Chapter 3 88 is relatively short compared to the expected long-term use of GAHT. During the first year of GAHT, people experience physical changes, but also changes in their social environment, such as changes in support from loved ones or confrontation with discrimination and stigmatization (Collet, Kiyar, et al., 2023). It is likely that getting used to both the physical and social changes during this transition may take more than just one year, and longer followup is needed. Second, there may be selection bias in participant inclusion and follow-up, possibly impacting our findings. The exclusion of individuals with diagnosed sleep disorders in the RESTED study could lead to underestimation of insomnia symptoms. Furthermore, participants with mental health issues might be less likely to participate. However, when comparing the psychiatric history of the current study participants, we see that 47% of our current participants had a lifetime mood disorder, compared to 60% of participants in Heylens et al. (2014) and and 34.5% in Defreyne et al. (2019), showing no indications of selection bias from psychiatric history. There also is loss to follow up in the study, and GAHT users who experienced more or less depression-related side effects could have been more likely to drop out. Addressing this bias, we compared IDS-SR score severity and baseline IDS-SR scores between participants who were and were not lost to follow-up, as displayed in Supplementary Table 5. These comparisons reveal no significant differences between the groups, indicating no selective loss to follow-up. Nonetheless, the loss of power due to loss to follow-up should be taken into account in the interpretation of the study results, and further replication is necessary. Third, as shown in the EFA results, not all items from the IDS-SR could be included in the resulting subscales due to low factor loadings. An EFA analysis is based on associations between questionnaire items, and low factor loadings are generally seen in items that lack strong associations with the other questionnaire items. This can be caused by low variability in the item, lack of specificity in the item or lack of associations with other items. For this study, we opted to exclude items with factor loadings below 0.4, to obtain more valid and reliable subscales, but this also led to the exclusion of numerous items from the subscales.
RkJQdWJsaXNoZXIy MTk4NDMw