The influence of recall and timing of assessment on the estimation of the QALY when health fluctuates recurrently

Objectives: When health fluctuates recurrently estimating quality-of-life is challenging, risking over/underestimation due to measures' recall periods and timing. To inform how/when to capture quality-of-life, we compare responses using different recall periods and assessment timings.

Methods: For one 3-week chemotherapy cycle, cancer patients were randomly assigned to complete EQ-5D-5L or SF-12v2 (daily with a daily recall, and weekly with a weekly recall, and at 3-weeks with a 3-week recall); a third group completed SF-12v2 daily with a 3-week recall. EQ-5D-5L and SF-6D utilities (anchored at 1(full-health) and 0(dead)) were generated and repeated measures ANOVA, t-tests and effect sizes calculated to compare recall.

Results: 503 patients consented; all 21 daily questionnaires were completed by 84(50%), 67(40%), 72(43%) in the groups respectively. Both measures captured fluctuations in quality-of-life suggesting differences are due to recall effects. Mean daily scores were greater than scores for the past week at days 7, 14 and 21 (p<0.0001), Utility was underestimated (by 0.0782, 0.0374, 0.0437) for EQ-5D-5L and (0.0387, 0.0266, 0.0304) for SF-6D, with the EQ-5D-5L comparison at day 7 reaching a minimally important difference. The 'past 3-weeks' generated the lowest scores (p<0.0001), with utility underestimated by 0.0746 (EQ-5D-5L) and 0.0310 (SF-6D), heavily skewed by the first treatment week.

Conclusions: Current practice of using a single estimate at the beginning or end of a cycle with a daily (EQ-5D-5L) or longer (SF-12/SF-36) recall could bias cost-effectiveness estimates. Quality-of-life should be captured frequently with short recall when fluctuations are likely, and less frequently with longer recall in stable periods.