AI Intervention Noninferior to Therapist Phone Calls for Chronic Back Pain

Artificial intelligence (AI)-delivered cognitive behavioral therapy for chronic pain (CBT-CP) was found to be noninferior to CBT-CP delivered by therapists over the telephone, according to the results of a study published in JAMA Internal Medicine.

A randomized noninferiority comparative effectiveness was conducted at 2 Department of Veterans Affairs (VA) medical centers. Patients (N=278) with chronic back pain were identified through health records between 2017 and 2019 and were randomly assigned in a 1.4:1.0 ratio to receive AI-CBT-CP (n=168) or CBT-CP delivered via telephone by a therapist (n=110). Both groups received 10 weekly CBT-CP sessions addressing pain coping skills. The telephone intervention lasted 45 minutes per session. For participants in the AI-CBT-CP group, the AI recommended 1 of 3 options: an individualized, recorded, interactive voice response (IVR) message, a 15-minute live telephone session with a therapist, or a 45-minute live session with a therapist. The primary outcome was change in Roland Morris Disability Questionnaire (RMDQ) through month 6.

The AI and telephone cohorts comprised 87.5% and 91.8% men, mean age was 62.9 (standard deviation [SD], 13.1) and 65.5 (SD, 10.6) years, 80.1% and 84.4% were White, 66.7% and 60.0% lived at least 20 miles from their VA medical center, 23.4% and 22.9% had more than 20 outpatient visits the previous year, and 31.7% and 23.1% had at least 2 emergency department visits in the previous year, respectively.

Compared with patients in the telephone group, patients in the AI group completed more treatment weeks (P <.001) and more modules (81.5% vs 57.3%). Of all AI-CBT-CP sessions, 45.8% were delivered via IVR, 41.6% were delivered via 15-minute live sessions, and 12.6% were delivered via 45-minute live sessions. Assuming all study participants completed the entire treatment protocol, the AI recipients would have received 26% as much live therapist contact as the control-group participants.

Compared between groups, RMDQ scores differed by -0.72 (95% CI, -2.06 to 0.62; P <.001) at 3 months and by -1.24 (95% CI, -2.48 to 0; P <.05) at 6 months, meeting the criteria for noninferiority.

The threshold for noninferiority was not met for change in the Numerical Rating Scale (NRS) for pain, Brief Pain Inventory-Interference, Patient Health Questionnaire-9, Physical Composite summary, or Mental Composite summary (all P ≥.12).

At 6 months, more of the AI recipients had a clinically meaningful improvement in RMDQ (37% vs 19%; P =.01) and NRS (29% vs 17%; P =.03).

The results of this study may not be generalizable, as patients were recruited from the VA and this patient population tends to be older with more male and White patients.

This study found that an AI-delivered CBT-CP intervention was noninferior to a telephone-based intervention. The investigators propose that “more patients may achieve clinically meaningful improvements in pain control with AICBT-CP than with standard CBT-CP approaches.”

Disclosure: Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.