How are you? Do people with inflammatory bowel disease experience response shift on this question?
© Mayo et al.; licensee BioMed Central. 2015
Received: 16 May 2014
Accepted: 6 March 2015
Published: 6 May 2015
As individuals experience changes in their health, they may alter the way they evaluate health and quality of life. The purpose of this study is to estimate the extent to which individuals with IBD change their rating of health over time because of response shift (RS).
This is a reanalysis of a population-based longitudinal study of IBD in Manitoba, Canada (n = 388). RS was examined using trajectories of the difference between observed and predicted health. Logistic regression and dual trajectories were used to identify predictors of RS.
Disease activity, vitality, pain, somatization, and physical and social function explained 51% of the variation in general health over two years with no evidence of RS in 82% of the sample. Negative RS was found for 8%, who initially rated health better than predicted; positive RS was found for 6%. The positive RS group was younger and had better baseline scores on measures of general health, hostility, pain, mental health and social and role function; less pain and better social function scores at baseline were predictors of negative RS.
In conclusion, the majority of people with IBD did not demonstrate a RS indicating that the health rating over time was stable in relation to that predicted by known time varying clinical variables. This adds to the evidence that the single question on self-rated health is useful for monitoring individuals over time.
Health-related quality of life (HRQL) is a broad concept that encompasses multiple domains, including physical, psychological, social, and emotional health, as well as general or global perceptions of health [1,2]. Longitudinal studies about HRQL and other patient-reported outcomes rest on the assumption that the respondent’s interpretation of a construct remains constant over time. Increasingly, HRQL measures are being used in clinical trials to evaluate the efficacy of new treatments or interventions, population-based surveys to describe population health, and chronic disease cohort studies to understand the impact of health challenges and treatment over the disease course. In longitudinal studies, a question of primary interest is: “Is there a significant improvement or deterioration in global HRQL and its domains?”
As individuals experience changes in their health, they may change the way that they evaluate their health and quality of life. Response shift (RS) is defined as a change in an individual’s internal standards (recalibration), values (reprioritization), or conceptualizations of health (reconceptualization) . It is theorized that RS occurs when individuals experience a significant health event (i.e., catalyst), such as a stroke, cancer treatment, or chronic disease diagnosis . When RS is present in HRQL, conventional statistical analyses may not reveal any evidence of an observed change even when a true change exists. Moreover, RS may not affect all HRQL domains or all samples, sub-samples, or individuals equally, which can complicate longitudinal data analysis.
Methods to detect RS include: (a) Design-based methods that collect supplementary data to estimate the magnitude of RS, (b) Interview or focus group methods that collect qualitative information about individuals’ experiences of RS, (c) Preference-based methods that collect comparative data about changes in the rank-order of domains, and (d) Model-based methods based on measurement error, multivariate, and longitudinal statistical models .
The most common design based method is the ‘then’ test . Individuals evaluate HRQL at the baseline occasion and then again at the post-test (or follow-up) measurement occasion. At the post-test, individuals are also asked to re-evaluate their HRQL at the pretest (baseline) assessment occasion. The difference between the original pretest score and the retrospective pre-test score estimates the magnitude of RS. The ‘then’ test has several limitations: it must be built in at the design stage, it can only be used to estimate recalibration, it is sensitive to recall bias and social acceptability bias, and it is problematic to use when there are multiple measurement occasions [7,8]. The other design based methods have similar challenges with respect to data collection and respondent burden.
Model-based methods include: (a) structural equation models (SEM) [9-13]; (b) random-effects models, (c) relative important measures, and (d) classification models. Of these methods, random-effects models are more easily applicable in the case of multiple time points. We previously  described a subject-specific model-based method of testing for RS in data collected at multiple occasions. This is a useful method when the pattern of change over time may be influenced by the course of progression of a health condition. The method focuses on an analysis of the pattern of residuals (the difference between predicted and observed HRQL scores) over time, classifying people into discrete RS groups. This method has been used in an inception cohort of individuals with stroke  and in a large volunteer study of individuals with prevalent multiple sclerosis (MS) . These two samples yielded, not surprisingly, different prevalences of response shift: 33% for stroke and 1% for MS. The high rates found in the stroke cohort are in keeping with the concept of RS resulting from a catalyst; incident stroke is a strong catalyst with a definitive onset. MS, which shares commonalities with other chronic diseases such as inflammatory bowel disease (IBD) and rheumatoid arthritis, develops over a longer time period and severity fluctuates. However, for the MS sample studied for RS, the timing of measurements was determined by the individual and not a study protocol. Thus, individuals may have chosen to report only when they were feeling well, and the picture of disease impact over time may have been minimized.
RS is an important consideration when evaluating changes in disease outcomes because it can result in an overestimating of problems or benefits in some individuals, and an underestimation in others. In a research context, these effects could cancel, yielding a conclusion of no change [16,17]. IBD is a chronic inflammatory disease with two distinct entities: Crohn’s disease and ulcerative colitis. However, regardless of the type of disease, over its course, individuals will experience changes in the intensity and severity of their symptoms. When there is increased inflammation, the disease is considered to be in an active stage and the individual will experience a flare-up of the condition, which may severely affect participation in work and social functions [18,19]. Symptoms include pain, fatigue and diarrhea; IBD is also associated with mastery, distress and lower psychological well-being [20,21]. When the degree of inflammation is low, the individual usually has mild symptoms or is symptom-free. Medical and surgical management has varying success, with up to half of individuals with IBD experiencing relapses every year. All of these aspects take a toll on health and quality of life of persons with IBD [20,22,23].
In order to accurately interpret IBD impact over time, from both the perspective of an individual and for group comparisons, it is important to have methods to identify RS. The area of IBD is understudied for this phenomenon.
The purpose of this study is to estimate the extent to which RS occurs among individuals with IBD, and to identify predictors of RS.
Data for this analysis is from the Manitoba IBD Cohort Study, a prospective longitudinal cohort study that is investigating the determinants of disease outcomes. Initiated in 2002, the Cohort Study includes persons from a population-based registry, which was established in 1995 at the University of Manitoba. The formation of this cohort is described in previous publications [20,21]. The present study is a reanalysis of the data, which consists of 388 respondents, 18 years of age or older and diagnosed within seven years prior to enrollment (mean 4.3 standard deviation 2.1), with data collected by mail questionnaire at 5 measurement occasions between study entry and 24-months. During this period, 34 respondents (8.8%) dropped out.
Respondents in the Manitoba IBD Cohort Study provided demographic information and information on a wide spectrum of health outcomes related to disease activity, physical and psychological symptoms, physical, role and social function, perceived health, and social support. Physical symptoms cover pain, fatigue and vitality, as well as those relating to the gastrointestinal system. Psychological symptoms include anxiety, mood, and somatic and systemic symptoms.
The outcome for this study is the single-item from the SF-36 General Health Perception (GHP) subscale: “In general would you say your health is”, with response options Excellent, Very Good, Good, Fair, Poor (EVGGFP). The other sub-scales of the SF-36 assess physical and social function, pain, vitality, mental health, and physical and emotional role impact, and were used as potential predictors of EVGGFP. All subscales are scored from 0 to 100, with higher values indicating better health. Some questions refer to the past four months, others, a typical day. The SF-36 has been widely validated as a measure of perceived health status [24,25]. Disease activity was assessed using the Manitoba IBD Index , a single-item index for symptom persistence based on the previous six-month period, shown to have good validity. IBD-specific predictors were selected from the Inflammatory Bowel Disease Questionnaire (IBDQ) : gastrointestinal symptoms, systemic problems (fatigue, energy, feeling unwell, sleep, weight), emotional dysfunction, and social difficulties during the last two weeks. The IBDQ is a commonly used and extensively validated measure of HRQL in IBD .
Negative psychological functioning was assessed using the Cohen Perceived Stress Scale (CPSS), which asks about feelings and thoughts during the last month. It is a validated tool for measuring the role of stress in disease [29,30].
The Brief Symptom Inventory (BSI) measures depressive symptoms from the past seven days in nine dimensions: somatization, obsession-compulsion, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychoticism .
The Multidimensional Scale of Perceived Social Support (MSPSS) , with 12 items, assesses the degree of support from family and friends, presumably currently.
The detection of RS is based on the patterns of residuals over time, which are the differences between predicted EVGGFP and the observed value for each respondent at each time period. These residual values were centered by subtracting each respondent’s mean residual from each time-specific residual to focus on change over time rather than on each respondent’s deviation from expected. Persons with fewer than three residuals were excluded from further analysis.
Nagin’s group-based trajectory method  was used to identify respondents considered likely to have experienced a RS based on the patterns of residuals over time. Large fluctuations in a respondent’s observed and predicted health over time suggest that a RS has occurred, while a pattern of consistent residuals over time (though not necessarily close to predicted) suggests no response shift. GBTA is a form of latent class analysis for continuous outcomes based on the assumption that the population is a mixture of distinct groups defined by common change over time, while recognizing uncertainty in group membership. RS trajectories are qualitatively classified based on their shape and direction, reflecting the change in the centered residuals.
Posterior probabilities, representing a respondent’s likelihood of belonging to each of the trajectory groups, are provided for each respondent for each trajectory. They are used to calculate theoretical and assigned proportions. The theoretical proportion is an expected value calculated as the sum of posterior probabilities for that trajectory group over all respondents; the assigned proportion is calculated from each respondent’s highest posterior probability. Similarity between the theoretical and assigned probabilities is considered to indicate good model fit, as are respondents having a high posterior probability for a single group (measured using means of the group with the highest posterior probability). Other indicators of model fit are based on the AIC and BIC statistics, where values closer to zero indicate better fit. Two BIC statistics are produced, one based on the number of respondents, the other on the number of observations, with the correct but incalculable BIC lying between the two. Coherence with theory and all fit indices are considered in determining the number of trajectory groups and the number of parameters per trajectory (intercept, linear, quadratic, etc.).
Four response shift patterns were hypothesized based on other work : none, positive, negative, or fluctuating. No response shift is likely when the centered residuals form a flat trajectory at 0: the difference between predicted and observed is consistent across time. Positive RS is deemed present when the magnitude of the residual increases over time in a direction of a more positive rating. This includes respondents who start with health ratings worse than predicted but who increase their rating over time to be better than predicted, as well as respondents who may begin better than expected but, over time, increase to a rating that is even more positive than expected. Negative RS is deemed present when respondents rate their health better than (or closer to) expected early on, but whose ratings decrease over time compared to expected.
To answer the question as to whether response shift depends on perceived health, we modeled the trajectories of RS conditional on the trajectories of EVGGFP using dual trajectory analysis .
To answer the question as to who is likely to experience a RS, we created two logistic regression models one to identify predictors of positive RS and one to predict negative RS. Although trajectory analysis is a probabilistic method, in that respondents receive a posterior probability of being in each trajectory group, respondents can be assigned deterministically to a specific group based on their highest probability. Odds ratios and 95% confidence intervals (CI) were estimated for each variable measured at baseline. With five levels and a reasonably normal distribution, EVGGFP can be modeled as continuous with little repercussion [14,15,35].
Ethics approval for the Cohort Study was obtained from the University of Manitoba Health Research Ethics Board.
Description of the cohort
Characteristics of IBD cohort study participants (n = 388)
N or Mean (SD)
Age at baseline: Mean (SD)
Age at diagnosis: Mean (SD)
Ulcerative colitis 
Education: High School or less/Trade School or Diploma/University
Working full time
Worked in last year
Missed work in last 6 months due to IBD
Ratings of health
SF-36 General health question (EVGGFP) at study entry and 2 years later among 289 (74%) with ratings at both measurement occasions
Rating at Cohort entry (n)
Rating at 2 years
Concordant (n = 144, 49.8% )
Increased (n = 78, 27.0% )
Decreased (n = 67, 23.2% )
Very good (90)
Disease activity, symptoms, functioning and social support at study entry and at one- and two-year follow-ups
Predictor [norm when available]
Year 1 follow-up
Year 2 follow-up
SF-36 subscales: scored 0–100, higher scores correspond to better HRQL
General health perception 
Physical function 
Role physical 
Bodily pain 
Social function 
Role emotional 
Mental health 
IBDQ subscales: scored 0 to 7, higher scores correspond to better HRQL
Brief symptom inventory: scaled 0 to 100, higher scores indicate worse symptoms
Somatization excluding nausea
MSPSS social support: higher scores correspond to greater support. Total is scored 12–84
Significant other (1–7)
CPSS Stress scale: scored 0–56, higher scores correspond to greater perceived stress
Predictive model of health
Best predictive model of EVGGFP over time
(30–85) vs (0–25)
SF-36 physical function
IBDQ social function
6.5 v 7
5.5–6.5 v 7
<4.5 v 7
Group-based trajectory model of residuals
Trajectory groups of RS categories
No RS detected
N (%) based on assignment
Range of centered residuals
Posterior probability of belonging to RS group (>0.7 considered good fit)
Observed versus predicted health
Dual trajectory analysis
Results from the dual trajectory analysis of EVGGFP modeled conditionally on RS group indicate that only four trajectories of health could be modeled jointly, with some small redistribution of people into trajectories. These are: three flat groups, one between poor and fair (20%) and two others at good (51%) and at VG (24%), and the remaining 5% of the sample with improving health ratings from F to VG over the two year period.
Predictors of RS
Baseline predictors of response shift
Negative (n = 31) v no RS (n = 294)
Positive (n = 23) v no RS (n = 294)
Odds ratio (95% CI) [per unit of predictor]
Mean (RS-) versus mean (RS 0 ) (pooled SD) p-value of the difference in means
Odds ratio (95% CI) [per unit of predictor]
Mean (RS-) versus mean (RS 0 ) (pooled SD) p-value of the difference in means
Age at diagnosis: older age at diagnosis reduces the risk of RS+
0.96 (0.93, <1.00)
30 v 38 (14.8) p = 0.0216
BSI Hostility: more hostility symptoms (score ≥35) at baseline increases the risk of RS+
3.63 (1.11, 11.94)
17% v 5% scored above 35 P = 0.0477^
SF-36 (OR based on a 10-pt difference)
Higher GH at baseline reduces risk of RS+
41.6 v 55.5 (21.5) p = 0.0072
Lower pain at baseline reduces risk for RS- and RS+
51.1 v 60.3 (2.4)
45.5 v 60.3 (24.9) p = 0.0080
p = 0.0593
Better MH at baseline reduces risk of RS+
0.76 (0.60, 0.95)
59.6 v 68.7 (17.3) p = 0.0152
Better social function at baseline reduces the risk of RS- and RS+
58.9 v 69.3 (27.7) p = 0.0465
57.6 v 69.3 (27.9) p = 0.0533
Better role physical at base-line reduces the risk of RS+
56.8 v 70.1 (29.8) p = 0.0398
This study found no evidence of RS in 82% of the cohort on one of the most frequently asked questions in health. This suggests that this is a good question to ask over time in this population.
Positive RS was experienced by 6% of the IBD sample and negative RS by 8% over a two-year period. These RS rates are lower than were found over one year in a sample of incident stroke (14% positive and 15% negative), but higher than were found in a volunteer study of prevalent MS. While the diagnosis of a chronic disease such as IBD would be expected to be a sufficient catalyst to initiate a RS, it is not surprising that stroke, an event with a very precise onset that can produce serious health consequences in the blink of an eye, would be a stronger catalyst, resulting in more RS, even over a shorter time period. Also, while this IBD cohort was established within seven years of diagnosis (mean years since diagnosis = 4, SD = 2), it may be that RS occurs early and was missed in some respondents. Both the positive and negative RS found in this study occurred by 6 months of follow-up. We previously detected almost no RS in a group with prevalent MS . It was impossible to determine in that sample whether that was because no RS had occurred or whether it was just not detected because it had occurred prior to the measurement occasions, or even because of arbitrary measurement timing, chosen by the volunteer.
It is not possible, using this method, to identify which type of RS (reconceptualization, reprioritization, or recalibration) occurred and it would take qualitative work to untangle what drove the residual value. In our previous work , we found support for recalibration because of validation with the then-test.
RS was found to be associated with health perception. Several other predictors of RS were identified. Worse baseline pain and social function scores were found to be associated with the presence, but not direction, of RS: RS was more likely among those in either the positive or negative RS groups compared to those without RS. Those identified with positive response shift were also more likely to be younger, as well as having worse mental health, role physical functioning and general health perception scores and more hostility symptoms at baseline. These results are concordant with the finding from Lix et al. using discriminative analysis and logistic regression; this method identified pain and social function as predictors of reprioritization RS .
The EVGGFP question was selected as the outcome rather than the general health subscale for reasons of predictors to evaluate RS. While reports of symptoms, functioning, and perceived stress and social support may be good predictors for the rating of heath [20,22,23], they would not be expected to predict some of the other constructs in the GH subscale such as the extent to which people expect their health to worsen.
Foundational work on the meaning of self-rated health [36,37] indicates that people do not use the same frame of reference when answering the global health status item. Most commonly the rating is influenced by physical health including physical function and general physical condition; presence of positive health behaviours and absence of negative behaviours also play role as do health comparisons with social group. This early research has also shown that the frame of reference used did not depend on the person's rating of their health. In longitudinal studies, it would be relevant to ask whether the person’s frame of reference changes over time, however, change in frame of reference is a mechanism of response shift . Further research on the propensity for people to make a response shift is warranted.
A limitation of this analysis is that all predictors were obtained by self-report and could themselves have been influenced by RS. Although measured variables were included as part of the Cohort Study data collection, none were observed at all measurement occasions. While fixed predictors may increase the variation in the outcome explained by the predictive model, they do not contribute to the determination of RS as they do not contribute to discrepancies over time between observed and predicted scores.
This is not the first study to identify the presence of RS in individuals with IBD. Previous research to develop a new method of detecting reprioritization RS  also used data from the IBD Cohort Study, at the baseline and six month measurement occasions. RS was identified in some of the SF-36 subscales.
These results, combined with the current ones, suggest that some individuals with chronic conditions that are associated with exacerbations and remissions of symptoms over time are susceptible to RS. These flares in activity may serve as catalysts for RS . However, even individuals not currently experiencing disease exacerbations have been shown to be susceptible to RS , perhaps because their expectations for high symptom burden were not realized.
In conclusion, the majority of people with IBD did not demonstrate a RS indicating that the health rating over time was stable in relation to that predicted by known time varying clinical variables. This adds to the evidence that the single question on self-rated health is useful for monitoring individuals over time .
- Berzon R, Hays RD, Shumaker SA. International use, application and performance of health-related quality of life instruments. Qual Life Res. 1993;2(6):367–8.View ArticlePubMedGoogle Scholar
- Revicki DA. Health-related quality of life in the evaluation of medical therapy for chronic illness. J Fam Pract. 1989;29(4):377–80.PubMedGoogle Scholar
- Sprangers MA, Schwartz CE. Integrating response shift into health-related quality of life research: a theoretical model. Soc Sci Med. 1999;48(11):1507–15.View ArticlePubMedGoogle Scholar
- Schwartz CE, Sprangers MA. Reflections on genes and sustainable change: toward a trait and state conceptualization of response shift. J Clin Epidemiol. 2009;62:1118–23.View ArticlePubMedGoogle Scholar
- Schwartz CE, Ahmed S, Sawatzky R, Sajobi T, Mayo N, Finkelstein J, et al. Guidelines for secondary analysis in search of response shift. Qual Life Res. 2013;22(10):2663–73.View ArticlePubMedGoogle Scholar
- Schwartz CE, Bode R, Repucci N, Becker J, Sprangers MAG, Fayers PM. The clinical significance of adaptation to changing health: A meta-analysis of response shift. Qual Life Res. 2006;15:1533–50.View ArticlePubMedGoogle Scholar
- Nolte S, Elsworth G, Sinclair A, Osborne R. Tests of measurement invariance failed to support the application of the "then-test". J Clin Epidemiol. 2009;62:1173–80.View ArticlePubMedGoogle Scholar
- Brossart DF, Clay DL, Willson VL. Methodological and statistical considerations for threats to internal validity in pediatric outcome Data: Response Shift in Self-Report Outcomes. J Pediatr Psychol. 2002;27(1):97–107.View ArticlePubMedGoogle Scholar
- Schmitt N, Pulakos E, Lieblein A. Comparison of three techniques to assess group-level beta and gamma change. Appl Psychol Meas. 1984;8:249–60.View ArticleGoogle Scholar
- Schmitt N. The use of analysis of covariance structures to assess beta and gamma change. Source. Multivariate Behavioral Research 1982. 1982;17(3):343–58.View ArticleGoogle Scholar
- Oort FJ, Visser MRM, Sprangers MAG. An application of structural equation modeling to detect response shifts and true change in quality of life data from cancer patients undergoing invasive surgery. Qual Life Res. 2005;14:599–609.View ArticlePubMedGoogle Scholar
- Oort FJ. Using structural equation modeling to detect response shifts and true change. Qual Life Res. 2005;14(3):587–98.View ArticlePubMedGoogle Scholar
- Barclay-Goddard R, Lix LM, Tate R, Weinberg L, Mayo NE. Response shift was identified over multiple occasions with a structural equation modeling framework. J Clin Epidemiol 2009.Google Scholar
- Mayo NE, Scott SC, Dendukuri N, Ahmed S, Wood-Dauphinee S. Identifying response shift statistically at the individual level. Qual Life Res 2008.Google Scholar
- Ahmed S, Mayo N, Scott S, Kuspinar A, Schwartz C. Using latent trajectory analysis of residuals to detect response shift in general health among patients with multiple sclerosis. [corrected]. Qual Life Res. 2011;20(10):1555–60.View ArticlePubMedGoogle Scholar
- Mayo NE, Nadeau L, Ahmed S, White C, Grad R, Huang A, et al. Bridging the gap: the effectiveness of teaming a stroke coordinator with patient's personal physician on the outcome of stroke. Age Ageing. 2007;37(1):32–8.View ArticlePubMedGoogle Scholar
- Mayo NE, Scott SC, Ahmed S. Case management poststroke did not induce response shift: the value of residuals. J Clin Epidemiol. 2009;62(11):1148–56.View ArticlePubMedGoogle Scholar
- Li Y, Rapkin B. Classification and regression tree uncovered hierarchy of psychosocial determinants underlying quality-of-life response shift in HIV/AIDS. J Clin Epidemiol. 2009;62(11):1138–47.View ArticlePubMed CentralPubMedGoogle Scholar
- King-Kallimanis BL, Oort FJ, Visser MR, Sprangers MA. Structural equation modeling of health-related quality-of-life data illustrates the measurement and conceptual perspectives on response shift. J Clin Epidemiol. 2009;62(11):1157–64.View ArticlePubMedGoogle Scholar
- Graff LA, Walker JR, Clara I, Lix L, Miller N, Rogala L, et al. Stress, coping, distress and health perceptions in inflammatory bowel disease and community controls. Am J Gastroenterol. 2009;104:2959–69.View ArticlePubMedGoogle Scholar
- Graff LA, Walker JR, Lix L, Clara I, Rawsthorne P, Rogala L, et al. The relationship of inflammatory bowel disease type and activity to psychological functioning and quality of life. Clin Gerontol. 2012;4:1491–501.Google Scholar
- Walker JR, Ediger JP, Graff LA, Greenfeld JM, Clara I, Lea L. The Manitoba IBD cohort study: a population-based study of the prevalence of lifetime and 12-month anxiety and mood disorders. Am J Gastroenterol. 2008;103:1989–97.View ArticlePubMedGoogle Scholar
- Lix LM, Graff LA, Walker JR, Clara I, Rawsthorne P, Rogala L, et al. Longitudinal study of quality of life and psychological functioning for active, fluctuating, and inactive disease patterns in inflammatory bowel disease. Inflamm Bowel Dis. 2008;14(11):1575–84.View ArticlePubMedGoogle Scholar
- Ware Jr JE, Kosinski M, Keller SD. SF-36 Physical & Mental Scales: A User's Manual. Boston, Massachusetts: The Health Institute, New England Medical Center; 1994.Google Scholar
- Ware JE, Keller SD. Interpreting general health measures. In: Spilker B, editor. Quality of Life and Pharmacoeconomics in Clinical Trials. 2nd ed. Philadelphia: Lippincott-Raven; 1996. p. 445–60.Google Scholar
- Clara I, Lix LM, Walker JR, Graff LA, Miller N, Rogala L, et al. The Manitoba IBD Index: evidence for a new and simple indicator of IBD activity. Am J Gastroenterol. 2009;104:1754–63.View ArticlePubMedGoogle Scholar
- Guyatt GH, Mitchell A, Irvine EJ, Singer J, Williams N, Goodacre R, et al. A new measure of health status for clinical trials in inflammatory bowel disease. Gastroenterology. 1989;96:804–10.PubMedGoogle Scholar
- Sainsbury A, Heatley RV. Review article: Psychosocial factors in the quality of life of patients with inflammatory bowel disease. Aliment Pharmacol Ther. 2005;21:499–508.View ArticlePubMedGoogle Scholar
- Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav. 1983;24:385–96.View ArticlePubMedGoogle Scholar
- Cohen S, Williamson GM. Perceived stress in a probability sample of the Unisted States. In: Spacapan SS, Oskemp S, editors. The social psychology of health. Sage; 1988:31–67.Google Scholar
- Derogatis LR. Administratoin, scoring and procedures manual for the brief symptom inventory. 4th ed. Minneapolis, MN: National Computer Systems; 1993.Google Scholar
- Zimet GD, Dahlem NW, Zimet SG, Farley GK. The multidimensional scale of perceived social support. J Pers Assess. 1988;52:30–41.View ArticleGoogle Scholar
- Singer JD, Willett JB. Applied Longitudinal Data Analysis Modeling Change and Event Occurrence. New York: Oxford University Press; 2003.View ArticleGoogle Scholar
- Nagin DS. Group-Based Modeling of Development. Cambridge, Mass.: Harvard University Press; 2005.Google Scholar
- Bollen KA, Barb KH. Pearson’s R and Coarsely Categorized Measures. Am Sociol Rev. 1981;46:232–9.View ArticleGoogle Scholar
- Measuring Functioning and Well-Being: The Medical Outcomes Study Approach. Durham and London: Duke University Press; 1992.Google Scholar
- Krause NM, Jay GM. What do global self-rated health items measure? Med Care. 1994;32(9):930–42.View ArticlePubMedGoogle Scholar
- Rapkin BD, Schwartz CE. Toward a theoretical model of quality-of-life appraisal: Implications of findings from studies of response shift. Health Qual Life Outcomes. 2004;2:14.View ArticlePubMed CentralPubMedGoogle Scholar
- Lix LM, Sajobi TT, Sawatzky R, Liu J, Mayo NE, Huang Y, et al. Relative importance measures for reprioritization response shift. Qual Life Res. 2013;22(4):695–703.View ArticlePubMedGoogle Scholar
- King-Kallimanis BL, Oort FJ, Garst GJA. Using structural equation modelling to detect measurement bias and response shift in longitudinal data. Adv Stat Anal. 2010;94:139–56.View ArticleGoogle Scholar
- Rosenzveig A, Kuspinar A, Daskalopoulou SS, Mayo NE. Toward patient-centered care: a systematic review of how to ask questions that matter to patients. Medicine (Baltimore). 2014;93(22):e120.View ArticleGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.