JMIR Publications

ABSTRACT

Neurocognitive, affective, and psychosocial health in older adults requires specialized focus due to the interplay of multiple comorbidities, reduced social interactions, and age-related physical decline. Traditional methods for early detection and diagnosis of these disorders rely on screening tools, surveys, interviews, assessments, and brain imaging. Emerging artificial intelligence (AI) approaches have been developed to assist clinicians with screening, diagnosis and intervention, for example, analyzing patterns in electronic health records (EHR), speech recognition, and digital phenotyping. This scoping review aims to describe the current landscape of AI applications in neurocognitive, affective, and psychosocial health for older adults. Specifically, we examine publication trends over the past decade, characterize the AI technologies and models being applied, and assess how clinical tasks such as detection, identification, diagnosis, prediction, monitoring, and treatment benefit from AI technologies. We followed a process that includes formulating the research question, defining eligibility criteria, searching for literature, screening against eligibility criteria, extracting and coding data, and synthesizing findings. Inclusion criteria include two databases (Scopus and PubMed) for papers published between January 1st, 2015, and July 31st, 2025, discussing neurocognitive, affective, and psychosocial health issues in older adults (≥ 65 years) and AI-based applications. Data extraction focused on psychological conditions (e.g., dementia, depression, anxiety), AI techniques, and the clinical context of AI usage. Descriptive statistics and thematic analysis were used to summarize results. We identified 268 relevant publications. There is a clear increasing trend in the number of papers per year, with a notable spike around 2020–2022 corresponding to surging interest in AI and the introduction of GenAI tools. A variety of methodologies were identified, with a more utilization of “shallow” machine learning models such as Support Vector Machine (SVM), Random Forest, and Logistic Regression than “deep” learning models like XGBoost, CNN, or LSTM during the time period studied. AI techniques were predominantly used for screening and diagnosis tasks, such as early detection of cognitive impairment and automated classification of dementia or depression, with relatively fewer studies focusing on interventions or monitoring. Reported model performance was generally high, with the majority of studies achieving good accuracy and area under the curve values. The literature demonstrated a growing body of literature focused on enhancing the neurocognitive, affective, and psychosocial health care of older adults through technology applications. AI shows promise for improving early identification of at-risk individuals and aiding diagnosis through analysis of complex data, which could enable timelier and more personalized interventions. However, most applications to date emphasize diagnostic prediction and risk assessment, whereas AI-driven intervention and monitoring tools remain scarce. AI applications are likely to become an increasingly integral part of mental health practice, but careful implementation and oversight will be required to ensure these tools augment rather than replace person-centered care.