Factors that predict a mental health diagnosis

Factors that predict mental health diagnoses can be found in occupational health care data with machine learning models. Review the probability of being diagnosed with a mental health issue within different population groups. There are three indicators for subjective well-being, and background variables related to gender and age.

Description

The mental health diagnosis prediction chart is based on analyses that, on the basis of sociodemographic background factors and responses to a well-being at work survey, predicted the probability of 18–64-year-old occupational health care clients to receive their (first) mental health diagnosis within two years from taking the survey. The data is based on the answers of 13,883 employees who took a survey of well-being at work. During the monitoring period, 24 per cent of them received a diagnosis related to mental health. The mental health diagnoses included in the data covered 1) mood disorders, 2) neurotic, stress-related and somatoform disorders, 3) non-organic sleep disorders and 4) job burnout. The pseudonymized data were analysed with the XGBoost classification model by using the responses to the well-being at work survey, diagnosis data and sociodemographic background factors.

The data is related to the Predictive methods for improving sustainability of mental well-being (Paremmalla ennakoinnilla kestävämpään mielen hyvinvointiin) research project co-ordinated by Finnish Institute of Occupational Health and funded by Finnish Work Environment Fund and Finnish Institute of Occupational Health. The project was carried out between 1 February 2020 and 31 July 2022. The co-operative partners of the project were Terveystalo and the University of Helsinki.

What the indicators indicate

Based on the analyses, five factors that best predicted a mental health diagnosis with nearly as good a classification ability as all potential factors used in the analyses were selected. These five factors were about age and gender and questions of perceived stress, sadness and daytime tiredness.

The chart can be filtered by gender or age group. Each figure represents a matrix with perceived sadness on the X axis and perceived stress on the Y axis. Different figures indicate alternative responses to the question about daytime sadness. Light grey-blue indicates a low likelihood for mental health diagnosis, whereas red suggest a high likelihood.

It is important to pay attention to the fact that the colours of the map depict the average risk on a population level, but the map cannot be used to predict mental health on an individual level. The colours depict things such as how different mental health risks are unevenly distributed within different population groups. Mental unwellness can be depicted with indicators other than occupational health care diagnoses. Therefore, the map has its limitations, but it aims to increase awareness of the dimensions relevant to the mental health of employed people.