A Systematic Review of the impact of Response Shift on Reporting Patient-Reported Outcomes (PRO) Change over Time in Randomized Clinical Trials
Author(s)
Shahraz S1, Mohebi F2, Khanmohammadi S3, Nolte S1
1ICON PLC, SAN FRANCISCO, CA, USA, 2Berkley University, BERKLEY, CA, USA, 3TEHRAN UNIVERSITY, TEHRAN, Iran (Islamic Republic of)
OBJECTIVES To explore the existence and magnitude of the potential impact of response shift on patient-reported outcomes (PRO) change over time in randomized controlled trials (RCTs). METHODS We included RCTs in which the measurement of response shift in at least one PRO for any two time points was reported. PubMed, CINAHL, EMBASE, and PsycINFO databases were explored (English only, no time limit). We assessed the risk of bias regarding the adequacy of subgroup analysis and assessment times, sample size in general and across time points, follow-up time, the PRO selection, and sensitivity to the measurement method. RESULTS We included 20 trials involving 5542 participants having a diverse set of medical conditions and evaluated various parameters, including the type of condition and its severity, length of follow-up, type of intervention, type of the PRO instruments, analysis methods, the existence and impact of response shift on PRO change. Response shift was reported across time points in 17 of 20 studies. The most commonly used methods to measure response shift were structural equation models and then-test, each used in 8 studies. The longest and most common time point (6 of 20 papers) was 12 months, at which the possibility of response shift was assessed. The invariance due to demographic factors was evaluated in three papers. The most frequent reported limitation was small sample size, indicated in 9 papers. CONCLUSIONS Across the identified studies, the occurrence of response shift was indicated as a nontrivial factor confounding the interpretation of PRO change over time. Therefore, we highly recommend the assessment of response shift as part of psychometric evaluation of PROs in RCTs. In contrast, if response shift effects are not assessed, there is a risk of invalid inferences of results obtained through patient self-report.
Conference/Value in Health Info
2021-05, ISPOR 2021, Montreal, Canada
Value in Health, Volume 24, Issue 5, S1 (May 2021)
Code
PMU65
Topic
Clinical Outcomes, Health Policy & Regulatory, Organizational Practices, Patient-Centered Research
Topic Subcategory
Approval & Labeling, Best Research Practices, Clinical Outcomes Assessment, Patient-reported Outcomes & Quality of Life Outcomes
Disease
No Specific Disease