Machine Learning Improves Schizophrenia Treatment Success


A new algorithm using a 70% training had a 51.7% success rate in predicting individualized treatments for first-episode schizophrenia patients.

Chi-Shin Wu, MD, PhD

Chi-Shin Wu, MD, PhD

Schizophrenia remains 1 of the more difficult mental health disorders to treat, with many first-choice medication selections resulting in relapses or non-adherence.

A team, led by Chi-Shin Wu, MD, PhD, Department of Psychiatry at the National Taiwan University Hospital & College of Medicine, examined whether an individualized treatment rule can be developed using administrative date to improve the antipsychotic medication selection for patients with first-episode schizophrenia.

There is currently little guidance as to how to select antipsychotic medications for patients with first-episode schizophrenia.

In the prognostic study, the investigators examined the records of 32,777 Taiwanese patients with first-episode schizophrenia from Taiwan’s National Health Insurance Research Database on patients with prescribed antipsychotic medications, ambulatory claims, or discharge diagnoses of a schizophrenia between January 2005 and December 2011.

Using this data, they created an ensemble machine learning technique for an individualized treatment rule by applying a targeted minimum loss-based ensemble machine learning method to predict treatment success from baseline clinical and demographic data.

The rule, which was developed in a 70% training sample had an estimated treatment success rate of 51.7% when applied to the remaining 30% of the patients in the holdout sample. This was substantially higher than the observed 44.5% treatment success rate in the holdout sample.

The investigators sought main outcomes of treatment success as defined by not switching medication and not being hospitalized for 12 months.

The estimated treatment success if all patients were given a prescription for 1 medication was significantly lower for each of the 13 medications in the study than under the individualized treatment rule (Z =&thinsp;4.2-16.8; all P&thinsp;<&thinsp;.001).

Aripiprazole (3088 [31.9%]) and amisulpride (2920 [30.2%]) were the medications most often recommended using the new machine learning method. However, only 1054 patients (10.9%) received individualized treatment rule-recommended medications.

Observed treatment success was significantly higher than if medications had been randomized (44.5% [SE, 0.55%] vs 41.3% [SE, 0.4%]; Z&thinsp;=&thinsp;6.9; P&thinsp;<.001). However, it was only marginally higher than if medications had been randomized in their observed population proportions (44.5% [SE, 0.5%] vs 43.5% [SE, 0.4%]; Z&thinsp;=&thinsp;2.2; P&thinsp;= .03]).

“These results suggest that an [individualized treatment rule] may be associated with an increase in the treatment success rate among patients with first-episode schizophrenia, but experimental evaluation is needed to confirm this possibility,” the authors wrote. “If confirmed, model refinement that investigates biomarkers, clinical observations, and patient reports as additional predictors in iterative pragmatic trials would be needed before clinical implementation.”

For schizophrenia patients, antipsychotic medications can often reduce symptoms and decrease relapse rates. However, recently investigators found that only 60% of patients with first-episode schizophrenia achieved significant symptom reduction with their first antipsychotic medication.

Between 33-72% of this patient population changed medications or discontinued treatment in the 12 months following initiation due to insufficient efficacy or adverse effects. Inadequate treatment increases the risk of both relapse and poor long-term functional outcomes.

The study, “Development and Validation of a Machine Learning Individualized Treatment Rule in First-Episode Schizophrenia,” was published online in JAMA Network Open.

Related Videos
Insight on the Promising 52-Week KarXT Data with Rishi Kakar, MD
Sunny Rai, PhD: “I” Language Markers Do Not Detect Depression in Black Individuals
Rebecca A. Andrews, MD: Issues and Steps to Improve MDD Performance Measures
A Voice Detecting Depression? Lindsey Venesky, PhD, Discusses New Data
Daniel Karlin, MD: FDA Grants Breakthrough Designation to MM120 for Anxiety
Leesha Ellis-Cox: Steps to Closing the Bipolar Disorder Diagnosis Gap for Blacks
© 2024 MJH Life Sciences

All rights reserved.