Self-Reported Outcomes (SRO) such as anxiety or quality of life are frequently assessed with questionnaires to better understand the subjective experience of individuals and to inform healthcare decision-making. Faced with a major life event (e.g. COVID-19 crisis), people may change their perception of the SRO and of the questionnaire, a phenomenon called response shift (RS). If RS is not accounted for, the estimation of longitudinal changes in SRO may be biased and the clinical interpretation may be obfuscated. Furthermore, RS may provide more insight into changes in SRO in the face of a major event and may be linked with psychological adaptation. Most of RS analyses based on statistical methods consider two measurement occasions (before and after a major event). Yet, changes of SRO over time are often assessed over multiple time points. In this context, the trajectory of the construct is continuous by essence and RS could also be envisioned as a continuous process. Furthermore, the exact timing of assessments may also vary across individuals even if visits are usually planned. Thus, the timescale of the study has often to be treated as continuous rather than discrete with equally spaced time points. RS analyses also generally assume that most individuals experience RS in the same way regardless of individual characteristics. However, it is likely that RS may occur for some individuals only, in different ways and at different times. As such, trajectories of RS are likely to be heterogeneous within a sample. Last, RS detection methods most often perform RS analysis at the domain level using subscale scores combining several items. These methods cannot distinguish which items are specifically affected by RS and domain-level RS analysis might not always appropriately reflect what is going on at the item level, especially if RS has opposite effects depending on the item. Thus, item-level RS analyses would give more insight on RS and on the interpretation of psychological adaptation. Methods able to detect and adjust for item-level RS over multiple time points in studies with continuous time while accounting for heterogeneity in trajectories of RS are lacking. The RESCUE project aims to develop a method for RS analysis at the item level in longitudinal SRO studies across multiple time points. RS has never been defined nor operationalized in a continuous timescale. We plan to define RS over multiple time points in a multidisciplinary framework. Rasch models have shown good performances for item-level RS detection between two time points with heterogeneous RS. RS over multiple time points will be operationalized in an appropriate continuous time model, an innovative combination of linear mixed models and Rasch models. An item-level RS detection method using this continuous-time Rasch model will then be proposed. It will aim at investigating RS in depth in order to detect RS, examine items expected to be affected by RS, type and size of RS effects. To ensure that the proposed method for RS analysis is valid and reliable, different simulation studies will be performed all along the project to assess the performances of the method. The method will be automated in a module of a statistical software. To overcome practical issues related to the analysis of real datasets, we will apply the proposed method to study the changes and RS in anxiety and depression of healthcare workers of comprehensive cancer centers during the Covid-19 pandemic. This application will also illustrate how RS analyses allow getting more insight into the adaptive processes in face of a health crisis. A work on graphical representations of results of longitudinal SRO changes and RS analysis will aim at providing understandable and meaningful reports of the results to promote knowledge translation. Indeed, a survey will assess data visualization preferences and understanding of the results in a multidisciplinary framework.