JMIR Publications

ABSTRACT

Background:

Stress is a key determinant of health outcomes and may influence work performance. Questionnaire-based assessments of stress are typically broad and retrospective.. Daily stress measurements via smartphones offer more granular, real-time data but suffer from adherence issues. Using an already established communication medium (WhatsApp) and a more conversational style assessment might improve adherence and help collect more detailed insights in (work) stress, underlying stressors and countering energy sources.

Objective:

This study focuses on the usability and feasibility of semi-automated voice- and text-messages (with and without emojis) via WhatsApp as a method to collect daily data on experienced work stress, stressors and energy sources.

Methods:

A sample of 210 online-recruited workers participated in a 10-workday diary study using semi-automated WhatsApp messages to rate daily stress, stressors and energy sources. Questions (with and without emojis) were presented by a chatbot as text messages with clickable buttons (multiple-choice questions; MC) or with instructions to answer with either a voice or a text message. The study employed an experimental design with 4 groups: 1) week 1 voice, week 2 text/MC with emojis; 2) week 1 voice, week 2 text/MC without emojis; 3) week 1 text/MC , week 2 voice with emojis; 4) week 1 text/MC , week 2 voice without emojis. Pre- and post-study online questionnaires assessed demographics, familiarity with voice messages, and usability including participants’ preference. Open answers were coded using artificial intelligence (AI) and number of stressors or energy sources were compared across the three collection methods (MC, voice and text messages) to determine if the amount and quality of information collected differs per method within subjects.

Results:

158 workers completed at least 80% of scheduled conversations. The sample was predominantly female (81%), highly educated (82%), and a slight majority worked part-time (52%). Mean adherence to daily schedule was very high (M=95%). The post-questionnaire revealed a strong preference for MC and text over voice messages, mostly due to ease and convenience in a variety of situations. The number of stressors per week was approximately three times higher in the MC-condition than in the voice condition, even though average stress levels per week did not differ significantly within participants. The number of energy sources was comparable between open answers in the voice and text condition, but voice messages consisted of more words.

Conclusions:

Collecting (work) stress data via semi-automatic WhatsApp messages is a feasible method with low effort for participants. Usability ratings indicated a strong preference among participants for MC and text messages over voice messages. Future research should explore usability in more diverse samples and in direct comparison to traditional assessment methods.