Computer-based Stress Measurement

Stress is a major risk factor for the development of physical and mental illnesses, including hypertension, depression and anxiety disorders.
To study stress and its effects, psychologists use classical stress induction paradigms (stress tests) such as the Trier Social Stress Test (TSST), which are cost- and time-intensive. As scalable solutions to induce stress are still missing, we have developed a Digital Stress Test (DST).
Previous methods for the (early) detection of acute stress reactions mostly focus on measuring subjective stress or individual parameters of the physiological reaction, such as heart rate.
In various studies, we record the non-verbal behaviour of test persons together with hormonal, cardiovascular (e.g. heart rate) and subjective markers in form of questionnaires.
Based on these multimodal data, we develop algorithms for automatic stress detection and validate them in natural environments.
In the long term, we hope to be able to objectively and individually measure and feedback acute stress in an everyday setting and thus enable better handling of individual stressors.

This research was funded by the "Empathische Künstliche Intelligenz" (EKI) grant, FKZ 01IS20046 from 01.12.2020-30.11.2024.

Theoretical considerations

Valuable insights into the pathophysiology and consequences of acute psychosocial stress have been gained using standardized stress induction experiments[1,2]. Additionally, controlled stress induction procedures play an important role for the development of objective stress detection methods as they strongly rely on high-qualitative and representative datasets obtained through stress induction experiments[3].

However, most protocols are limited to laboratory settings, labor-intensive and cannot be scaled to bigger cohorts nor transferred to daily life scenarios.

The DST leverages the abundance of smartphones in daily lifes and fits principles of classical stress paradigms (e.g. TSST, MIST, IMPRESS [4,5,6]) into a scalable web-application.

Elements of social-evaluative threat and uncontrollability [7] build the core of the two implemented tasks and are further enhanced by stress-inducing framings.

The potentials of a digital stress paradigm for the standardized induction and recording of acute stress responses are manyfold. First, general stress effects, and individual stress reactivity, could be studied outside the lab and with larger sample sizes. This would allow the effects to be reproduced, compared, and adapted for different cohorts (e.g. stress in back pain patients) and contexts (e.g. work-related stress) from any internet-connected location. Stress experiments might be performed to study the effect of stress prevention (e.g. exercise) and intervention (e.g. meditation) strategies in daily life settings and with individuals from diverse cultural, ethnical and geographical backgrounds.

Secondly, the recording of multiple individual controlled stress responses could be used to further analyse and develop stress-detection algorithms.

Procedure

The whole procedure takes place on the screen of the participants’ smartphone and does not take longer than 5-8 minutes. A presentation version without any data saving can be tried out on www.digitalstresstest.org. The stress induction paradigm consists of several framings as well as a mental arithmetic and a verbal answering task:

• Framing:

The DST is introduced as a research tool that tests the participants’ individual cognitive-verbal performance and analyzes their behavior recorded through the front-cameras of their devices. Framings can be adjusted for specific study designs.

• Mental Arithmetic Task:

In the mental arithmetic task, the participants solve arithmetic tasks of varying difficulty. The difficulty exceeds the ability of the participants by shortening the time provided for solving the single tasks. Other elements such as negative feedback, negative social comparison and live broadcasting of their own video intensify the psychosocial stress.

• Verbal Answering Task:

In the verbal task of the DST, the participants respond orally to standardised questions on personally relevant topics. The psychosocial stress is again increased by "stressful" elements in the design and the feedback, as well as by the active video broadcast.

Control – DST (C-DST)

We also developed a control version of the DST that resembles its structure and procedure but differs regarding the stress induction elements. Participants get informed that they take part in a control condition and that no videos will be recorded. The arithmetic tasks is designed to be easier and the verbal answering scenarios are chosen to be less critical. Additional stressful design elements like in the DST are left out on purpose.

A presentation version without any data saving can be tried out on www.digitalstresstest.org/control. In randomized online study with 284 participants, we could already show that participants of the DST manifested significantly higher perceived stress indices compared to C-DST participants [8}. A subsequent validation study including physiological markers is pre-registered (https://osf.io/xwu9a) and currently on-going.

Development

DST and C-DST have been developed as single page web applications using the JavaScript framework React.js. The apps are designed to run in standard browsers and are hosted on public university servers. In order to start the apps, access from mobile devices is required. Both versions can be used in German or English. Design, wordings and tasks can be adapted according to study needs.

Additionally, we have constructed a secure technical infrastructure for saving data on a separate virtual machine and have received ethical approval for our data storage concept.

References

[1] Dedovic K, D’Aguiar C, Pruessner JC. What stress does to your brain: a review of neuroimaging studies. Can J Psychiatry. 2009;54(1):6-15. doi:10.1177/070674370905400104

[2] Chen X, Gianferante D, Hanlin L, et al. HPA-axis and inflammatory reactivity to acute stress is related with basal HPA-axis activity. Psychoneuroendocrinology. 2017;78:168-176. doi:10.1016/j.psyneuen.2017.01.035

[3] Mahesh B, Hassan T, Prassler E, Garbas J. Requirements for a Reference Dataset for Multimodal Human Stress Detection. In: 2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops). ; 2019:492-498. doi:10.1109/PERCOMW.2019.8730884

[4] Kirschbaum C, Pirke KM, Hellhammer DH. The ’Trier Social Stress Test’--a tool for investigating psychobiological stress responses in a laboratory setting. Neuropsychobiology. 1993;28(1-2):76-81. doi:10.1159/000119004

[5] Dedovic K, Renwick R, Mahani NK, Engert V, Lupien SJ, Pruessner JC. The Montreal Imaging Stress Task: using functional imaging to investigate the effects of perceiving and processing psychosocial stress in the human brain. J Psychiatry Neurosci. 2005;30(5):319-325.

[6] Fehlner P, Bilek E, Harneit A, et al. Neural responses to social evaluative threat in the absence of negative investigator feedback and provoked performance failures. Human Brain Mapping. 2020;41(8):2092-2103. doi:https://doi.org/10.1002/hbm.24932

[7] Dickerson SS, Kemeny ME. Acute stressors and cortisol responses: a theoretical integration and synthesis of laboratory research. Psychol Bull. 2004;130(3):355-391. doi:10.1037/0033-2909.130.3.355

[8] Norden M, Hofmann A, Meier M, Balzer F, Wolf O, Böttinger E, Drimalla H Inducing and Recording Acute Stress Responses on a Large Scale With the Digital Stress Test (DST): Development and Evaluation Study J Med Internet Res 2022;24(7):e32280 URL: https://www.jmir.org/2022/7/e32280 DOI: 10.2196/32280