Monday, February 10, 2020

Computer Model Might Help Identify Patients at Risk of Not Taking Their Antidepressants

Using electronic health records, researchers have developed a computer program that can predict which patients are at risk of not taking their prescribed antidepressants with about 70% accuracy. The study was published in Translational Psychiatry.

“Treatment discontinuation may reflect a range of features, from depression-associated amotivation and hopelessness to failure to perceive a benefit to concerns about cost,” wrote Melanie Pradier, Ph.D., of Harvard University and colleagues. “However heterogeneous, the consequences of treatment discontinuation are substantial, contributing to poor treatment outcomes and depression chronicity.”

To build the computer program, Pradier and colleagues analyzed electronic health record data of adult patients aged 18 to 80 who received at least one antidepressant prescription between 2008 and 2014. The study included 51,683 patients who had a diagnosis of a depressive disorder and began treatment with one of nine common antidepressants (bupropion, citalopram, duloxetine, escitalopram, fluoxetine, mirtazapine, paroxetine, sertraline, or venlafaxine) and at least one follow-up visit 90 days or more after their first prescription.

The final sample included 70,121 prescription initiations (as many patients switched antidepressants during treatment). Of these prescriptions, 23.77% were associated with a discontinuation of treatment (for example, no prescription refill and no evidence in the medical record of any nonpharmacological depression treatment). Paroxetine was associated with the highest discontinuation rate (27.71%) while venlafaxine was associated with the lowest (20.78%).

The computer program then analyzed patients’ sociodemographic features and medical history to predict the risk of discontinuation for each of the nine antidepressants. Overall, the program was able to predict the patients who would discontinue an antidepressant with 69% accuracy. For individual medications, the model’s accuracy ranged from 62% for paroxetine to 80% for escitalopram.

Pradier and colleagues highlighted two possible applications for their program: “[T]he risk for discontinuation predicted by the machine-learned model might help in prioritizing interventions aimed at retention in treatment and adherence, including making follow-up phone calls, deploying mobile applications, … or simply scheduling an earlier return visit,” they wrote. Alternatively, these models might be applied in settings where multiple medications would be equally reasonable choices for a patient. “Here, all other things being equal, the clinician might prefer the medication with the lowest risk of treatment discontinuation for that patient.”

For related information, see the Psychiatric News article “Tips for Recognizing, Treating Symptoms of SSRI Discontinuation.”


(image: iStock/Shidlovski)


Follow Psychiatric News on Twitter!

And check out the new Psychiatric News Brief on Alexa-enabled devices.

Disclaimer

The content of Psychiatric News does not necessarily reflect the views of APA or the editors. Unless so stated, neither Psychiatric News nor APA guarantees, warrants, or endorses information or advertising in this newspaper. Clinical opinions are not peer reviewed and thus should be independently verified.