Teknologi Kesehatan Mental Digital Berbasis Kecerdasan Buatan: Tinjauan Bukti Biomedis dan Psikologis

Abstract

The global prevalence of depression and anxiety increased substantially, placing significant pressure on conventional mental health services. Although digital mental health technologies expanded access to care, many systems remained static and lacked adaptive personalization. Artificial intelligence was introduced to enhance real-time monitoring, personalization, and automated decision-making within digital interventions. This study aimed to systematically evaluate empirical evidence regarding psychological and biomedical outcomes of AI-powered digital mental health technologies and to synthesize methodological patterns influencing clinical applicability. A systematic review was conducted following PRISMA 2020 guidelines. Literature searches were performed in PubMed, Scopus, and SpringerLink for studies published between 2020 and 2026. Studies were included if they implemented learning-based artificial intelligence as a core intervention component and reported quantitative psychological or biomedical outcomes. Sixteen studies met eligibility criteria. The findings indicated that 62.5% of included studies employed randomized controlled trial designs. Fourteen of sixteen studies reported statistically significant short-term reductions in depression or anxiety symptoms. Supervised machine learning and natural language processing were the most frequently applied approaches. Biomedical indicators such as heart rate variability and sleep metrics were typically used as secondary exploratory outcomes rather than primary endpoints, and external validation of artificial intelligence models was uncommon. AI-powered digital mental health systems demonstrated consistent short-term psychological benefits; however, biomedical integration and long-term validation remained limited. Strengthening methodological rigor and multimodal integration was necessary to enhance clinical applicability