The project team succeeded in screening treatment records and occupational health survey answers automatically to identify factors that predict a mental health diagnosis or the prolongation of the treatment period. Factors predicting a diagnosis included, in particular, intensive stress, repeated fatigue, sadness and being female. The approach can be used as a new kind of predictive tool for mental health challenges at the population level.
The deterioration of employees’ mental health can be predicted with machine learning methods. In the Finnish Institute of Occupational Health’s research project, both the first mental health diagnosis and the prolongation of treatment could be predicted fairly well with the aid of machine learning.
In occupational health survey answers, seven factors were identified as increasing the likelihood of receiving a diagnosis related to a mental health problem or a sleep disorder during the two-year monitoring period.
Factors predicting the deterioration of mental health included intensive stress, repeated fatigue and exhaustion during the day, sadness and melancholy as well as anxiety symptoms. In addition self-assessed well-being, the person’s sex is also a key factor. On the basis of the project, women have an increased likelihood of receiving a diagnosis.
“We identified factors that predict a diagnosis among more than one hundred questions. The predictive ability of these seven questions was nearly as good as that of the entire set of questions,” says Pekka Varje, Research Manager at the Finnish Institute of Occupational Health.
Younger employees are more likely to receive one of the diagnoses covered in the study but the age was not among the strongest predictive factors.
Modelling not suitable for planning the treatment of an individual
The results of the “Predictive methods for improving sustainability of mental well-being at work” research project served as a basis for the publication of a predictive map for mental health diagnosis in the Work-Life Knowledge Service. A tool like the predictive map could serve as an aid in the planning of occupational health care resources and in preventive mental health work, for example.
“At the level of individuals, the modelling contains too much uncertainty. It is possible that this kind of tool may help identify risk groups or development trends more broadly in the working-age population,” says Research Professor Ari Väänänen.
The prolongation of treatment can be predicted on the basis of treatment records
In the second part of the project, the researched topic was the prolongation of the mental health treatment period. Treatment records written by physicians were analysed automatically to identify what kind of topics predicted a treatment period of over four physician visits.
The prolongation of the treatment period was predicted especially by treatment record entries related to depression, its medical treatment and exhaustion. However, the key predictive factor was a diagnosis received at the beginning of the treatment period. Certain diagnoses related to depression and anxiety disorder predicted a longer treatment period in occupational health care.
Machine learning makes it possible to analyse massive datasets
The analyses of the project were based on machine learning methods. Two very extensive research populations were monitored for several years. The populations, consisting of working-age people, represented different sectors and occupational groups comprehensively.
“Machine learning methods and advanced computing performance have made it possible to use research approaches that apply big data. As recently as ten years ago, this kind of research could not be carried out,” says Pekka Varje.
Predictive methods for improving sustainability of mental well-being at work
- Predictive map of mental health diagnosis in the Work-Life Knowledge Service: Factors that predict a mental health diagnosis | Work-life knowledge service | www.worklifedata.fi (tyoelamatieto.fi)
- Analysis page in the Work-Life Knowledge Service: Mental health diagnosis prediction chart – background and interpretation | Work-life knowledge service | www.worklifedata.fi (tyoelamatieto.fi)
- Final report of the project (in Finnish): Kohti kestävämpää mielen hyvinvointia työssä : Koneoppiminen ja mielenterveystapahtumien ennakointi (julkari.fi)(Improving sustainability of mental well-being at work – machine learning and the prediction of mental health-related events)
- Project: Predictive methods for improving sustainability of mental well-being at work (ENNAKKO) | Finnish Institute of Occupational Health (ttl.fi)
- The project was funded by the Finnish Work Environment Fund. The co-operative partners of the project were Terveystalo and the University of Helsinki.
Source: The Finnish Institute of Occupational Health (FIOH)