Smart ring-based assessment of physical activity intensity and sleep disturbances in older adults with mild cognitive impairment

Abstract

Objective

Sleep disturbance is common in older adults with mild cognitive impairment (MCI) and is linked to poorer cognitive and overall health, but prior evidence is mixed and often based on self-report. Using validated smart ring data, this study quantifies physical activity (PA) intensity as metabolic equivalents of task (MET) and examines its association with objectively measured sleep disturbance in this population.

Methods

This was a 14-day observational study in a long-term care facility using a smart ring. Fifteen participants were enrolled, 11 met eligibility at baseline, and seven completed the exit visit. The Oura Ring measured PA and classified in vigorous, moderate, and light intensities from METs. Sleep disturbance was measured as any 5-min segment with movement measured by the accelerometer during sleep that coincided with changes in heart rate measured by the photoplethysmography sensor and changes in skin temperature measured by the temperature sensor.

Results

Vigorous PA was significantly related to reduced sleep disturbance, with each additional second linked to a 0.18-s decrease in disturbance (B = −0.18, 95% CI [−0.29, −0.07]). Moderate PA had a small, nonsignificant positive coefficient (B = 0.01, p > .05). Light PA showed a significant negative association, describing slightly reduced sleep disturbance with increased time in light PA (B = −0.01, 95% CI [−0.01, −0.01]).

Conclusion

Light and vigorous PA were associated with lower sleep disturbance in older adults with MCI, suggesting that intensity-targeted programs emphasizing light (e.g. walking) and vigorous (e.g. swimming) activity may help reduce sleep disturbance in this population.

Keywords

Digital health dementia well-being wearable self-monitoring physical activity

Introduction

Sleep quality is a significant determinant of overall health in older adults. Research evidence suggests that poor sleep contributes to the acceleration of neurodegenerative processes (e.g. cognitive decline and dementia) and diminished capacity to recover from intercurrent health risks in later life.^1–3 Challenges to effective sleep are especially prevalent among older adults living with mild cognitive impairment (MCI). Recent studies estimate the prevalence of sleep-related challenges to be approximately 35.8% based on self-report and 46.3% when assessed using objective measures.⁴ According to Wei et al.,⁵ older adults living with MCI demonstrate poorer sleep quality with about 34 min less total sleep time, lower sleep efficiency, and greater sleep latency than their peers without MCI.⁵ Such sleep patterns that are present in older adults living with MCI are a significant public health issue.

Within the spectrum of sleep problems, sleep disturbances include sleep fragmentation, insomnia, increased waking after sleep onset, and brief arousals.^6,7 Systematic reviews indicate that sleep disturbance is associated with subsequent cognitive decline and more rapid progression of cognitive impairment.^1,8 For example, over 36% of community-dwelling adults living with MCI experience sleep disturbance that is associated with poorer cognitive function and increased dementia risk.^1–3 Thus, it is of critical importance to reduce sleep disturbance in older adults living with MCI to improve their cognitive health and overall health quality.

Substantial evidence has been presented that participation in physical activity (PA) can reduce the sleep disturbances of older adults living with MCI, especially those who are experiencing a substantial burden of poor sleep.⁴ Meta-analyses have provided evidence that regular PA participation improves overall sleep quality, shortens sleep latency, and reduces nocturnal awakenings.^9,10 These studies have stressed the importance of PA participation for reducing sleep disturbances and improving the sleep quality and cognitive function of older adults living with MCI.

Given the importance of PA participation for improving sleep quality, researchers have investigated the relationship between different PA levels and sleep outcomes. Some studies have highlighted the effects of light intensity PA such as walking or stretching for sleep quality,^11–13 whereas others have demonstrated that moderate and vigorous PA are more beneficial for sleep quality.^14–16 In contrast, other studies suggest that vigorous PA adversely affects sleep quality, reflecting contradictory findings in the literature. These inconsistencies may be due to differences in PA intensity and study populations,^17,18 as most prior studies focused on younger adults or individuals with insomnia. However, there remains a significant gap in understanding which levels of PA can be instrumental in reducing sleep disturbances among older adults with MCI. This investigation can provide critical evidence to guide the design and implementation of PA programs tailored to the needs of older adults with MCI.

In addition, there are critical research gaps that have been caused by methodological inconsistencies in the measurement of both PA and sleep. Prior studies that assessed PA and sleep in older adults have primarily relied on self-reported measures. However, these self-reported measures for older adults living with MCI may limit the credibility of such data due to cognitive challenges and biases inherent to self-reported assessments. For example, analyses using the Health and Retirement Study (HRS) used self-reported sleep disturbance questionnaires. However, reliability studies that analyzed the self-reported sleep disturbance questionnaires used by the HRS found inconsistencies between self-report and sensor-based assessments of both PA and sleep.^19–21 Such methodological limitations weaken inferences about PA intensity-specific associations with sleep disturbance and decrease our ability to accurately estimate health benefits. To address these limitations, recent health behavior studies have employed biometric sensors to obtain objective, continuous measures of PA and sleep that capture more accurate patterns of PA and sleep in this vulnerable population.

Multi sensor smart rings that have been validated against polysomnography show high reliability and sensitivity for sleep detection and acceptable epoch level agreement for sleep stage classification.^22–25 Smart rings quantify daily PA volume and intensity and accurately capture sleep metrics that allow metabolic equivalents of task (METs) based PA intensity to be paired with disturbance measures to assess their association with sleep disturbances in older adults living with MCI. To control for potential confounding effects on the relationship between PA and sleep outcomes, age and sex were included as covariates in the regression model. Thus, we investigated the intensity specific association between PA and sleep disturbance among older adults living with MCI. The purpose of our study was to quantify PA intensity in METs and examine its relationship with objectively measured sleep disturbances, using validated sensor-derived data from smart rings. Based on this purpose, the following hypotheses were proposed:

Hypothesis 1: Sleep disturbance differ across PA intensity levels (light, moderate, and vigorous) among older adults with MCI.

Hypothesis 2: Moderate PA is associated with lower levels of sleep disturbance than light PA.

Hypothesis 3: Vigorous PA is associated with lower levels of sleep disturbance than moderate PA.

Methods

Participants

Study participants were recruited from a long-term care facility in southern Mississippi (USA) that provides comprehensive healthcare, nutrition coaching, therapeutic activities, and sleep care, but not memory care programs. For study inclusion, participant eligibility required being older than 60 years of age, having MCI as assessed by the Montreal Cognitive Assessment (MoCA), being able to wear a smart ring given limited size options, and providing written informed consent. Individuals were excluded if they had substantial communication difficulties, functional limitations that impeded independent living (e.g. notable visual or auditory deficits, a recent knee replacement), current use of medications for sleep disorders or dementia.

Study design

This observational study tracked participant biomarkers using smart rings over 14 days to examine the relationship between sleep disturbance and PA intensity as assessed by MET score. Fifteen individuals were initially recruited and enrolled, 11 of whom met the study inclusion criteria (Figure 1). Before monitoring began (T0), baseline data was collected at an intake session at the facility that including administering the MoCA to confirm an MCI score of between 18 and 25 out of 30,²⁶ and finger measurements to assign custom ring sizes (12 sizes available). Custom-fitted rings were delivered within a few days, and the research team confirmed proper fit and paired each ring with a study tablet configured to upload data to Oura on the Web via Bluetooth.^24,27,28 To minimize self-serving bias, participants did not have access to their stored data. During the monitoring period (T1–T14), the research team met participants for 15–20 min twice weekly to confirm adherence, synchronize data from rings to tablets, and recharge batteries. No additional study interventions were delivered during the study period. One day after monitoring ended, a 30-min exit visit (T15) was conducted. During this visit, participants completed a survey, returned the devices, and received a $20 gift card. Four participants withdrew due to ring discomfort, initiation of dementia, sleep-specific care, and other health issues. Seven participants were retained for the final analysis set.

Figure 1.

Research design.

Ethical consideration

All procedures followed the protocol approved by the Institutional Review Board of the sponsoring institution (IRB #23-1035). Written informed consent was obtained, and participants were informed of the study purpose, duration, procedures, and expectations before engaging in any study activities. All activities adhered to the IRB-approved protocol with careful protection of study participant rights.

Technology application

Biomarker data were collected using the Oura Ring, version 4 (Oura Health, Finland), a multisensor wearable with clinically validated algorithms that capture sleep and PA feedback.^24,28,29 This device integrates an eight-path, multiwavelength photoplethysmography (PPG) module using infrared light to assess blood oxygen saturation, heart rate variability (HRV), skin temperature alongside 3D accelerometers and a gyroscope to quantify movement.²⁹ Sensors are housed beneath a hypoallergenic titanium exterior to reduce discomfort and skin irritation. When the paired mobile device is within 33 feet of the tablet, data are synchronized via Bluetooth without a physical connection. Battery endurance is up to eight days per charge, which enables continuous wear during daily activities, exercise, water exposure, and sleep patterns. Onboard memory preserves data for up to 15 days even if the battery is depleted, which prevents data loss prior to synchronization with tablets. These capabilities enable uninterrupted 24/7 monitoring and support the detection of longitudinal variations in sleep patterns and PA that improves data reliability.^24,28–30

Measures

Sleep disturbance

Sleep disturbance was measured using the Oura Ring that provides a disturbance score and continuously tracks movement via a 3D accelerometer, and heart rate and HRV via a PPG sensor, and skin temperature using an onboard temperature sensor.^27,31 Data streams from these sensors are processed by Oura's sleep staging algorithm (OSSA 2.0) that has been demonstrated in validation studies to have a high degree of agreement with gold-standard polysomnography, and to have the ability to report data with a high degree of accuracy.^24,28–30 For sleep disturbance analysis, accelerometer data are aggregated into fixed time segments (i.e. 5 min) to evaluate the magnitude, frequency, and intensity of movements. Movements during sleep accompanied by changes in heart rate or skin temperature were defined as disturbance episodes. The OSSA captured the frequency and magnitude of these episodes in 5-min intervals. The algorithm applies higher weights to movements occurring during light sleep or wake periods and lower weights to those in deep or rapid eye movement sleep stages to generate a composite disturbance score. Sleep disturbance scores indicated an aggregated measure of all sleep episodes recorded while the ring was worn, including both nocturnal sleep and daytime sleep episodes (e.g. naps), rather than being limited to nighttime sleep only. Participants primarily slept alone in private rooms at night, consistent with the nursing home setting. Daytime naps, when they occurred, typically took place in shared common areas. All detected sleep episodes were combined into a single composite disturbance metric for analysis.

PA intensity

The Oura Ring classifies each second of daily PA into one of three intensity categories: light, moderate, or vigorous. PA levels represent the total amount of time participants engaged in each intensity level per day and indicate daily averages rather than cumulative activity across the 14-day monitoring period. This classification is based on a machine learning algorithm designed to measure and categorize PA levels.^24,28–30 The algorithm computes the MET using data collected from its onboard sensors.²⁹ Movement data are collected from a 3D accelerometer, that records the frequency, amplitude, speed, and direction of motion. A PPG sensor measures heart rate and HRV, while a temperature sensor monitors skin temperature to help identify PA patterns. Energy expenditure is computed by combining accelerometer-derived movement patterns with heart rate data, adjusting values upward when heart rate exceeds the expected level for a predefined movement pattern and downward when it is lower.³² For MET classification, the device calculates energy cost in kcal/min and divides it by the resting energy cost, defined as 1 MET (3.5 ml O₂ per kg per minute).^33,34 Activities below 3 METs are classified as light (e.g. sitting, walking), between 3 and 6 METs as moderate (e.g. cycling), and above 6 METs as vigorous (e.g. running, interval training). The Oura Ring then averages MET values over fixed 1-s intervals and assigns each interval to one of the three intensity levels.

HRV and balance

The Oura Ring assesses HRV by tracking blood flow with 250 Hz infrared PPG sensors and detecting interbeat intervals. During sleep, the OSSA computes the root mean square of successive differences (RMSSD), a key marker of parasympathetic activity, in consecutive five-min windows and summarizes sleep patterns using average and maximum RMSSD values that serve as a sleep quality indicator.³⁵ HRV balance measures how recent HRV compares to the individual's baseline and serves as an indicator of overall changes in sleep quality. A low RMSSD with below-baseline HRV balance indicates impaired recovery and an elevated risk of sleep disturbances, whereas higher values suggest adequate recovery and preserved sleep quality. Validation studies have provided evidence of moderate agreement between Oura's HRV measurements and ECG-derived metrics, although short 5-min readings and frequency-domain measures are less reliable.³⁶

Data analysis

Data analysis followed sequential steps to examine the longitudinal association between PA intensity and sleep disturbance in older adults living with MCI. First, descriptive statistics summarized participant demographics and Q-Q plots were used to assess the distribution and normality of residuals. As some daily measurements were missing, it was necessary to address missingness before model fitting. Missing rates were below 8% for both study variables, which is considered acceptable for imputation without introducing substantial bias.³⁷ Thus, we used multiple imputation by chained equations to replace missing entries with estimated values derived from observed data patterns. This method was chosen because it uses all available information from correlated variables to generate estimates, which minimizes the residual differences between observed and imputed values while preserving the variability and structure of the dataset.³⁸ Following data preparation, hypotheses were tested using Generalized Estimating Equations (GEEs). GEE was selected because it is specifically designed for repeated measures longitudinal data as it accounts for within subject correlations without requiring full likelihood specification.^39,40 This makes GEE suitable for analyzing relationships between varying PA intensities and sleep disturbances across multiple days while controlling for the nonindependence of repeated observations. Further, GEE focuses on estimating population averaged effects, aligning with our goal of identifying group level trends rather than subject specific trajectories, and provides stable estimates even with small sample sizes (n < 20).^41,42

\begin{aligned} Sleep disturbance \\ = β_{0} + {(a g e β_{1} + s e x β_{2}) + \\ H R V β_{3} + H R V b a l a n c e β_{4} + \\ l i g h t P A β_{5} + m o d e r a t e P A β_{6} + \\ v i g o r o u s P A β_{7}} \end{aligned}

Results

Table 1 outlines the demographic characteristics and study variables for the seven participants included in the analysis. The age of participants ranged from 73 to 92 years, with an average of 83.01 years (SD = 5.79). The sample consisted of four men (57.0%) and three women (43.0%). With respect to marital status, two individuals (28.0%) reported being married, three (44.0%) were separated, and two (28.0%) were widowed. In terms of educational background, two participants (28.0%) had attained some high school education or less, four (58.0%) had completed high school, and one (14.0%) reported some college experience. Regarding self-perceived general health, one person (14.0%) rated their health as very good, one (14.0%) as good, three (44.0%) as fair, one (14.0%) as poor, and one (14.0%) chose not to provide a rating. Cognitive status, assessed using the MoCA, had a mean score of 20.35 (SD = 3.96). The dependent measure, sleep disturbance, averaged 191.68 s (SD = 183.71). PA averaged 343.92 s (SD = 340.05) for vigorous PA, 1905.65 s (SD = 1441.94) for moderate PA, and 15,785.97 s (SD = 9553.66) for PA. Indicators of cardiac autonomic function included a mean HRV of 37.75 (SD = 15.43) and a mean HRV balance score of 85.04 (SD = 5.12).

Table 1.

Demographic characteristics.

	M	SD	n	%
Age
73–92 years old (M = 83.01, SD = 5.79)			7	100
Sex
Male			4	57.0
Female			3	43.0
Marital status
Married			2	28.0
Separated			3	44.0
Widowed			2	28.0
Education
Some high school or lower			2	28.0
High school graduate			4	58.0
Some college			1	14.0
General health
Very good			1	14.0
Good			1	14.0
Fair			3	44.0
Poor			1	14.0
Prefer not to answer			1	14.0
Cognitive function	20.35	3.96
Dependent variable
Sleep disturbance	191.68	183.71
Independent variable
Vigorous PA	343.92	340.05
Moderate PA	1905.65	1441.94
Light PA	15785.97	9553.66
HRV	37.75	15.43
HRV balance	85.04	5.12

Notes: HRV: heart rate variability.

Sample: 7, Index: 14. Sleep disturbance was measured in seconds.

Physical activity (PA) participation was measured in seconds.

Cognitive function was assessed using the Montreal Cognitive Assessment.

Table 2 shows the Wald Chi-Square test results for the GEE model predicting sleep disturbance. The intercept was not statistically significant (Wald χ² = 0.00, df = 1, p = .95). Age (Wald χ² = 1.79, df = 1, p = .18) and sex (Wald χ² = 0.01, df = 1, p = .91) were also not significant predictors. For different intensities of PA, vigorous PA (Wald χ² = 11.25, df = 1, p < .01) and light PA (Wald χ² = 12.19, df = 1, p < .01) were significantly associated with sleep disturbance, while moderate PA (Wald χ² = 3.04, df = 1, p = .08) showed nonsignificant. Regarding physiological measures, HRV was significantly related to sleep disturbance (Wald χ² = 7.06, df = 1, p = .01), whereas HRV balance was not (Wald χ² = 0.14, df = 1, p = .70).

Table 2.

Wald chi-square test for GEE model.

Type III
Independent variables	Wald chi-square	df	Sig.
Intercept	0.00	1	.95
Age	1.79	1	.18
Sex	0.01	1	.91
HRV	7.06	1	.01*
HRV balance	0.14	1	.70
Vigorous PA	11.25	1	.00*
Moderate PA	3.04	1	.08
Light PA	12.19	1	.00*

Notes: GEE: generalized estimated equation; HRV: heart rate variability.

Dependent variable: sleep disturbance measured in seconds.

Physical activity (PA) was measured in seconds.

*p < .05.

Table 3 summarizes the estimated effects of PA in different levels and physiological factors on sleep disturbance. Neither age (B = 4.45, SD = 3.32, p = .18, 95% CI [−2.06, 10.97]) nor sex (B = −3.92, SD = 35.73, p = .91, 95% CI [−73.96, 66.12]) presented significant relationships with sleep disturbance. Vigorous PA was significantly related to reduced sleep disturbance, with each additional second linked to a 0.18-s decrease in disturbance (B = −0.18, SD = 0.05, p < .01, 95% CI [−0.29, −0.07]). Moderate PA had a small, nonsignificant positive coefficient (B = 0.01, SD = 0.01, p = .08, 95% CI [−0.01, 0.03]). Light PA showed a significant negative association, describing slightly reduced sleep disturbance with increased time in light PA (B = −0.01, SD = 0.01, p < .01, 95% CI [−0.01, −0.01]). HRV was also significantly associated with greater sleep disturbance (B = 3.37, SD = 1.27, p = .01, 95% CI [0.88, 5.86]), whereas HRV balance was not significant (B = −1.46, SD = 3.80, p = .70, 95% CI [−8.92, 5.99]).

Table 3.

Relationship between physical activity intensity levels and sleep disturbance.

		Hypothesis test				95% Wald CI
Independentvariables(parameter)	B	SD	Waldchi-square	df	Sig.	Lower	Upper
Intercept	−23.14	443.72	0.01	1	.95	−892.82	846.83
Age	4.45	3.32	1.79	1	.18	−2.06	10.97
Sex	−3.92	35.73	0.01	1	.91	−73.96	66.12
HRV	3.37	1.27	7.06	1	.01*	0.88	5.86
HRV balance	−1.46	3.80	0.14	1	.70	−8.92	5.99
Vigorous PA	−0.18	0.05	11.25	1	.00*	−0.29	−0.07
Moderate PA	0.01	0.01	3.04	1	.08	−0.01	0.03
Light PA	−0.01	0.01	12.19	1	.00*	−0.01	−0.01
Scale	20130.39

Notes: CI: confidence interval; HRV: heart rate variability.

Dependent variable: sleep disturbance measured in seconds.

Physical activity (PA) was measured in seconds.

*p < .05.

Table 4 summarizes the post hoc power analysis and required sample sizes. Power was estimated using an approximation based on Fisher's z transformation with standardized effects, assuming an intraclass correlation coefficient (ICC) of 0.50, indicating an effective sample size of n_eff = 13.1. With the standardized effects (light PA, r = −0.67, vigorous PA, r = −0.53, HRV, r = 0.47), the current design provides moderate power for light PA but low power for vigorous PA and HRV at ICC = 0.50. The reported sample size estimates required to achieve 80% power at α = 0.05 (approximately 9, 14, and 18 participants for light PA, vigorous PA, and HRV, respectively) should be interpreted within the context of a within-person, repeated-measures design and do not represent mutually exclusive activity groups. Thus, these estimates reflect the number of participants required to detect the observed effects given repeated observations within individuals. Given limited power for moderate PA and HRV balance, future trials should extend follow up, enroll multisite cohorts, and include a randomized comparator arm.

Table 4.

Achieved power at current design (fisher-z power approximation).

ICC	n_eff	LightPA (r = −.67)	ModeratePA (r = .20)	VigorousPA (r = −.53)	HRV(r = .37)	HRVbalance (r = −.05)
.30	20.0	.92	.13	.68	.36	.06
.50	13.1	.73	.10	.47	.23	.05
.70	9.7	.56	.08	.33	.17	.05

Notes. ICC: intraclass correlation coefficients; HRV: heart rate variability.

Seven participants with 14 days of monitoring.

n_eff = (n × m)/[1 + (m − 1)ρ].

Effect sizes (r) approximate standardized slopes.

Discussion

In this study, we investigated the association between the PA intensity and sleep disturbances of older adults living with MCI, and our findings indicate that participation in vigorous and light PA was significantly associated with the occurrence of reduced sleep disturbances in this population. Our findings suggest that certain levels of PA participation can be instrumental in reducing sleep disturbances and lead to improved sleep quality and overall health in older adults living with MCI.

Prior studies have suggested that moderate PA participation is associated with longer sleep duration, decreased sleep latency, and improved sleep quality in young adults,^9,10 with these studies stressing the positive impact of moderate PA intensity on sleep outcomes in this population. In contrast, the findings of our study reveal no significant association between moderate levels of PA and sleep disturbances in older adults with MCI and suggest that the relationship between PA intensity and sleep outcomes may differ by age group. This finding indicates that PA engagement intensity plays a vital role in influencing the sleep quality of older adults with MCI.

Systematic analyses have indicated that high intensity PA does contribute to poor sleep quality including difficulties initiating sleep and potential risks for insomnia, and that moderate PA is often reported as more effective for improving overall sleep quality.^14–16 These findings of these studies suggest that the excessive physical demands of high intensity PA may disrupt sleep patterns. Inversely, the findings of our study provide evidence that high intensity PA has been associated with reduced sleep disturbances in older adults living with MCI, which suggests that vigorous PA can be beneficial for reducing sleep disturbances and potentially improving sleep quality.

Previous research has revealed the positive impact of light intensity PA on the sleep quality of adolescents,^11–13 including providing evidence that light-intensity activities such as walking or stretching, decrease light sleep and nighttime awakening and improve deep sleep by regulating circadian rhythms and promoting relaxation. The findings of our research are aligned with the previous literature suggesting that light intensity PA can reduce the occurrence of sleep disturbances in older adults living with MCI, and that this level of PA can help promote sleep efficiency.

Limitations and suggestions for future study

Despite the importance of light and vigorous PA to sleep quality, there are several limitations to the application of our findings that need to be addressed and suggestions to make for future studies. First, we employed a small sample (n = 7) can inherently reduce statistical power to investigate the relationship between the PA levels and sleep disturbances of older adults who are living with MCI, therefor, a study with a larger sample size is needed to fully characterize their relationship. Second, a key limitation of this study is the use of broad intensity-based categories of PA (light, moderate, vigorous) based on the Oura activity score. We focused on the relationship between the PA intensity and sleep disturbances of older adults with MCI but did not examine the impact of the types of PA chosen or the time of day in which PA is pursued, factors that can potentially influence the sleep outcomes experienced by older adults with MCI. The impacts of how the type of PA pursued or the timing during the day of PA participation contributes to sleep quality should therefore be investigated. For example, while this measure captures cardiovascular activity well, it may underrepresent strength-based exercise, leading to potential measurement bias. Grouping different types of PA by intensity alone have limited reliability in examining activity-specific effects and may have weakened the observed associations with sleep outcomes, as aerobic and resistance activities influence sleep differently. This limitation should be considered when interpreting the findings. Third, sleep disturbances were examined as an overall outcome without distinguishing between nocturnal sleep and daytime naps. As a result, we were unable to account for potential differences in the underlying factors, types, and intensity of sleep disturbances that may occur during nighttime versus daytime sleep. Future studies should attempt to isolate and analyze sleep disturbance episodes occurring specifically during nighttime sleep. Fourth, there are several limitations in using the Oura Ring as an assessment device in this study. Although Oura documentation suggests that a calibration period of approximately two weeks may be required to provide reliable metrics, no formal calibration period was implemented in this study, which may have affected the reliability of the data. Furthermore, cross-device comparisons with other wearables and, where feasible, polysomnography, should be used to assess measurement reliability and strengthen external validity. Last, potential confounders including residential settings (home vs assisted living), chronic comorbidities, functional limitations, current medications, mood symptoms, and other physiological biomarkers were not fully integrated into our study. Future work should measure and integrate these covariates more comprehensively into the study design, model time-varying confounding factors, and report sensitivity analyses.

Conclusion

Our study provides evidence that light and vigorous PA participation are associated with reduced levels of sleep disturbance in older adults living with MCI. This finding suggests that intensity-targeted PA that emphasizes light (e.g. walking) and vigorous (e.g. swimming) PA intensity levels may help reduce sleep disturbances in this population. Our findings also provide evidence of the importance of adopting intensity considerations into community-based and PA/exercise programs for older adults who are living with MCI. Designing and implementing group-based walking programs (light PA) and tailored vigorous PA opportunities has the potential to provide practical and sustainable ways to improve the sleep outcomes of older adults living with MCI. By encouraging regular participation in light and vigorous PA, older adults living with MCI may experience reduced levels of sleep disturbances, enhanced overall sleep quality, and experience secondary benefits in cognitive function, mental health, and overall quality of life.

Footnotes

ORCID iDs

Jungjoo Lee

Junhyoung Kim

Ethical consideration

All study procedures were approved by the Institutional Review Board (IRB #23-1035). This study is based on secondary data analysis, which does not require an additional IRB process.

Consent for publication

Written informed consent for publication of anonymized data was obtained from all participants (or their legally authorized representatives). Consent documentation is held by the authors and available to the journal on request. No identifiable personal information, images, or videos are included.

Contribution

Jungjoo Lee: conceptualization, data curation, formal analysis, methodology, and writing—original draft; and Junhyoung Kim: project administration, validation, writing—original draft, and writing—review and editing.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy and/or ethical restrictions.

References

Wennberg

AMV

Rosenberg

, et al. Sleep disturbance, cognitive decline, and dementia: a review. Semin Neurol 2017; 37: 395–406.

Salari

Lotfi

Abdolmaleki

, et al. The global prevalence of mild cognitive impairment in geriatric population with emphasis on influential factors: a systematic review and meta-analysis. BMC Geriatr 2025; 25: 13.

Bubu

Brannick

Mortimer

, et al. Sleep, cognitive impairment, and Alzheimer's disease: a systematic review and meta-analysis. Sleep 2017; 40: zsw032.

Zhang

Liu

, et al. Prevalence of sleep disturbances in people with mild cognitive impairment: a systematic review and meta-analysis. J Affect Disord 2024; 315: 1–10.

Wei

Wang

Guo

, et al. Sleep structure assessed by objective measurement in patients with mild cognitive impairment: a meta-analysis. Sleep Med 2024; 113: 397–405.

Qaseem

Kansagara

Forciea

, et al. Clinical guidelines committee of the American College of Physicians. Management of chronic insomnia disorder in adults: a clinical practice guideline from the American College of Physicians. Ann Intern Med 2016; 165: 125–133.

Sateia

Buysse

Krystal

, et al. Clinical practice guideline for the pharmacologic treatment of chronic insomnia in adults: an American academy of sleep medicine clinical practice guideline. J Clin Sleep Med 2017; 13: 307–349.

Mander

Winer

Jagust

, et al. Sleep and human aging. Neuron 2017; 94: 19–36.

Dolezal

Neufeld

Boland

, et al. Interrelationship between sleep and exercise: a systematic review. Adv Prev Med 2017; 2017: 1364387.

10.

Xie

Liu

Chen

, et al. Effects of exercise on sleep quality and insomnia in adults: a systematic review and meta-analysis of randomized controlled trials. Front Psychiatry 2021; 12: 664499.

11.

Chen

Tsai

, et al. Effect of walking on circadian rhythms and sleep quality of patients with lung cancer: a randomized controlled trial. Br J Cancer 2016; 115: 1304–1312.

12.

D’Aurea

CVR

Poyares

Passos

, et al. Effects of resistance exercise training and stretching on chronic insomnia. Braz J Psychiatry 2019; 41: 51–57.

13.

Seol

Lee

Park

, et al. Bidirectional associations between physical activity and sleep in older adults: a multilevel analysis using polysomnography. Sci Rep 2022; 12: 15399.

14.

Kline

. Physical activity and sleep: an updated umbrella review of the 2018 physical activity guidelines advisory committee report. Sleep Med Rev 2021; 58: 101489.

15.

Alnawwar

, et al. The effect of physical activity on sleep quality and sleep disorder: a systematic review. Healthcare 2023; 11: 2516.

16.

Zapalac

Naufel

Mychasiuk

, et al. The effects of physical activity on sleep architecture and mood in naturalistic environments. Sci Rep 2024; 14: 3112.

17.

Alkhaldi

Alshammari

Alqahtani

, et al. Effect of nighttime exercise on sleep quality among the general population in Riyadh, Saudi Arabia. Cureus 2023; 15: e43218.

18.

Dubinina

Korostovtseva

Rotar

, et al. Physical activity is associated with sleep quality: results of the ESSE-RF epidemiological study. Front Psychol 2021; 12: 705212.

19.

Whibley

Guyer

Swanson

, et al. Sleep disturbance as a moderator of the association between physical activity and later pain onset among American adults aged 50 and over: evidence from the health and retirement study. BMJ Open 2020; 10: e036219.

20.

Prince

Cardilli

Reed

, et al. A comparison of self-reported and device measured sedentary behaviour in adults: a systematic review and meta-analysis. Int J Behav Nutr Phys Act 2020; 17: 31.

21.

Miner

Stone

Zeitzer

, et al. Self-reported and actigraphic short sleep duration in older adults. J Clin Sleep Med 2022; 18: 403–413.

22.

de Zambotti

Rosas

Colrain

, et al. The sleep of the ring: comparison of the ŌURA sleep tracker against polysomnography. Behav Sleep Med 2019; 17: 124–136.

23.

Altini

Kinnunen

. The promise of sleep: a multi-sensor approach for accurate sleep stage detection using the Oura Ring. Sensors 2021; 21: 4302.

24.

Svensson

Gislason

Kristiansson

, et al. Validation of the Oura Ring generation 3 with OSSA 2.0 against multi-night ambulatory polysomnography. Sleep Med 2024; 96: 104–115.

25.

Chee

NIYN

Ghorbani

Golkashani

, et al. Multi-night validation of a sleep tracking ring in adolescents compared with a research actigraph and polysomnography. Nat Sci Sleep 2021; 13: 177–190.

26.

Ciesielska

Sokołowski

Mazur

, et al. Is the Montreal cognitive assessment (MoCA) test better suited than the mini-mental state examination (MMSE) in mild cognitive impairment (MCI) detection among people aged over 60? Meta-analysis. Psychiatr Pol 2016; 50: 1039–1052.

27.

Ong

Golkashani

Ghorbani

, et al. Selecting a sleep tracker from EEG-based, iteratively improved, low-cost multisensor, and actigraphy-only devices. Sleep Health 2024; 10: 9–23.

28.

Svensson

Madhawa

, et al. Validity and reliability of the Oura Ring generation 3 (Gen3) with Oura sleep staging algorithm 2.0 (OSSA 2.0) when compared to multi-night ambulatory polysomnography: a validation study of 96 participants and 421,045 epochs. Sleep Med 2024; 115: 251–263.

29.

Kristiansson

Fridolfsson

Arvidsson

, et al. Validation of Oura Ring energy expenditure and steps in laboratory and free-living. BMC Med Res Methodol 2023; 23: 50.

30.

Sharifi-Heris

Feli

Jasinski

, et al. Preterm birth risk stratification through longitudinal heart rate and HRV monitoring in daily life. Sci Rep 2024; 14: 1–12.

31.

Bruce

Kasl

Soltani

, et al. Correction: variability of temperature measurements recorded by a wearable device by biological sex. Biol Sex Differ 2023; 14: 82.

32.

Basavaraj

Grant

Aras

, et al. Deep learning model using continuous skin temperature data predicts labor onset. BMC Pregnancy Childbirth 2024; 24: 777.

33.

Mendes

da Silva

Ramires

, et al. Metabolic equivalent of task (METs) thresholds as an indicator of physical activity intensity. PLoS One 2018; 13: e0200701.

34.

U.S. Department of Health and Human Services. 2018 physical activity guidelines advisory committee scientific report. Washington (DC): Office of Disease Prevention and Health Promotion; 2018. Available from: https://odphp.health.gov/sites/default/files/2019-09/PAG_Advisory_Committee_Report.pdf

35.

Cao

Azimi

Sarhaddi

, et al. Accuracy assessment of Oura Ring nocturnal heart rate and heart rate variability in comparison with electrocardiography in time and frequency domains: comprehensive analysis. J Med Internet Res 2022; 24: e27487.

36.

Liang

Yilmaz

Soon

. Deriving accurate nocturnal heart rate, rMSSD and frequency HRV from the Oura Ring. Sensors 2024; 24: 7475.

37.

Kazijevs

Samad

. Deep imputation of missing values in time series health data: a review with benchmarking. J Biomed Inform 2023; 144: 104440.

38.

Mera-Gaona

Neumann

Vargas-Canas

, et al. Evaluating the impact of multivariate imputation by MICE in feature selection. PLoS One 2021; 16: e0254720.

39.

Hardin

Hilbe

. Generalized estimating equations. Boca Raton: Chapman and Hall/CRC, 2002.

40.

Vens

Ziegler

. Generalized estimating equations and regression diagnostics for longitudinal controlled clinical trials: a case study. Comput Stat Data Anal 2012; 56: 1232–1242.

41.

Ziegler

. Generalized estimating equations. Berlin: Springer Science & Business Media, 2011.

42.

Ballinger

. Using generalized estimating equations for longitudinal data analysis. Organ Res Methods 2004; 7: 127–150.