1107 Facilitating Accurate Sleep Diary Data Collection and Processing Based on Common Participant Data Entry Errors

Abstract Introduction Sleep diaries are the gold-standard measure of self-reported sleep parameters; however, data utility is contingent upon accurate data entry and clarity of standardized diary terminology/calculations between participants and researchers. Informed by the RESTING study, an effectiveness-implementation RCT of cognitive-behavioral therapy for insomnia, sleep diary dataset, the current study elucidates common participant sleep diary entry errors to inform standardizable research strategies for minimizing post hoc inaccuracies. Methods RESTING Study participants (N=245) with insomnia disorder completed Consensus Sleep Diaries through REDCap at baseline. Embedded survey data validation prompted participants to check unexpected values (e.g., calculated TIB >10 hours). Researchers reviewed diaries weekly, correcting immediately apparent errors. Post hoc data cleaning scripts flagged sleep diaries with potential inaccuracies based on: (1) 9 hours< Calculated Total Sleep Time (TST)< 4.5 hours, (2) Difference between calculated and self-reported TST >2 hours, (3) 100%< Sleep Efficiency< 45%, (4) Difference between lights out (LO) and bedtime (BT)< 0, or (5) Difference between rise time (RT) and wake time (WT)< 0. Researchers examined internal consistency within flagged diaries and across all entries by a participant. Results 3,382 days of sleep diary data were submitted at baseline (98% of expected). During active data collection, researchers corrected errors within 61 (1.8%) unique diaries, completed by 37 participants. Of 503 (14.9%) distinct sleep diaries (among 173 participants) flagged by post hoc scripts, 173 (5.1%) contained at least one error. Examining internal consistency across diary dates when reviewing flagged entries revealed 20 additional diaries with errors, among 2 participants. In sum, 348 errors were made among 119 participants. The most common errors were: confusion between WT and RT (n=150), confusion between BT and LO (n=45), inclusion of multiple sleep parameters within self-reported “wake after sleep onset” (n=29), unambiguous typos (n=29), and AM/PM mix-ups (n=28). Conclusion Despite training participants on sleep diary completion and utilizing survey data validation, 5.7% of diaries contained errors corrected post hoc. This highlights the importance of standardizing use of post hoc data cleaning scripts. Common errors identified can inform best practices for embedded sleep diary data validation and development of sensitive, precise protocols for reviewing diaries throughout data collection. Support (if any) 1R01AG057500