Uncovering Problem Opioid Use In Chronic Opioid Therapy With Patient Medical Records

What the Study Found

More than 22,000 patients on COT were enrolled in the study. The patients were from Group Health, a large mixed-model, nonprofit health plan in Washington State. A computer-assisted and manually reviewed search of the medical records using NLP identified 9.4% of patients with problem use. In this same group of patients, 10.1% had an ICD-9 diagnosis of opioid abuse and dependence documented in their medical records.1

“This would seem to be a good correlation between diagnostic coding and NLP, but when we dug deeper we found that about one-third of patients with NLP indicated problem use did not have an ICD-9 diagnosis, and about one-third of patients with an ICD-9 diagnosis were not identified using NLP,” said Palmer. “We can speculate about this discrepancy. It may be due to incomplete documentation or an unwillingness to stigmatize patients with a suggestion or diagnosis of opioid abuse. We don’t know.”

The study also identified 14 risk indicators to assess patients’ relationship to problem opioid use. More than 90% of the patients in the study had at least one indicator. As the number of indicators went up, so did the risk of problem use. The risk of problem use was about 10 percent for patients with three to four indicators, and greater than 50 percent for patients with seven or more indicators.1

Higher rates of problem opioid use were found in patients age 35 and younger who had sustained COT (lasting longer than one year) and in those receiving higher opioid doses. “There was a high correlation between three or more risk indicators and problem use. Patients who had three or more prescribers outside the primary clinic in two or more calendar quarters had a risk for problem use of about 40%,” said Palmer.