Endocrine Abstracts

Introduction: Some methodologists have raised concern about the validity of the standard method of assessing patient-reported outcomes, i.e. retrospective questionnaires. Repeated momentary measurements have been introduced to avoid recall bias and provide ecological validity. Although having theoretical advantages, the measurement properties remain unsubstantiated. This study examines the relationship between the retrospective thyroid-related quality of life patient-reported outcome measure ThyPRO and a momentary version of ThyPRO.

Methods: Eighty-three hyperthyroid patients were included to answer questions about their thyroid-related quality of life. Twelve momentary items from four multi-item scales were administered thrice daily via a smartphone application during 28 days. On day 28, the original retrospective ThyPRO was administered, thus the same period was covered by two different measurement methods, enabling head-to-head comparison. Correlations, differences, and levels of agreement were examined across all four scales. Further, it was explored if the two most substantiated forms of recall bias were evident (the peak effect and the end effect).

Results: Retrospective and averaged momentary ThyPRO ratings were highly correlated with Pearson correlations of 0.74–0.88. However, retrospective ratings provided significantly higher results, i.e. worse quality of life, on all scales with varying magnitude. Bland-Altman plots showed a skewed distribution, indicating low levels of agreement. Results supported the presence of the peak effect when measuring tiredness, but not on the remaining scales. There was partial support for the presence of an end effect, which was found in two of four scales.

Conclusion: Retrospective ThyPRO ratings and the average of momentary ThyPRO ratings were highly correlated, but provided significantly different results on the four tested scales, with higher scores in the retrospective rating. Only limited evidence was found to support the peak effect and the end effect. Thus, other possible explanations for the observed differences should be examined. The differences in scores were of clinically relevant magnitudes, why the two measures should not be interchanged. In most clinical studies the retrospective rating will be useable, however when designing clinical studies the measurement method should be carefully selected depending on the aim of measurement. To identify the most valid measurement method prospective analysis will compare responsiveness to change.