- Email: [email protected]
The National Sleep Foundation's Sleep Health Index
Sleep Health xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Sleep Health Journal of the National Sleep Foundation journal homepage: sleephealthjournal.org
The National Sleep Foundation's Sleep Health Index Kristen L Knutson, PhD a,⁎,1, Julie Phelan, PhD b,1, Michael J. Paskow, MPH c, Anita Roach, MS c, Kaitlyn Whiton, MHS c,d, Gary Langer, BA b, D. Sunshine Hillygus, PhD e, Michael Mokrzycki, BS f, William A. Broughton, MD g, Sudhansu Chokroverty, MD, FRCP h,i, Kenneth L. Lichstein, PhD j, Terri E. Weaver, PHD, RN, FAAN k, Max Hirshkowitz, PhD, DABSM c,l,m a
Northwestern University, Chicago, IL Langer Research Associates, New York, NY c National Sleep Foundation, Arlington, VA d The Hilltop Institute, University of Maryland, Baltimore County, Baltimore, MD e Duke University, Durham, NC f Mokrzycki Survey Research Services, West Newbury, MA g University of South Alabama School of Medicine, Mobile, AL h JFK New Jersey Neuroscience Institute, Edison, NJ i Seton Hall University School of Graduate Medical Education, South Orange, NJ j University of Alabama, Tuscaloosa, AL k University of Illinois at Chicago, Chicago, IL l Baylor College of Medicine, Houston, TX m Stanford University, Stanford, CA b
a r t i c l e
i n f o
Article history: Received 30 January 2017 Received in revised form 26 May 2017 Accepted 30 May 2017 Available online xxxx
a b s t r a c t Objectives: A validated survey instrument to assess general sleep health would be a useful research tool, particularly when objective measures of sleep are not feasible. Thus, the National Sleep Foundation spearheaded the development of the Sleep Health Index (SHI). Design: The development of the SHI began with a task force of experts who identified key sleep domains and questions. An initial draft of the survey was created and questions were refined using cognitive testing and pretesting. The resulting 28-question survey was administered via random-sample telephone interviews to nationally representative samples of adults in 2014 (n = 1253) and 2015 (n = 1250). These data were combined to create the index. A factor analysis linked 14 questions to 3 discrete domains: sleep quality, sleep duration, and disordered sleep. These were assembled as sub-indices, then combined to form the overall SHI, with scores ranging from 0 to 100 (higher score reflects better sleep health). Results: Americans earned an overall SHI score of 76/100, with sub-index scores of 81/100 in disordered sleep, 79/100 in sleep duration, and 68/100 in sleep quality. In regression analyses, the strongest independent predictors of sleep health were self-reported stress (β = −0.26) and overall health (β = 0.26), which were also the strongest predictors of sleep quality (β = −0.32 and β = 0.27 respectively). Conclusions: The current 12-item SHI is a valid, reliable research tool that robustly measures 3 separate but related elements of sleep health—duration, quality, and disorders—and assesses the sleep health status of adults in the United States. © 2017 The Authors. Published by Elsevier Inc. on behalf of National Sleep Foundation. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Introduction Sleep is essential for optimal cognitive performance, physiological processes, emotional regulation, and quality of life.1–6 Research consistently demonstrates that sleep is a significant component of ⁎ Corresponding author. E-mail address: [email protected] (K.L. Knutson). 1 Co-primary authors.
physical and mental health, as well as overall well-being. Therefore, a comprehensive evaluation of an individual's health and wellness necessitates an assessment of sleep health. Unfortunately, objective measures of sleep, such as the gold-standard polysomnography, can be impractical and expensive and therefore infeasible for many large-scale studies, especially when the research is not primarily focused on sleep. A valid and reliable self-reported sleep health measure would therefore be a useful research tool, benefiting ongoing and future research studies across multiple fields of health.
http://dx.doi.org/10.1016/j.sleh.2017.05.011 2352-7218/© 2017 The Authors. Published by Elsevier Inc. on behalf of National Sleep Foundation. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
2
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
Sleep surveys and questions certainly have been used previously. Several large research studies of sleep and health relied on subjective reports of sleep; however, substantial heterogeneity exists in the questions asked to assess sleep duration and quality, which limits the ability to compare results. Moreover, many existing questionnaires that assess sleep focus on the clinical and sleep-disturbance dimensions of sleep health. This includes widely used instruments such as the Pittsburgh Sleep Quality Index (PSQI), 7 whose goal was to measure sleep quality in clinical populations, the Epworth Sleepiness Scale, 8 which assesses only daytime sleepiness, and the National Institutes of Health Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance scale, which focuses on disturbed sleep and is part of the set of patient-centered measures developed by the National Institutes of Health. 9 Currently, no uniform definition of “sleep health” exists, although the National Sleep Foundation (NSF) is working on developing its own definition. Sleep health is not merely the absence of a sleep disorder or problem, and therefore, to adequately assess general sleep health, multiple dimensions of sleep should be assessed. 10 For example, overall sleep health may include the quantity, quality, and impact of sleep, which are all essential, especially for the large proportion of the population that does not suffer from sleep pathologies. The National Sleep Foundation (NSF) set out to develop a single survey to gauge “sleep health” among adults in the United States.11 A critical mission of the NSF is to improve health and well-being through improving sleep health in the United States and abroad; consequently, it was important for the instrument to assess sleep health in the general population. Historically, the NSF conducted annual polls on sleep-related topics; however, the polls' topics and samples varied each year, which did not allow for analysis of trends across time. Thus, the NSF developed a poll to benchmark sleep health trends longitudinally. This poll, the Sleep Health Index (SHI), was developed and validated so that researchers and the public alike will better understand sleep health in the general population. The ultimate goal is to extend sleep health research, improve wellbeing, and empower individuals to understand the importance of sleep. Methods Development of the Sleep Health Index The SHI was developed as a poll to assess multiple dimensions of sleep health to generate an overall SHI score. The development of the SHI, over the course of 4 years, included assembling a task force, identifying specific topic areas, developing and testing the questions, surveying a beta version of the instrument, refining the instrument, and developing and testing the index. First, a task force was established in October 2013 as an initial step in this process. This task force included experts in sleep, polling, and survey methodology, as well as members of the polling firm,
Princeton Survey Research Associates International. The task force members were initially asked to provide questions to address several domains related to sleep and health: general health, sleep habits, sleep schedules, sleep quality, sleep problems, sleep environment, sleep knowledge, and sleep beliefs. From an extensive list of suggested questions, a first draft of the poll was created and questions were subsequently refined between December 2013 and June 2014 using an iterative process involving conference calls and refining poll questions among task force members. Per best-practice guidelines,12 an important goal was to minimize respondent burden by retaining the minimum number of questions necessary to achieve our goals. One significant area of discussion among the task force was appropriate recall period. To minimize error due to poor recall, the task force selected a period of 7 days. Another decision was to use time in bed instead of “actual sleep” to minimize cognitive burden due to the amount of mental calculations and estimation the respondent would need to complete. 13 Recalling a clock time requires a smaller cognitive burden than calculating time in bed and adjusting for sleeplessness. The instrument was further refined after cognitive testing and pretesting in July 2014. Ten (4 women/6 men) cognitive interviews were conducted in-person using the verbal probing technique. 14 The cognitive interviews were used to determine if respondents understood the questions, both consistently across subjects and in the way intended. Next, a pilot test of the questionnaire was fielded to an online convenience sample of 167 respondents to test question wording alternatives and to inform questionnaire reduction. After the cognitive testing and initial online pilot test, a pretest of the revised poll was conducted in August 2014 over the telephone with 24 subjects to determine the length of the questionnaire administration and to further identify questions that needed refinement. This led to a penultimate version of the sleep health survey with 28 substantive questions addressing all the topics listed above. The penultimate version of the instrument was administered via random-sample telephone interviews to a nationally representative sample of adults in 2014 (n = 1253) and 2015 (n = 1250). The basic disclosure elements per the American Association for Public Opinion Research 12 are provided in Table 1. These 2 surveys were used to further refine the instrument and create the SHI score. We eliminated questions that did not pertain directly to individual sleep health based on conceptual grounds because they were less directly related to the experience of sleep. For example, the questions about general health and quality of life were eliminated because although sleep is associated with general health and quality of life, it is not a direct measure of sleep itself. Similarly, questions about the sleep environment, electronic use, beliefs, and knowledge about sleep were eliminated because although they likely impact sleep, they, too, are not direct measures of the experience of sleep itself. We also excluded a question about napping behavior because it is not clear whether a nap is an indicator of good or poor sleep health, and indeed the answer likely varies. Finally, because our goal was to create a score for each response, we wanted to compare the time
Table 1 Basic disclosure elements details
Survey sponsor Survey/data collection supplier Population represented Sample size Mode of data collection Type of sample Start and end dates of data collection Margin of sampling error for total sample Data weighting
2014 SHI
2015 SHI
National Sleep Foundation Princeton Survey Research Associates International (PSRAI) US adults 1253 Telephone (RDD landline and cell) Probability September 8-29, 2014 ±3.1 percentage points Weighted to correct known demographic discrepancies
National Sleep Foundation PSRAI US adults 1250 Telephone (RDD landline and cell) Probability April 27-May 17, 2015 ±3.1 percentage points Weighted to correct known demographic discrepancies
SHI, Sleep Health Index.
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
spent in bed to the times recommended by an expert panel.15 To do this, we fielded a nationally representative survey question in March 2016 to obtain self-reported average time asleep in the previous 7 days. The average was 6.5 hours, compared with an average of 7.7 hours in bed. We used this 1.2-hour difference to score the sleep duration items in this instrument; this value is reevaluated annually for appropriate time adjustments. The 14-question SHI instrument is provided as supplemental material. Covariates The survey used in 2014 and 2015 included additional questions that were used in these analyses. These included demographic information such as age and sex. Educational level was assessed with the question, “What is the highest level of school you have completed or the highest degree?” and 8 responses were available ranging from “Less than high school” to “Post-graduate or professional degree.” Income was based on the question, “Last year—that is, in 2013 [or 2014] —what was your total family income from all sources, before taxes?” and 8 responses were provided ranging from “Less than $20,000” to “$100,000 or more.” The surveys also asked for the number of children 6 years and younger currently living in the household. Respondents were asked to rate their overall health as excellent, very good, good, only fair, or poor. The surveys also asked about life satisfaction based on the question, “All things considered, how satisfied are you with your life as a whole these days?” with 5 possible responses, including extremely satisfied, very satisfied, fairly satisfied, only a little satisfied, or not at all satisfied. The surveys asked about stress using the question, “Overall, how stressed have you felt in the past month?” and responses included extremely stressed, very stressed, fairly stressed, only a little stressed, and not at all stressed. Respondents were asked how often they snore and responses included every night or almost every night, a few nights a week, rarely, or never/I don't snore. The surveys asked the respondent to, “Imagine if you had an extra hour in the day, would you want to…” and the response options were exercise, read, spend time with friends or family, sleep, or work/housework. The survey assessed sleep environment by asking respondents to rate level of light, level of noise, and mattress comfort, and 4 responses were provided ranging from very dark/quiet/comfortable to not at all dark/quiet/ comfortable. The survey also asked how many nights the respondent sent or read text messages, e-mails or other electronic communications after falling asleep, and response options were none, 1-2 nights, 3-4 nights, 5 or more nights. Finally, the survey asked for the number of days in the past week the respondent shared a bed with a partner, child, or pet and if he and/or she texted before or after falling asleep. Analyses We used factor analysis using direct oblimin rotation and principal axis extraction to evaluate the items included in the survey and to identify sub-indices. The cutoff considered for inclusion was 0.4. We tested the sub-indices and the full index for internal consistency, reliability, convergent validity, and known-groups validity. We evaluated the reliability of the SHI using Cronbach α, a test for internal consistency that measures whether items intended to reflect the same construct actually do so. A Cronbach α score of 1 reflects an exact match among all items (indicating redundancy) and 0 indicates no relationship among them. Construct validity is the extent to which measurements reflect the concept they are intended to assess (eg, whether knowledge tests actually measure knowledge). We assessed the construct validity of the SHI by evaluating convergent and known-groups validity. Convergent validity is demonstrated when measures that theoretically should be correlated with the index are in fact correlated. Known-groups validity tests whether groups that are expected to differ actually do so. The “known groups” in these analyses are based on the questions
3
about general health, stress, and life satisfaction. We also compare the SHI scores to other factors expected to be associated with sleep health, including how respondents would spend an extra hour in the day, use of sleep disruptors, sleep beliefs, the sleeping environment, and the presence of a sleep disorder. We calculated the scores for the overall SHI as well as the subindices for 2014 and 2015 individually and combined. Scores range from 0 to 100, and the scoring algorithm will be evaluated and adjusted as necessary every quarter. Thus, anyone interested in using this survey should contact the NSF ([email protected]). We tested for differences between the years using Student's t test. Student's t test and regression models were used to test for differences in SHI scores across demographic categories. Pearson correlation coefficients were used to test for associations between the SHI scores and other measures thought to be associated with sleep health. Finally, linear regression analyses were performed to determine which factors are predictive of the SHI overall score and its sub-indices. The covariates included in the regression models were based on the available questions in the full 2014 and 2015 survey, including sex, age, race, marital status, employment status, education, income, number of children, subjective health, life satisfaction, amount of stress, frequency of snoring, mattress comfort, bedroom darkness, and quietness of bedroom, as well as (separately) the number of days in the past 7 days the respondent shared a bed with a partner, child, or pet, and if he and/or she used a computer/tablet before bed and texted/e-mailed after falling asleep. A P value of .001 was selected as significant to be conservative given the large sample size. Statistical analyses were performed using SPSS (v19; IBM SPSS Statistics, Chicago, IL) and Stata SE survey commands (v14; StataCorp, College Station, TX).
Results A description of respondent characteristics for both the 2014 and 2015 surveys, separately and combined, is provided in Table 2. The Table 2 Sample composition (unweighted sample size and weighted % of population) combined and by year Combined
Total Men Women 18-29 y 30-39 y 40-49 y 50-64 y 65+ y White Black Latino Other Married/partnered Not married/partnered Employed Not employeda No high school diploma High school graduate Some college College graduate b$20,000 $20,000-$39,999 $40,000-$59,999 $60,000-$79,999 $80,000-$99,999 $100,000+
2014
n
%
n
2503 1178 1325 463 258 322 679 761 1726 280 344 120 1445 1045 1261 1225 241 772 650 820 454 464 371 266 166 360
100 48 52 22 15 17 27 18 66 12 15 7 60 39 57 43 10 31 31 28 20 19 15 10 7 14
1253 580 673 223 114 165 355 387 871 135 169 60 722 526 628 616 104 383 337 419 228 244 172 139 77 175
2015 %
n
%
48 52 22 14 18 28 18 66 12 15 7 59 40 57 43 10 32 31 28 20 20 14 11 7 14
1250 598 652 240 144 157 324 374 855 145 175 60 723 519 633 609 137 389 313 401 226 220 199 127 89 185
49 51 22 16 17 26 18 65 12 15 7 62 38 56 44 11 30 30 28 20 18 16 10 7 14
a Not employed includes respondents who are unemployed, students, retired, on disability or a stay-at-home parent.
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
4
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
table provides both the unweighted sample sizes within each group and proportion weighted to be nationally representative. Factor analysis Factor analysis led to the exclusion of a question about snoring frequency. Snoring was not included in the index because factor analysis indicated that it did not load strongly onto any of the 3 factors, and it reduced the reliability of the index overall. There were 2 questions whose factor loadings were below 0.4 but which were retained because of expert opinion favoring their inclusion as important components. These included unintentionally dozing and use of sleep medication. Factor analysis of the final 14 questions was used to identify 3 discrete domains: sleep quality, sleep duration, and disordered sleep. These were assembled as sub-indices that were then combined to form the overall SHI. These sub-indices are described in detail below. The factor loadings of the final 12 items are presented in Table 3. Index composition The final sub-indices were sleep quality, sleep duration, and disordered sleep. The sleep quality sub-index includes 6 items: • Respondents' overall ratings of their sleep quality • The number of days in the past 7 days that respondents: ○ felt well-rested, ○ had trouble falling asleep, ○ had trouble staying asleep, ○ were negatively impacted by lack of sleep, and ○ dozed unintentionally. The sleep duration sub-index includes 3 items: • A weekday sleep score reflecting how well respondents' selfreported time in bed aligned with NSF's published recommendations for optimal sleep.15 Time in bed was calculated based on the time respondents reported that they most often went to bed and woke up on weekdays in the previous 7 days. Time in bed was scored against expert recommendations for optimal sleep adjusting for the difference between time in bed and sleep duration. • A sleep deficit score based on the difference between weekday sleep and the amount of sleep respondents said they needed to feel their best • A “sleep variability” score reflecting the extent to which respondents' weekday time in bed differed from their weekend time in bed. This is a measure of variability in duration. The disordered sleep index includes 3 items:
Table 3 Factor loadings of each question on each of the 3 sub-indices
Sleep quality rating Well-rested Trouble falling asleep Trouble staying asleep Negative impact Unintentional doze Weekday sleep Sleep deficit Sleep variability Sleep medication Diagnosed sleep disorder Sleep doctor
Quality
Duration
Disorder
.70 .68 .63 .66 .67 .26 .25 .31 .11 .29 .25 .35
.15 .34 .13 .10 .27 .16 .75 .74 .42 −.06 .07 .01
.33 .19 .30 .30 .23 .11 .03 .08 −.01 .35 .64 .85
Bolded numbers show items included in each sub-index.
• The number of days in the past 7 days in which the respondent took a sleep medication • Whether or not the respondent had been told by a doctor that they have a sleep disorder • Whether or not the respondent had discussed his/her sleep with a doctor or medical professional Reliability The factor analysis helped to identify 3 separate but related constructs (Table 3). Cronbach α results indicated high levels of internal consistency, ranging from α = .63 for both disordered sleep and sleep duration to α = .77 for sleep quality in the combined data set. The 12item Cronbach α was .76 for 2014 and .75 for 2015. To assess the reliability of the SHI over time, we repeated this assessment separately for the 2014 and 2015 data and the results were virtually identical. Construct validity We found statistically significant correlations (P b .001) between the SHI and respondents' ratings of their overall health (r = 0.38), stress (r = −0.37), and life satisfaction (r = 0.36). The SHI score was 67 for those who reported having been extremely or very stressed in the previous month, compared with 81 for those who felt little or no stress. Comparative sub-index scores were 54 vs 77, respectively, for sleep quality; 73 vs 82, respectively, for sleep duration; and 74 vs 84, respectively, for disordered sleep. Those who rated their overall health as excellent or very good had an SHI score of 81, in comparison to 64 among those who said that their health was just fair or poor. This reflected large differences in sleep quality (76 vs 51) and disordered sleep (88 vs 66), but only a small difference in sleep duration (80 vs 77). Those who were extremely or very satisfied with their lives scored significantly higher on the SHI than did those who were only a little or not at all satisfied with their lives (80 vs 60), including on sleep quality (74 vs 46), disordered sleep (84 vs 66), and sleep duration (81 vs 69). All differences are statistically significant (P b .001). Respondents were asked what they would do with an extra hour in the day. As would be expected, those who said they would use it to sleep, scored lower on the SHI overall (67 vs 78) and all 3 of its subindices—sleep quality (56 vs 71), sleep duration (71 vs 80), and disordered sleep (74 vs 82)—compared with those who would use that hour to work, exercise, read, or socialize. We also examined the relationships between sleep health and measures of possible sleep disruptors (ie, electronic use, and bed partners), sleep beliefs, and sleeping environment (ie, noise, light, and mattress comfort). These 3 constructs were not related to sleep health (all r values b 0.2). The strongest correlation was between sleep quality and comfort of mattress (r = 0.24). Mattress comfort also was linked to longer sleep duration (r = 0.13) and less disordered sleep (r = 0.07), but more weakly. The correlation with the overall SHI was r = 0.20. The SHI was 77 for adults who reported having more comfortable mattresses vs 65 for those with less comfortable ones. Comparative sub-indices were 70 vs 53 for sleep quality, 80 vs 73 for sleep duration, and 82 vs 72 for disordered sleep. There were group differences based on the presence of a sleep disorder. The SHI score was 52 among those who had been told by a doctor that they had a sleep disorder, compared with 80 among those without a sleep disorder. Those with a sleep disorder scored less positively for sleep duration (75 vs 80) and sleep quality (56 vs 71). They also scored considerably worse on the disordered sleep sub-index (28 vs 91), but that was largely because this sub-index includes having a sleep disorder. There were also differences based on the use of sleep medication. The overall SHI was 59 among adults who had taken a sleep
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
medication at least once in the past 7 days, compared with 79 among those who had not, including 54 vs 71 for sleep quality. There were no differences in sleep duration between those who used and those who did not use sleep medication. SHI scores Americans earned an average SHI score of 76 of a possible score of 100, with underlying scores of 81 in disordered sleep, 79 in sleep duration, and 68 in sleep quality (Table 4). Actual scores ranged from 1 to 100 for the SHI score and 0 to 100 for each sub-index. The 2014 and 2015 surveys produced virtually identical results. For the 2014 SHI, the public scored an overall SHI of 75, including scores of 80 for disordered sleep, 79 for sleep duration, and 68 for sleep quality (a higher score reflects greater sleep health). The 2015 SHI scores were 77 overall, 82 for disordered sleep, 80 for sleep duration, and 69 for sleep quality. None of the 1- or 2-point differences between 2014 and 2015 data were statistically significant. The SHI scores stratified by demographic groups are presented in Table 5. Mean SHI was significantly (P b .05) higher in men, among older (65+ years) and younger adults (18-29, 30-39 years), among whites and Hispanic/ Latinos, among married respondents, among employed respondents, among college graduates, and among those with higher income. Sleep quality sub-index The sleep quality sub-index includes scores for self-reported sleep quality as well as scores for how many days out of the past 7 the respondent felt well rested, had trouble falling asleep, had trouble staying asleep, was negatively impacted by lack of sleep, or dozed unintentionally. Overall, Americans had a score of 68 on this sub-index, including a score of only 49 for self-reported sleep quality (Table 4). Just 3 in 10 rated their sleep quality as excellent (11%) or very good (19%). Thirty-five percent described their sleep quality as “good,” 22% as “fair,” and 12% as “poor.” Scores on the other items in the sub-index ranged from a score of 55, for the number of days the respondent felt well rested in the last 7 days, to a score of 82, for the number of days the respondent dozed unintentionally. Respondents indicated feeling well rested an average of 3.9 days of the previous 7 days. They had trouble falling asleep 1.7 nights and staying asleep 2.1 nights of the previous 7 days on Table 4 SHI overall scores, sub-indices scores, and the scores by individual questions in the combined sample and by year Combined years
2014
2015
76 (75-77)
75 (74-76)
77 (76-78)
68 (67-69) 49 (48-51) 55 (54-57) 69 (68-71) 76 (75-78) 78 (77-80) 82 (81-83)
68 (66-69) 49 (47-51) 54 (52-56) 69 (66-71) 76 (74-78) 78 (76-80) 82 (80-83)
69 (67-70) 50 (48-51) 57 (54-59) 70 (68-73) 77 (74-79) 79 (77-81) 83 (81-84)
79 (78-80) 67 (65-68) 84 (83-84) 88 (87-89)
79 (77-80) 66 (64-68) 84 (82-85) 87 (85-88)
80 (78-81) 68 (66-70) 84 (82-85) 89 (87-90)
81 (79-82) 69 (67-71) 85 (83-86) 89 (87-90)
80 (78-82) 68 (66-71) 83 (81-85) 88 (86-90)
82 (80-83) 70 (67-73) 86 (84-89) 89 (87-91)
Mean (95% CI) SHI Sleep quality score Sub-index overall Sleep quality rating Well-rested Trouble staying asleep Trouble falling asleep Negative impact Unintentional dozing Sleep duration score Sub-index overall Weekday sleep Sleep deficit Sleep variability Disordered sleep score Sub-index overall Sleep doctor Diagnosed disorder Sleep medication
CI, confidence interval; SHI, Sleep Health Index.
5
Table 5 Scores on the SHI and its sub-indices by groupsa
Total Men Women P 18-29 30-39 40-49 50-64 65+ P White Black Latino Other P Married/partnered Not married/partnered P Employed Not employedb P No high school diploma High school graduate. Some college College graduate P b$20,000 $20,000-$39,999 $40,000-$59,999 $60,000-$79,999 $80,000-$99,999 $100,000+ P
SHI
Quality
Duration
Disorder
76 77 75
68 70 66
79 78 80
81 82 80
.18 76 78 75 80 86 b.0001 81 72 78 78 b.0001 80 78 .026 77 82 b.0001 76 79 79 81 .05 77 78 79 78 79 83 .008
.05 89 86 77 74 79 b.0001 79 81 85 89 .0001 81 80 .44 84 76 b.0001 81 82 78 82 .07 78 80 82 81 80 81 .11
.035 78 77 73 73 79 b.0001 76 72 76 80 .002 77 75 .004 77 74 .0001 74 76 75 78 .0004 71 75 77 77 77 78 b.0001
.0002 70 67 66 66 72 .0005 69 63 67 73 .0001 70 66 .0005 71 65 b.0001 64 66 68 72 b.0001 59 65 70 73 73 73 b.0001
SHI, Sleep Health Index a Student t tests were used to compare dichotomous variables and regression analyses tested for differences among N2 categories. b Not employed includes respondents who are unemployed, students, retired, on disability, or a stay-at-home parent.
average. Furthermore, they were impacted by lack of sleep an average of 1.5 days and dozed unintentionally on 1.3 days of the previous 7 days. Mean scores on the sleep quality sub-index were compared across demographic groups (Table 5). Mean scores were significantly (P b .05) higher for men, younger (18-29 years) and older (65+ years) adults, respondents from “other” race/ethnicity, married respondents, employed respondents, college graduates, and those with higher incomes ($60,000/y or more). Sleep duration sub-index The average sleep duration sub-index score was a 79, with underlying scores including 67 for weekday sleep duration, 84 for sleep deficit, and 88 for sleep variability (Table 4). Americans reported spending an average of 7.7 hours in bed on weekdays. Roughly half reported spending between 7 and 9 hours in bed—a quarter spent less time than that, and 23% averaged more than 9 hours in bed on weeknights. Individual weekday sleep scores were assigned relative to how far respondents were from sleep duration recommendations established by an expert panel convened by the NSF. 15 Twenty-six percent of respondents fell into the recommended range. The majority—68%—spent less time in bed than recommended. Respondents reported needing an average of 7.1 hours of sleep to feel their best. Forty-four percent of these individuals fell short of the average by an hour or more. Among the remainder, 28% got as much (or more) sleep than they felt they needed, and 29% had a deficit of less than an hour. Sleep deficit scores were assigned based on how
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
6
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
close respondents' approximate weekday sleep was to meeting their self-assessed sleep needs. In terms of sleep variability, the average difference between weekday and weekend time in bed was just more than 1 hour. Half of the respondents had a difference of less than an hour, but 22% of respondents had a difference of 2 or more hours between weekday and weekend time in bed. Mean scores on this sub-index were also compared across demographic groups (Table 5) and these scores were significantly (P b .05) higher, indicating better sleep health, for older adults, whites, married respondents, respondents who were not employed, those with greater education, and those with more income. Disordered sleep index The overall score of 81 on the disordered sleep sub-index included subcategory scores of 69 for talking to a doctor about their sleep, 85 for diagnosed sleep disorders, and 89 for sleep medication use (Table 4). Three in 10 respondents had discussed sleep issues with a doctor or medical professional, whereas the 69% who had not discussed sleep problems with a doctor scored 100 on this item. Similarly, the score of 85 for disordered sleep reflects the fact that more than 8 in 10 people indicated that they do not have a diagnosed sleep disorder. Prescription or over-the-counter sleep aids were taken an average of 0.8 of 7 nights. Eighty-three percent had not taken a sleep aid in the past 7 days. Among those who had taken medication, most (51%) did so on all 7 nights of the previous week. Finally, we compared the scores on this sub-index across demographic groups (Table 5), and scores were significantly (P b .05) higher for younger adults, respondents who identify as Latino/Hispanic and other race/ethnicity, and employed respondents. Regressions and group differences In regression analyses (Table 6), the strongest independent predictors of sleep health were self-reported stress—a strong predictor of sleep quality–and overall health. Weaker predictors included life satisfaction, frequency of texting or e-mailing after initially falling asleep, frequency of snoring, and mattress comfort. As would be expected, life satisfaction and mattress comfort positively predicted the SHI, whereas texting and snoring were related to poorer sleep health. The separate regression models predicting sleep quality, sleep duration, and disordered sleep offer further evidence that these subindices tap into distinct constructs related to sleep health. Stress was a negative predictor of all 3 of the sub-indices, and a particularly strong predictor of sleep quality. Overall health was associated with both sleep quality (β = 0.27) and disordered sleep (β = 0.22), but Table 6 Results from regression analyses Outcome SHI
Quality
Duration
Disordered
Regression coefficient (β) Stress Health Life satisfaction Frequency text after sleep Frequency snore Mattress comfort Employed Age Adjusted R2
−0.25 0.25 0.14 −0.11 −0.10 0.08 – – 0.29
SHI, Sleep Health Index – = not statistically significant (P N .001)
−0.31 0.26 0.14 −0.12 −0.08 0.11 – – 0.37
−0.13 – 0.10 −0.11 – – −0.11 – 0.10
−0.11 0.21 – – −0.10 – – −0.08 0.13
not sleep duration. Frequency of texting after falling asleep was negatively associated with sleep quality and duration, whereas frequency of snoring was negatively associated with sleep quality and disordered sleep, but not duration. Mattress comfort was positively related to sleep quality, but not to sleep duration or disordered sleep. People with an uncomfortable mattress stayed in bed just as long as those who did not report having an uncomfortable mattress, but the data suggest that the quality of sleep while in bed was poorer. Employment negatively predicted sleep duration, but not the other 2 sub-indices. Those with a full- or part-time job spent less time in bed on weekdays than did those without a job. Older age predicted disordered sleep, but not sleep quality or duration. Older adults were more apt than younger adults to take sleep aids, have a diagnosed sleep disorder, and talk to their doctor about sleep. Conclusions Given its inclusion of 3 separate but related elements of sleep health—duration, disorders, and quality, each measured in multiple survey questions—the SHI is a robust, valid measure of sleep health. The statistical testing described in this report indicates validity and reliability in the underlying data and index construction alike. Furthermore, the SHI score varies between groups expected to demonstrate differences, including higher scores (better sleep health) among those reporting lower stress, better general health, and greater life satisfaction. Those who reported that they would spend an extra hour in the day sleeping had lower SHI scores. Sleep disruptors (such as electronic use or bed partners), beliefs about sleep, and the sleeping environment were not related to SHI scores. This may reflect the nature of the items. For example, sleeping with a partner, a pet, or a child may be positively or negatively linked to sleep health, depending on the sleeper's preferences and the sleep partner's characteristics. This leads to a weak correlation overall, although the factors may be strongly linked at the individual level. Another possible explanation is that these factors do not affect sleep health strongly, at least as assessed by the questions in the survey. The overall score and the sub-indices also showed several consistent differences between demographic groups, particularly higher scores among men, among those of higher socioeconomic status (ie, higher education, income, or those employed), and among some racial/ethnic groups. Whether these differences in SHI scores reflect meaningful differences in sleep health, or health in general, warrants further investigation. The SHI instrument provides a new robust tool for assessing general sleep health in the population. As mentioned, other instruments designed to assess sleep do exist and are widely used in a variety of studies, for example, the PSQI, which was designed as a clinical tool to assess sleep quality and discriminate between good and poor sleepers. 7 It includes several factors that disturb sleep, subjective sleep quality, medication use, and 2 questions pertaining to daytime dysfunction. The PSQI differs from the SHI in 2 distinct ways. First, although it also includes a measure of sleep duration, it does not distinguish between work days/weekdays and non-workdays/weekends. Second, the SHI uses a shorter recall period (7 days vs 1 month) to improve accuracy due to recall error. 13 Another commonly used sleep questionnaire is the Epworth Sleepiness Scale, 8,16 which is an 8-item questionnaire designed to assess only general level of daytime sleepiness and does not assess quality, duration, or disturbed sleep. More recently, sleep-related instruments have been developed for the National Institutes of Health PROMIS health measures. 17,18 There are PROMIS instruments for “sleep disturbances” and “sleeprelated impairment,” and as these names indicate, they focus on disturbances in sleep, including sleep quality, or daytime dysfunction, but do not include sleep duration. The PROMIS sleep disturbance scale includes questions are that similar to 4 of the 6 items in the SHI sleep quality sub-index, specifically questions about overall
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
sleep quality, restful sleep, and difficulty falling or staying asleep, but does not include questions about impact on daily activities or dozing unintentionally. The selection of a sleep instrument will ultimately depend on the goals of the research. For those interested in general sleep health, rather than a clinical population, the SHI may be the better option, particularly if interested in comparing results to the nationally representative results obtained and published by the NSF (sleepfoundation.org). The strengths of this instrument include the multidimensional aspects of sleep health that have been incorporated and the multidisciplinary task force that comprised both sleep and polling/survey experts. Another important strength is that because the NSF is administering this instrument to a nationally representative US sample at least every year, SHI scores collected by other individuals or groups can compare their data to nationally representative data. Finally, NSF and the task force took care to follow best practices in tool development, with a focus on avoiding respondent fatigue. There are also some limitations to acknowledge. First, the survey asks about diagnosed sleep disorders; however, many sleep disorders go undiagnosed and therefore sleep health may be overestimated in these individuals. The original survey did ask about snoring, a symptom of obstructive sleep apnea. Snoring was not included in the index because factor analysis indicated it did not load strongly onto any of the 3 factors, and it reduced the reliability of the index overall. Furthermore, the inclusion of several sleep quality questions may improve the ability of the instrument to capture underlying sleep disorders. Another possible limitation is that the instrument asks for bed time and wake time rather than actual sleep duration. This will overestimate sleep duration in individuals who spend a lot of time in bed but not sleeping, but the task force of sleep and survey experts determined that the recall accuracy and lower cognitive burden of reporting clock times outweighed this issue. In addition, we adjust the index score to account for differences in sleep duration and time in bed (as described above). As mentioned, the instrument obtains sleep quality information that may help to mitigate this error. Finally, as with any survey, these data are all self-reported; however, there is value in obtaining subjective perceptions of sleep health. Sleep health is a multifaceted concept, and a person's perception of his or her sleep will have important implications for their overall well-being and quality of life. Supported by this empirical testing, the SHI presents a simple 14question tool, in the form of a numerical score out of 100, with which to assess sleep health in large populations. The NSF continues to administer the SHI in nationally representative samples of adults in the United States. As such, the SHI sheds unique light on the state of sleep health in our nation today and over time. Future projects using SHI will enable the general public to compare their individual sleep health with that of the nation, as well as with individuals of similar demographics, backgrounds, and locale. In addition, others interested in assessing sleep health have the opportunity to incorporate this validated measure of sleep health into their own work. In conclusion, the NSF's SHI provides a robust, clear, and concise assessment of the American public's sleep health, a critical component of well-being. Based on nationally-representative public opinion survey
7
data, the SHI demonstrates strong validity and reliability, marking its value for researchers and policymakers as well as the general public. Disclosures G.L. and J.P. provided data analysis for this project. M.H. has a consulting agreement with Jazz Pharma. K.L. is a consultant to Merck. D.S.H. was a consultant for this project. The National Sleep Foundation (NSF), a 501(c)3 charitable and scientific organization, was the sole funder of this project. All other authors have no disclosures. Appendix A. Supplementary data Supplementary data to this article can be found online at http:// dx.doi.org/10.1016/j.sleh.2017.05.011. References 1. Van Dongen HP, Maislin G, Mullington JM, Dinges DF. The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep. 2003;26:117–126. 2. O'Leary K, Small BJ, Panaite V, Bylsma LM, Rottenberg J. Sleep quality in healthy and mood-disordered persons predicts daily life emotional reactivity. Cogn Emot. 2017;31:435–443. 3. Tartar J, Fins A, Lopez A, et al. Sleep restriction and delayed sleep associate with psychological health and biomarkers of stress and inflammation in women. Sleep Health. 2015;1:249–256. 4. Benca RM, Quintas J. Sleep and host defenses: a review. Sleep. 1997;20: 1027–1037. 5. Rangaraj VR, Knutson KL. Association between sleep deficiency and cardiometabolic disease: implications for health disparities. Sleep Med. 2015;18:19–35. 6. Ford D, Cooper-Patrick L. Sleep disturbances and mood disorders: an epidemiologic perspective. Depress Anxiety. 2001;14:3–6. 7. Buysse DJ, Reynolds III CF, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28:193–213. 8. Johns MW. Reliability and factor analysis of the Epworth sleepiness scale. Sleep. 1992;15:376–381. 9. Buysse DJ, Yu L, Moul DE, et al. Development and validation of patient-reported outcome measures for sleep disturbance and sleep-related impairments. Sleep. 2010;33:781–792. 10. Buysse DJ. Sleep health: can we define it? Does it matter? Sleep. 2014;37:9–17. 11. Hirshkowitz M. Editorial. Sleep Health. 2017-this issue. 12. American Association for Public Opinion Research. Best practices for survey research. [last accessed May 23, 2017]; Available from: http://www.aapor.org/Standards-Ethics/Best-Practices.aspx. 13. Krosnick JA, Presser S. Question and questionnaire design. In: Wright JD, Marsdent PV, editors. Handbook of Survey Research. West Yorkshire, England: Emerald Group Publishing Ltd; 2010. p. 263–313. 14. Willis GB. Cognitive Interviewing: A Tool for Improving Questionnaire Design. Sage Publications; 2004. 15. Hirshkowitz M, Whiton K, Albert S, et al. National Sleep Foundation's updated sleep duration recommendations: final report. Sleep Health. 2015;1:233–243. 16. Johns MW. A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep. 1991;14:540–545. 17. Cella D, Riley W, Stone A, et al. The Patient-Reported Outcomes Measurement Information System (PROMIS) developed and tested its first wave of adult self-reported health outcome item banks: 2005-2008. J Clin Epidemiol. 2010;63:1179–1194. 18. Yu L, Buysse DJ, Germain A, et al. Development of short forms from the PROMIS (TM) sleep disturbance and sleep-related impairment item banks. Behav Sleep Med. 2012;10:6–24.
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
Contents lists available at ScienceDirect
Sleep Health Journal of the National Sleep Foundation journal homepage: sleephealthjournal.org
The National Sleep Foundation's Sleep Health Index Kristen L Knutson, PhD a,⁎,1, Julie Phelan, PhD b,1, Michael J. Paskow, MPH c, Anita Roach, MS c, Kaitlyn Whiton, MHS c,d, Gary Langer, BA b, D. Sunshine Hillygus, PhD e, Michael Mokrzycki, BS f, William A. Broughton, MD g, Sudhansu Chokroverty, MD, FRCP h,i, Kenneth L. Lichstein, PhD j, Terri E. Weaver, PHD, RN, FAAN k, Max Hirshkowitz, PhD, DABSM c,l,m a
Northwestern University, Chicago, IL Langer Research Associates, New York, NY c National Sleep Foundation, Arlington, VA d The Hilltop Institute, University of Maryland, Baltimore County, Baltimore, MD e Duke University, Durham, NC f Mokrzycki Survey Research Services, West Newbury, MA g University of South Alabama School of Medicine, Mobile, AL h JFK New Jersey Neuroscience Institute, Edison, NJ i Seton Hall University School of Graduate Medical Education, South Orange, NJ j University of Alabama, Tuscaloosa, AL k University of Illinois at Chicago, Chicago, IL l Baylor College of Medicine, Houston, TX m Stanford University, Stanford, CA b
a r t i c l e
i n f o
Article history: Received 30 January 2017 Received in revised form 26 May 2017 Accepted 30 May 2017 Available online xxxx
a b s t r a c t Objectives: A validated survey instrument to assess general sleep health would be a useful research tool, particularly when objective measures of sleep are not feasible. Thus, the National Sleep Foundation spearheaded the development of the Sleep Health Index (SHI). Design: The development of the SHI began with a task force of experts who identified key sleep domains and questions. An initial draft of the survey was created and questions were refined using cognitive testing and pretesting. The resulting 28-question survey was administered via random-sample telephone interviews to nationally representative samples of adults in 2014 (n = 1253) and 2015 (n = 1250). These data were combined to create the index. A factor analysis linked 14 questions to 3 discrete domains: sleep quality, sleep duration, and disordered sleep. These were assembled as sub-indices, then combined to form the overall SHI, with scores ranging from 0 to 100 (higher score reflects better sleep health). Results: Americans earned an overall SHI score of 76/100, with sub-index scores of 81/100 in disordered sleep, 79/100 in sleep duration, and 68/100 in sleep quality. In regression analyses, the strongest independent predictors of sleep health were self-reported stress (β = −0.26) and overall health (β = 0.26), which were also the strongest predictors of sleep quality (β = −0.32 and β = 0.27 respectively). Conclusions: The current 12-item SHI is a valid, reliable research tool that robustly measures 3 separate but related elements of sleep health—duration, quality, and disorders—and assesses the sleep health status of adults in the United States. © 2017 The Authors. Published by Elsevier Inc. on behalf of National Sleep Foundation. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Introduction Sleep is essential for optimal cognitive performance, physiological processes, emotional regulation, and quality of life.1–6 Research consistently demonstrates that sleep is a significant component of ⁎ Corresponding author. E-mail address: [email protected] (K.L. Knutson). 1 Co-primary authors.
physical and mental health, as well as overall well-being. Therefore, a comprehensive evaluation of an individual's health and wellness necessitates an assessment of sleep health. Unfortunately, objective measures of sleep, such as the gold-standard polysomnography, can be impractical and expensive and therefore infeasible for many large-scale studies, especially when the research is not primarily focused on sleep. A valid and reliable self-reported sleep health measure would therefore be a useful research tool, benefiting ongoing and future research studies across multiple fields of health.
http://dx.doi.org/10.1016/j.sleh.2017.05.011 2352-7218/© 2017 The Authors. Published by Elsevier Inc. on behalf of National Sleep Foundation. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
2
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
Sleep surveys and questions certainly have been used previously. Several large research studies of sleep and health relied on subjective reports of sleep; however, substantial heterogeneity exists in the questions asked to assess sleep duration and quality, which limits the ability to compare results. Moreover, many existing questionnaires that assess sleep focus on the clinical and sleep-disturbance dimensions of sleep health. This includes widely used instruments such as the Pittsburgh Sleep Quality Index (PSQI), 7 whose goal was to measure sleep quality in clinical populations, the Epworth Sleepiness Scale, 8 which assesses only daytime sleepiness, and the National Institutes of Health Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance scale, which focuses on disturbed sleep and is part of the set of patient-centered measures developed by the National Institutes of Health. 9 Currently, no uniform definition of “sleep health” exists, although the National Sleep Foundation (NSF) is working on developing its own definition. Sleep health is not merely the absence of a sleep disorder or problem, and therefore, to adequately assess general sleep health, multiple dimensions of sleep should be assessed. 10 For example, overall sleep health may include the quantity, quality, and impact of sleep, which are all essential, especially for the large proportion of the population that does not suffer from sleep pathologies. The National Sleep Foundation (NSF) set out to develop a single survey to gauge “sleep health” among adults in the United States.11 A critical mission of the NSF is to improve health and well-being through improving sleep health in the United States and abroad; consequently, it was important for the instrument to assess sleep health in the general population. Historically, the NSF conducted annual polls on sleep-related topics; however, the polls' topics and samples varied each year, which did not allow for analysis of trends across time. Thus, the NSF developed a poll to benchmark sleep health trends longitudinally. This poll, the Sleep Health Index (SHI), was developed and validated so that researchers and the public alike will better understand sleep health in the general population. The ultimate goal is to extend sleep health research, improve wellbeing, and empower individuals to understand the importance of sleep. Methods Development of the Sleep Health Index The SHI was developed as a poll to assess multiple dimensions of sleep health to generate an overall SHI score. The development of the SHI, over the course of 4 years, included assembling a task force, identifying specific topic areas, developing and testing the questions, surveying a beta version of the instrument, refining the instrument, and developing and testing the index. First, a task force was established in October 2013 as an initial step in this process. This task force included experts in sleep, polling, and survey methodology, as well as members of the polling firm,
Princeton Survey Research Associates International. The task force members were initially asked to provide questions to address several domains related to sleep and health: general health, sleep habits, sleep schedules, sleep quality, sleep problems, sleep environment, sleep knowledge, and sleep beliefs. From an extensive list of suggested questions, a first draft of the poll was created and questions were subsequently refined between December 2013 and June 2014 using an iterative process involving conference calls and refining poll questions among task force members. Per best-practice guidelines,12 an important goal was to minimize respondent burden by retaining the minimum number of questions necessary to achieve our goals. One significant area of discussion among the task force was appropriate recall period. To minimize error due to poor recall, the task force selected a period of 7 days. Another decision was to use time in bed instead of “actual sleep” to minimize cognitive burden due to the amount of mental calculations and estimation the respondent would need to complete. 13 Recalling a clock time requires a smaller cognitive burden than calculating time in bed and adjusting for sleeplessness. The instrument was further refined after cognitive testing and pretesting in July 2014. Ten (4 women/6 men) cognitive interviews were conducted in-person using the verbal probing technique. 14 The cognitive interviews were used to determine if respondents understood the questions, both consistently across subjects and in the way intended. Next, a pilot test of the questionnaire was fielded to an online convenience sample of 167 respondents to test question wording alternatives and to inform questionnaire reduction. After the cognitive testing and initial online pilot test, a pretest of the revised poll was conducted in August 2014 over the telephone with 24 subjects to determine the length of the questionnaire administration and to further identify questions that needed refinement. This led to a penultimate version of the sleep health survey with 28 substantive questions addressing all the topics listed above. The penultimate version of the instrument was administered via random-sample telephone interviews to a nationally representative sample of adults in 2014 (n = 1253) and 2015 (n = 1250). The basic disclosure elements per the American Association for Public Opinion Research 12 are provided in Table 1. These 2 surveys were used to further refine the instrument and create the SHI score. We eliminated questions that did not pertain directly to individual sleep health based on conceptual grounds because they were less directly related to the experience of sleep. For example, the questions about general health and quality of life were eliminated because although sleep is associated with general health and quality of life, it is not a direct measure of sleep itself. Similarly, questions about the sleep environment, electronic use, beliefs, and knowledge about sleep were eliminated because although they likely impact sleep, they, too, are not direct measures of the experience of sleep itself. We also excluded a question about napping behavior because it is not clear whether a nap is an indicator of good or poor sleep health, and indeed the answer likely varies. Finally, because our goal was to create a score for each response, we wanted to compare the time
Table 1 Basic disclosure elements details
Survey sponsor Survey/data collection supplier Population represented Sample size Mode of data collection Type of sample Start and end dates of data collection Margin of sampling error for total sample Data weighting
2014 SHI
2015 SHI
National Sleep Foundation Princeton Survey Research Associates International (PSRAI) US adults 1253 Telephone (RDD landline and cell) Probability September 8-29, 2014 ±3.1 percentage points Weighted to correct known demographic discrepancies
National Sleep Foundation PSRAI US adults 1250 Telephone (RDD landline and cell) Probability April 27-May 17, 2015 ±3.1 percentage points Weighted to correct known demographic discrepancies
SHI, Sleep Health Index.
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
spent in bed to the times recommended by an expert panel.15 To do this, we fielded a nationally representative survey question in March 2016 to obtain self-reported average time asleep in the previous 7 days. The average was 6.5 hours, compared with an average of 7.7 hours in bed. We used this 1.2-hour difference to score the sleep duration items in this instrument; this value is reevaluated annually for appropriate time adjustments. The 14-question SHI instrument is provided as supplemental material. Covariates The survey used in 2014 and 2015 included additional questions that were used in these analyses. These included demographic information such as age and sex. Educational level was assessed with the question, “What is the highest level of school you have completed or the highest degree?” and 8 responses were available ranging from “Less than high school” to “Post-graduate or professional degree.” Income was based on the question, “Last year—that is, in 2013 [or 2014] —what was your total family income from all sources, before taxes?” and 8 responses were provided ranging from “Less than $20,000” to “$100,000 or more.” The surveys also asked for the number of children 6 years and younger currently living in the household. Respondents were asked to rate their overall health as excellent, very good, good, only fair, or poor. The surveys also asked about life satisfaction based on the question, “All things considered, how satisfied are you with your life as a whole these days?” with 5 possible responses, including extremely satisfied, very satisfied, fairly satisfied, only a little satisfied, or not at all satisfied. The surveys asked about stress using the question, “Overall, how stressed have you felt in the past month?” and responses included extremely stressed, very stressed, fairly stressed, only a little stressed, and not at all stressed. Respondents were asked how often they snore and responses included every night or almost every night, a few nights a week, rarely, or never/I don't snore. The surveys asked the respondent to, “Imagine if you had an extra hour in the day, would you want to…” and the response options were exercise, read, spend time with friends or family, sleep, or work/housework. The survey assessed sleep environment by asking respondents to rate level of light, level of noise, and mattress comfort, and 4 responses were provided ranging from very dark/quiet/comfortable to not at all dark/quiet/ comfortable. The survey also asked how many nights the respondent sent or read text messages, e-mails or other electronic communications after falling asleep, and response options were none, 1-2 nights, 3-4 nights, 5 or more nights. Finally, the survey asked for the number of days in the past week the respondent shared a bed with a partner, child, or pet and if he and/or she texted before or after falling asleep. Analyses We used factor analysis using direct oblimin rotation and principal axis extraction to evaluate the items included in the survey and to identify sub-indices. The cutoff considered for inclusion was 0.4. We tested the sub-indices and the full index for internal consistency, reliability, convergent validity, and known-groups validity. We evaluated the reliability of the SHI using Cronbach α, a test for internal consistency that measures whether items intended to reflect the same construct actually do so. A Cronbach α score of 1 reflects an exact match among all items (indicating redundancy) and 0 indicates no relationship among them. Construct validity is the extent to which measurements reflect the concept they are intended to assess (eg, whether knowledge tests actually measure knowledge). We assessed the construct validity of the SHI by evaluating convergent and known-groups validity. Convergent validity is demonstrated when measures that theoretically should be correlated with the index are in fact correlated. Known-groups validity tests whether groups that are expected to differ actually do so. The “known groups” in these analyses are based on the questions
3
about general health, stress, and life satisfaction. We also compare the SHI scores to other factors expected to be associated with sleep health, including how respondents would spend an extra hour in the day, use of sleep disruptors, sleep beliefs, the sleeping environment, and the presence of a sleep disorder. We calculated the scores for the overall SHI as well as the subindices for 2014 and 2015 individually and combined. Scores range from 0 to 100, and the scoring algorithm will be evaluated and adjusted as necessary every quarter. Thus, anyone interested in using this survey should contact the NSF ([email protected]). We tested for differences between the years using Student's t test. Student's t test and regression models were used to test for differences in SHI scores across demographic categories. Pearson correlation coefficients were used to test for associations between the SHI scores and other measures thought to be associated with sleep health. Finally, linear regression analyses were performed to determine which factors are predictive of the SHI overall score and its sub-indices. The covariates included in the regression models were based on the available questions in the full 2014 and 2015 survey, including sex, age, race, marital status, employment status, education, income, number of children, subjective health, life satisfaction, amount of stress, frequency of snoring, mattress comfort, bedroom darkness, and quietness of bedroom, as well as (separately) the number of days in the past 7 days the respondent shared a bed with a partner, child, or pet, and if he and/or she used a computer/tablet before bed and texted/e-mailed after falling asleep. A P value of .001 was selected as significant to be conservative given the large sample size. Statistical analyses were performed using SPSS (v19; IBM SPSS Statistics, Chicago, IL) and Stata SE survey commands (v14; StataCorp, College Station, TX).
Results A description of respondent characteristics for both the 2014 and 2015 surveys, separately and combined, is provided in Table 2. The Table 2 Sample composition (unweighted sample size and weighted % of population) combined and by year Combined
Total Men Women 18-29 y 30-39 y 40-49 y 50-64 y 65+ y White Black Latino Other Married/partnered Not married/partnered Employed Not employeda No high school diploma High school graduate Some college College graduate b$20,000 $20,000-$39,999 $40,000-$59,999 $60,000-$79,999 $80,000-$99,999 $100,000+
2014
n
%
n
2503 1178 1325 463 258 322 679 761 1726 280 344 120 1445 1045 1261 1225 241 772 650 820 454 464 371 266 166 360
100 48 52 22 15 17 27 18 66 12 15 7 60 39 57 43 10 31 31 28 20 19 15 10 7 14
1253 580 673 223 114 165 355 387 871 135 169 60 722 526 628 616 104 383 337 419 228 244 172 139 77 175
2015 %
n
%
48 52 22 14 18 28 18 66 12 15 7 59 40 57 43 10 32 31 28 20 20 14 11 7 14
1250 598 652 240 144 157 324 374 855 145 175 60 723 519 633 609 137 389 313 401 226 220 199 127 89 185
49 51 22 16 17 26 18 65 12 15 7 62 38 56 44 11 30 30 28 20 18 16 10 7 14
a Not employed includes respondents who are unemployed, students, retired, on disability or a stay-at-home parent.
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
4
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
table provides both the unweighted sample sizes within each group and proportion weighted to be nationally representative. Factor analysis Factor analysis led to the exclusion of a question about snoring frequency. Snoring was not included in the index because factor analysis indicated that it did not load strongly onto any of the 3 factors, and it reduced the reliability of the index overall. There were 2 questions whose factor loadings were below 0.4 but which were retained because of expert opinion favoring their inclusion as important components. These included unintentionally dozing and use of sleep medication. Factor analysis of the final 14 questions was used to identify 3 discrete domains: sleep quality, sleep duration, and disordered sleep. These were assembled as sub-indices that were then combined to form the overall SHI. These sub-indices are described in detail below. The factor loadings of the final 12 items are presented in Table 3. Index composition The final sub-indices were sleep quality, sleep duration, and disordered sleep. The sleep quality sub-index includes 6 items: • Respondents' overall ratings of their sleep quality • The number of days in the past 7 days that respondents: ○ felt well-rested, ○ had trouble falling asleep, ○ had trouble staying asleep, ○ were negatively impacted by lack of sleep, and ○ dozed unintentionally. The sleep duration sub-index includes 3 items: • A weekday sleep score reflecting how well respondents' selfreported time in bed aligned with NSF's published recommendations for optimal sleep.15 Time in bed was calculated based on the time respondents reported that they most often went to bed and woke up on weekdays in the previous 7 days. Time in bed was scored against expert recommendations for optimal sleep adjusting for the difference between time in bed and sleep duration. • A sleep deficit score based on the difference between weekday sleep and the amount of sleep respondents said they needed to feel their best • A “sleep variability” score reflecting the extent to which respondents' weekday time in bed differed from their weekend time in bed. This is a measure of variability in duration. The disordered sleep index includes 3 items:
Table 3 Factor loadings of each question on each of the 3 sub-indices
Sleep quality rating Well-rested Trouble falling asleep Trouble staying asleep Negative impact Unintentional doze Weekday sleep Sleep deficit Sleep variability Sleep medication Diagnosed sleep disorder Sleep doctor
Quality
Duration
Disorder
.70 .68 .63 .66 .67 .26 .25 .31 .11 .29 .25 .35
.15 .34 .13 .10 .27 .16 .75 .74 .42 −.06 .07 .01
.33 .19 .30 .30 .23 .11 .03 .08 −.01 .35 .64 .85
Bolded numbers show items included in each sub-index.
• The number of days in the past 7 days in which the respondent took a sleep medication • Whether or not the respondent had been told by a doctor that they have a sleep disorder • Whether or not the respondent had discussed his/her sleep with a doctor or medical professional Reliability The factor analysis helped to identify 3 separate but related constructs (Table 3). Cronbach α results indicated high levels of internal consistency, ranging from α = .63 for both disordered sleep and sleep duration to α = .77 for sleep quality in the combined data set. The 12item Cronbach α was .76 for 2014 and .75 for 2015. To assess the reliability of the SHI over time, we repeated this assessment separately for the 2014 and 2015 data and the results were virtually identical. Construct validity We found statistically significant correlations (P b .001) between the SHI and respondents' ratings of their overall health (r = 0.38), stress (r = −0.37), and life satisfaction (r = 0.36). The SHI score was 67 for those who reported having been extremely or very stressed in the previous month, compared with 81 for those who felt little or no stress. Comparative sub-index scores were 54 vs 77, respectively, for sleep quality; 73 vs 82, respectively, for sleep duration; and 74 vs 84, respectively, for disordered sleep. Those who rated their overall health as excellent or very good had an SHI score of 81, in comparison to 64 among those who said that their health was just fair or poor. This reflected large differences in sleep quality (76 vs 51) and disordered sleep (88 vs 66), but only a small difference in sleep duration (80 vs 77). Those who were extremely or very satisfied with their lives scored significantly higher on the SHI than did those who were only a little or not at all satisfied with their lives (80 vs 60), including on sleep quality (74 vs 46), disordered sleep (84 vs 66), and sleep duration (81 vs 69). All differences are statistically significant (P b .001). Respondents were asked what they would do with an extra hour in the day. As would be expected, those who said they would use it to sleep, scored lower on the SHI overall (67 vs 78) and all 3 of its subindices—sleep quality (56 vs 71), sleep duration (71 vs 80), and disordered sleep (74 vs 82)—compared with those who would use that hour to work, exercise, read, or socialize. We also examined the relationships between sleep health and measures of possible sleep disruptors (ie, electronic use, and bed partners), sleep beliefs, and sleeping environment (ie, noise, light, and mattress comfort). These 3 constructs were not related to sleep health (all r values b 0.2). The strongest correlation was between sleep quality and comfort of mattress (r = 0.24). Mattress comfort also was linked to longer sleep duration (r = 0.13) and less disordered sleep (r = 0.07), but more weakly. The correlation with the overall SHI was r = 0.20. The SHI was 77 for adults who reported having more comfortable mattresses vs 65 for those with less comfortable ones. Comparative sub-indices were 70 vs 53 for sleep quality, 80 vs 73 for sleep duration, and 82 vs 72 for disordered sleep. There were group differences based on the presence of a sleep disorder. The SHI score was 52 among those who had been told by a doctor that they had a sleep disorder, compared with 80 among those without a sleep disorder. Those with a sleep disorder scored less positively for sleep duration (75 vs 80) and sleep quality (56 vs 71). They also scored considerably worse on the disordered sleep sub-index (28 vs 91), but that was largely because this sub-index includes having a sleep disorder. There were also differences based on the use of sleep medication. The overall SHI was 59 among adults who had taken a sleep
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
medication at least once in the past 7 days, compared with 79 among those who had not, including 54 vs 71 for sleep quality. There were no differences in sleep duration between those who used and those who did not use sleep medication. SHI scores Americans earned an average SHI score of 76 of a possible score of 100, with underlying scores of 81 in disordered sleep, 79 in sleep duration, and 68 in sleep quality (Table 4). Actual scores ranged from 1 to 100 for the SHI score and 0 to 100 for each sub-index. The 2014 and 2015 surveys produced virtually identical results. For the 2014 SHI, the public scored an overall SHI of 75, including scores of 80 for disordered sleep, 79 for sleep duration, and 68 for sleep quality (a higher score reflects greater sleep health). The 2015 SHI scores were 77 overall, 82 for disordered sleep, 80 for sleep duration, and 69 for sleep quality. None of the 1- or 2-point differences between 2014 and 2015 data were statistically significant. The SHI scores stratified by demographic groups are presented in Table 5. Mean SHI was significantly (P b .05) higher in men, among older (65+ years) and younger adults (18-29, 30-39 years), among whites and Hispanic/ Latinos, among married respondents, among employed respondents, among college graduates, and among those with higher income. Sleep quality sub-index The sleep quality sub-index includes scores for self-reported sleep quality as well as scores for how many days out of the past 7 the respondent felt well rested, had trouble falling asleep, had trouble staying asleep, was negatively impacted by lack of sleep, or dozed unintentionally. Overall, Americans had a score of 68 on this sub-index, including a score of only 49 for self-reported sleep quality (Table 4). Just 3 in 10 rated their sleep quality as excellent (11%) or very good (19%). Thirty-five percent described their sleep quality as “good,” 22% as “fair,” and 12% as “poor.” Scores on the other items in the sub-index ranged from a score of 55, for the number of days the respondent felt well rested in the last 7 days, to a score of 82, for the number of days the respondent dozed unintentionally. Respondents indicated feeling well rested an average of 3.9 days of the previous 7 days. They had trouble falling asleep 1.7 nights and staying asleep 2.1 nights of the previous 7 days on Table 4 SHI overall scores, sub-indices scores, and the scores by individual questions in the combined sample and by year Combined years
2014
2015
76 (75-77)
75 (74-76)
77 (76-78)
68 (67-69) 49 (48-51) 55 (54-57) 69 (68-71) 76 (75-78) 78 (77-80) 82 (81-83)
68 (66-69) 49 (47-51) 54 (52-56) 69 (66-71) 76 (74-78) 78 (76-80) 82 (80-83)
69 (67-70) 50 (48-51) 57 (54-59) 70 (68-73) 77 (74-79) 79 (77-81) 83 (81-84)
79 (78-80) 67 (65-68) 84 (83-84) 88 (87-89)
79 (77-80) 66 (64-68) 84 (82-85) 87 (85-88)
80 (78-81) 68 (66-70) 84 (82-85) 89 (87-90)
81 (79-82) 69 (67-71) 85 (83-86) 89 (87-90)
80 (78-82) 68 (66-71) 83 (81-85) 88 (86-90)
82 (80-83) 70 (67-73) 86 (84-89) 89 (87-91)
Mean (95% CI) SHI Sleep quality score Sub-index overall Sleep quality rating Well-rested Trouble staying asleep Trouble falling asleep Negative impact Unintentional dozing Sleep duration score Sub-index overall Weekday sleep Sleep deficit Sleep variability Disordered sleep score Sub-index overall Sleep doctor Diagnosed disorder Sleep medication
CI, confidence interval; SHI, Sleep Health Index.
5
Table 5 Scores on the SHI and its sub-indices by groupsa
Total Men Women P 18-29 30-39 40-49 50-64 65+ P White Black Latino Other P Married/partnered Not married/partnered P Employed Not employedb P No high school diploma High school graduate. Some college College graduate P b$20,000 $20,000-$39,999 $40,000-$59,999 $60,000-$79,999 $80,000-$99,999 $100,000+ P
SHI
Quality
Duration
Disorder
76 77 75
68 70 66
79 78 80
81 82 80
.18 76 78 75 80 86 b.0001 81 72 78 78 b.0001 80 78 .026 77 82 b.0001 76 79 79 81 .05 77 78 79 78 79 83 .008
.05 89 86 77 74 79 b.0001 79 81 85 89 .0001 81 80 .44 84 76 b.0001 81 82 78 82 .07 78 80 82 81 80 81 .11
.035 78 77 73 73 79 b.0001 76 72 76 80 .002 77 75 .004 77 74 .0001 74 76 75 78 .0004 71 75 77 77 77 78 b.0001
.0002 70 67 66 66 72 .0005 69 63 67 73 .0001 70 66 .0005 71 65 b.0001 64 66 68 72 b.0001 59 65 70 73 73 73 b.0001
SHI, Sleep Health Index a Student t tests were used to compare dichotomous variables and regression analyses tested for differences among N2 categories. b Not employed includes respondents who are unemployed, students, retired, on disability, or a stay-at-home parent.
average. Furthermore, they were impacted by lack of sleep an average of 1.5 days and dozed unintentionally on 1.3 days of the previous 7 days. Mean scores on the sleep quality sub-index were compared across demographic groups (Table 5). Mean scores were significantly (P b .05) higher for men, younger (18-29 years) and older (65+ years) adults, respondents from “other” race/ethnicity, married respondents, employed respondents, college graduates, and those with higher incomes ($60,000/y or more). Sleep duration sub-index The average sleep duration sub-index score was a 79, with underlying scores including 67 for weekday sleep duration, 84 for sleep deficit, and 88 for sleep variability (Table 4). Americans reported spending an average of 7.7 hours in bed on weekdays. Roughly half reported spending between 7 and 9 hours in bed—a quarter spent less time than that, and 23% averaged more than 9 hours in bed on weeknights. Individual weekday sleep scores were assigned relative to how far respondents were from sleep duration recommendations established by an expert panel convened by the NSF. 15 Twenty-six percent of respondents fell into the recommended range. The majority—68%—spent less time in bed than recommended. Respondents reported needing an average of 7.1 hours of sleep to feel their best. Forty-four percent of these individuals fell short of the average by an hour or more. Among the remainder, 28% got as much (or more) sleep than they felt they needed, and 29% had a deficit of less than an hour. Sleep deficit scores were assigned based on how
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
6
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
close respondents' approximate weekday sleep was to meeting their self-assessed sleep needs. In terms of sleep variability, the average difference between weekday and weekend time in bed was just more than 1 hour. Half of the respondents had a difference of less than an hour, but 22% of respondents had a difference of 2 or more hours between weekday and weekend time in bed. Mean scores on this sub-index were also compared across demographic groups (Table 5) and these scores were significantly (P b .05) higher, indicating better sleep health, for older adults, whites, married respondents, respondents who were not employed, those with greater education, and those with more income. Disordered sleep index The overall score of 81 on the disordered sleep sub-index included subcategory scores of 69 for talking to a doctor about their sleep, 85 for diagnosed sleep disorders, and 89 for sleep medication use (Table 4). Three in 10 respondents had discussed sleep issues with a doctor or medical professional, whereas the 69% who had not discussed sleep problems with a doctor scored 100 on this item. Similarly, the score of 85 for disordered sleep reflects the fact that more than 8 in 10 people indicated that they do not have a diagnosed sleep disorder. Prescription or over-the-counter sleep aids were taken an average of 0.8 of 7 nights. Eighty-three percent had not taken a sleep aid in the past 7 days. Among those who had taken medication, most (51%) did so on all 7 nights of the previous week. Finally, we compared the scores on this sub-index across demographic groups (Table 5), and scores were significantly (P b .05) higher for younger adults, respondents who identify as Latino/Hispanic and other race/ethnicity, and employed respondents. Regressions and group differences In regression analyses (Table 6), the strongest independent predictors of sleep health were self-reported stress—a strong predictor of sleep quality–and overall health. Weaker predictors included life satisfaction, frequency of texting or e-mailing after initially falling asleep, frequency of snoring, and mattress comfort. As would be expected, life satisfaction and mattress comfort positively predicted the SHI, whereas texting and snoring were related to poorer sleep health. The separate regression models predicting sleep quality, sleep duration, and disordered sleep offer further evidence that these subindices tap into distinct constructs related to sleep health. Stress was a negative predictor of all 3 of the sub-indices, and a particularly strong predictor of sleep quality. Overall health was associated with both sleep quality (β = 0.27) and disordered sleep (β = 0.22), but Table 6 Results from regression analyses Outcome SHI
Quality
Duration
Disordered
Regression coefficient (β) Stress Health Life satisfaction Frequency text after sleep Frequency snore Mattress comfort Employed Age Adjusted R2
−0.25 0.25 0.14 −0.11 −0.10 0.08 – – 0.29
SHI, Sleep Health Index – = not statistically significant (P N .001)
−0.31 0.26 0.14 −0.12 −0.08 0.11 – – 0.37
−0.13 – 0.10 −0.11 – – −0.11 – 0.10
−0.11 0.21 – – −0.10 – – −0.08 0.13
not sleep duration. Frequency of texting after falling asleep was negatively associated with sleep quality and duration, whereas frequency of snoring was negatively associated with sleep quality and disordered sleep, but not duration. Mattress comfort was positively related to sleep quality, but not to sleep duration or disordered sleep. People with an uncomfortable mattress stayed in bed just as long as those who did not report having an uncomfortable mattress, but the data suggest that the quality of sleep while in bed was poorer. Employment negatively predicted sleep duration, but not the other 2 sub-indices. Those with a full- or part-time job spent less time in bed on weekdays than did those without a job. Older age predicted disordered sleep, but not sleep quality or duration. Older adults were more apt than younger adults to take sleep aids, have a diagnosed sleep disorder, and talk to their doctor about sleep. Conclusions Given its inclusion of 3 separate but related elements of sleep health—duration, disorders, and quality, each measured in multiple survey questions—the SHI is a robust, valid measure of sleep health. The statistical testing described in this report indicates validity and reliability in the underlying data and index construction alike. Furthermore, the SHI score varies between groups expected to demonstrate differences, including higher scores (better sleep health) among those reporting lower stress, better general health, and greater life satisfaction. Those who reported that they would spend an extra hour in the day sleeping had lower SHI scores. Sleep disruptors (such as electronic use or bed partners), beliefs about sleep, and the sleeping environment were not related to SHI scores. This may reflect the nature of the items. For example, sleeping with a partner, a pet, or a child may be positively or negatively linked to sleep health, depending on the sleeper's preferences and the sleep partner's characteristics. This leads to a weak correlation overall, although the factors may be strongly linked at the individual level. Another possible explanation is that these factors do not affect sleep health strongly, at least as assessed by the questions in the survey. The overall score and the sub-indices also showed several consistent differences between demographic groups, particularly higher scores among men, among those of higher socioeconomic status (ie, higher education, income, or those employed), and among some racial/ethnic groups. Whether these differences in SHI scores reflect meaningful differences in sleep health, or health in general, warrants further investigation. The SHI instrument provides a new robust tool for assessing general sleep health in the population. As mentioned, other instruments designed to assess sleep do exist and are widely used in a variety of studies, for example, the PSQI, which was designed as a clinical tool to assess sleep quality and discriminate between good and poor sleepers. 7 It includes several factors that disturb sleep, subjective sleep quality, medication use, and 2 questions pertaining to daytime dysfunction. The PSQI differs from the SHI in 2 distinct ways. First, although it also includes a measure of sleep duration, it does not distinguish between work days/weekdays and non-workdays/weekends. Second, the SHI uses a shorter recall period (7 days vs 1 month) to improve accuracy due to recall error. 13 Another commonly used sleep questionnaire is the Epworth Sleepiness Scale, 8,16 which is an 8-item questionnaire designed to assess only general level of daytime sleepiness and does not assess quality, duration, or disturbed sleep. More recently, sleep-related instruments have been developed for the National Institutes of Health PROMIS health measures. 17,18 There are PROMIS instruments for “sleep disturbances” and “sleeprelated impairment,” and as these names indicate, they focus on disturbances in sleep, including sleep quality, or daytime dysfunction, but do not include sleep duration. The PROMIS sleep disturbance scale includes questions are that similar to 4 of the 6 items in the SHI sleep quality sub-index, specifically questions about overall
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011
K.L. Knutson et al. / Sleep Health xxx (2017) xxx–xxx
sleep quality, restful sleep, and difficulty falling or staying asleep, but does not include questions about impact on daily activities or dozing unintentionally. The selection of a sleep instrument will ultimately depend on the goals of the research. For those interested in general sleep health, rather than a clinical population, the SHI may be the better option, particularly if interested in comparing results to the nationally representative results obtained and published by the NSF (sleepfoundation.org). The strengths of this instrument include the multidimensional aspects of sleep health that have been incorporated and the multidisciplinary task force that comprised both sleep and polling/survey experts. Another important strength is that because the NSF is administering this instrument to a nationally representative US sample at least every year, SHI scores collected by other individuals or groups can compare their data to nationally representative data. Finally, NSF and the task force took care to follow best practices in tool development, with a focus on avoiding respondent fatigue. There are also some limitations to acknowledge. First, the survey asks about diagnosed sleep disorders; however, many sleep disorders go undiagnosed and therefore sleep health may be overestimated in these individuals. The original survey did ask about snoring, a symptom of obstructive sleep apnea. Snoring was not included in the index because factor analysis indicated it did not load strongly onto any of the 3 factors, and it reduced the reliability of the index overall. Furthermore, the inclusion of several sleep quality questions may improve the ability of the instrument to capture underlying sleep disorders. Another possible limitation is that the instrument asks for bed time and wake time rather than actual sleep duration. This will overestimate sleep duration in individuals who spend a lot of time in bed but not sleeping, but the task force of sleep and survey experts determined that the recall accuracy and lower cognitive burden of reporting clock times outweighed this issue. In addition, we adjust the index score to account for differences in sleep duration and time in bed (as described above). As mentioned, the instrument obtains sleep quality information that may help to mitigate this error. Finally, as with any survey, these data are all self-reported; however, there is value in obtaining subjective perceptions of sleep health. Sleep health is a multifaceted concept, and a person's perception of his or her sleep will have important implications for their overall well-being and quality of life. Supported by this empirical testing, the SHI presents a simple 14question tool, in the form of a numerical score out of 100, with which to assess sleep health in large populations. The NSF continues to administer the SHI in nationally representative samples of adults in the United States. As such, the SHI sheds unique light on the state of sleep health in our nation today and over time. Future projects using SHI will enable the general public to compare their individual sleep health with that of the nation, as well as with individuals of similar demographics, backgrounds, and locale. In addition, others interested in assessing sleep health have the opportunity to incorporate this validated measure of sleep health into their own work. In conclusion, the NSF's SHI provides a robust, clear, and concise assessment of the American public's sleep health, a critical component of well-being. Based on nationally-representative public opinion survey
7
data, the SHI demonstrates strong validity and reliability, marking its value for researchers and policymakers as well as the general public. Disclosures G.L. and J.P. provided data analysis for this project. M.H. has a consulting agreement with Jazz Pharma. K.L. is a consultant to Merck. D.S.H. was a consultant for this project. The National Sleep Foundation (NSF), a 501(c)3 charitable and scientific organization, was the sole funder of this project. All other authors have no disclosures. Appendix A. Supplementary data Supplementary data to this article can be found online at http:// dx.doi.org/10.1016/j.sleh.2017.05.011. References 1. Van Dongen HP, Maislin G, Mullington JM, Dinges DF. The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep. 2003;26:117–126. 2. O'Leary K, Small BJ, Panaite V, Bylsma LM, Rottenberg J. Sleep quality in healthy and mood-disordered persons predicts daily life emotional reactivity. Cogn Emot. 2017;31:435–443. 3. Tartar J, Fins A, Lopez A, et al. Sleep restriction and delayed sleep associate with psychological health and biomarkers of stress and inflammation in women. Sleep Health. 2015;1:249–256. 4. Benca RM, Quintas J. Sleep and host defenses: a review. Sleep. 1997;20: 1027–1037. 5. Rangaraj VR, Knutson KL. Association between sleep deficiency and cardiometabolic disease: implications for health disparities. Sleep Med. 2015;18:19–35. 6. Ford D, Cooper-Patrick L. Sleep disturbances and mood disorders: an epidemiologic perspective. Depress Anxiety. 2001;14:3–6. 7. Buysse DJ, Reynolds III CF, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28:193–213. 8. Johns MW. Reliability and factor analysis of the Epworth sleepiness scale. Sleep. 1992;15:376–381. 9. Buysse DJ, Yu L, Moul DE, et al. Development and validation of patient-reported outcome measures for sleep disturbance and sleep-related impairments. Sleep. 2010;33:781–792. 10. Buysse DJ. Sleep health: can we define it? Does it matter? Sleep. 2014;37:9–17. 11. Hirshkowitz M. Editorial. Sleep Health. 2017-this issue. 12. American Association for Public Opinion Research. Best practices for survey research. [last accessed May 23, 2017]; Available from: http://www.aapor.org/Standards-Ethics/Best-Practices.aspx. 13. Krosnick JA, Presser S. Question and questionnaire design. In: Wright JD, Marsdent PV, editors. Handbook of Survey Research. West Yorkshire, England: Emerald Group Publishing Ltd; 2010. p. 263–313. 14. Willis GB. Cognitive Interviewing: A Tool for Improving Questionnaire Design. Sage Publications; 2004. 15. Hirshkowitz M, Whiton K, Albert S, et al. National Sleep Foundation's updated sleep duration recommendations: final report. Sleep Health. 2015;1:233–243. 16. Johns MW. A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep. 1991;14:540–545. 17. Cella D, Riley W, Stone A, et al. The Patient-Reported Outcomes Measurement Information System (PROMIS) developed and tested its first wave of adult self-reported health outcome item banks: 2005-2008. J Clin Epidemiol. 2010;63:1179–1194. 18. Yu L, Buysse DJ, Germain A, et al. Development of short forms from the PROMIS (TM) sleep disturbance and sleep-related impairment item banks. Behav Sleep Med. 2012;10:6–24.
Please cite this article as: Knutson KL, et al, The National Sleep Foundation's Sleep Health Index, Sleep Health (2017), http://dx.doi.org/ 10.1016/j.sleh.2017.05.011