Helplines for mental healthcare differ from other call centers in various aspects. Many agents are volunteers, the conversations are often more complex and emotional, and many helplines use a triage system. In this paper, we first propose a call center model that includes the specifics of online mental health helplines, including features such as a triage system for chats and service times consisting of a warm-up, conversation, and wrap-up cool-down periods. The model is validated using a trace-driven simulation based on real-life (anonymous) data provided by 113 Suicide Prevention. The results show that the model can simulate the waiting-time performance of the helpline accurately. Second, we focus on forecasting the number of chats and telephone calls. Our results show that (S)ARIMA models trained on historical data perform better than other models in the case of short-term forecasting (five weeks or less ahead), while using linear regression works best for long-term forecasts (longer than five weeks).

, , , , ,
International Journal on Advances in Life Sciences
Het dectecteren van (corona-gerelateerde) problemen bij de hulpvragers van 113 Zelfmoordpreventie voor en de uitbraak van de coronapandamie in Nederland middels state-of-the-art Natural Language Processing (NLP)
Stochastics

de Boer, T., Mérelle, S., Bhulai, S., & van der Mei, R. (2022). Performance modeling for call centers providing online mental health support. International Journal on Advances in Life Sciences, 14, 120–129.