Algorithmic biases in mental health diagnoses and their impact on vulnerable populations: a documentary review of advances and challenges

Algorithmic biases in mental health diagnostic systems represent a critical challenge, particularly for vulnerable populations, as they perpetuate inequities in access to and quality of care. This article aims to analyze advances and challenges in identifying and mitigating these biases through a documentary review of Spanish and English articles indexed in Scopus between 2018 and 2022. The methodology involved a systematic analysis of 50 selected studies, classified into four thematic areas: types of algorithmic biases, clinical impact on vulnerable populations, technical limitations in algorithm development, and proposed mitigation strategies. The results demonstrate that biases are deeply rooted in training data and the unequal representation of marginalized groups, leading to less accurate diagnoses for women, racialized communities, and low-income individuals. Although technical and ethical approaches have been proposed, gaps persist in their practical implementation. The study concludes that without multidisciplinary intervention integrating public health, ethics, and data science perspectives, algorithms will continue to reproduce structural inequalities. This research underscores the urgency of inclusive policies and robust regulatory frameworks to ensure equity in digital mental health.