Purpose: To simulate the influence that the frequency of data collection and the recall period (i.e., the length of time the subject is asked to consider) for health utility assessment might have on the statistical properties of calculated QALYs using dysfunctional uterine bleeding (DUB) (a condition with intermittent symptoms of varying severity) as an example. Design: Using literature-based parameters, data from a previous study, and testable assumptions, we estimated a mixture distribution of health utilities at a random day in the menstrual cycle and a beta distribution for the month average utility. These were used to simulate sets of 12 monthly observations for 1,000 replications of 120 women (approximately the number in each arm in the original clinical trial) at each of ten levels of first order correlations of health utility values (0-0.9 in 0.1 increments). QALYs were calculated using one day and month average utilities from all12 months; from months 3, 6, 9, and 12; and from months 6 and 12. The replications were used to test for differences in variances of QALYs calculated the six different ways (one day and month long data with each of the three intervals described) and to test for differences in means between the original distribution and a shifted distribution. Results: QALYs calculated using 12 one day utility measurements had lower variances and more frequently rejected the null hypotheses of no differences in means than QALYs calculated with fewer month average utility measurements at low first order correlations; the opposite held at high correlations. Otherwise, QALYs calculated with more frequent data or with month average rather than one day utilities had lower variance and rejected the null hypothesis of no different in means more frequently. Conclusion: Data collection strategies operating within a budget may have to choose between emphasizing a longer recall period or more frequent interviews. When studying patients with an intermittently symptomatic condition of varying several, which stratgy results in QALY calculations that appropriately reject the null hypothesis of no differences in means a threshold proportion of times may depend on the correlation of health utility data over time.