] Clin Epidemiol Vol. 49, No. 4, pp. 455-460, 1996 Copyright 0 1996 Elsevier Science Inc. ELSEVIER 0895-4356/96/$15.00 SSDI 0895-4356(95)00529-D A Comparison of Prospective and Retrospective Assessments of Sleep Harvey Babkojf, * Aron Weller, and Michal Luvidor DEPARTMENTOF PSYCHOLOGY, BAR-WAN UNIVERSITY, RAMAT-GAN, ISRAEL ABSTRACT. The main purpose of this study was to evaluate the correlations between retrospective and prospective assessments of sleep. Subjective reports of sleep characteristics and sleep quality were obtained from 146 young (20-40 years, average age 29.15) healthy men (n = 43), nonpregnant (n = 70) and pregnant women (n = 33). Three types of subjective sleep reports were obtained: (1) general estimates of sleep habits (retrospective measures); (2) sleep logs completed upon awakening on three successive workdays (prospective measures); and (3) general sleep complaints. Men showed the greatest stability in the sleep log reports, followed by nonpregnant women, with pregnant women showing the least stability. Significant positive correlations between prospective (log reported) and retrospective (general estimates) sleep measures were found only for men. However, for measures of sleep quality, such as feelings of fatigue upon awakening, women showed greater agreement between their general estimates and three nightly reports. The limitations of using retrospective estimates in sleep surveys are discussed. J CLIN EPIDEMIOL 49;4:455-460, 1996. KEY WORDS. Sleep logs, sleep latency, night awakenings, sleep duration INTRODUCTION Sleep patterns are a common topic in community surveys of self- reported health problems. Relations between sleep quality and health complaints are well documented (see, for example, Morison et al. [l], or Hoch et al. [2]). Most of the sleep-health studies in normal popula- tions use the same general methodology: an epidemiological survey based on self-report questionnaires [3,4], which are mailed to a large number of people. For practical reasons the questionnaires contain a small set of sleep and health questions, in which subjects are asked to describe their usual sleep habits. For example, Habte-Gabr et al. [5], asked their 3097 subjects four sleep-related questions and several per- sonality and health questions. The sleep questions included the sub- ject’s usual bedtime, usual sleep duration, latency to fall asleep and feelings of fatigue. The other common use of epidemiological surveys is to describe sleep patterns, either in a general population [4] or in selected subsectors, such as teenagers [l]. One inherent limitation of the typical survey-type questions should be noted. The sleep patterns, in an epidemiological survey, are ob- tained retrospectively. Nisbet and Wilson [6] noted that people are often unable to report accurately on their cognitions and perceptions. Since retrospective reports are mediated via the cognitions and percep- tions of the individual, they may be biased. An alternative approach to the study of subjective reports of sleep patterns is the use of sleep logs [7,8] filled upon awakening each mom- ing. Although also dependent upon memory, the sleep log procedure is more of a prospective measure as it depends upon short-term memory only and the subject is aware before going to sleep that he or she must report on the sleep experience immediately upon awakening [9]. Sleep logs can be used to measure daily fluctuations in the quality and quan- tity of sleep. The method is best suited to a small number of subjects, preferably when some control over the subject’s participation is possi- ble. Sleep logs are the most commonlv used method for sleep assess- ‘All correspondence should be addressed to: Harvey Babkoff, Department of Psychology, Bar-llan University, 52900 Ramat-Can, Israel. Received in revised form. ment in insomnia research, while sleep estimates are used in general epidemiological surveys [lo]. Mullington et al. [9] studied the degree of correspondence between sleep duration as reported by sleep logs and by survey-type question- naires. Their subjects (22 healthy students, ages 22-41, 12 men and 10 women) were asked to assess their usual sleep duration. For 42 days (six weeks) the students reported sleep duration via sleep logs. Pearson correlations computed among the 42 reported sleep durations showed that seven days are sufficient to produce high significant correlations (T = 0.85) between the first seven nights and all of the 42 nights. Measurements over 14 days increased the correlation slightly (r = 0.87). Such high correlations imply that sleep duration, as reported in sleep logs, is a very stable measure. However, correlation of the average duration of the 42 nights reported by daily sleep logs to the general (retrospective) estimate of sleep duration, reported prior to filling out the sleep logs, was r = 0.58, much lower than the intercor- relations among the sleep logs. Mullington et al. [9] found that the general estimates of sleep duration tend to be underestimated for 25% of the subjects in relation to the daily log reports. Subjects tend to report longer sleep durations on the sleep logs than they estimate as their usual sleep duration. Mullington et al. [9] suggested that estimated duration reflected a typical night rather than the average duration. Consequently, they correlated the mode of the estimate with the mode of the duration of the 42 sleep logs. The results were even poorer (r = 0.3 1, nonsignificant). Johns [1 1] compared estimates of the latency to fall alseep with actual sleep latency as measured over six nights in a sleep laboratory. Although the mean estimated latency of the group did not differ sig- nificantly from the mean latency measured in the laboratory, there were large individual differences. The above results challenge the meaning of sleep data based only on the survey methodology, which consists of a few questions about the “usual night’s” sleep. Since such surveys have shown some in- sightful links between sleep measures and health, it is important to find out how reliable such subjective estimates may be when compared to other subjective techniques.