MATHematics RESEARCHERS ANALYZE SUICIDE PREVENTION HELP CALLS DURING PANDEMIC

Stichting 113 Suicide Prevention helps people that are in mental distress and have thoughts about suicide through supporting conversations via telephone and chats. PhD student Salim Salmi at the Mathematics Department in the Analytics and Optimization group, analyzed conversations between Stichting 113 caregivers and clients before and during the COVID-19 pandemic, together with experts from Stichting 113, CWI and Vrije Universiteit Amsterdam (Rob van der Mei and Sandjai Bhulai). The research was partly funded by the COVID-19 programme of ZonMW.

01/21/2021 | 1:03 PM

The research shows that people need a conversation during the corona crisis, but find it difficult to talk about feelings in their own environment. Since the start of the corona crisis, men have suffered more from anxiety and young people experience less self-confidence.

Goal of the research

The main goal of the research was to determine whether the subject of help calls has changed under influence of the COVID-19 pandemic. The most common conversation topics during this period were mapped as well. The researchers also wanted to gain insight in differences between conversations of different target groups (young/old, man/woman, living alone). The study analyzed 8,589 chat conversations: 5,179 conversations prior to the corona crisis and 3,410 conversations after the introduction of government measures for corona. For the analysis, specific Machine Learning and Natural Language Processing techniques were used.

Machine Learning method BERT

The machine learning method Bidirectional Encoder Representations from Transformers (BERT) was used to find topics in the chats. This model converts texts into numerical representations that can be used for calculations. This means that each chat message is converted to a numeric representation and clustered. For each cluster the researchers used the corresponding chat messages to arrive at a number of words that describe the group. For this, a selection was made of the 2000 most used words. With help of the so-called term frequency–inverse document frequency (TF-IDF) statistic the most important words were determined. With this selection of words and accompanying chat messages, the researchers found a topic of conversation for each cluster, for example ‘panic and anxiety’, ‘suicide thoughts’, and ‘friendship’.

Conclusions

In fourteen percent of the requests for help, "Corona" was a topic of discussion, after introduction of corona measures. Problems experienced by help seekers that mentioned corona were more loneliness, fear, little distraction, less assistance, threat of unemployment and drug abuse.

Most of the changes in topics of conversation in the 113 Suicide Prevention Helpline, following the implementation of the Corona measures, are as expected. Worrying changes are the increase in panic and anxiety in men, the increase in lack of confidence in oneself or others among younger people, and the increase in suicide plans among people seeking help who live alone.

It is especially striking that gratitude for the conversation itself as well as for listening to the person in distress have increased after the introduction of the Corona measures, especially among men and youth. This means that the 113 Suicide Prevention helpline seems to meet the need for contact with these groups of clients. Fortunately, no increase in the total number of suicides in the Netherlands has been observed over the period from March 2020 up to and including November 2020. In order to ensure that the number of suicides does not rise in 2021, it is crucial to stay alert and discuss feelings together.

More information