Study: ChatGPT shows promise in managing depression

Key takeaways:

• ChatGPT 3.5 and 4 prescribed psychotherapy for mild depression in over 95% of cases, adhering to guidance.
• Researchers stressed that AI should supplement human clinical judgement, not supersede it.

ChatGPT performed better at managing mild and severe depression in accordance with current clinical guidance without showing gender or socioeconomic bias compared with primary care physicians, according to researchers in Israel.

Despite artificial intelligence’s (AI) potential in clinical settings, “its potency in addressing critical clinical mental health challenges has yet to be definitively determined,” Inbar Levkovich, PhD, a senior lecturer at Oranim Academic College of Education, and Zohar Elyoseph, PhD, a senior lecturer at Max Stern Yezreel Valley College, wrote in Family Medicine and Community Health.

They also highlighted an absence of research regarding ChatGPT and depression.

The researchers compared how ChatGPT evaluated depressive episodes and made suggestions for treatments in hypothetical scenarios with recommendations made by 1,249 human PCPs.

They found that psychotherapy was recommended for mild depression in:

• 95% of cases by ChatGPT 3.5;
• 97.5% of cases by ChatGPT 4; and
• 4.3% of cases by PCPs.

PCPs instead often recommended prescription of a pharmacological treatment (48.3%) or a combination of pharmacological treatment and psychotherapy (32.5%).

Levkovich and Elyoseph explained 2-week psychotherapy is often sufficient for mild depression, and that antidepressants should be considered only if there is no improvement, the patient does not quality for psychotherapy or if symptoms are moderate to severe.

Thus, “unlike the recommendations of the PCPs, the therapeutic proposals of ChatGPT are in line with the accepted guidelines for managing mild depression,” they wrote.

For pharmacological treatment of severe depression, ChatGPT 3.5 and 4 recommended an exclusive use of antidepressants in 74% and 68% of cases, respectively.

Physicians recommended a mix of antidepressants and anxiolytics/hypnotics in 67.4% of cases, and exclusive use of antidepressants in 17.7% of cases.

“Again, the therapeutic recommendations of ChatGPT were in line with the accepted guidelines for managing severe depression,” the researchers wrote.

They also pointed out ChatGPT 3.5 and 4 showed no biases regarding gender or socioeconomic status [SES] in their recommendations. In contrast, previous research suggests PCPs can be susceptible to those types of biases — for example, being significantly less likely to prescribe antidepressants to women vs. men.

Levkovich and Elyoseph explained that this finding “offers great potential.”

“Indeed, the ability to adjust treatment without ‘falling into the trap’ of gender or SES bias has the potential to promote both quality and equity in mental health care,” they wrote.

However, the researchers stressed that "ChatGPT should not supersede human clinical judgment in the diagnosis or treatment of depression. It should instead serve as a supportive instrument that fortifies professionals in their attempt to make well-informed clinical decisions."