When an Algorithm Guides Pain Management: The Growing Backlash Against NarxCare Scores
This article is the first in a series on prescription drug monitoring programs and analysis tools designed to give clinician a view of how likely their patient is to abuse or overdose on a prescribed controlled substance. The second part dives into the lack of government oversight of these tools.
When patients are scheduled for surgery at Cleveland Clinic in Cleveland, clinicians check the person’s prescription drug history embedded in the electronic health record. This history provides scores intended to predict the likelihood a patient will misuse or abuse opioids.
Experts agree the algorithm-generated scores may have some benefits. But clinicians, researchers, and patient advocates fear its adoption has come without adequate scientific validation of those benefits. Critics say the lack of demonstrated value has raised the possibility clinicians may be withholding appropriate pain medication from some patients.
Scott Weiner, MD, MPH
Other tools and algorithms “are all very rigorously developed and validated; we know the literature well, and we rely upon that, and that really doesn’t exist for this score,” said Scott Weiner, MD, MPH, the director of emergency department–based substance use disorder initiatives at Brigham and Women’s Hospital in Boston. “If someone isn’t familiar with the literature, they might assume that this is a well validated tool, and unfortunately, it’s not.”
The NarxCare scores are intended to indicate the likelihood a patient is to misuse, abuse, or overdose on opioid analgesics and other controlled substances. Around 20 states use the product, according to the company that sells the software.
Nicolas Piuzzi, MD, director of the Adult Reconstruction Research Center and an orthopedic surgeon at Cleveland Clinic, Cleveland, found in his research that patients with scores indicating risk for misuse and overdose of opioids and stimulants had higher risk for longer hospital stays and 90-day readmissions following certain orthopedic procedures.
In part based on his research, patients in his department with higher scores now receive a pain management consult prior to surgery. Clinicians ask why patients were on previous drugs and find out details that a number alone cannot provide, enabling better planning of pain management after surgery.
Nicolas Piuzzi, MD
Piuzzi said the scores should help clinicians provide more resources for patients, not serve as “stop gates” that take treatment options away.
Bamboo Health, a Louisville, Kentucky–based behavioral health company that developed the widely-used algorithm, said in a statement that NarxCare “should never replace medical decision-making. NarxCare is a valuable tool that healthcare professionals can use, in addition to other patient-centric factors, to help better understand and improve the management of opioid exposure,” the company, formerly known as Appriss Health, said.
Bamboo Health declined requests for interviews and referred Medscape Medical News to an outside spokesperson, who said the company conducts “internal evaluations of NarxCare’s usefulness in clinical workflows in collaboration with state prescription drug monitoring program administrators (PDMPs).”
The spokesperson also said the company also has a clinical advisory council for the NarxCare platform. Reviews are conducted with external clinicians quarterly to gather feedback, the spokesperson said.
Searching for Validation
More than 2.6 million physicians and other health professionals were registered on their states’ drug monitoring program in 2023, sixfold more than that in 2014, when fewer than 500,000 were on the rolls, according to the American Medical Association.
Most states – 48 – now have laws and rules requiring clinicians to check PDMP data before writing opioid prescriptions, albeit with great variation among them in exemptions to this rule.
About 40 states, including Massachusetts, Colorado, and Florida, contract with Bamboo Health to handle various technical services for their PDMPs. About 20 states use the NarxCare platform.
NarxCare scores are not backed by a robust body of publicly-available data that exist for other widely used clinical support tools, and more research is needed to validate the ability of the scores to predict risk for abuse, experts said. Ranging from 000 to 900, the scores are generated in part on previous opioid prescriptions.
“Because the score is presented in big bold font as if it was definitive, it just naturally tends to mislead any viewer to think it has more precision and utility than it really does,” said Stefan G. Kertesz, MD, MSc, a professor of medicine at Marnix E. Heersink School of Medicine, The University of Alabama at Birmingham.
Only one published study appears to validate the scores— a trial conducted at pharmacies in Ohio and Indiana owned by a single company that compared scores to the results of more than 1400 surveys of patients on their experiences with opioids. The study did not involve physician prescribing. The authors of the study wrote the algorithm could be a “useful broad-based universal screen for risky opioid medication use among community pharmacy patients”
But they also found 17.2% of patients were classified as false positives: Their high scores suggested potential for problems related to substance use, but they had not self-reported misuse. A little over 13.4% of people had false negative scores: Their scores did not suggest a potential for abuse but they had self-reported previous misuse of drugs.
Kertesz said the scores can provide an incomplete picture of a patient’s previous experience with opioids, resulting in higher scores even when drugs were appropriately used.
“It’s a little bit like using a bad credit score, where the score would mislead you to a significant degree about a person’s ability to cover their expenses,” he said.
Weiner and others say clinicians should keep in mind the limits of NarxCare scores when making prescribing decisions. If a patient has a score of 000, they lack a reported history of opioids and other controlled substances.
“Anything apart from 000, it just means I have to look a little bit more at the prescription history,” Weiner said. A score “is nuanced. It depends on what other prescriptions they are taking, how long ago they were, which prescribers gave it to them, why it was prescribed. There’s all these things that are just not present in the PDMP data that the score has no access to.”
Filed under: General Problems
Leave a Reply