What is problematic about binary questions on gender in health surveys – a missing answer analysis

Background: In many studies, participants who do not state their gender are excluded from the analysis. This may be appropriate if they do not answer the questionnaire seriously. However, some participants may have understandable reasons for not reporting their gender, e.g. questioning their gender identity. Research question: How many students and which students do not answer the question on gender? Methods: We analyzed data of the Health Behaviour in School-aged Children study from Ireland, France, Hungary, Scotland, Belgium (Flemish) and Luxembourg (n = 40,053). To explore the reasons for non-response, we divided the participants into 3 groups: 1. Responders answered both socio- demographic questions (age and gender) 2. age non-responders did not answer the question on age. 3. Gender non-responders answered the question on age, but not the one on gender. Results: 311 out of 40,053 (0.8%) pupils aged 11–18 did not report their gender. About 40% of them did not answer the age question either. However, the other 60% belong to the group of gender non-responders and this group is disadvantaged compared to responders: they report lower self-rated health, more health complaints, less family support and more substance use (alcohol, tobacco, cannabis). 1.9% of pupils did not answer the question about age. These age non-responders answered the questionnaire more selectively overall and skipped more questions. Conclusion: The data suggest that the reasons for age non-response and gender non-response are different. For age non-responders, the fear of de-anonymization seems to be the reason for not indicating their age. Not answering the question on gender is rare. If the participants answered the question on age, but not the question on gender, then the variable gender is missing not at random. The health problems of gender non-responders correspond to the health problems of gender non-conforming adolescents. Thus, the question arises if the group of gender non-responders should be included in the analysis and if the question on gender should be asked differently in the future