQ) scores.

Although the difference in RMDQ scores was not enough to be clinically meaningful at the 3-month mark (between-group difference = -0.72; 95% CI, 2.06 to 0.62), there was a greater difference at 6 months (between-group difference -1.24; 95% CI, -2.48 to 0). At both points, the results met noninferiority criterion (P < .001).

Overall, more patients receiving AI-CBT-CP had clinically meaningful improvements at 6 months for both pain intensity scores (29% vs. 17%; P = .03) and RMDQ (37% vs. 19%; P = .01), according to the researchers.

“Given the distribution of session types, AI-CBT-CP achieved these outcomes with only 30% of the clinician time required for the comparison program of weekly 45-minute therapist sessions,” the researchers wrote.

The patients in the intervention group also had an 82% completion rate for all weekly sessions compared with 57% in the control group.

“The AI algorithm used to drive decision-making in the current intervention reflected a large number of design features and decisions informed by a panel of CBT-CP experts and best practices in reinforcement learning,” the authors wrote. “Some of these decisions were made with incomplete information, and different results may have been obtained with different features.”

In light of the results, the researchers wrote that “patients may find the intervention more convenient, and health systems could use it to treat more patients without additional clinical resources.”