Abstract
In recent years, the number of people using antidepressants in Taiwan has continued to rise, but clinical identification and intervention strategies for the risk of suicidal behavior in the "early post-drug discon tinuation stage" are still insufficient. This study intends to integrate artificial intelligence technology and meta-analysis methods to explore the risk of suicidal behavior in patients within three months after dis continuation of antidepressants, and establish a prediction model that can be applied clinically. Antide pressants have a clear effect on the treatment of major depression and related mood disorders. However, recent studies have pointed out that the short-term stage after discontinuation of antidepressants, espe cially within three months after discontinuation, may be accompanied by an increasing trend of suicidal behavior risk. This study aims to evaluate the changes in suicide risk during the discontinuation period through a systematic literature meta-analysis and an artificial intelligence (AI) risk prediction model, and establish an early warning model that can be applied to clinical decision-making. This study was con ducted in three stages: First, a systematic literature review and meta-analysis was conducted to compile research data related to suicidal behavior after antidepressant discontinuation over the past 20 years; second, antidepressant use and discontinuation cases were collected through the Taiwan National Health Insurance Database (NHIRD), and training and validation were performed based on variables such as medical history, age, gender, comorbidity, medication type, and discontinuation pattern; finally, a predic tion model was established using machine learning algorithms (such as XGBoost, Random Forest, Logis tic Regression) to screen out high-risk group characteristics and perform risk stratification. The expected results will provide a set of highly accurate clinical auxiliary judgment tools to improve the efficiency of follow-up and psychological support intervention after discontinuation of medication, and serve as a reference for psychiatric outpatient clinics and public health policies. A systematic review of 200 related articles from 2020 to 2025 was conducted to screen randomized controlled trials (RCTs), cohort studies, and case-control studies that showed a correlation between discontinuation and suicidal behavior, and a total of 40 quality-qualified articles were included for integrated analysis.The definition of suicidal be havior includes suicidal ideation, suicide attempt, and completed suicide. The AI model trains multiple al gorithms (including random forest, XGBoost, LSTM, etc.) based on these data, and imports some predic tion parameters (age, gender, comorbidity, medical history, drug type and dosage, psychotherapy record, etc.) from the Taiwan National Health Insurance Database. The results of the meta-analysis showed that within three months after discontinuation of medication, the risk of suicidal behavior increased signifi cantly compared with those who continued to take medication, with a combined odds ratio (OR) of 1.78 (95% CI: 1.43–2.21). The overall accuracy of the AI prediction model reached 87%, among which the sensitivity of the XGBoost model in identifying high-risk suicide was 0.81 and the specificity was 0.89. The important predictors were: past suicide attempt record, rapid dose change within two weeks before dis continuation, no psychotherapy, young male population, and comorbid anxiety disorder. The conclusion is that the risk of suicide does increase in the short term after discontinuation of antidepressant drugs. Unplanned discontinuation should be avoided in clinical practice, and follow-up and psychological sup port should be strengthened. The AI model constructed in this study has high accuracy and early warning potential. It is recommended that it can be incorporated into the electronic medical record system in the future for real-time risk assessment as a decision-making aid in psychiatric outpatient clinics. Furthermore, depression is caused by the lack of dopamine secretion and the lack of dopamine enzyme secretion, which leads to the interruption of dopamine secretion. In order to make the enzyme secretion normal, in addition to nutritional therapy, iron and B vitamins are given to supplement tyrosine; MAO-B inhibitors are supplemented to enhance the dopamine enzyme pathway. Moreover, some antidepressants directly supplement dopamine enzymes such as Tyrosine Hydroxylase, and indirectly regulate the con centration or mechanism of action of neurotransmitters to restore dopamine secretion to normal. Such molecular medical mechanism of action enables us to understand that normal enzyme function can lead to normal dopamine secretion; this helps the production of new antidepressant drugs; further, it is a great blessing for patients with dementia and severe depression in psychiatry; and dopamine is a happiness factor, and its normal secretion can also prevent cancer; in addition to benefiting the public, it is also a big step forward in molecular psychiatry.
To Read or Download the Article PDF