Single momentary assessments are not reliable outcomes for clinical trials

Patient reported outcomes (PROs) play an essential role in clinical trials, though questions have been raised about the accuracy of PROs using long recall periods. This paper examines the utility of a PRO employing a single momentary assessment of pain in a sample of community rheumatology patients. We explore the accuracy and reliability of a single assessment versus the average of multiple assessments taken over 1-week, which is considered a common outcome reporting period. A secondary analysis of 128 patients who monitored their pain intensity with momentary data collections several times a day for a week and 3 months later for another week allowed a comparison of randomly-selected single momentary assessments with the average of many assessments from the week. Results from cross-sectional analyses of the first week were that levels of pain measured by single points were not significantly different than the week average in 4 of 5 analyses, but these single-point assessments had much higher variance. Correlations of single-point and week averages were below 0.70. Longitudinal analysis of change scores across 3 months also demonstrated considerable unreliability of single-point measures, thus the statistical power generated by single-point assessments was considerably less than the more reliable week average. Our conclusion is that single momentary assessments, at least for representing an outcome over a period of a week, are not ideal measures. We discuss alternative measurement strategies for efficiently collecting PRO data for a 1-week period using end-of-day diaries or 7-day recall measures.