A new multi-institution research study shows how Artificial Intelligence (AI) machine learning combined with Electronic Health Records (Electronic Health Records) can predict antidepressant treatment outcomes. “These investigations have the potential to drive the development of a clinical decision‐making tool for personalized management of depression,” wrote researchers affiliated with Weill Cornell Medicine, Temple University, the University of Washington, Mayo Clinic, Northwestern University, and the University of Florida who conducted the study, which was funded in part by the U.S. National Institutes of Health.
An estimated 280 million people worldwide, or 3.8 percent of the global population, experience depression, according to the World Health Organization (WHO). Fortunately, there are effective treatments for depression. Symptoms of depression may include persistent feelings of sadness, the loss of interest or pleasure in things and activities once enjoyed, feelings of guilt or worthlessness, thoughts of suicide or death, slowed movements or speech, difficulty thinking or making decisions, concentration challenges, changes in appetite, too much or too little sleep, loss of energy or increased fatigue, and loss of energy according to the American Psychiatric Association (APA).
“While antidepressants are commonly prescribed to patients suffering from depression, due to the complex etiology and heterogeneous symptomatology of depression, prior studies suggest that antidepressant treatment efficacy is usually low, with as few as 11–30 percent of depressed patients obtaining remission after initial treatment,” the scientists wrote.
Antidepressants used to treat depression include Selective Serotonin Reuptake Inhibitors (SSRI), Serotonin and Norepinephrine Reuptake Inhibitors (SNRI), Monoamine Oxidase Inhibitors (MAOIs), Neuroactive Steroid Gamma-Aminobutyric Acid (GABA)-A Receptor Positive Modulator, Tricyclic and Tetracyclic Antidepressants, Atypical Antidepressants, and N-methyl D-aspartate (NMDA) Antagonist, according to the U.S. Food and Drug Administration (FDA).
The researchers used a variety of AI machine learning (ML) algorithms, such as gradient boosting decision tree (GBDT), Naïve Bayes (NB), random forest (RF), and L2 norm regularized Logistic Regression (LR) to predict outcomes to antidepressant treatment. The study used the XGBoost software library for the gradient boosting decision tree algorithm, and scikit-learn software library, a machine learning library for Python, for the other algorithms. The gradient boosting decision tree algorithm performed the best in predicting antidepressant treatment outcomes.
The study used fully de-identified data from over 800 adults who received at least one antidepressant prescription from an outpatient behavioral health clinic at a New York City academic medical center. To train the AI models, the scientists used a variety of data from electronic health records, such as prescription medications, procedures, demographic information, baseline depression severity, and comorbidities.
Comorbidities are the presence of more than one disease or disorder that may impact physical or mental health that happens at the same time in a person. Up to 90 percent of patients with anxiety disorders have comorbid depression, according to research by Jack M. Gorman M.D. published in Depression and Anxiety.
Patients with chronic medical diseases, such as cancer, neurological, metabolic, and cardiovascular disorders often also have depression as a comorbidity. Up to 25 percent of cancer patients have depression as a comorbidity, according to the National Cancer Institute. The statistics are higher for heart failure patients, of whom up to 30 percent also suffer from depression, according to “Depression and heart failure: the lonely comorbidity” published in the European Society of Cardiology.
The researchers demonstrated that their machine learning algorithms predicted antidepressant treatment outcomes using the patient’s medical history, which may help clinicians in the future. “Beyond informing targeted treatments, these predictive models may potentially contribute to the design of a new generation of EHR‐linked clinical trials,” the Researchers suggested. “For example, clinicians can stratify the patients into “high‐risk” and “low risk” groups based on predictive results (“worsening” or “recovering”) and pay closer attention to the treatments and prognosis of the “high‐risk” group.” Additionally, the researchers believe that their AI algorithms predicted results may be useful in developing more targeted psychiatric treatment plans. “Our predictive tool holds the promise to accelerate personalized medical management in patients with psychiatric illnesses,” the Researchers concluded.
REFERENCE: Psychology Today; 05 APR 2023; Tyler Woods