Sleep Following a Stroke: Multimodal Evaluation of Sleep Health and Disruptions and Impact on Recovery During Acute Inpatient Rehabilitation

Abstract

Background

Our understanding of sleep during early stroke care and its impact on rehabilitation outcomes remains limited. The objectives of this work were to (1) evaluate multidimensional sleep health and disruptions during acute inpatient rehabilitation for individuals with stroke, and (2) explore the relationship between sleep health/disruptions and functional recovery.

Methods

Data from 103 individuals with stroke were analyzed during acute inpatient rehabilitation. Sleep health/disruptions were assessed via patient reports, actigraphy, and biometric sensors. Functional outcomes were measured at admission and discharge. Generalized Linear Models (GLMs) were used to describe changes in sleep health over time, and multivariate regressions analyzed sleep disruptions and sleep-related predictors of functional recovery.

Results

Over inpatient stays, sleep improved with a 23% reduction in wake after sleep onset and 15% fewer multiple overnight disruptions. GLMs revealed that improved sleep quality was associated with reduced overnight activity and increased heart rate over time. Poor initial sleep quality and cognitive status were associated with more overnight disruptions. Lastly, minimal associations were found between sleep health and functional recovery.

Conclusions

Sleep health during inpatient stroke rehabilitation is generally poor, though improves over time. Sleep is affected by neurological recovery and hospital environment. Overnight activity and autonomic biomarkers were associated with perceived sleep health, and both physiological and environmental factors triggered disruptions. The association between functional recovery and indirect indicators of sleep health requires further investigation. These findings reveal new insights about inpatient sleep which can inform early, targeted sleep interventions to optimize post-stroke outcomes.

SIESTA, ClinicalTrials.gov (NCT04254484).

Keywords

sleep disorders sleep monitoring stroke neurorehabilitation actigraphy wearable sensors

Introduction

Good sleep is critical for health and neural recovery^1,2 but often overlooked during standard rehabilitation care. Following neurological injury, such as stroke, the brain undergoes reorganization to prompt some degree of spontaneous physical and cognitive recovery.³ This early neuroplasticity is harnessed during inpatient rehabilitation to maximize functional recovery.^3,4 Importantly, sleep health also impacts neuroplasticity, with good sleep facilitating motor learning, attention, and executive function, and ultimately influencing motor and cognitive outcomes.⁴ Preliminary evidence in both humans and animals has shown that better sleep during the early stages after a stoke can improve long-term recovery.⁵ However, post-stroke sleep interventions are currently scarce,^1,2 with only a few examples in the outpatient⁶ and community settings,⁷ and even fewer during the acute stages of recovery.⁸

Stroke survivors tend to have overwhelmingly poor sleep,⁴ which can result in devastating long-term consequences such as functional and cognitive decline, depression and anxiety, disease progression, and morbidity.^9-11 Causes of poor sleep could be external (ie, from the environment) or internal (ie, within the patient). Externally, environmental factors such as noise or light can disturb normal sleep patterns. Such disruptions occur frequently during inpatient care due to overnight medications and vitals monitoring, ambient hospital noise or light, and staff interruptions.^2,12 Internally, stroke causes somatic and physiological changes to the autonomic nervous system, such as pain,¹³ body temperature,¹⁴ heart rate variability (HRV),¹⁵ and diurnal variations in blood pressure (BP),¹⁶ all of which have been linked to fragmented sleep, reduced daily physical activity, and additional cardiovascular risks.¹⁷ Thus, there is a critical need to identify and address barriers to good sleep as early as possible to enhance health and neural recovery after stroke.

Our understanding of sleep quality during early stroke recovery is extremely limited, relying predominantly on small samples or low-resolution measurements, such as from subjective patient reports or imprecise staff observations. Polysomnography (PSG), which is the gold standard technique for obtaining detailed and objective sleep measurements, is resource-intensive and difficult to deploy at scale in the inpatient setting. For example, Huang et al¹⁸ have reported on one of the largest cohorts for post-stroke sleep measurements (123 participants with subacute stroke); however, they could perform only 1 night of PSG and were unable to comment on longitudinal sleep patterns during inpatient rehabilitation. Wireless, wearable sensors offer an ecological method to obtain objective, longitudinal sleep data, including from actigraphy^19,20 and multimodal biometric sensors.^21,22 Actigraphy uses motion to estimate sleep health, such as the duration and timing of sleep and wake, sleep fragmentation, and sleep efficiency. Biometric sensors can record additional autonomic descriptors such as heart rate, blood oxygen saturation (SPO2), and skin temperature, which serve as indicators of the circadian rhythms regulating sleep patterns. These data are then interpreted by custom machine learning algorithms or trained clinicians.²³ At present, multidimensional sleep measurement techniques have not been applied together to characterize post-stroke sleep during inpatient rehabilitation and recovery. The need for multidimensional evaluation of post-stroke sleep is further highlighted in the work by Williams-Cooke et al,²⁴ who found poor sleep in over half of an inpatient cohort (37 individuals with acute stroke) using actigraphy but were not able to determine the underlying correlates of good or poor sleep during the rehabilitation process.

This study sought to address the gaps in our understanding of post-stroke sleep, by systematically and comprehensively examining inpatient sleep during early stroke rehabilitation and its implications for recovery. The primary objective of this research was to explore sleep health and sleep disruptions for individuals with stroke undergoing treatment in an acute Inpatient Rehabilitation Facility (IRF). The secondary objective was to investigate associations between sleep health and functional recovery of patients in the IRF setting, under the hypothesis that individuals with better sleep health would demonstrate greater functional recovery during inpatient stroke rehabilitation.^25,26 Understanding multimodal sleep trends in the inpatient setting would pave the way for the design, implementation, and tracking of evidence-based sleep health interventions during a critical stage after stroke, thereby improving long-term patient health and recovery.

Methods

Setting and Participants

This prospective observational study, following Strengthening the Reporting of Observational Studies in Epidemiology guidelines, enrolled individuals with stroke from the inpatient units of the Shirley Ryan AbilityLab (Chicago, IL, USA) as part of a larger clinical trial testing the effectiveness of a sleep-promoting intervention during inpatient rehabilitation (ClinicalTrials.gov: NCT04254484). Participants enrolled in the trial’s control arm from July 2020 until December 2022 (prior to intervention initiation, or in an inpatient unit separate from the intervention) were considered for this analysis. The study was approved by the Institutional Review Board at Northwestern University (STU00211695), and all participants provided written informed consent. Enrollment started on July 21st, 2020.

Inclusion criteria were: at least 18 years of age; primary diagnosis of stroke; and able and willing to give written consent and comply with study procedures. Exclusion criteria were: serious cardiac conditions or neurological degenerative pathologies as co-morbidities (ie, multiple sclerosis, Alzheimer’s disease, and Parkinson’s disease); use of an implantable cardiac device that would affect biometric sensor measurements; previous diagnosis of sleep disorders (eg, sleep apnea), which would indicate serious sleep issues prior to their stroke; open wounds, which would limit the use of sleep measurement tools; or pregnant or nursing.

Data Acquisition

Study data were derived from clinical assessments, patient reports, and sensor recordings (Figure 1, Supplemental Method SM1). These data were collected at a single time point (static), between 2 and 6 discrete time points (semi-dynamic), or every day and night throughout their IRF stay (dynamic).

Figure 1.

Data acquisition for multimodal sleep evaluation during inpatient stroke rehabilitation. Timing of different data types acquired during the inpatient stay, from the initial evaluation (after admission and study enrollment) to discharge.

Static data included demographic information, stroke characteristics (history, time between stroke event and enrollment [days; Time Post Onset], stroke type, stroke side of the lesion, inpatient length of stay [LoS]), a cognitive screening tool (Montreal Cognitive Assessment), and 6 sleep-related patient reports which were administered after enrollment (Pittsburgh Quality Sleep Index; Berlin Obstructive Sleep Apnea Scale; Sleep Self-Efficacy Scale; Insomnia Severity Index; Epworth Sleepiness Scale; and Functional Assessment of Chronic Illness Therapy Fatigue Scale).

Semi-dynamic data included an initial sleep evaluation and a discharge sleep evaluation, with 3 daily patient reports to obtain per-night perceptions of sleep health (Functional Outcomes of Sleep Questionnaire [FOSQ] and Karolinska Sleepiness Log [KSL]) and disruptions (Potential Hospital Sleep Disruptions [PHSD]). Semi-dynamic data also included biometric data from wireless, wearable sensors during overnight sleep (Advanced NeoNatal Epidermal sleep system; Sibel Health, Inc.; Chicago, IL, USA). Sensors were placed on the chest and finger (second or third digit on the hand less affected by stroke) to record heart rate from chest-Electrocardiogram (HR) and finger-Photoplethysmogram (PR), SPO2, and chest/finger skin temperature. These data are indicators of circadian rhythm, providing indirect measures of sleep health.²⁷ Additional semi-dynamic data included standardized assessments of patient functional levels at the initial and discharge time points (Quality Indicator [QI] Mobility domain; QI Self-Care domain; Action Research Arm Test [ARAT]; 10-m Walk Test [10MWT]; and 6-minute Walk Test [6MWT]).

Lastly, dynamic data were obtained from 2 common actigraphy devices both worn on the less-affected wrist, the Actiwatch Spectrum (Philips Respironics; Murrysville, PA, USA) and the ActiGraph wGT3X-BT (ActiGraph; Pensacola, FL, USA). These devices computed sleep health metrics throughout the IRF stay, including sleep duration (Total Sleep Time [TST]), overnight activity (Total Activity Count [Total AC]), sleep efficiency, sleep fragmentation, and wakefulness (Wake After Sleep Onset [WASO]). Each device extracts these metrics using different algorithms (Actiwatch: Autoscore Automatic Minor Rest Interval Algorithm, Actiware [version 6.2.0.39]; ActiGraph: Cole-Kripke algorithm, ActiLife [version 6.13.4]). Given that the literature does not indicate that 1 device provides more accurate metrics than the other in absolute terms,²⁸ both ActiGraph and Actiwatch were included to capture a broader range of sleep-related data and indirectly gain insights into the degree of similarity between their respective sleep variables.

Data Analyses

Prior to analysis, data were imputed to mitigate potential issues with missing values in this longitudinal and multimodal dataset and to facilitate analyses across as many participants as possible.²⁹ Imputation used a weighted k-Nearest Neighbors technique, specifically designed for static and dynamic data imputation.³⁰ The data preprocessing and imputation technique, as well as the imputation accuracy and reliability are described in Supplemental Methods SM2 and SM3 and Supplemental Results SR1.

In the final dataset, both dynamic and semi dynamic data (excluding functional recovery) were discretized to represent overall values during initial and discharge time points. Thus, the 3 assessments within the initial window and 3 assessments in the discharge window were averaged into a single value across all dynamic and semi-dynamic (excluding functional recovery) data types for use in analysis (Figure 1). The full set of dynamic data from actigraphy across patient stay was only used to evaluate imputation performance, as described in the Supplemental Method SM2.

Descriptive analyses, time comparisons, and group comparisons were performed to examine sleep health and sleep disruptions for this cohort, and subsequently, the relationship between sleep health and functional recovery. An overview of the data analyses to address each study objective is presented in Figure 2 and described below. Data imputation was performed using R (version 4.3.3). All other analyses were performed using Python (version 3.11.7) and IBM SPSS Statistics (version 20.0).

Figure 2.

Data analysis overview. Summary of analyses conducted to evaluate sleep health, sleep disruptions, and their relation to functional recovery during inpatient stroke rehabilitation. Relevant figures and tables for each analysis are indicated on the lower-right side of its box.

Sleep Health

Sleep health was investigated through an exploratory approach, primarily aiming at the description of sleep health of participants during rehabilitation after stroke. We sought to understand initial sleep health, if it changes over time, if it is different across clinical groups, and what factors affect its evolution in time.

To understand initial sleep health and any changes over time, descriptive analyses and time comparisons were performed. Distribution normality was tested using the Shapiro–Wilk test. Descriptive analyses were provided in terms of mean (standard deviation [SD]) for continuous, normally distributed variables, median [interquartile range] for continuous, non-normally distributed variables, and absolute and percentage frequencies for categorical variables. Time comparisons (ie, comparing variables at initial and discharge) were performed with the McNemar test (dichotomous), Maxwell’s test (categorical, more than 2 groups), and paired t-test (continuous, normally distributed) or Wilcoxon test (continuous, non-normally distributed). Cohen’s d (calculated on raw data for both paired t-test and Wilcoxon) was used to quantify effect size in time comparison analyses. The significance threshold was 0.05 for all tests.

To evaluate sleep health across clinical groups, comparisons were based on factors known to be one of the main indicators of sleep quality (PSQI³¹), to affect sleep (age,³² sex,³³ and BMI³⁴), and attributes of stroke (type³⁵ and region³⁶). Differences between participants with higher/lower LoS were investigated as a proxy of participants’ recovery. Patient subgroups were dichotomized as follows: age (age ≥ 65 vs age < 65), sex (male vs female), BMI (BMI < 25 vs BMI ≥ 25), PSQI scores (PSQI < 5 vs PSQI ≥ 5³¹), stroke type (ischemic vs hemorrhagic), stroke region (left vs right brain region), and LoS (LoS < median LoS vs LoS ≥ median LoS). A chi-squared test (categorical), t-test (continuous, normally distributed), and Mann–Whitney U test (continuous, non-normally distributed) were used to determine differences among groups with a significance threshold of 0.05 (not corrected for multiple comparisons due to the exploratory nature of the analyses).

To understand factors affecting sleep health evolution, multivariate analyses of sleep health over time were obtained using Generalized Linear Models (GLM) for repeated measures. Dependent variables for the GLM models were selected to represent the main sleep dimensions, namely sleep satisfaction, duration, continuity, alertness, and timing.³⁷ Individual models were developed for each of the sleep health measures, namely the KSL, sleep duration (TST), and nighttime activity (Total AC; for both Actiwatch and ActiGraph). For each model, demographics, autonomic biomarkers (temperature and SPO2), and the remaining sleep health measures (those not assigned as the dependent variable in that model) served as covariates. The dichotomized PSQI categorization was included as a between-subject factor to account for the patient’s perceived sleep health prior to hospitalization. This analysis explored the effect of each covariate on the dependent variable, and how it evolved between the initial and discharge evaluations.

Sleep Disruptions

Like sleep health, sleep disruptions were investigated following an exploratory approach. We aimed to understand sleep disruptions across patients, if the number of disturbances is changing over time, if there are differences between nights with or without disturbances, and which factors are associated with disturbances. Statistical methods for descriptive analyses, time comparisons, and group comparisons were performed as described above in Sleep health.

Sleep disruptions were captured using the PHSD survey,³⁸ administered during the initial and discharge evaluations to record patient perceptions about the severity (1: not disruptive, 5: extremely disruptive) and frequency (total number of reported disruptions in a night with severity ≥2) of 14 different types of disruptions to their previous night of sleep (vitals, medications, testing/drawing blood, turning, toileting, feeding, pain, feeling anxious, noise, staff conversation, alarms, bed comfort, cleaning staff, and room temperature).

To determine factors associated with IRF sleep disruptions during the inpatient stay, a binary logistic multivariate regression was performed. The dependent variable was frequency of disruption, simplified into “None or 1 disruptions” or “2 or more disruptions” considering the average number of reported nightly disruptions during the entire inpatient stay. Independent variables included demographics, stroke history, clinical assessment scores, and initial sleep health. Each independent variable was first analyzed singularly through group comparisons; then, all variables significantly associated with the disruption (P < .05) were analyzed in a multivariate model.

To explore differences between disrupted and non-disrupted nights on measures of sleep health, a separate analysis was conducted focusing on the impact of individual disruptions. Unlike the primary analysis, which uses imputed data averaged into initial and discharge time points to determine changes in disruption over time, this analysis averages all disrupted nights (specific disruption severity ≥2) and non-disrupted nights (specific disruption severity <2) to determine sleep health in the presence of a particular disruption, independent of time. Only participants with at least 2 disrupted and 2 non-disrupted nights per disruption type were included, along with all night-matched sleep health metrics. Since each disruption was analyzed independently, the number of participants included varied, and multiple disruptions could occur on the same night.

Functional Recovery

Functional recovery was evaluated based on changes in standardized functional levels from initial to discharge. We sought to changes in function during the inpatient stay, whether sleep health differs across patients with and without improved function, and whether sleep health is a predictor of functional recovery. Measurements of functional outcomes included: the Mobility and Self-Care domains of QI (derived from section GG of the IRF-Patient Assessment Inventory³⁹), ARAT, 6MWT, and 10MWT. Statistical methods for descriptive analyses, time comparisons, and group comparisons were performed as described above in Sleep health.

Predictors of functional recovery were analyzed using group comparisons and multivariate logistic regressions, utilizing as independent variables those significantly associated to the outcome in the group comparison (P < .05). The dependent variables were the grouping of patients as responders or non-responders for each functional outcome. Specifically, patients with improvements greater than or equal to the Minimal Clinically Important Difference (MCID) for each outcome were considered “responders,” otherwise “non-responders.” The MCID values for the ARAT, 10MWT, and 6MWT were 15,⁴⁰ 0.16 m/s,⁴¹ and 54 m,⁴² respectively. The QI Mobility MCID value was estimated using Youden’s J statistic on the cohort data, using the responder/non-responder groups on the 10MWT as the reference for the calculation, while the QI Self-Care MCID was estimated with the median value over the delta (discharge QI Self-Care-initial QI Self-Care). MCID for the QI Mobility and Self-Care outcomes were estimated at scores of 33 and 14, respectively.

Results

Data Availability and Processing

A total of 164 patients were recruited for this study. Of these, 43 withdrew prior to or during data collection due to personal or medical reasons, resulting in 121 patients who completed the study and 1714 nights of observation (final dataset). An additional 18 patients had fewer than 2 nights of data available at the initial and discharge time points and were excluded from the analyses, resulting in 103 patients and 1475 nights of observation (final imputed and non-imputed datasets). Imputation resulted in a low normalized Root Mean Square Deviation (<0.4), with few differences compared to the non-imputed data (Supplemental Results SR1), indicating the validity of the imputation. As such, results are presented below for the imputed dataset. Results for the non-imputed dataset are reported in the Supplemental Tables S2 and S3b to S10b.

Sleep Health and Sleep Disruptions

Descriptive statistics of the 121 participants in the final dataset and the 103 participants included for analysis (final imputed and non-imputed datasets) are presented in Supplemental Table S1, Tables 1 and 2, and Supplemental Table S2, respectively.

Table 1.

Demographics and Functional Outcomes for Inpatient Stroke Cohort (N = 103; Static and Semi-dynamic Data).

Variables	Initial measure	Discharge measure	Statistic
Age	60.86 (15.85)	—	—
Sex	Male: 54 {52.43}
BMI	26.6 [7.85]	—	—
Race	Black/African American: 55 {53.4}; White: 41 {39.81};Other: 7 {6.80}	—	—
Stroke type	Hemorrhagic: 29 {28.16};Ischemic: 66 {64.08};Both: 8 {7.77}	—	—
Stroke region	Left: 37 {35.92};Right: 47 {45.63}Bilateral/both: 19 {18.45}	—	—
TPO, days	10 [9.00]	—	—
ESS	Low risk (0,1): 60 {58.25}	—	—
BOSA	Low risk (0): 18 {17.48}	—	—
ISI	Low risk (0,1): 66 {64.08}	—	—
PSQI	6 [6.00]	—	—
SSES	33 [12.50]	—	—
FACIT-F	37 [16.00]	—	—
MoCA	22 [6.00]	—	—
LoS, days	—	21 [10.00]	—
Days in study	16 [9.00]	—	—
Days until enrollment in study after admission	5 [2.00]	—	—
QI Mobility	31 [18.50]	66 [28.50]	1.5 [<0.001]
10MWT, m/s	0.250 [0.430]	0.580 [0.610]	314.5 [<0.001]
6MWT, m	58.52 [140.9]	172.2 [233.9]	70 [<0.001]
QI Self-Care	20 [8.00]	36 [14.00]	2.0 [<0.001]
ARAT	29.5 [35.90]	38.0 [14.12]	491.5 [<0.001]

Abbreviations: 10MWT, 10-Meter Walk Test (in m/s); 6MWT, 6-Minute Walk Test (in m); ARAT, Action Research Arm Test (ara total score); BMI, Body Mass Index; BOSA, Berlin Obstructive Sleep Apnea Scale; ESS, Epworth Sleepiness Scale; FACIT-F, Functional Assessment of Chronic Illness Therapy Fatigue Scale; FOSQ, Functional Outcomes of Sleep Questionnaire; IQR, Interquartile Range; ISI, Insomnia Severity Index; LoS,Length of Stay; MoCA, Montreal Cognitive Assessment; N, Number; PSQI, Pittsburgh Sleep Quality Index; QI, Quality Indicator; SSES, Sleep Self-Efficacy Scale; std, Standard Deviation.

N = 103. Except for: Disruption Frequency, 10MWT, 6MWT, ARAT.

Measure: mean (std), median [IQR], N {%}.

Statistic: Paired t-test statistic (p), Wilcoxon signed-rank test statistic [p], McNemar test statistic {p}.

Table 2.

Descriptives and Time Comparisons for Dynamic Variables and Semi-dynamic Sleep Quality and Disruptions.

Variables	Initial measure	Discharge measure	Statistic (p-value), effect size
Mean total AC, counts	9324.41 [6133.17]	7250.67 [5177.31]	1749 [0.002], −0.296
Mean total AC^*, counts	40485.48 [32631.97]	34150.84 [23737.51]	2194 [0.111], −0.220
Mean latency, min	6.17 [9.71]	8.07 [11.33]	2116.5 [0.065], 0.196
Mean latency^*, min	0.86 [1.33]	0.67 [1]	1777.5 [0.086], −0.150
Mean efficiency, %	84.14 [7.42]	85.43 [7.58]	2123 [0.068], 0.187
Mean efficiency^*, %	86.89 [9.53]	88.81 [6.56]	1819 [0.005], 0.333
Mean WASO, min	66.5 [36.42]	51.17 [31.47]	1632.5 [0.001], −0.375
Mean WASO^*, min	69.03 [57.76]	61.5 [41.71]	2021.5 [0.031], −0.285
Mean TST, min	541.06 (68.77)	543.37 (82.33)	2545 [0.662], −0.089
Mean TST^*, min	472.24 (77.19)	486.09 (67.42)	2139 [0.076], 0.156
Mean Frag. Index, %	27.28 [8.7]	25.32 (8.42)	1737 [0.002], −0.350
Mean Frag. index^*, %	13.9 (6.18)	13.35 (5.58)	0.67 (0.506), −0.066
Mean duration day after, min	861.24 (55.28)	853.43 [92.72]	2281 [0.192], −0.239
Mean total AC day after, counts	86494.12 [42 945.5]	84263.72 [43 804.69]	2281 [0.192], −0.029
Mean HR, BPM	73.81 (10.14)	72.48 (9.33)	2.23 (0.028), −0.220
Mean PR, BPM	72.75 (10.06)	71.69 (9.12)	1.68 (0.097), −0.165
Mean SPO2, %	94.78 [3.23]	94.64 (2.47)	2518 [0.599], 0.032
Mean SPO2< 90%, %	3.83 [13.38]	5.38 [13.86]	2442 [0.651], 0.056
Mean SPO2 < 88%, %	0.97 [6.17]	1.34 [6.73]	2244 [0.520], −0.028
Mean finger temperature, °C	33.31 [2.21]	33.66 [1.35]	1741 [0.002], 0.341
Mean chest temperature, °C	34.62 (0.75)	34.62 (0.74)	0.01 (0.992), −0.001
Mean KSL	18.19 (5.05)	16.33 [6.67]	1974.5 [0.030], −0.251
Mean FOSQ	12 [1], 12.82 (2.07)	12 [1], 12.29 (1.67)	1051.5 [<0.001], −0.335
Mean PHSD category	None or 1: 32 {31.1}	None or 1: 47 {45.6}	206.5 [0.023], −0.242
VitalVitals: Severity	2 [1.5]	2 [2.0]	867.5 [0.422], −0.094
Vitals: Number	0.50 [0.33]	0.33 [0.67]	737 [0.027]
Medication: Severity	2 [1.5]	2 [2.0]	773 [0.827], 0.047
Medication: Number	0.33 [0.67]	0.33 [0.67]	850.5 [0.632]
Tests:Severity	2 [1.0]	1 [1.0]	451.5 [0.054],−0.193
Tests: Number	0.33 [0.67]	0.33 [0.33]	836.5 [0.081]
Turning: Severity	2 [1.0]	1 [1.0]	225 [0.115], −0.165
Turning: Number	0.00 [0.33]	0.00 [0.33]	516 [0.229]
Toileting: Severity	1 [1.0]	1 [0.0]	628 [0.564], −0.052
Toileting: Number	0.33 [0.67]	0.13 [0.63]	968 [0.781]
Feeding: Severity	1 [1.0]	1 [1.0]	48.5 [0.790], 0.037
Feeding: Number	0.00 [0.00]	0.00 [0.00]	98.5 [0.804]
Pain: Severity	1 [0.0]	1 [0.0]	293.5 [0.157], −0.152
Pain: Number	0.33 [0.67]	0.00 [0.33]	408.5 [0.095]
Anxiety: Severity	1 [1.0]	1 [0.0]	230 [0.218], −0.120
Anxiety: Number	0.00 [0.33]	0.00 [0.31]	514 [0.967]
Noise: Severity	1 [1.0]	1 [0.0]	113 [0.249], −0.120
Noise: Number	0.00 [0.33]	0.00 [0.00]	232.5 [0.168]
Staff: Severity	1 [1.0]	1 [0.0]	77.5 [0.448], −0.074
Staff: Number	0.00 [0.00]	0.00 [0.00]	67 [0.083]
Alarm: Severity	1 [0.0]	1 [0.0]	59.5 [0.976], −0.018
Alarm: Number	0.00 [0.00]	0.00 [0.00]	98.5 [0.545]
Bed: Severity	1 [0]	1 [0.0]	351 [0.562], −0.041
Bed: Number	0.00 [0.58]	0.00 [0.33]	495.5 [0.338]
Cleaning staff: Severity	1 [1.0]	1 [1.0]	52.5 [1.000], 0
Cleaning staff: Number	0.00 [0.00]	0.00 [0.00]	82 [0.375]
Room temperature: Severity	1 [0.0]	1 [0.0]	274 [0.199], −0.135
Room temperature: Number	10.00 [0.33]	10.00 [0.31]	274402 [0.048]

Abbreviations: AC, Activity Count; FOSQ: Functional Outcomes of Sleep Questionnaire; Frag., Fragmentation; HR, Heart Rate; IQR, Interquartile Range; KSL, Karolinska Sleepiness Log; PHSD, Potential Hospital Sleep Disruptions; PSQI, Pittsburg Sleep Quality Index; PR, Pulse Rate; std, Standard Deviation; TST, Total Sleep Time; WASO, Wake After Sleep Onset.

N = 103.

Measure: mean (std), median [IQR], N {%}.

Statistic: Paired t-test statistic (p-value), Wilcoxon signed-rank test statistic [p-value], McNemar test statistic {p-value}. Note: p-values are reported uncorrected.

Severity: Severity of the disruption following the PHSD scale (1: not at all disruptive – 5: extremely disruptive).

Number: Number of disruptions reported per patient at initial and discharge, non-imputed. Number scale of 0 to 3, given the maximum of 3 initial and 3 discharge nights.

From ActiGraph, else from Actiwatch.

Sleep Health

Generally, sleep health improved between the initial and discharge evaluations in the IRF, though with high wakefulness (WASO) and sleep fragmentation throughout their stay, thus still demonstrating poor sleep as defined by Nelson et al¹² (Table 2, Figure 3). Objective sleep metrics revealed significantly better sleep efficiency and shorter wakefulness (WASO) by discharge (Actiwatch and ActiGraph, Figure 3, top and middle). For Actiwatch only, a reduction in overnight activity (Total AC) and fragmentation over this period (Figure 3 top) was also found. Patients perceived significantly improved sleep quality (KSL) and improved function after sleep (FOSQ), and they exhibited reduced overnight heart rate and higher overnight finger temperature between initial and discharge.

Figure 3.

Sleep health. Violin plots displaying sleep health metrics from Actiwatch (top row), ActiGraph (middle row) and the Karolinska Sleep Log (KSL; bottom row), comparing initial and discharge time points. Shaded regions are based on the recommendations from Nelson et al,¹² showing values that are considered representative of poor sleep. The bottom right box represents the multivariate results for nighttime activity (Total AC; Actiwatch) and heart rate (HR, PR). Arrows represent groups based on increasing or decreasing values of the mentioned sleep metric between initial and discharge.

Dichotomized group comparisons are provided in Supplemental Tables S3a and S3b. Very few significant differences were found between subgroups based on stroke descriptors (ie, stroke type, ischemic vs hemorrhagic, and stroke brain region [left vs right]). BMI subgroups (BMI < 25 vs BMI ≥ 25) showed statistically significant differences between autonomic sleep metrics (ie, increased initial and discharge overnight SPO2, higher initial and discharge overnight chest temperature, and reduced discharge hypoxemia for the lower BMI group). Age (age < 65 vs age ≥ 65) and PSQI (PSQI < 5 vs PSQI ≥ 5) subgroups had significantly different actigraphy-based sleep metrics (ie, reduced discharge sleep fragmentation and higher discharge daytime activity from Actiwatch, and shorter discharge sleep duration [Actiwatch and ActGraph] for the younger age group; shorter discharge sleep latency [ActiGraph]; and increased discharge sleep fragmentation [Actiwatch] for the better perceived sleep quality group).

The results of the GLM analyses are provided in Supplemental Tables S4a and S4b (with demographic covariates only) and Supplemental Tables S5a and S5b (demographics and the other sleep health metrics not used as dependent variables as covariates). Changes over time of sleep duration (ActiGraph Mean TST) and overnight activity (ActiGraph Mean Total AC) were significantly associated (P < .001 in both models with outcome ActiGraph Mean TST and Mean Total AC, respectively). The variables showed an inverse relationship—that is, from initial to discharge time points, when overnight activity is lower, sleep duration is shorter. Further, perceived sleep quality (Mean KSL) was significantly associated with overnight activity (Actiwatch Mean Total AC) and heart rate (Mean HR and Mean PR; P = .015 for Time × Mean Total AC delta, P = .025 for Time × Mean HR delta, and P = .024 for Time × Mean PR delta). In all cases, there was a steeper improvement in sleep quality between initial and discharge in the presence of lower overnight activity and higher heart rate over time (Figure 3, bottom).

Sleep Disruptions

Of the 590 nights with available associated sleep disruption surveys across 103 participants, 77% were reported as having a sleep disruption. Specifically, 374 nights (63%) had 2 or more disruptions, 80 nights (14%) had 1 disruption, and 136 nights (23%) had no disruptions. The most frequently reported sleep disruptions were vitals measurement, medication, toileting, and tests, whereas disruptions due to feeding or cleaning staff were rarely reported (Figure 4). From initial to discharge, there was a significant reduction in the total number of reported sleep disruptions (PHSD), with 69% reporting multiple disruptions during initial evaluation and 54% reporting multiple disruptions during discharge evaluation. Individual disruption severity did not significantly change between initial and discharge, although the proportion of nights with disruptions per patient due to vitals and room temperature significantly decreased (Table 2). Detailed changes in sleep disruptions from initial to discharge time points are summarized in Table 2 and Supplemental Table S2, including the total number of reported disruptions and the severity and number of each disruption type. Reduced sleep disruptions during the inpatient hospitalization were significantly associated with improved perceived sleep quality (Mean KSL initial, R² = 36.9%) through binary logistic regression (Supplemental Tables S6a and S7b).

Figure 4.

Sleep disruptions. Percentage of participants who reported various disruptions on each night of sleep at initial and discharge evaluation, from the Potential Hospital Sleep Disruption (PHSD) survey. Darker colors indicate patients who reported greater severity of the disruption to their sleep (1: not disruptive, 5: extremely disruptive).

Associations related to specific sleep disruptions, disrupted versus non-disrupted sleep, are provided in Supplemental Table S8. Of the 14 types of disruptions, 9 (excluding feeding, staff, alarm, cleaning, and room temperature) revealed significant differences when comparing nights with and without that disruption. Eight of these resulted in significantly worsened perceived sleep quality (Mean KSL). Nights with vitals disruptions had significantly longer wakefulness (Mean WASO, ActiGraph), while nights with tests and toileting disruptions had significantly longer wakefulness, higher initial overnight activity (Actiwatch and ActiGraph: tests; ActiGraph: toileting), and increased sleep fragmentation (Actiwatch). Participants who experienced noise and bed discomfort disruptions had higher overnight finger and chest temperatures compared to their nights of non-disrupted sleep.

Functional Recovery and the Relationship to Sleep

All functional outcomes demonstrated a statistically significant improvement between the initial and discharge time points (Table 1; Supplemental Table S1), with improvement beyond the MCID for 48.5% of patients for QI Mobility, 58.3% for 6MWT, 53.4% for 10MWT, 55.3% for QI Self-Care, and 27.2% for ARAT scores.

In general, functional responders had improved daytime activity (Actiwatch; QI Mobility) and functional outcomes near discharge; however, minimal significant differences were found between functional responders and non-responders regarding sleep. Of note, functional responders had higher overnight chest and finger temperature for 6MWT, QI Self-Care, ARAT groups, and better PSQI with QI-Self-Care groups, but no overnight measures of actigraphy were significantly different amongst groups. Results for group comparisons, univariate, and multivariate regressions are displayed in Supplemental Tables S9a and S10b.

Discussion

In this work, we performed an extensive and multimodal description of sleep health for individuals with stroke in the inpatient rehabilitation setting, considering both subjective (patient reports) and objective (sensors for actigraphy and autonomic monitoring) sleep data. Four key findings from this work are: (1) patients’ sleep health improved over time, but remained poor; (2) sleep disruptions fluctuated throughout hospitalization, highlighting the variable nature of environmental and physiological barriers to poor sleep; (3) biomarkers such as heart rate and activity provided additional value to capture sleep changes over time; and (4) no significant association between sleep health and functional recovery was found for this cohort. Each of these findings is discussed below.

Our analysis revealed generally improved sleep health over the IRF stay (Figure 3). Average sleep duration (TST) at both initial and discharge was above the minimum of 7 hours (420 minutes) suggested for adults.¹² However, both the initial and discharge wakefulness (WASO) exceeded the normative range of 20 to 51 minutes¹² (average 14 minute improvement). Additionally, patients’ reported sleep remained poor with worse perceived sleep quality (KSL < 0) and suboptimal function after sleep (FOSQ < 7; Table 2). Comparatively, Huang et al¹⁸ reported poor sleep health at admission to post-stroke inpatient rehabilitation (mean (SD); sleep duration, 259 (71) minutes; WASO, 93.1 (74.2) minutes), and Williams-Cooke et al²⁴ found that poor sleep can persist or develop over time in the inpatient setting (57% of their acute stroke cohort sleeping outside of 7-9 hours). Our study found that age affected sleep health, with patients younger than 65 experiencing less fragmentation and higher daytime activity commensurate with age and sleep literature,^32,43 but contrasted with lower total sleep time.⁴⁴ There were no clear associations between stroke etiology and sleep health; previous literature has found conflicting evidence and no consensus yet on this topic.^35,45,46 Overall, additional studies are needed to determine whether improvements in post-stroke sleep health are due to patient characteristics, neural recovery, and acclimation to the hospital setting, or a combination of these factors.

Poor sleep and frequent disruptions persisted throughout inpatient rehabilitation with most disruptions being accredited to hospital-specific environmental factors such as vitals, medication, testing, bed comfort, and room temperature. Reports of multiple disruptions saw a significant reduction from initial to discharge evaluation, but remained on more than 50% of recorded nights. Hence, policy or behavioral interventions might be considered to reduce unintended disruptions, which could lead to fewer awakenings and longer sleep durations.

Multivariate analyses of sleep health demonstrated that the improvement of perceived sleep (KSL) relied on a reduced overnight activity (Total AC) and changes in biometric signals (HR and PR). These findings align with existing literature linking sleep health to circadian rhythms; particularly the well-documented decrease in HRV during non-REM sleep and its increase during REM sleep and wakefulness.^47,48 Hence, given the observed reductions in overnight activity, WASO, and fragmentation, along with increased sleep efficiency, suggest a prolonged duration of non-REM sleep and a corresponding decrease in HR.

Previous work has argued that sleep is an essential factor in functional recovery during neurological rehabilitation.^1,2 Fleming et al⁴⁹ found more disrupted sleep to be significantly associated with a slower functional recovery (Functional Independence Measure) in a cohort of 59 participants with neurological injury, including stroke. In contrast, sleep health was not strongly associated with functional outcomes in our cohort. However, there were numerous methodological differences which may account for these different findings including our shorter median time since stroke (10 days vs 29 days) and shorter median length of stay (21 days vs 71 days) compared to Fleming et al. Factors such as spontaneous recovery and therapy intensity may overshadow the effect of sleep health, since the first 3 weeks post-stroke is the period of greatest improvements.³ Other differences, such as the types of functional assessments and neurological conditions, warrant the need for additional studies to expand upon the relationship between early functional recovery and sleep health.

When interpreting the findings of this study, certain limitations should be considered. Data were collected at a specific IRF site, and patients were excluded if they had prior diagnoses of sleep disorders, which can also impact post-stroke recovery.⁵⁰ As such, results may not generalize to other sites and patient subgroups. Missing data also posed a challenge in the analysis, since not all participants completed every measurement during their stay. Although we made rigorous efforts to maximize the sample size for analysis through a validated data imputation process, there is still some uncertainty resulting from missing data that should be considered. Furthermore, the multiple comparisons performed in this analysis may increase the risk of type I errors. Although methods such as Bonferroni correction are commonly applied in such contexts, we chose to report uncorrected p-values for this moderately large sample and exploratory analysis.⁵¹ Since many of these comparisons were used to screen variables for subsequent multivariate analyses,⁵² we chose an approach that prioritized minimizing the risk of type II errors, since correcting for multiple comparisons could otherwise lead to missing significant associations.⁵³ Finally, sleep metrics from the Actiwatch and ActiGraph devices have not been validated for an acute or subacute stroke population. Thus, we prioritized analysis of within-individual sleep changes during the inpatient stay, relying on the devices to capture general sleep trends. Interestingly, we found few differences in sleep metrics computed by ActiGraph and Actiwatch, with the exception of activity counts (Figure 3), which is most likely due to the difference in algorithms and activity thresholds. Despite this, the remaining sleep metrics suggest that both devices capture similar information and supporting the stability of measures computed by these devices.

Overall, this study illuminates the relationship between sleep health and underlying physiological mechanisms during inpatient stroke rehabilitation. These findings have promising implications for the design of targeted, sleep-improving interventions to alleviate poor sleep and facilitate early stroke recovery in the inpatient setting. Helping patients sleep better during early post-stroke recovery may facilitate recovery and long-term outcomes as well as mitigate the detrimental impact of poor sleep on overall health.¹⁰

Supplemental Material

sj-docx-1-nnr-10.1177_15459683251335332 – Supplemental material for Sleep Following a Stroke: Multimodal Evaluation of Sleep Health and Disruptions and Impact on Recovery During Acute Inpatient Rehabilitation

Supplemental material, sj-docx-1-nnr-10.1177_15459683251335332 for Sleep Following a Stroke: Multimodal Evaluation of Sleep Health and Disruptions and Impact on Recovery During Acute Inpatient Rehabilitation by Jacob Sindorf, Silvia Campagnini, Megan K. O’Brien, Aashna Sunderrajan, Kristen L. Knutson, Phyllis C. Zee, Lisa Wolfe, Vineet M. Arora and Arun Jayaraman in Neurorehabilitation and Neural Repair

Footnotes

Acknowledgements

The authors thank the research and data collection teams.

Author Contributions

Jacob Sindorf: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Software; Validation; Visualization; Writing—original draft; and Writing—review & editing. Silvia Campagnini: Conceptualization; Data curation; Formal analysis; Methodology; Software; Validation; Visualization; Writing—original draft; and Writing—review & editing. Megan K. O’Brien: Conceptualization; Methodology; Resources; Supervision; Writing—original draft; and Writing—review & editing. Aashna Sunderrajan: Conceptualization; Investigation; Writing—original draft; and Writing—review & editing. Kristen L. Knutson: Conceptualization; Software; and Writing—review & editing. Phyllis C. Zee: Conceptualization; Resources; and Writing—review & editing. Lisa Wolfe: Conceptualization; Resources; and Writing—review & editing. Vineet M. Arora: Conceptualization; Funding acquisition; Resources; Supervision; and Writing—review & editing. Arun Jayaraman: Conceptualization; Funding acquisition; Resources; Supervision; and Writing—review & editing.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project is supported by the National Institutes of Health (R01HD097786 and T32HD007418 [J.S.])

Ethical Considerations

The study was approved by the Institutional Review Board at Northwestern University (STU00211695)

Consent to Participate

All participants provided informed, written consent.

Consent for Publication

Not applicable.

ORCID iDs

Jacob Sindorf

Silvia Campagnini

Megan K. O’Brien

Arun Jayaraman

Supplementary material for this article is available on the Neurorehabilitation & Neural Repair website along with the online version of this article.

References

Tramontano

De Angelis

Galeoto

, et al. Physical therapy exercises for sleep disorders in a rehabilitation setting for neurological patients: a systematic review and meta-analysis. Brain Sci. 2021;11:1176.

Arora

Machado

Anderson

, et al. Effectiveness of SIESTA on objective and subjective metrics of nighttime hospital sleep disruptors. J Hosp Med. 2019;14:38-41.

Grefkes

Fink

GR.

Recovery from stroke: current concepts and future perspectives. Neurol Res Pract. 2020;2:17.

Fulk

Boyne

Hauger

, et al. The impact of sleep disorders on functional recovery and participation following stroke: a systematic review and meta-analysis. Neurorehabil Neural Repair. 2020;34:1050-1061.

Duss

Seiler

Schmidt

, et al. The role of sleep in recovery following ischemic stroke: a review of human and animal data. Neurobiol Sleep Circadian Rhythms. 2016;2:94-105.

Weightman

Robinson

Fallows

, et al. Improving sleep and learning in rehabilitation after stroke, part 2 (INSPIRES2): study protocol for a home-based randomised control trial of digital cognitive behavioural therapy (dCBT) for insomnia. BMJ Open. 2023;13:e071764.

Fleming

Smejka

Macey

, et al. Improving sleep after stroke: a randomised controlled trial of digital cognitive behavioural therapy for insomnia. J Sleep Res. 2024;33:e13971.

Sterr

Kuhn

Nissen

, et al. Post-stroke insomnia in community-dwelling patients with chronic motor stroke: physiological evidence and implications for stroke care. Sci Rep. 2018;8:8409.

Colten

Altevogt

, Institute of Medicine (US) Committee on Sleep Medicine and Research. Extent and health consequences of chronic sleep loss and sleep disorders. In: Colten

Altevogt

, eds. Sleep Disorders and Sleep Deprivation: An Unmet Public Health Problem. National Academies Press; 2006: 137-172. Accessed August 7, 2024. https://www.ncbi.nlm.nih.gov/books/NBK19961/

10.

Braun

Gooneratne

NS.

Rehabilitating sleep. Sleep. 2011;34:1627-1628.

11.

Cai

Wang

X-P

Yang

G-Y.

Sleep disorders in stroke: an update on management. Aging Dis. 2021;12:570-585.

12.

Nelson

Davis

Corbett

CF.

Sleep quality: an evolutionary concept analysis. Nurs Forum. 2022;57:144-151.

13.

Treister

Hatch

Cramer

Chang

EY.

Demystifying post-stroke pain: from etiology to treatment. PM R. 2017;9:63-75.

14.

Alonso-Alonso

Sampedro-Viana

Rodríguez-Yáñez

, et al. Influence of temperature chronobiology on stroke outcome. Int J Mol Sci. 2023;24:3746.

15.

Thosar

Butler

Shea

. Role of the circadian system in cardiovascular disease. J Clin Invest. 128:2157-2167.

16.

Stergiou

Vemmos

Pliarchopoulou

Synetos

Roussias

Mountokalakis

TD.

Parallel morning and evening surge in stroke onset, blood pressure, and physical activity. Stroke. 2002;33:1480-1486.

17.

Pérez-Carbonell

Bashir

Narrative review of sleep and stroke. J Thorac Dis. 2020;12:S176-S190.

18.

Huang

R-J

Lai

C-H

Lee

S-D

, et al. Objective sleep measures in inpatients with subacute stroke associated with levels and improvements in activities of daily living. Arch Phys Med Rehabil. 2018;99:699-706.

19.

Cavalcanti

Campos

Araujo

Actigraphic analysis of the sleep-wake cycle and physical activity level in patients with stroke: implications for clinical practice. Chronobiol Int. 2012;29:1267-1272.

20.

Smith

McCrae

Cheung

, et al. Use of actigraphy for the evaluation of sleep disorders and circadian rhythm sleep-wake disorders: an American Academy of Sleep Medicine systematic review, meta-analysis, and GRADE assessment. J Clin Sleep Med. 2018;14:1209-1230.

21.

Chen

P-W

O’Brien

Horin

, et al. Sleep monitoring during acute stroke rehabilitation: toward automated measurement using multimodal wireless sensors. Sensors. 2022;22:6190.

22.

Davies

Lee

Walter

, et al. A single-arm, open-label, multicenter, and comparative study of the ANNE sleep system vs polysomnography to diagnose obstructive sleep apnea. J Clin Sleep Med. 2022;18:2703-2712.

23.

Ferre

Ribó

Rodríguez-Luna

, et al. Strokes and their relationship with sleep and sleep disorders. Neurol Barc Spain. 2013;28:103-118.

24.

Williams-Cooke

Watts

Bonnett

Alshehri

Siengsukon

Association between sleep duration and functional disability in inpatient stroke rehabilitation: a pilot observational study. Arch Rehabil Res Clin Transl. 2021;3:100150.

25.

Bakken

Kim

Finset

Lerdal

Stroke patients’ functions in personal activities of daily living in relation to sleep and socio-demographic and clinical variables in the acute phase after first-time stroke and at six months of follow-up. J Clin Nurs. 2012;21:1886-1895.

26.

Sonmez

Karasel

Poor sleep quality I related to impaired functional status following stroke. J Stroke Cerebrovasc Dis. 2019;28:104349.

27.

Kim

Jung

Kim

D-J.

Autonomic dysfunction in sleep disorders: from neurobiological basis to potential therapeutic approaches. J Clin Neurol Seoul Korea. 2022;18:140-151.

28.

Liu

Schrack

Wanigatunga

, et al. Comparison of sleep parameters from wrist-worn ActiGraph and Actiwatch devices. Sleep. 2024;47:zsad155.

29.

Heymans

Twisk

JWR

. Handling missing data in clinical research. J Clin Epidemiol. 2022;151:185-188.

30.

Tavazzi

Daberdaku

Vasta

Calvo

Chiò

Di Camillo

Exploiting mutual information for the imputation of static and dynamic mixed-type clinical data with an adaptive k-nearest neighbours approach. BMC Med Inform Decis Mak. 2020;20:174.

31.

Luo

Liu

Zhuge

Zhu

Sleep quality after stroke: a systematic review and meta-analysis. Medicine. 2023;102:e33777.

32.

Lavoie

Zeidler

Martin

JL.

Sleep and aging. Sleep Sci. Pract. 2018;2:3.

33.

Krishnan

Collop

NA.

Gender differences in sleep disorders. Curr Opin Pulm Med. 2006;12:383-389.

34.

Gupta

Srivastava

Gupta

Tiwari

Banerjee

Association of sleep duration and sleep quality with body mass index among young adults. J Fam Med Prim Care. 2022;11:3251-3256.

35.

Khazaei

Ayubi

Khazaei

Afrookhteh

Sleep quality and related determinants among stroke patients: a cross-sectional study. Iran J Psychiatry. 2022;17:84-90.

36.

Hepburn

Bollu

French

Sahota

Sleep medicine: stroke and sleep. Mo Med. 2018;115:527-532.

37.

Buysse

DJ.

Sleep health: can we define it? Does it matter?

Sleep. 2014;37:9-17.

38.

Grossman

Anderson

Worku

, et al. Awakenings? Patient and hospital staff perceptions of nighttime disruptions and their effect on patient sleep. J Clin Sleep Med. JCSM. 2017;13:301-306.

39.

Deutsch

Palmer

Vaughan

Schwartz

McMullen

Inpatient rehabilitation facility patients’ functional abilities and validity testing of the standardized self-care and mobility data elements. Arch Phys Med Rehabil. 2022;103:1070-1084.e3.

40.

Lang

Edwards

Birkenmeier

Dromerick

AW.

Estimating minimal clinically important differences of upper extremity measures early after stroke. Arch Phys Med. Rehabil. 2008;89:1693.

41.

Tilson

Sullivan

Cen

, et al. Meaningful gait speed improvement during the first 60 days poststroke: minimal clinically important difference. Phys Ther. 2010;90:196-208.

42.

Rasekaba

Lee

Naughton

Williams

Holland

AE.

The six-minute walk test: a useful metric for the cardiopulmonary patient. Intern Med J. 2009;39:495-501.

43.

Martin

Koster

Murphy

, et al. Changes in daily activity patterns with age in U.S. men and women: national health and nutrition examination survey 2003–04 and 2005–06. J Am Geriatr Soc. 2014;62:1263-1271.

44.

Hirshkowitz

Whiton

Albert

, et al. National Sleep Foundation’s updated sleep duration recommendations: final report. Sleep Health. 2015;1:233-243.

45.

Karaca

Factors affecting poststroke sleep disorders. J Stroke Cerebrovasc Dis. 2016;25:727-732.

46.

Pasic

Smajlovic

Dostovic

Kojic

Selmanovic

Incidence and types of sleep disorders in patients with stroke. Med Arh. 2011;65:225-227.

47.

Faust

Feldman

Mattingly

Hachen

V. Chawla

Deviations from normal bedtimes are associated with short-term increases in resting heart rate. Npj Digit Med. 2020;3:1-9.

48.

Sajjadieh

Shahsavari

Safaei

, et al. The association of sleep duration and quality with heart rate variability and blood pressure. Tanaffos. 2020;19:135-143.

49.

Fleming

Smejka

Henderson Slater

, et al. Sleep disruption after brain injury is associated with worse motor outcomes and slower functional recovery. Neurorehabil Neural Repair. 2020;34:661-671.

50.

Hermann

Bassetti

CL.

Role of sleep-disordered breathing and sleep-wake disturbances for stroke and stroke recovery. Neurology. 2016;87:1407-1416.

51.

Streiner

Norman

GR.

Correction for multiple testing: is there a resolution?

Chest. 2011;140:16-18.

52.

Sauerbrei

Perperoglou

Schmid

, et al. State of the art in selection of variables and functional forms in multivariable analysis—outstanding issues. Diagn Progn Res. 2020;4:3.

53.

Armstrong

RA.

When to use the Bonferroni correction. Ophthalmic Physiol Opt J Br Coll Ophthalmic Opt Optom. 2014;34:502-508.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.62 MB