OBJECTIVES: Technological advances allow increased use of electronic diaries to capture the patient experience in clinical trials. While increasing the amount of Patient Reported Outcome data, increased data complexities and analytical considerations, including determining acceptable amounts of missingness, occur. Limited evidence exists for missing data rules and an empirical method establishing acceptable missingness in diary studies is needed. The purpose of this study is to explore the impact of missing days on average scores from diary collection.

METHODS: Subjects with Eosinophilic Esophagitis completed a daily diary assessing symptom burden over a 14-day period. A 14-day average score was calculated for each subject. Score ranges from 0 to 8.5. For subjects with complete data, the 14-day average was computed, along with the group mean and standard deviation (SD) over these subjects. A simulation study was conducted randomly removing days of observation (1 up to 13), with 1,000 iterations per day removed. Days were removed under the missing completely at random (MCAR) assumption. For each iteration, the subject-level average was calculated, as well as the group mean and SD. Performance was assessed by means of absolute bias and root mean square error (RMSE) for each simulation scenario.

RESULTS: Only 38 of 106 participants had complete 14 days of data. The overall mean symptom score was 2.58±1.77. The observed bias was minimal, on the order of 0.001. The RMSE increased as the number of days removed increased, ranging from 0.016 for 1 day removed to 2.646 for 13 days removed.

CONCLUSIONS: MCAR days in diaries may not produce biased average scores, however, their impact on the variance can be substantial. This can lead to true treatment differences or instrument psychometric properties being obscured. Further research on the number of non-missing days needed under different assumptions is warranted.