Original Research••

Sleep patterns and long-term mortality among older Israeli adults: a population-based study

Abstract

Introduction The joint association of night-time sleep duration and daytime napping (siesta) with mortality remains elusive. We explored sleep patterns and long-term mortality in older adults and tested whether the relationship is modified by cognitive function.

Methods We analysed data from 1519 participants in the National Health and Nutrition Survey of older adults aged 65+ years (‘Mabat Zahav’), conducted by the Israel Center for Disease Control during 2005–2006. A detailed questionnaire was administered at study entry to gather information on sleeping habits, including night-time sleep duration and siesta. A Mini-Mental State Examination was administered to assess cognitive status (score <27 considered impaired). Mortality data were obtained from the Ministry of Health (last follow-up: June 2019; 782 deaths). Cox models were constructed to estimate the HRs for mortality associated with sleep patterns, defined according to night sleep duration (>8 vs ≤8 hours) and siesta (Y/N). Spline regression models were constructed to examine the linearity of the association across cognitive statuses.

Results Sleep categories among participants (mean age 75; 53% women) included 291 (19.2%) with long night sleep and siesta, 139 (9.1%) with long night sleep and no siesta, 806 (53.1%) with short night sleep and siesta, and 283 (18.6%) with short night sleep and no siesta. HRs for mortality were 2.07 (95% CI: 1.63 to 2.62), 1.63 (95% CI: 1.22 to 2.18) and 1.43 (95% CI: 1.16 to 1.76) in the former three versus latter sleep patterns, respectively. Multivariable adjustment for sociodemographic, behavioural and clinical covariates attenuated the HRs to 1.27–1.41 (all p<0.05). The relationship between night sleep duration and mortality was linear (plinearity=0.047) among cognitively preserved individuals and U-shaped (pnon-linearity<0.001) among cognitively impaired ones.

Conclusions Prolonged night sleep and siesta were associated with increased mortality, a relationship that varied by cognitive performance.

What is already known on this topic

  • Sleep patterns are associated with adverse outcomes, especially among vulnerable populations.

What this study adds

  • In a national cohort of older adults (mean age 75 years), we found the lowest mortality risk associated with sleep duration ≤8 hours without a siesta.

  • We observed a linear relationship between sleep duration and mortality among cognitively preserved individuals and a U-shaped pattern in cognitively impaired individuals.

  • Siesta was associated with increased mortality among cognitively preserved individuals only.

How this study might affect research, practice or policy

  • These findings could influence sleep recommendations and inform clinical practice targeting vulnerable populations.

Introduction

Sleeping habits are a fundamental part of overall health and well-being, and their significance in public health, especially among ageing populations, has become increasingly apparent.1

It is well established that sleep patterns are associated with various health outcomes.2 Emerging evidence shows an association between indicators of suboptimal sleeping habits and a multitude of health outcomes, including hypertension,3 metabolic disorders,4 stroke,5 coronary artery disease,5 respiratory diseases,6 chronic pain,6 and all-cause and cause-specific7 mortality. These findings underscore the critical role of sleep patterns in the burden of chronic morbidity on public health, especially among vulnerable populations such as older adults.8

Siesta, a short nap typically taken during the afternoon, is attributed to lifestyle and cultural customs. Napping prevalence is regionally variable globally,9 stressing the need for nation-specific studies. Recent studies have shown an association between daytime napping and morbidity and mortality,9 10 including among older Israeli adults.11 Although siesta and night-time sleep duration are arguably complementary behaviours, their interaction in the context of mortality among older adults has yet to be elucidated.

Sleep patterns correlate with lifestyle, habits and sociodemographic factors,11 12 potential confounders for the association with mortality. This emphasises the importance of rich datasets, measuring and accounting for different aspects of daily living. Cognitive impairment, a prevalent condition among older adults, is also associated with changes in sleep patterns13 and may, therefore, hold a moderating effect on the sleep–mortality relationship. This possible effect modification has yet to be studied.

We explored the joint association of sleep duration and siesta with long-term mortality in older Israeli adults (65+ years). We also investigated the independent associations of sleep duration and siesta with mortality and whether cognitive function modifies these relationships.

Methods

Study design and setting

This cohort comprises participants interviewed in the first National Health and Nutrition Survey of the Elderly in Israel (‘Mabat Zahav’), conducted between 2005 and 2006 by the Israel Center for Disease Control and the Nutrition Department of the Israel Ministry of Health (MOH). The survey was conducted among members of the two major health maintenance organisations in Israel (Clalit Health Services and Maccabi Health Services), representing approximately 86% of all elderly individuals in Israel. Both organisations randomly sampled community-dwelling Israeli citizens aged 65 years or older residing in Israel for at least 1 year. Personal interview was conducted at baseline in the interviewee’s place of residence (own home or retirement home).

Among the 1852 participants interviewed, 66 were excluded due to severe cognitive impairment (n=46) or incomplete questionnaires (n=20). As detailed below, an additional 267 participants lacked data on sleep habits and were excluded from the analysis, leaving 1519 participants. Participants were surveyed for information on sociodemographic status, pre-existing medical conditions, disabilities, mental health, cognitive function, anthropometric measures and nutritional status. All questionnaires and summarised survey data are detailed on the MOH website14 (English version available).

Assessment of sleep habits

Sleeping habits were self-reported and included the usual time of sleep at night, waking up in the morning and the tendency to nap during the day (a dichotomous indicator). Night-time sleep duration was calculated as the duration between the usual time of going to sleep and the regular time of waking up. A reported night-time sleep duration shorter than 4 or longer than 14 hours was considered erroneous and was treated as missing. Night-time sleep duration longer than 8 hours was considered long, and shorter than or equal to 8 hours was considered short, in line with previous studies, including in the Israeli population.11 15 To examine the joint effect of night-time sleep duration and siesta, we categorised the subjects into four mutually exclusive groups (online supplemental table 1): short night sleep without siesta (short NS-no siesta), short night sleep with siesta (short NS-siesta), long night sleep without siesta (long NS-no siesta) and long night sleep with siesta (long NS-siesta).

Mortality assessment

Data on all-cause mortality (available through June 2019) were obtained by linking the cohort via national identification numbers to the nationwide death registry managed by the MOH.

Assessment of cognitive function

Cognitive function was assessed using the Mini-Mental State Examination (MMSE), with scores ranging from 0 to 30.16 Higher scores indicate better cognitive function. Cognitive impairment was defined as unadjusted scores of 26 and below, aligned with a previously suggested cut-off.17

Baseline covariates

Baseline covariates included demographic measures such as age, sex, ethnicity (Jewish and Arab) and years of education. Neighbourhood socioeconomic status (nSES) was determined by the Central Bureau of Statistics based on a 2008 census, ranging between 1 and 20, with lower values representing deprived residential areas. Behavioural measures were collected, including physical activity level (PAL, based on reported total weekly time of leisure-time physical activity and intensity; participants were classified into three categories: sufficiently active, insufficiently active or inactive, as described previously)18 and smoking status (smoking, previously smoking or never smoking). Clinical measures included functional level (measured via Katz’s activities of daily living score (ADL))19 and multimorbidity index (calculated as the number of pre-existing chronic diseases, including stroke, kidney disease, chronic respiratory diseases, diabetes, hypercholesterolaemia, osteoporosis and hypertension, and analysed categorically (0=no morbidity; 1–3=moderate morbidity; and 4+=severe morbidity)). Mental health was assessed via the 12-item General Health Questionnaire (GHQ) score, ranging between 0 and 36, with lower scores indicating better psychological well-being.20 Self-rated health (SRH) was assessed using a single question (scale 1–4).21

Statistical analyses

Baseline characteristics by sleep patterns were summarised as frequency and percentage for categorical variables and as mean (SD) for continuous variables; they were compared in univariable analyses by Χ2 test and one-way analysis of variance test, respectively. Sleep duration was treated on categorical and continuous scales. Kaplan-Meier estimators were constructed to assess the long-term survival probability across sleep patterns and were compared using the log-rank test. Cox proportional hazards regression models were constructed with time since baseline interview as the time scale to estimate the HRs and 95% CIs for all-cause mortality across sleep patterns. An unadjusted analysis was initially conducted, followed by adjustment for sociodemographic variables (age, sex, ethnicity, years of education and nSES). Finally, models were further adjusted for clinical and behavioural risk factors (PAL, smoking status, ADL, cognitive performance and multimorbidity index). The independent associations of sleep duration and siesta were similarly explored in the entire cohort and stratified by cognitive function, with further adjusting for either siesta or sleep duration, as appropriate. Model covariates were selected a priori if they were well-established mortality risk factors among older adults.22 The proportional hazards assumption was tested using the Schoenfeld residuals, with no violations found in any of the models. Sensitivity analyses for the sleep pattern–mortality associations included further stratification based on clinical recommendations for appropriate night-time sleep duration23 and additional adjustment for GHQ and SRH as possible confounders.24

We applied penalised smoothing splines with four knots to assess the non-linear relationship between sleep duration and mortality in a Cox regression model, adjusted for all sociodemographic, clinical and behavioural covariates included in the final model, including siesta.

Analyses were performed using R software, V.4.3.0 (R Development Core Team).

Results

At study entry, the mean (SD) age of the 1519 participants was 74.6 (6.2) years, and 812 (53%) were women. Overall, 695 participants (46%) were current or past smokers, and 340 (23%) had mild ADL impairment. At baseline, the median (IQR) night-time sleep duration was 7.5 (6.5–8.5) hours, and 430 (28%) of the participants slept longer than 8 hours per night. In addition, 1097 (72%) participants reported napping. Among the cohort, 283 (19%) were in short NS-no siesta group, 806 (53%) were in short NS-siesta groups, 139 (9%) were in long NS-no siesta group and 291 (19%) were in long NS-siesta group. Baseline characteristics differed across exposure groups (table 1). Age differed significantly across the groups, with the long NS-siesta group having the highest average age. The proportion of females was higher in the groups with longer night-time sleep. In general, longer night-time sleep and siesta were associated with worse socioeconomic, clinical and functional status.

Table 1
|
Characteristics of participants in the Mabat Zahav cohort by sleep patterns, Israel, 2005–2019

Over a median (IQR) follow-up period of 13.3 (13.0–13.5) years (among survivors), 782 (51%) participants died. Long-term mortality risk differed across sleep patterns (p<0.001), with the lowest risk among short NS-no siesta, followed by short NS-siesta, long NS-no siesta and long NS-siesta (figure 1).

Figure 1
Request permissionrequest permission icon
Figure 1

Kaplan-Meier curves for probability of survival in the Mabat Zahav population-based cohort of older Israeli adults initiated in Israel in 2005 with follow-up through 2019, according to sleep patterns: short night sleep (≤8 hours per night) without siesta (short NS-no siesta), short night sleep with siesta (short NS-siesta), long night sleep (>8 hours per night) without siesta (long NS-no siesta) and long night sleep with siesta (long NS-siesta). Minor ticks on the curves represent censored individuals. The p value is derived from the log-rank test. The table at the bottom indicates the number of individuals at risk.

Individuals in short NS-siesta, long NS-no siesta and long NS-no siesta groups had a higher risk of all-cause mortality compared with short NS-no siesta (table 2). The HRs (95% CIs) were 1.43 (1.16, 1.76), 1.63 (1.22, 2.18) and 2.07 (1.63, 2.62), respectively. Multivariable adjustment for potential confounders attenuated the associations, with the adjusted HRs being 1.27 (1.02, 1.57), 1.41 (1.05, 1.9) and 1.32 (1.02, 1.69), respectively. Additional stratification by clinical recommendations revealed similar patterns (online supplemental table 2). Adjustments for GHQ and SRH scores did not further attenuate the association between sleep patterns and mortality (online supplemental table 3).

Table 2
|
HRs (95% CIs) for mortality associated with sleep patterns in the Mabat Zahav cohort, Israel, 2005–2006

Applying spline methodology to examine the association between night-time sleep duration (on a continuous scale) and mortality, the multivariable-adjusted risk appeared to be relatively constant up to approximately 8 hours, above which we observed a substantial increase (figure 2A; p for linearity<0.01, p for non-linearity=0.06). Spline analysis further revealed distinct patterns in the relationship between sleep duration and mortality by cognitive status (figure 2B,C). Among cognitively preserved individuals (n=1091, mean (SD) MMSE=28.8 (1.07)), we observed a linear relationship across almost the entire sleep duration distribution (p for linearity=0.047, p for non-linearity=0.53). On the other hand, the relationship among cognitively impaired individuals (n=428, mean (SD) MMSE=22.3 (3.63)) was U-shaped (p for non-linearity<0.001), with an optimal duration set around 7.5 hours of sleep duration.

Figure 2
Request permissionrequest permission icon
Figure 2

Penalised smoothing spline-based HRs (95% CIs) for all-cause mortality associated with self-reported night-time sleep duration in the Mabat Zahav population-based cohort of older Israeli adults initiated in Israel in 2005 with follow-up through 2019 (A) and stratified by preserved cognition (B) and impaired cognition (C). The curve is based on a spline-based Cox model, adjusted for baseline covariates including age, sex, ethnicity, education, neighbourhood socioeconomic status, physical activity, smoking status, activities of daily living score, Mini-Mental State Examination score, multimorbidity index and daytime napping. The y-axes are presented on a logarithmic scale. The histogram at the bottom shows the relative overall distribution of sleep duration.

Siesta was also associated with mortality. The HR (95% CI) was 1.37 (1.18, 1.59), compared with no siesta. Multivariable adjustment for potential confounders and sleep duration attenuated the association (HR (95% CI); 1.15 (0.97, 1.36)). The association differed by cognitive status. After adjustment for multiple potential confounders, the HR (95% CI) for death was 0.98 (0.70, 1.35) among cognitively impaired and 1.26 (1.03, 1.54) among cognitively preserved participants (table 3).

Table 3
|
HRs (95% CIs) for mortality associated with siesta by cognitive function status in the Mabat Zahav cohort, Israel, 2005–2006

Discussion

The primary objective of our study was to investigate the association between night-time sleep duration, siesta and long-term all-cause mortality risk in an older Israeli adult population while trying to understand the joint relationship between sleep duration and siesta. We found higher adjusted mortality risk in long NS-siesta, long NS-no siesta and short NS-siesta compared with short NS-no siesta. This finding suggests that 8 hours or shorter sleep without siesta is associated with the lowest risk of all four sleeping patterns examined. Sleep duration was positively and linearly associated with adjusted mortality in cognitively preserved individuals. However, a U-shaped relationship was observed in cognitively impaired individuals, with sleeping less than 7 hours or more than 8 hours associated with increased risk. Siesta was significantly associated with increased mortality among cognitively preserved individuals only. Our study presents a unique perspective on how night-time and daytime sleep behaviours are jointly associated with mortality risk.

Sleep duration and siesta have long been linked to mortality. A meta-analysis conducted by da Silva et al25 involving 27 cohort studies of older adults showed that both short and long sleep duration and daytime napping are associated with increased mortality risk. The joint association of sleep duration and siesta was not assessed. Using data from the UK Biobank (n≈400 000 adults), Tao et al26 found that sleep duration was associated with mortality (risk ratio (RR)=1.25, 95% CI: 1.16 to 1.34, for sleep duration ≤5 hours and RR=1.30, 95% CI: 1.22 to 1.38, for sleep duration ≥9 hours, compared with sleep duration of 7–9 hours). Siesta has also emerged as a risk factor for all-cause and cause-specific mortality,10 11 adjusted for night-time sleep duration. We aimed to assess the joint association of sleep duration and siesta by investigating four distinct sleep patterns. Our results support the prevailing thought that sleep patterns may be a potential marker for poor physical condition.1 2 27

Furthermore, the entire cohort shows a consistent association between night-time sleep duration and mortality risk. An earlier study by Wang et al9 has demonstrated a distinct J-shaped relationship between sleep duration and mortality among younger adults from 21 countries (Israel not included). However, our cohort is much older. Indeed, a linear trend has previously been described among older Israeli adults.28 29 For example, Cohen-Mansfield and Perach,28 analysing data from 1166 participants aged 75–94 years recruited in the late 1980s, reported HRs of 0.98 (95% CI 0.84 to 1.13) for sleep duration shorter than 7 hours and 1.32 (95% CI 1.09 to 1.58) for sleep duration longer than 9 hours compared with 7–9 hours.

We observed distinct exposure–response relationships by cognitive status: while we observed a nearly linear association among cognitively preserved participants, the association among cognitively impaired subjects was non-linear, with the optimal duration set around 7.5 hours of sleep duration. Individuals with cognitive impairment often have disrupted sleep patterns related to neurodegenerative changes.30 Our results suggest that cognitively impaired individuals require a specific amount of sleep to maintain health, while insufficient and excessive sleep duration could increase mortality risk. Alternatively, the observed association could reflect a bidirectional relationship, where cognitive function alters sleep habits.31

Furthermore, we only found an association between siesta and mortality risk in cognitively preserved individuals. Daytime lethargy may be an early sign of subclinical disease progression32 and may be a reliable marker for physical deterioration among cognitively preserved individuals.

Sleep patterns may act as markers and drivers of physical and functional alterations. Certain sleep behaviours, such as longer sleep duration or siesta, may indicate underlying health conditions such as chronic cardiovascular, metabolic and mental health issues,33 as these conditions often lead to fatigue. As we controlled for some of these conditions, the observed associations partly account for these individual factors. Yet, reverse causality cannot be ruled out. Concurrently, alteration in sleep patterns can disrupt physiological processes such as hormonal and metabolic balances,34 35 thereby worsening health outcomes over time. In addition, sleep alterations are associated with poorer lifestyle indicators.36

Social factors have been associated with variations in sleep habits.37 38 Therefore, the results should be contextualised to the uniqueness of the Israeli population.11 28 As the Israeli population comprises Jewish and Arab individuals, we adjusted for ethnicity in our models to account in part for ethnic differences. Additionally, we focused on older adults, as they present a prone population with many mortality risk factors compared with other age groups.39

As the causal nature of the sleep–mortality association is biologically plausible, interventions to sustain optimal sleep duration and daytime wakefulness could reduce the associated risk. A recent target trial emulation study40 has shown that interventions designed to address inappropriate sleep duration can potentially reduce incident cognitive impairment with an RR of 0.88 (95% CI: 0.80 to 0.95), a promising finding for reducing the risk of other adverse outcomes.

An important consideration should be given to sleep duration classification. Indeed, the existing literature lacks a uniform definition, and several classification systems have been used—a methodological flaw that limits the generalisability of these studies. Moreover, the optimal night-time sleep duration definition for adults is between 7 and 9 hours per night per the American Academy of Sleep Medicine and Sleep Research Society,23 a priori considers the relationship between sleep duration and mortality to be J-shaped or U-shaped. However, this is not the case in our cohort. Consequently, we opted to stratify this older Israeli population based on an established cut-off point among older Israelis, which is 8 hours per night.11 29 Nevertheless, we conducted a sensitivity analysis based on the clinically recommended classification, with the results being generally comparable with the primary analysis (online supplemental table 2).

While our study provides important insights, it is essential to consider several limitations. First, our study’s observational nature makes it difficult to establish causality. Second, although we controlled for various potential confounders, including PAL,41 socioeconomic status42 and pre-existing comorbidities, residual confounding is likely due to other unknown or unmeasured factors. Possible confounders are depression and subjective health,24 28 for which we adjusted (online supplemental table 3) without attenuating the observed associations. Data on the potential use of sleep aids, however, were not collected during the interviews, and therefore, we were unable to account for them. Additionally, our reliance on self-reported sleep habits at a singular time may introduce bias due to inaccurate recall or subjective assessment and not account for changes throughout the follow-up period. Recent studies were inconclusive regarding the validity of personal sleep duration assessment, with some showing a correlation between objective measurement and self-reported sleep habits,43 44 suggesting that our data are generally valid, and others showing a discrepancy.45 The quality of self-reported sleep habits can vary depending on cognitive status, posing a challenge among cognitively impaired individuals. However, we tried to mitigate these biases by excluding participants with severe cognitive impairment and adjusting for cognitive function in our multivariable regression models. Information about siesta in our study was also limited, as no data about the duration of daytime napping were available. Furthermore, we did not consider sleep quality, which may contribute to mortality risk.24 26 46 Another possible limiting factor is the potential underpowering of our analysis in the cognitively impaired cohort due to its smaller sample size than the cognitively preserved group. Indeed, future studies should employ a larger sample size in order to analyse sleep habits according to cognitive status or clinically recommended classification with sufficient power. Despite this, the higher mortality rate within the cognitively impaired group yielded a reasonable number of events.

In conclusion, our study unveils an essential aspect of the complex relationship between sleep behaviours and mortality risk in an older Israeli population. We have identified distinct sleep patterns associated with increased mortality risk compared with short night-time sleep without a siesta. We also pointed out cognitive performance as an essential effect modifier of the sleep–mortality relationship. These findings underscore the potential of sleep behaviours as a marker of health and an indicator of adverse outcomes among older adults.

Overview

Article metrics
Altmetric data not available for this article.
Dimensionsopen-url