A framework for comparing data collection methods for measuring costs and other outcomes that are aggregated over time.

Melissa A. Clark, PhD Department of Community Health, Brown University, Program in Public Health, Providence RI The paper by Chou et al. (“Methods for measuring personal costs associated with vision impairment.”)1 in this issue of Ophthalmic Epidemiology is their second paper on this subject, now using a full year of diary data on the cost of blindness, collected in Australia. Their first paper2 discussed whether any bias resulted if less than the full year of continuous data were used. In the second paper, Chou et al. conclude that diary methods are feasible and valid and that they can serve a useful purpose in future research on the costs of blindness. While Chou et al. have made a convincing case for the feasibility of diary data collection in a visually impaired population, the advantages and disadvantages of using diary data to measure costs or other constructs that are aggregated over time (e.g., health utility being aggregated in a quality-adjusted life-year calculation) need to be considered carefully. There is a broad general literature on diary data collection. Advantageously, diary data collection methods permit the examination of reported experiences and events in their natural, spontaneous context.3 Diaries can be used to gather (1) reliable person-level data; (2) estimates of within-person changes over time; and (3) individual differences in change over time.3 The presumed advantages of diaries include greater accuracy, fewer recall problems, improved ability to determine the order of events, and the ability to capture events of low salience.4 In addition, diaries do not require participants to summarize or average experiences or behaviors and can elicit more comprehensive descriptions of specific experiences.5 Further, diary data can reduce three potential problems with retrospective reports, particularly when a visual impairment condition is associated with variable symptomatology and experience. First, when using retrospective data collection methods, limited recall ability may result in biased responses in which the study subject incorrectly aggregates over the recall period. Second, people tend to use the current state of health to cue retrospective reports (e.g., using visual problems on the day in question to estimate average visual problems during the past week). Third, respondents tend to rely on peak or extreme levels as well as the most recent levels of experience, rather than weighting each experience equally (e.g., reporting visual problems on the worst day rather than average visual problems during a reporting period).3 Received 30 July 2006 Accepted 10 August 2006