Agreement between directly reported and recalled EQ-5D-5 L outcomes after an acute health event: a prospective cohort...

To assess the agreement between directly reported and retrospectively recalled health-related quality of life (HRQoL) in patients with diabetic ketoacidosis (DKA), based on reports from both patients and their proxies using the EQ-5D-5 L.

A prospective cohort of 200 patients with DKA and 200 corresponding proxies (a family member or caregiver who lived with or regularly cared for the patient and accompanied them during hospitalization) was enrolled on a staggered, consecutive basis (not simultaneously) as eligible patients presented over the 4-month enrolment period, from April to July 2025. HRQoL was assessed using the Amharic EQ-5D-5 L, self-completed by participants once clinically stabilized (approximately 4-7 days after DKA onset, and always before hospital discharge) (T1) and participants were again given a further set of self-administered questionnaires and instructed to complete them independently at home at one month (T2) and two months (T3) after the acute event, with telephone reminders to prompt timely completion of the T2 and T3 questionnaires and subsequent retrieval of completed forms by the study team. Of the 400 enrolled pairs, 388 individuals (194 patient-proxy pairs) completed assessments at all-time points and were included in the final analysis. Agreement between directly reported and recalled EQ-5D utility and EQ-VAS scores at T1-T2, T1-T3, and T2-T3 was evaluated using paired t-tests, intraclass correlation coefficients (ICCs), and Bland-Altman analyses.

Patients were on average 44.5 years old (SD 14.2), 58.8% male, and predominantly urban residents (73.7%), with moderate DKA severity in about half of the sample (49.5%). Recalled EQ-5D utility and EQ-VAS scores were generally lower than directly reported values for baseline assessments in both patients and proxies, with larger differences at the longer recall interval (T3) than at the shorter interval (T2). Differences between directly reported and recalled scores at later follow-up were smaller. Despite statistically significant mean differences, ICCs indicated good agreement (ICC > 0.75) for both EQ-5D utility and EQ-VAS across patients and proxies. Bland-Altman analyses supported acceptable agreement. Agreement was weaker among patients with mild or recurrent DKA and stronger among those with tertiary education. Dimension-level agreement, assessed using Wilcoxon signed-rank tests and Cohen's kappa, was lowest for anxiety/depression and pain/discomfort among EQ-5D dimensions. EQ-VAS demonstrated slightly stronger agreement than EQ-5D utility scores.

Recalled EQ-5D-5 L scores showed good relative (rank-order) agreement with directly reported scores, but were systematically lower by margins that reached or exceeded commonly used minimal important difference thresholds, particularly over longer recall intervals. Recalled data may therefore support comparative, group-level research when prospective data are unavailable, provided this systematic bias is accounted for, but are not recommended as a substitute for direct assessment at the individual level.