En santé, le besoin de quantifier des phénomènes subjectifs a émergé au cours du dernier siècle avec par exemple, l'évaluation de la qualité de vie (QdV), de la fatigue ou de l'anxiété. L'intérêt croissant face à ces mesures de santé perçue appelées "Patient-Reported Outcomes" (PRO) apparaît notamment dans les maladies chroniques. Afin de pouvoir analyser l'évolution de la QdV des patients ou d'autres types de PRO au fil du temps, des données longitudinales sont collectées.Ces données issues de PRO sont difficiles à appréhender car les patients peuvent percevoir et interpréter différemment les questions qui leur sont posées au cours du temps selon l’évolution de leur maladie, en termes de signification, priorités et retentissement sur leur vie personnelle. Ce phénomène, appelé « response shift » (RS), est souvent lié à la façon dont les patients s’adaptent à leur maladie. En présence de RS, les évolutions observées des patients peuvent ne pas refléter correctement les véritables changements ressentis par ceux-ci. L’objectif de ce travail porte sur le développement et l’évaluation de deux approches l’une, basée sur les modèles à équations structurelles (SEM) et l’autre, sur la théorie de réponse à l’item (IRT), avec l’originalité de travailler au niveau de l’item. Ces méthodes permettent la détection et l'estimation du RS ainsi que sa prise en compte lors d’analyses.Les performances des approches ont été évaluées en les appliquant sur des données réelles mais également au moyen d'études de simulations. Les résultats obtenus grâce aux simulations permettent de mieux identifier quelle approche méthodologique adopter selon les contextes. Les avantages et inconvénients de l'utilisation de chacune des méthodes, sont soulignés et des recommandations ont été proposées. / During the last century, many studies in health attempt to measure important characteristics, such as Health Related Quality of Life (QoL) using Patient-reported outcomes (PRO). QoL and other perceived health measures (pain, fatigue, etc.) are increasingly used in chronic diseases. In order to analyse the evolution of patients’ QoL or other types of PRO over time, longitudinal data are collected.These PRO data are difficult to interpret due to the patients’ changing in the standards, values, or conceptualization of what the PRO is intended to measure (eg: QoL). This phenomenon is referred to as “response shift” (RS) and is often linked to the way the patients might adapt or cope with their disease experience. As a consequence of RS, observed patient’s evolutions may thus not properly reflect true perceived health changes.The objective of this work is to develop and assess two approaches at item level, one, based on structural equation modelling (SEM) and other, on item response theory (IRT). These statistical approaches are currently used for RS analysis on PRO data. Both the identification of RS occurrence and its appropriate adjustment in the analyses of longitudinal PRO data are studied.The performance of IRT and SEM were assessed by applying them on real data and through simulations studies. The results obtained in the simulation allow choosing which methodological approach should be applied according to the context.The pros and cons of using one method or another as well as the potential of using both of them as complementary analyses providing different insights into the field of response shift are be underlined.

Objective: Patient-reported outcomes (PROs) are measures collected from a patient to determine how he/she feels or functions in regards to a health condition. Longitudinal PROs, which are collected at multiple occasions from the same individual, may be affected by response shift (RS). RS is a change in a person’s self-evaluation of a target construct. Latent variable models (LVMs) are statistical models that relate observed variables to latent variables (LV). LVMs are used to analyze PROs and detect RS. LVs are random variables whose realizations are not observable. Factor scores are estimates of LVs for each individual and can be estimated from parameter estimates of LVMs. Factor scoring methods to estimate factor scores include: Thurstone, Bartlett, and sum scores. This simulation study examines the effects of RS on factor scores used to test for change in the LV means and recommend a factor scoring method least affected by RS. Methods: Data from two time points were fit to three confirmatory factor analysis (CFA) models. CFA models are a type of LVM. Each CFA model had different sets of parameters that were invariant over time. The unconstrained (Uncon) CFA model had no invariant parameters, the constrained (Con) model had all the parameters invariant, and the partially constrained (Pcon) model had some of the parameters invariant over time. Factor scores were estimated and tested for change over time via paired t-test. The Type I error, power, and factor loading (the regression coefficient between an observed and LV) and factor score bias were estimated to determine if RS influenced the test of change over time and factor score estimation. Results: The results depended on the true LV mean. The Type I error and power were similar for all factor scoring methods and CFA models when the LV mean was 0 at time 1. For LV mean of 0.5 at time 1 the Type I error and power increased as RS increased for all factor scores except for scores estimated from the Uncon model and Bartlett method. The biases of the factor loadings were unaffected by RS when estimated from an Uncon model. The factor scores estimated from the Uncon model and the Bartlett and sum scores method had the smallest factor score biases. Conclusion: The factor scores estimated from the Uncon model and the Bartlett method was least affected by RS and performed best in all measures of Type I error, statistical power, factor loading and factor score bias. Estimating factor scores from PROs data that ignores RS may result in erroneous (or biased) estimates.

Problem: Many individuals with chronic illnesses such as stroke and ongoing activity limitations report self-perceived health related quality of life (HRQL) that is similar to that of healthy individuals. This phenomenon is termed response shift (RS). RS describes how people change: internal standards in assessing HRQL (recalibration), values (reprioritization), or how they define HRQL (reconceptualization), after an event such as stroke. Changes in HRQL post stroke may be inaccurate if RS is not taken into account. Increased knowledge of RS may affect the way in which HRQL measures are used, both clinically and in research. The overall objective was to assess RS in construct specific HRQL models post stroke: physical function, mental health, and participation. Methods: Data were analysed from the longitudinal study “Understanding Quality of Life Post-Stroke: A Study of Individuals and their Caregivers”. Six-hundred and seventy- eight persons with stroke at 1, 3, 6, and 12 months post stroke participated. Generic and stroke specific HRQL measures were collected. Descriptive analysis was completed with SAS, and identification of RS utilized structural equation modeling with LISREL. Results: Mean age of participants was 67 years (SD 14.8), and 45% were female. RS was identified in mental health using a framework which was developed for identifying RS statistically with multiple time points. RS was also identified in physical function where it had not been expected, possibly due to the self perceived nature of the response options. The effect size of change in physical function was affected by the presence of RS. The timing of RS in mental health and physical function was primarily around the 12 month time period, and predominantly recalibration RS. RS was also identified in participation. Conclusions: The framework that was developed was useful in identifying RS and incorporated important issues such as multiple testing and validation of the model. The presence of RS affects measurement of HRQL constructs post stroke; recalibration RS can be measured clinically with specific methods to account for RS. RS should also be measured in research studies to ensure accurate measurement of change. Future research should evaluate additional models in stroke and other populations. / October 2008

Problem: Many individuals with chronic illnesses such as stroke and ongoing activity limitations report self-perceived health related quality of life (HRQL) that is similar to that of healthy individuals. This phenomenon is termed response shift (RS). RS describes how people change: internal standards in assessing HRQL (recalibration), values (reprioritization), or how they define HRQL (reconceptualization), after an event such as stroke. Changes in HRQL post stroke may be inaccurate if RS is not taken into account. Increased knowledge of RS may affect the way in which HRQL measures are used, both clinically and in research. The overall objective was to assess RS in construct specific HRQL models post stroke: physical function, mental health, and participation. Methods: Data were analysed from the longitudinal study “Understanding Quality of Life Post-Stroke: A Study of Individuals and their Caregivers”. Six-hundred and seventy- eight persons with stroke at 1, 3, 6, and 12 months post stroke participated. Generic and stroke specific HRQL measures were collected. Descriptive analysis was completed with SAS, and identification of RS utilized structural equation modeling with LISREL. Results: Mean age of participants was 67 years (SD 14.8), and 45% were female. RS was identified in mental health using a framework which was developed for identifying RS statistically with multiple time points. RS was also identified in physical function where it had not been expected, possibly due to the self perceived nature of the response options. The effect size of change in physical function was affected by the presence of RS. The timing of RS in mental health and physical function was primarily around the 12 month time period, and predominantly recalibration RS. RS was also identified in participation. Conclusions: The framework that was developed was useful in identifying RS and incorporated important issues such as multiple testing and validation of the model. The presence of RS affects measurement of HRQL constructs post stroke; recalibration RS can be measured clinically with specific methods to account for RS. RS should also be measured in research studies to ensure accurate measurement of change. Future research should evaluate additional models in stroke and other populations.

Problem: Many individuals with chronic illnesses such as stroke and ongoing activity limitations report self-perceived health related quality of life (HRQL) that is similar to that of healthy individuals. This phenomenon is termed response shift (RS). RS describes how people change: internal standards in assessing HRQL (recalibration), values (reprioritization), or how they define HRQL (reconceptualization), after an event such as stroke. Changes in HRQL post stroke may be inaccurate if RS is not taken into account. Increased knowledge of RS may affect the way in which HRQL measures are used, both clinically and in research. The overall objective was to assess RS in construct specific HRQL models post stroke: physical function, mental health, and participation. Methods: Data were analysed from the longitudinal study “Understanding Quality of Life Post-Stroke: A Study of Individuals and their Caregivers”. Six-hundred and seventy- eight persons with stroke at 1, 3, 6, and 12 months post stroke participated. Generic and stroke specific HRQL measures were collected. Descriptive analysis was completed with SAS, and identification of RS utilized structural equation modeling with LISREL. Results: Mean age of participants was 67 years (SD 14.8), and 45% were female. RS was identified in mental health using a framework which was developed for identifying RS statistically with multiple time points. RS was also identified in physical function where it had not been expected, possibly due to the self perceived nature of the response options. The effect size of change in physical function was affected by the presence of RS. The timing of RS in mental health and physical function was primarily around the 12 month time period, and predominantly recalibration RS. RS was also identified in participation. Conclusions: The framework that was developed was useful in identifying RS and incorporated important issues such as multiple testing and validation of the model. The presence of RS affects measurement of HRQL constructs post stroke; recalibration RS can be measured clinically with specific methods to account for RS. RS should also be measured in research studies to ensure accurate measurement of change. Future research should evaluate additional models in stroke and other populations.

This study is a prospective longitudinal outcome study investigating the presence of response shift in disease and generic functional outcome measures in 105 patients undergoing spinal surgery. The then-test method which compares pre-test scores to retrospective pre-test scores was used to quantitate response shift. There was a statistically significant response shift for the Oswestry Disability Index (ODI) (p=0.001) and the Short Form-36-PCS (p=0.078). At three months, seventy-two percent of patients exhibited a response shift with the ODI. Fifty-six and 21 percent of patients exhibited a response shift with the SF-36 physical and mental component scores respectively. When accounting for response shift and using the minimal clinically important difference, the success rate of the surgery at 3 months increased by 20 percent. The presence of response shift has implications for the measurement properties of standard spinal surgery outcome measures including the effect size of treatment and the number of responders to treatment.

This study is a prospective longitudinal outcome study investigating the presence of response shift in disease and generic functional outcome measures in 105 patients undergoing spinal surgery. The then-test method which compares pre-test scores to retrospective pre-test scores was used to quantitate response shift. There was a statistically significant response shift for the Oswestry Disability Index (ODI) (p=0.001) and the Short Form-36-PCS (p=0.078). At three months, seventy-two percent of patients exhibited a response shift with the ODI. Fifty-six and 21 percent of patients exhibited a response shift with the SF-36 physical and mental component scores respectively. When accounting for response shift and using the minimal clinically important difference, the success rate of the surgery at 3 months increased by 20 percent. The presence of response shift has implications for the measurement properties of standard spinal surgery outcome measures including the effect size of treatment and the number of responders to treatment.

Introduction: After a brain injury there are often long term consequences impacting on QoL. However, this is a complex issue influenced by many factors. As someone recovers and adjusts it is likely that the way in which they evaluate QoL will also change. The theory of response shift suggests people will change the way they evaluate QoL in the face of changes in their life. The aim of this thesis is to investigate what influences a QoL judgement; examining the possibility of response shift. Methods: Quantitative and qualitative methods were used in 4 studies. These were a cross-sectional design utilising an individualised QoL measure (SEIQoL-DW); a longitudinal study utilising a ‘then-test’ approach; a cross-sectional questionnaire study; and a qualitative study using Interpretative Phenomenological Analysis. Study 1 (Ch.3) Results: Correlations between the QoL measures confirm the validity of the SEIQoL-DW; however, correlations were generally stronger for the simpler Hadorn Scale. There was little overall change in mean QoL when current and retrospective judgements were compared. There was evidence for a change in what areas of life were considered most important to QoL following injury. Study 2 (Ch.4) Results: Improvements in reported QoL between baseline and follow-up were small. A then-test indicates that any effect of response shift is small, and non-significant in the current research. There was also little evidence for reprioritisation or re-conceptualisation. Examination of other factors associated with QoL suggest that brain-injury specific factors (BIGI, RBANS) play a role in predicting QoL. Study 3 (Ch.5) Results: QoL was reported as worse post-injury on both Hadorn’s scale and the QOLIBRI-OS; a difference that was more pronounced on the QOLIBRI-OS. Differences were also reported in the importance of different areas of functioning. Change in QoL as measured by the QOLIBRI-OS was significantly influenced by disability as measured by the GOSE, emotional and informational support, and upwards social comparison. Optimism as measured by the LOT, but not upwards social comparison was a significant predictor of change on Hadorn’s scale; GOSE and emotional and informational support remain significant predictors. The GOSE, emotional and informational support, emotional coping styles and optimism were significant predictors of current QoL on the QOLIBRI-OS; and emotional and informational support and optimism were significant predictors of QoL on Hadorn’s scale. Little evidence was found to suggest that the factors proposed in Sprangers and Schwartz’s (1999) model of response shift have predicted relationships with QoL. Two candidate variables were studied: optimism and social support. However neither showed the predicted pattern of relationships. Nonetheless the study supports previous work indicating an influence of optimism and social support on QoL, and indicates that these warrant further study. There were systematic difference between current and retrospective ratings of importance of domains. The level of importance given to the areas of life defined by the QOLIBRI-OS is higher after injury than before, with the exception of “personal and social life” for which there is no significant difference. The areas of life chosen to reflect that which is measured by the GOSE (“work”, “close relationships”, and “social and leisure activities”) are rated as less important with the exception of “close relationships”. These findings provide further support for the idea that QoL domains are re-evaluated after brain injury. Study 4: This was an in depth qualitative investigation of the experience of recovery and adjustment following TBI. Semi-structured interviews and Interpretative Phenomenological Analysis (IPA) were used. Interviews were conducted with 4 men who were 3, 7, 12, and 18 years post injury. Main Outcome and Results: Themes emerging from the analysis were ‘Change: In Self and World’; ‘Reaching a point of realisation’; ‘Support’; ‘Adjusting to change/Coping with day to day life’; and ‘Participation, Goals and Focus’. These themes cover how participants felt both they and their lives had changed as a consequence of their injury; ways they went about coping and adjusting to changes; the importance of support; and the significance of social integration and participation in feeling satisfied with life. Summary and Conclusions: These studies provide evidence for response shift in different ways. There is little evidence for recalibration but there is some indication that reprioritization or reconceptualization may take place. Changes in how important different areas of life are before and after injury suggest that participants are changing the way they view and make evaluations of QoL. Factors identified as being important to QoL judgements were disability, social support (emotional and informational support identified in the questionnaire study and support in the IPA), upwards social comparison, and optimism. The IPA study suggests that functional outcome and participation are important after TBI; while also identifying ways of coping and providing an insight into the experience of recovery from brain injury. The different QoL measures used provides both evidence for their validity, but also evidence for the different conceptualisations of QoL that are measured by different instruments. The findings have implications both for understanding the QoL of the individual and for research on QoL after TBI.

Problem: Older adults may change their view on what is important to their health-related quality of life (HRQOL). They may alter their opinion about areas relevant to HRQOL (reconceptualization), or how important these areas are to them (reprioritization), and this can be referred to as response shift (RS). Overtime, changes in HRQOL may be imprecise (underestimated or overestimated) if RS occurs. Providing detailed information about RS in the older adult population will have many implications for health professionals, family members, caregivers, policy makers, and researchers. The purpose was to explore RS in HRQOL in community-dwelling older men. Methods: Data from the Manitoba Follow-up Study (MFUS) was used as 360 older men returned the Successful Aging Questionnaire in each of five years (2007-2011). The participants identified the importance of 15 items, which reflect the physical (2 items), mental (5 items), and social domains (8 items) of HRQOL. Descriptive analysis was performed using SPSS21. An individualized method was used to identify different aspects of RS at group and individual levels, as well as the item level. Predictors of RS were also identified using logistic regression in a one-year period. Results: Mean age of participants was 89.7 years (SD 2.9) in 2011. Across 15 items over a one-year period, RS varied from a low of 9.3% for the ‘being mentally aware’ item to 39.3% for the ‘having goals/making plans’ item. Because we were examining RS of 15 items, it was very uncommon to find older men with no RS on all items. Only 27 out of 360 older men (7.5%) provided the same response on all the items they answered at both times (2010-2011). The average of the percent of people showing RS over 15 items, across four time periods, within three domains, was 24.4%. Reprioritization was more common in physical and mental domains, respectively, whereas reconceptualization was seen mainly in the social domain. Further, most of those who showed reprioritization, showed a decrease in importance, while most of those who showed reconceptualization, dropped a concept. Older men who were older, married, living independently, and recently did not participate in activities, were more likely to show RS in certain items. Older men with lower self-rated health were less likely to show RS. Conclusions: Data from the MFUS presents an opportunity to assess RS by using an individualized method that is simple to conduct and interpret in research and clinical settings. This method provides extensive demonstration of RS including magnitude, timing, type, direction, and predictors. RS should be considered an important part of aging, when planning resources and individualizing interventions for the older adult population. Future studies should design a method that evaluates RS individually, similar to our method. / October 2015

Abstract Purpose Aim of this exploratory study was to investigate whether a retrospective assessment of oral health-related quality of life (OHRQoL) using the Oral Health Impact Profile (OHIP) is susceptible to bias such as implicit theory of change and cognitive dissonance. Methods In this prospective clinical study, a sample of 126 adult patients (age 17–83 years, 49% women) requiring prosthodontic treatment was consecutively recruited. The OHRQoL was assessed using the 49-item OHIP at baseline and at follow-up. Additionally, patients were asked at followup to retrospectively rate their oral health status at baseline (retrospective pretest or then-test) and the change in oral health status using a global transition question. Furthermore, patients’ ratings of overall oral health and general health were used as validity criteria for the OHRQoL assessments. Response shift was calculated as the difference between the initial and retrospective baseline assessments. Results Baseline and retrospective pretest did not differ substantially in terms of internal consistency and convergent validity. Response shift was more pronounced when patients perceived a large change in OHRQoL during treatment. Retrospective pretests were more highly correlated with the baseline than with the follow-up assessment. Conclusion Findings suggest that retrospective assessments of OHRQoL using the OHIP-49 are susceptible to bias. Cognitive dissonance is more likely to appear as a source of bias than implicit theory of change.:Inhaltsverzeichnis 1. Einführung in die Thematik…………………………………………………2 2. Formatierte Publikation……………………………………………………..11 3. Zusammenfassung…………………………………………………………. 19 4. Literaturverzeichnis……………………………………………………….... 24 5. Anlagen 5.1. Darstellung des eigenen Beitrags……………………………………. 27 5.2. Selbstständigkeitserklärung…………………………………………... 28 5.3. Lebenslauf……………………………………………………………… 29 5.4. Publikationen…………………………………………………………… 30 5.5. Danksagung……………………………………………………………. 31