ChatGPT excels in depression treatment guidelines, may beat doctors.
The study, which was published in the open-access journal Family Medicine and Community Health, suggests that ChatGPT may be able to deliver quick, unbiased insights drawn from data that can support conventional diagnostic techniques while maintaining anonymity and confidentiality.
In order to determine whether gender or socioeconomic class biases affected the suggested therapy approach for mild and severe major depression, the researchers examined the AI technology.
When asked, “What should a primary care physician suggest in this situation?” Prescription medications (for anxiety, depression, or sleep disorders), watchful waiting, referrals for psychotherapy, and referrals for psychotherapy in addition to prescribed medications were among the answers.
In accordance with clinical guidelines, the study found that only 4% of family physicians exclusively suggested referral for psychotherapy for mild cases. In contrast, ChatGPT-3.5 and ChatGPT-4 chose this course of action in 95% and 97.5 percent of cases, respectively.
The majority of doctors suggested either drug treatment just (48%) or psychotherapy along with prescription medication (32.5%). In extreme situations, the majority of physicians (44.5%) suggested both psychotherapy and prescription medication.
Additionally, in 67.5% of cases, the doctors prescribed a combination of antidepressants, anxiety medications, and sleeping pills, whereas ChatGPT’s was more likely than the doctors to prescribe antidepressants only in 74% of cases (version 3.5 and 68% in version 4).
Additionally, ChatGPT-3.5 (26%) and ChatGPT-4 (32) recommended combining antidepressants, anxiety medications, and sleeping aids.
“ChatGPT-4 demonstrated increased accuracy in therapy modification in accordance with clinical guidelines. Additionally, the researchers noted that they could not find any observable gender or socioeconomic level biases in the ChatGPT’s systems.
“ChatGPTs has the potential to enhance primary healthcare decision-making,” the study’s conclusion stated.
