Improving the Assessment of Depression and Anxiety in Children and Adolescents using Natural Language Processing

Research problem and specific questions The project is developing an alternative response format and method for assessing mental illness in children and adolescents.

The project addresses the need for assessments based on Longitudinal Expert All Data (LEAD), a method that is considered the best reference for assessing diagnostic precision, as well as the need to offer time-efficient support to facilitate clinicians´ work.

The solutions currently used in clinical practice have been criticized for being blunt or too time-consuming.Data and methodThe project mainly uses data in the form of written or spoken word answers.

The data is analyzed with the latest technologies in Artificial Intelligence: natural language processing and machine learning.

The project develops age-appropriate instructions, questions and response formats for children, young adolescents and older adolescents.

LEAD studies are carried out, where the participants are followed longitudinal and where senior experts review the data to reach the optimal reference for diagnostic precision. Based on this, we develop and test LEAD algorithms for each age group. Finally, a randomized experimental study is conducted as a basis for implementation.

The study investigates the type of training and data required to facilitate clinicians´ decision-making and create confidence in the results of the AI analysis.Plan for project realizationThe project is submitted by an international, interdisciplinary team with experienced researchers in psychology, medicine and computer science.

The steering group includes relevant professional clinics (doctors, psychologists) as well as user and interest organizations. Processes for co-creation are established early and permeate all parts of the research process. All participants are well positioned within their respective networks.

The total cost of 5.8 MSEK covers salaries, co-creation processes, carrying out the LEAD studies and dissemination of results.RelevanceThere is a great need to develop accurate and cost-effective solutions for identifying mental illness among children and young people.

Digitization and cross-sectional approaches where technology from computer science enables increased precision and streamlining of the assessment is crucial for the care process.

The solution also enables an effective increase in knowledge of which treatment works for which individual, which can seriously revolutionize the care of mental illness for children and young people.