Who has time to sleep?

Transcription

1 Journal of Public Health VoI. 27, No. 2, pp doi: /pubmed/fdi004 Advance Access Publication 7 March 2005 Who has time to sleep? Lauren Hale Abstract Background Both marriage and education appear to confer a protective effect on health. Few studies have examined the extent to which both undersleeping and oversleeping explain these relationships. I examined whether marital status, educational background, and other sociodemographic variables are associated with higher-risk sleep durations. Methods Over 7000 individuals aged collected timeuse diary data during a 24-h period. For both weekdays and weekend days, separate multinomial logistic regression models were estimated using three categories of sleep duration as the outcome category. Results Unmarried individuals are significantly more likely to sleep a short amount on both weekdays and weekends compared to married people. Single people also are significantly more likely to sleep a long amount. People with less than a college education are significantly more likely to sleep both a short amount and a long amount on weekdays relative to the college educated. Conclusions High-risk sleep durations (short sleeping and long sleeping) are positively associated with sociodemographic categories associated with poorer health. More research should investigate whether social inequalities in health can be explained in part by variation in sleep duration. Keywords: sleep duration, marital status, education, timeuse studies Introduction Although sleep comprises the largest single use of time for most adults, social scientists have rarely used time-use data to explore how sleep duration varies by education, marital status, and other sociodemographic characteristics. This paper explores these relationships in order to identify the characteristics of individuals that are at a higher risk of sleeping either a short amount or a long amount, controlling for other potentially confounding factors. This information can assist public health and health care professionals in identifying segments of the population that are at higher risk for sleep disorders or potential sleep-related disorders. Secondarily, an understanding of these correlations may provide help in explaining other areas of research that are not fully understood, such as social inequalities in health and test score gaps. This research may guide social scientists to incorporate sleep and other biological variables into future analyses. And finally, it may build additional awareness of the importance of sleep hygiene to the non-sleep research community. Sleep and health The relationship between sleep duration and health is complicated, because there are many confounding factors and because causality is likely to flow in both directions. Recent studies show that h of sleep on an average weeknight is associated with the lowest risk of all-cause mortality. 1 3 Controlling for demographic characteristics (e.g., age, race, education, occupation, marital status), health behaviours (e.g., exercise level, years of smoking, fat in diet), prior health conditions (e.g., body mass index, leg pain, and history of heart disease, hypertension, cancer, diabetes, stroke, bronchitis, emphysema, and kidney disease) and medication use, sleeping either a long or short amount increases the relative risk of all-cause mortality by up to 40 per cent. 1 The adverse effect of sleep deprivation on health may be due to disruption of circadian rhythms and impaired glucose metabolism. 4 6 Whether the effects of someone regularly sleeping 5 versus 8 h a night are physiologically similar to being sleep-deprived is not well understood. Similarly, the relationship between sleeping for a long time and health is not particularly clear. 1,6 Nonetheless, evidence suggests that having, on average, a long sleep duration is associated with as high, if not higher, a mortality risk factor than having midrange or short sleep duration on a regular basis. In addition, in a nationally representative US sample, both short and long sleepers report more sleep problems (i.e. wakening during the night, wakening too early, wakening unrefreshed, and daytime sleepiness) compared to the ers (7 or 8 h). 7 Health and sociodemographic factors An extensive literature investigates the positive associations between socioeconomic status and health and between marriage and health Some of this may be due to selection bias. For example, healthier people may be more likely to get married than unhealthy people. 11 In addition, health behaviours such as smoking and nutrition can explain some of these associations. 8,12 14 RAND Corporation Lauren Hale, PhD Address correspondence to Lauren Hale, RAND Corporation, 1776 Main Street, Santa Monica, CA 90407, USA. lhale@rand.org The Author 2005, Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved.

2 206 JOURNAL OF PUBLIC HEALTH Other likely explanations include psycho-social stressors and neighborhood factors, which may disproportionately threaten the well-being of the less-educated and unmarried segments of the population To a lesser extent, health care access and utilization play a role. 18,19 Sleep and sociodemographic characteristics Despite the clear associations between sleep and mortality described above, very few studies explicitly investigate how sleep patterns relate to sociodemographic characteristics. The few studies 5,20,21 that explore these relationships are limited by sample size, available variables, and generalizability. Further, they fail to consider the risks of short sleeping and long sleeping separately. Biddle and Hamermesh argue that sleep is subject to consumer choice, and is affected by the same economic variables that affect choices about other uses of time (ref. 20, p. 941). Their investigation of time-use data indicated that each additional hour of work costs between 7 and 10 min of sleep, based on a linear regression model that controls for age, education, marital status, children in household, religion, and hours worked. Their main focus was on the relationship between sleep duration and hours in the work force. They did not speculate about the implications of the correlations between sociodemographic factors and sleep. In another study, social psychologists investigated the roles that sleep quality and duration play in understanding the relationships between income/education and mental/physical health. They used the Detroit Area Study to show that self-reports of sleep quality are related to both mental and physical health, as well as to education and income. 21 While sleep duration was associated with health, it was not related to education or income in that study. 21 One major limitation of both of the above analyses is that they only consider sleep as a continuous linear variable. As a result, they do not allow for the possibility of a non-linear relationship between sociodemographic characteristics and sleep duration by exploring both ends of the sleep duration distribution. The current study In the current study, I use data from time-use diaries to identify the sociodemographic risk factors for being both short and long sleepers. To my knowledge, this study provides the only multivariate analysis of the sociodemographic correlates of sleep that allows for a trichotomy of sleep durations. This is particularly useful in light of recent research on the risks of both short sleeping and long sleeping. I hypothesize that the time someone goes to bed and rises is strongly related to his or her social experiences and behavioural demands. Hours spent in the labour force, for example, will be an important factor in determining the times for waking and sleeping. Those with stable full-time jobs will be more likely to have a structured schedule that does not allow for extremely late nights or sleeping in. Then again, people with a lot of financial uncertainty or family responsibilities may use late hours of the night to find the time to address their needs. Sleeplessness is likely to result from psychosocial problems and physiological responses to the chronic stressors of everyday life. I hypothesize that people who have greater chronic stressors are more likely to have reduced sleep time because of sleepless nights. People who work the night shift or other unconventional hours are also likely to have disrupted circadian rhythms that make sleeping difficult. An additional potential cause for difficulty sleeping is adverse environmental conditions, such as noisy living space in a city environment, or crowded living spaces or bedrooms (see Conley 22 for an analysis of the effect of household crowding on educational attainment). I expect that disruptive sleeping conditions that might result in short sleep durations would be more common among poorer and/or urban respondents. Materials and methods Design of the American Use of Time Studies This study analyses data from a subset of four cross-sectional time-use studies conducted in 1965, 1975, 1985, and While these studies all have the same principal investigator, they differ in sample size and somewhat in methodology. The first three of the studies are part of the Americans Use of Time Project. 23 In the fall of 1965, 1244 adults aged received a brief at-home warm-up interview along with a diary for the respondent to record their next day s activities. On the day after the activities were recorded, an interviewer returned to the respondent s home to ensure that the diary was filled out completely. In 1975, a similar survey was conducted of 2406 individuals aged 18 and over. This survey over-sampled Sundays and under-sampled Saturdays. The 1985 survey consisted of personal interviews, telephone interviews, and mail-back diaries. This survey was spread over the course of the year interviewing over 5000 people in total. The 1999 survey comes from the Family Interaction, Social Capital and Trends in Time Use 24 which interviewed 1151 individuals aged 18 and over. In every year, the individual recorded his or her activity between 12:00 AM until 12:00 AM using open-ended diary entries across the full 24-h periods using the same basic codes for diary activities. The subsample of individuals included in the current analysis was restricted to those people between age 25 and 64 to avoid confounding factors of college attendance and retirement in the analyses. Four additional individuals were excluded because no amount of sleep activity was recorded. In the analyses reported below, we consider 1096 individuals from 1965, 1710 from 1975, 3514 from 1985, and 825 from The analysis estimates a multinomial logistic regression equation on the amount of sleep reported for the 24-h period of the time-use diary. The three outcome categories are short sleep,, and long sleep. The definition of short sleep is less than 6.5 reported hours, whereas the definition of

3 WHO HAS TIME TO SLEEP? 207 long is greater than 8.5, with the being in between. In another study, 7 a range of 7 8 h was used as the midrange amount of sleep for a typical night. For that study, the data did not differentiate time sleeping into anything smaller than one-hour increments. The h time range allows for a more generous definition of a midrange duration of sleep, and is based on epidemiological studies that suggest this is range of sleep duration is associated with lower mortality risks. 1,2 The explanatory variables are dichotomous variables for calendar year, marital status, gender, educational status, employment, and minutes of television watched per day (in 1 h intervals). Age and age-squared are also included as continuous variables. The regression equations are estimated separately for reports taken during weekdays (Monday through Friday) and the weekend days (Saturday and Sunday). Saturday and Sunday diaries were used as the diaries for weekend and the Monday through Friday diaries were used as the weekday diaries. Since the diaries are recorded from 12:00 AM to 12:00 AM, Friday s diary starts on a weekday morning and ends on a weekend night, and conversely, Sunday s diary starts on a weekend morning but ends on a weekday night. This initially was a concern, but there were qualitatively similar results when the weekday model was run without Friday and the weekend model was run without Sunday. Another model was estimated to test for the interaction effects of weekend with all of the covariates. Results Table 1 shows the frequency of each personal characteristic for the entire sample, and then separately for the weekday sample Table 1 Proportions of dependent and independent variables in the entire sample, the weekday sample, and the weekend sample (standard deviations are shown in parentheses) Participant characteristic Entire sample (n 7095) Weekday diaries (n 5172) Weekend diaries (n 1923) Sleep duration 7.7 (1.8) 7.5 (1.7) 8.3 (1.9) Age 41.6 (11.2) 41.6 (11.2) 41.5 (11.1) Year Gender Female Male Marital status Married Separated/divorced Widow Single No information on marital status Educational status <High school High school graduate College graduate Graduate school * 0.098* No information on education Employment status No job Retired No work hours Work hours Work hours Work hours 50 plus Minutes/day watching TV No TV min *** 0.146*** min *** 0.156*** min > 180 min *** 0.309*** ***p < 0.001; **p < 0.01; *p < 0.05 for a comparison of proportions test between the weekday and weekend samples.

4 208 JOURNAL OF PUBLIC HEALTH and the weekend sample. On the weekdays, the average night recorded 7.5 h of sleep, whereas on the weekends, it is 8.3 (p<0.001). The rest of the table shows that there are few significant differences between the weekday and weekend populations. The differences that do exist are primarily with regard to televisionwatching behaviour. In addition, there are slightly more graduate school-educated respondents on the weekend compared to a weekday. Table 2 shows the distribution of personal characteristics by time slept, where each row sums to 1.0. The majority (54 per cent) of the sample slept within a range of h. Around 27 per cent were classified as long sleepers and 19 per cent as short sleepers, defined as more than 8.5 hours and less than 6.5 h of sleep, respectively. The pattern of Midrange Sleepers > Long Sleepers > Short Sleepers exists when the sample is stratified by each of the above personal characteristics, with one exception. This exception occurs among the people who work more than 50 h a week. Consistent with the economics literature, 20 among the long workers, the percentages of sleep duration are in the order of Midrange Sleepers > Short Sleepers > Long Sleepers. Table 3 provides the results from the multinomial logistic regression models, where the first two columns provide the results for the weekdays and the third and fourth columns show the results for the weekend days. The regression coefficients are exponentiated and presented as odds ratios to facilitate their interpretation. All groups of categorized variables significantly contribute to the model using a likelihood-ratio test. In the pooled model with both weekday and weekend diary data, the addition of interaction terms between weekend and all covariates contributed significantly to the model using a likelihood-ratio test. Thus, weekend sleep differs significantly from weekday sleep patterns. Table 2 Proportions of dependent and independent variables, by amount of sleep (standard deviations shown in parentheses) Short sleep (< 6.5 h/night) (n 1321 or 19%) Midrange sleep ( h/night) (n 3832 or 54%) Long sleep (> 8.5 h/night) (n 1942 or 27%) Sleep duration 5.2 (1.3) 7.5 (.6) 9.8 (1.2) Age 41.3 (11.0) 41.9 (11.1) 41.3 (11.6) Year*** Gender*** Female Male Marital Status*** Married Separated/divorced Widow Single Educational status + <High school High school graduate College graduate Graduate school Employment status*** No job Retired No work hours Work hours Work hours Work hours 50 plus Minutes/day watching TV*** No TV min min min > 180 min ***p < 0.001; **p < 0.01; *p < 0.05; + p < 0.10 for chi-squared test for the hypothesis that the rows and columns are independent

5 WHO HAS TIME TO SLEEP? 209 Table 3 Results of multinomial logistic regression equations, in odds ratios Weekday... Weekend Short versus Long versus Short versus Long versus Year *** * * * 1999 Gender Female.80** * 0.88 Male Marital status Married Separated/divorced c 1.29* 1.08 c 2.60*** 0.97 Widow 2.04*** Single 1.61*** 1.28* 2.06** 1.51* No information on marriage Age variables Age 0.98 b 0.91*** 1.04 b 1.00 Age squared b 1.001* b Educational status <High school 1.43** 1.61*** High school or more a 1.24* 1.19 a 1.91* 0.96 College graduate Graduate school No information school Employment status No job a * a Retired 0.94 b 1.90** b 0.95 Work hours < 36 Work hours a 0.75*** 0.98 a 0.93 Work hours 50 plus 1.45** 0.75* 1.71* 0.92 Minutes/day watching TV No TV 1 60 min 0.68*** c 0.73** 0.82 c min 0.63*** b 0.77* b min 0.61*** b b 1.24 > 180 min 0.78* a a 1.33* Log-likelihood Wald chi Persons ***p < 0.001; **p < 0.01; *p < 0.05; + p < A model pooling weekdays and weekend days was estimated to test whether the interaction between weekend and all other covariates are significant. Significant interaction effects are indicated with the letter symbol to the left of the coefficient: a p<0.10; b p <0.05; c p <0.01. Marital status is correlated with sleep duration. On the weekdays, relative to being married, the separated/divorced (OR = 1.29, p < 0.05), widowed (OR = 2.04, p < 0.001), and single people (OR = 1.61, p < 0.001) are more likely to be short sleepers over ers compared to married people. On the weekend, there is an increased risk of short sleeping for the separated/divorced (OR = 2.60, p < 0.001) and single (OR = 2.06, p < 0.01). Single people also have an increased risk of long sleeping on weekdays (OR = 1.28, p < 0.05) and weekends (OR = 1.51, p < 0.05) compared to married people. Finding that being single increases the risk of both short and long sleeping suggests that sleep duration should not be constructed as a linear outcome, as has been done in previous studies. 20,21 Education is also associated with sleep duration. People without a high school degree are both more likely to be short sleepers (OR = 1.43, p < 0.01) and long sleepers (OR = 1.61, p < 0.001) on the weekdays and they are also more likely to be long sleepers (OR = 1.77, p < 0.05) on the weekends, relative to people with a college degree. Relative to the college educated, high school graduates have an increased risk of short sleeping on the weekdays (OR = 1.24, p < 0.05) and on the weekends (OR = 1.91, p < 0.01).

6 210 JOURNAL OF PUBLIC HEALTH There is no consistent trend over the four years in which the data was collected. This may be due to variation in how the surveys were conducted. Out of the twelve odds ratios estimated for the year dummy variables, only four of them had a p < In 1965, the risk of long sleeping was much smaller on the weekdays (OR = 0.57, p < 0.001) and the weekends (OR = 0.65, p < 0.05) than it was in In 1975, the risk of short sleeping was reduced (OR = 0.71, p < 0.05) compared to 1999 on the weekdays. The risk of long sleeping on the weekend in 1985 compared to 1999 was also reduced (OR = 0.72, p < 0.05). For all four of these statistically significant relationships, the risk of sleeping outside of the h range is higher in This suggests that even after controlling for sociodemographic changes, there has been an increase in non-ers during these four decades. Women are less likely than men to be short sleepers on the weekdays (OR = 0.80, p < 0.01) and weekends (OR = 0.69, p < 0.05). There is no difference between men and women in terms of the risk of being a long sleeper. The age variables only show statistically significant differences with regard to long sleeping on the weekdays, even after controlling for retirement. There is no statistical significance of the age variables for the weekends, although there is a significant interaction between weekend and both of the age variables for sleeping too long (p < 0.01). Retirement and having no job both increase the risk of long sleeping on weekdays (OR = 1.90, p < 0.01 and OR = 1.43, p < 0.05, respectively). On the weekends, there is a reduced risk of short sleeping for the retired (OR = 0.39, p < 0.10). People who work more than 50 h per week have an increased risk for short sleeping and a decreased risk for long sleeping on the weekdays (OR = 1.45, p < 0.01 and OR = 0.75, p < 0.05, respectively) compared to people who work less than 35 h a week. These results confirm the finding that working longer hours increases the likelihood of sleeping too little. 20 Finally, relative to people who watch no television during the day on the weekdays, the risk of short sleeping and long sleeping is significantly smaller for all categories of television watching duration (with the important exception of the positive association between watching more than 180 min and long sleeping). One might expect that watching television might substitute for sleep, and would increase the risk of being a short sleeper. Instead people who watch no television at all are more likely to be short sleepers. Discussion The results described above provide an opportunity to discuss possible explanations. With regard to marriage, unmarried people may have more flexibility with their sleep schedule since they are less likely to share a bedroom with another person. They may use this flexibility to either sleep longer or to do other things. 14,15,19 Unmarried men, for example, binge drink and smoke marijuana more than their married counterparts. 14 On the other hand, unmarried people may require more time in order to manage the household (errands, financial responsibilities). Unmarried people also may have less social support and may have more difficulty sleeping appropriate amounts. Other reasons for the shortened (or lengthened) sleep schedule among the unmarried may be due to mourning a separation/divorce or the loss of a spouse, which may be physiologically disruptive. There are likely additional third-factor variables and reverse causality. For example, personality characteristics that are linked to sleep habits may also be linked to success on the marriage market. Explanations for the higher risk of sleeping a short or a long amount for the less educated people are less clear. Causality could flow in either direction: duration could cause high education or high education causes a midrange amount of sleep. In the first case, if poorer sleepers are always tired, this may affect their studying behaviour and educational attainment. If causality flows from education to sleep, a potential mechanism is that high levels of education may cause people to choose lifestyles that are healthier for them, including going to bed and waking at a regular hour each day. Another likely explanation is that people with greater education may have a fewer number of chronic stressors and thus maybe be able to sleep sounder at night. The finding with regard to television watching needs more research. According to the results described above, a moderate amount of television watching is associated with average duration sleeping. This may indicate that people who do not watch any television are too busy to watch television or to sleep enough. Whether watching television should be advised to people who lack good sleep habits is not clear from this result. The use of a trichotomy of sleep categories proved an effective way of categorizing people s sleep durations. Whereas previous research on sleep duration and education did not find a statistically significant association, these researchers did not look at the possibility that the relationship is nonlinear. The clue to this non-linearity comes from the sleep science literature that clearly shows a J-shaped curve in the relationship between sleep duration and mortality risk, controlling for a variety of other factors. 1,2 There are many implications of this in terms of understanding the larger story of social differences in health. Assuming that short and long sleepers have a higher mortality risk, the findings with regard to marriage may help explain the protective effect of marriage found in a number of studies. This study identifies that marriage confers a protective effect on sleep behaviours. Similarly, this study identifies that education also confers a protective effect on sleep behaviours, and shows that it would be useful for future studies of social inequalities to consider sleep duration. To the extent that sleep duration does explain the widely observed health inequalities, public health programmes may want to target sleep hygiene education programmes to populations (unmarried and less educated) that are observed to be at a higher risk of either short or long sleeping.

7 WHO HAS TIME TO SLEEP? 211 While I do not think that sleep duration provides a monocausal explanation of the social gradient with health, it is a likely candidate to explain some of the variation that has previously gone unexplained. In addition, it is an appealing candidate because sleep is a factor that can be improved through behavioural changes, therapy, and, when appropriate, medical treatment. There are several limitations and obvious next steps for this work. First, applying this analysis to explaining social inequalities in health assumes that there is a causal mechanism between short or long sleep, on the one hand, and mortality, on the other. At present, such relationships have been repeatedly observed, but the causal mechanism is less clear. These relationships need to be better understood. Secondly, this analysis relies on self-report of time-use data. In general, time-use diaries are relatively reliable since respondents need to account for all 24 h. However, recent work shows that objective sleep measures are not strongly correlated with subjective reports of sleep time. 25 If self-reports are less reliable from the unmarried or less educated, this is a possible source of error. Thus, future research should utilize more objective measures of sleep to the extent that it is possible. Also, this study does not consider measures of quality of sleep, but only looks at sleep duration. Sleep quality may be even more powerful in predicting adverse health outcomes than sleep duration. Lastly, more research on the causal mechanisms on the relationships between sociodemographic characteristics and sleep duration is needed. Longitudinal data that contain questions about reasons for various sleep behaviours would be a useful and important way to understand these relationships. References 1 Kripke DF, Garfinkel L, Wingard DL, Klauber MR, Marler, MR. Mortality associated with sleep duration and insomnia. Arch Gen Psychiatry 2002; 59: Tamakoshi A, Ohno Y. Self-reported sleep duration as a predictor of all-cause mortality: results from the JACC Study, Japan. Sleep 2004; 27: Wingard DL, Berkman LF. Mortality risk associated with sleeping patterns among adults. Sleep 1983; 6: Ayas NT, White DP, Manson JE, Stampfer MJ, Speizer FE, Malhotra A, Hu FB. A prospective study of sleep duration and coronary heart disease in women. Arch Int Med 2003; 163: Van Cauter E, Spiegel K. Sleep as a mediator of the relationship between socioeconomic status and health: a hypothesis. In: Adler N, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic status and health in industrial nations: social, psychological, and biological pathways. Annals of the New York Academy of Sciences, Volume 896. New York: The New York Academy of Sciences; Redwine L, Hauger RL, Gillin JC, Irwin M. Effects of sleep and sleep deprivation on interleukin-6, growth hormone, cortisol, and melatonin levels in humans. J Clin Endocrinol Metab 2000; 85: Grandner MA, Kripke DF. Self-reported sleep complaints with long and short sleep: A nationally representative sample. Psychosom Med 2004; 66: Adler NE, Ostrove, JM. Socioeconomic status and health: what we know and what we don t. In: Adler N, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic status and health in industrial nations: social, psychological, and biological pathways. Annals of the New York Academy of Sciences, Volume 896. New York: The New York Academy of Sciences; McGinnis JM, Williams-Russo, P, Knickman, JR. The case for more active policy attention to health promotion. Hlth Affairs 2002; 23: Coombs R. Marital status and personal well-being: a literature review. Fam Rel 1991; 40: Goldman N. Marriage selection and mortality patterns: inferences and fallacies. Demography 1993; 30: Lantz PM, House JS, Lepkowski JM, Williams DR, Mero RP, Chen J. Socioeconomic factors, health behaviors, and mortality: results from a nationally representative prospective study of US adults. J Am Med Assoc 1998; 279: Duncan GJ, Wilkerson B, England P. Cleaning up their act: the impacts of marriage and cohabitation on licit and illicit drug use. Northwestern University WP Umberson D. Family status and health behaviors: social control as a dimension of social integration. J Hlth Soc Behav 1987; 28: Kobrin, FE, Hendershot GE. Do family ties reduce mortality? Evidence from the United States, J Marriage Fam 1977; 39: McEwen, BS, Seeman T. Protective and damaging effects of mediators of stress: elaborating and testing the concepts of allostasis and allostatic load. In: Adler N, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic status and health in industrial nations: social, psychological, and biological pathways. Annals of the New York Academy of Sciences, Volume 896. New York: The New York Academy of Sciences; Diez Roux AV, Stein Merkin S, Arnett D et al. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med 2001; 345: Goodwin JS, Hunt WC, Key CR, Samet JM. The effect of marital status on stage, treatment and survival of cancer patients. J Am Med Assoc 1987; 258: Fiscella F, Franks P, Gold MR, Clancy CM. Inequality in quality: addressing socioeconomic, racial, and ethnic disparities in health care. J Am Med Assoc 2000; 283: Biddle JE, Hamermesh D. Sleep and the allocation of time. J Political Econ 1990; 98: Moore P, Adler NE, Williams DR, Jackson JS. Socioeconomic status and health: the role of sleep. Psychosom Med 2002; 64: Conley, D. A room with a view or a room of one s own? Housing and social stratification? Soc Forum 2001; 16: Robinson JP, Godbey G. Time for life: the surprising ways Americans use their time. University Park: Pennsylvania State University, Robinson JP, Bianchi SM, Presser S. Family interaction, social capital, and trends in time use (FISCT), Knutson KL, Yan LL, Rathouz PJ, Liu K, Lauderdale DS. Comparison of self-reported and objectively-measured sleep duration in a population-based cohort study. Sleep 2004; 27: A398.