Physiological sleep measures predict time to 15-year mortality in community adults: Application of a novel machine learning framework

Pioneering a Multi-Phase Framework to Harmonize Self-Reported Sleep Data Across Cohorts

STUDY OBJECTIVES

Harmonizing and aggregating data across studies enable pooled analyses that support external validation and enhance replicability and generalizability. However, the multidimensional nature of sleep poses challenges for data harmonization and aggregation. Here we describe and implement our process for harmonizing self-reported sleep data.

METHODS

We established a multi-phase framework to harmonize self-reported sleep data: (1) compile items; (2) group items into domains; (3) harmonize items; and (4) evaluate harmonizability. We applied this process to produce a pooled multi-cohort sample of five United States cohorts plus a separate yet fully harmonized sample from Rotterdam, Netherlands. Sleep and sociodemographic data are described and compared to demonstrate the utility of harmonization and aggregation.

RESULTS

We collected 190 unique self-reported sleep items and grouped them into 15 conceptual domains. Using these domains as guiderails, we developed 14 harmonized items measuring aspects of Satisfaction, Alertness/Sleepiness, Timing, Efficiency, Duration, Insomnia, and Sleep Apnea. External raters determined that 13 of these 14 items had moderate-to-high harmonizability. Alertness/Sleepiness items had lower harmonizability, while continuous, quantitative items (e.g., timing, total sleep time, efficiency) had higher harmonizability. Descriptive statistics identified features that are more consistent (e.g., wake-up time, duration) and more heterogeneous (e.g., time in bed, bedtime) across samples.

CONCLUSIONS

Our process can guide researchers and cohort stewards towards effective sleep harmonization and provides a foundation for further methodological development in this expanding field. Broader national and international initiatives promoting common data elements across cohorts are needed to enhance future harmonization and aggregation efforts.

Development and validation of a novel sleep health score in the sleep heart health study

BACKGROUND

There is a lack of consensus in evaluating multidimensional sleep health, especially concerning its implication for mortality. A validated multidimensional sleep health score is the foundation of effective interventions.

METHODS

We obtained data from 5706 participants in the Sleep Heart Health Study. First, random forest-recursive feature elimination algorithm was used to select potential predictive variables. Second, a sleep composite score was developed based on the regression coefficients from a Cox proportional hazards model evaluating the associations between selected sleep-related variables and mortality. Last, we validated the score by constructing Cox proportional hazards models to assess its association with mortality.

RESULTS

The mean age of participants was 63.2 years old, and 47.6% (2715/5706) were male. Six sleep variables, including average oxygen saturation (%), spindle density (C3), sleep efficiency (%), spindle density (C4), percentage of fast spindles (%) and percentage of rapid eye movement (%) were selected to construct this multidimensional sleep health score. The average sleep composite score in participants was 6.8 of 22 (lower is better). Participants with a one-point increase in sleep composite score had an 10% higher risk of death (hazard ratio = 1.10, 95% confidence interval: 1.08-1.12).

CONCLUSIONS

This study constructed and validated a novel multidimensional sleep health score to better predict death based on sleep, with significant associations between sleep composite score and all-cause mortality. Integrating questionnaire information and sleep microstructures, our sleep composite score is more appropriately applied for mortality risk stratification.

Achieving Better Understanding of Obstructive Sleep Apnea Treatment Effects on Cardiovascular Disease Outcomes through Machine Learning Approaches: A Narrative Review

Obstructive sleep apnea (OSA) affects almost a billion people worldwide and is associated with a myriad of adverse health outcomes. Among the most prevalent and morbid are cardiovascular diseases (CVDs). Nonetheless, randomized controlled trials (RCTs) of OSA treatment have failed to show improvements in CVD outcomes. A major limitation in our field is the lack of precision in defining OSA and specifically subgroups with the potential to benefit from therapy. Further, this has called into question the validity of using the time-honored apnea-hypopnea index as the ultimate defining criteria for OSA. Recent applications of advanced statistical methods and machine learning have brought to light a variety of OSA endotypes and phenotypes. These methods also provide an opportunity to understand the interaction between OSA and comorbid diseases for better CVD risk stratification. Lastly, machine learning and specifically heterogeneous treatment effects modeling can help uncover subgroups with differential outcomes after treatment initiation. In an era of data sharing and big data, these techniques will be at the forefront of OSA research. Advanced data science methods, such as machine-learning analyses and artificial intelligence, will improve our ability to determine the unique influence of OSA on CVD outcomes and ultimately allow us to better determine precision medicine approaches in OSA patients for CVD risk reduction. In this narrative review, we will highlight how team science via machine learning and artificial intelligence applied to existing clinical data, polysomnography, proteomics, and imaging can do just that.

Assessing Risk of Health Outcomes From Brain Activity in Sleep: A Retrospective Cohort Study

Background and Objectives

Patterns of electrical activity in the brain (EEG) during sleep are sensitive to various health conditions even at subclinical stages. The objective of this study was to estimate sleep EEG-predicted incidence of future neurologic, cardiovascular, psychiatric, and mortality outcomes.

Methods

This is a retrospective cohort study with 2 data sets. The Massachusetts General Hospital (MGH) sleep data set is a clinic-based cohort, used for model development. The Sleep Heart Health Study (SHHS) is a community-based cohort, used as the external validation cohort. Exposure is good, average, or poor sleep defined by quartiles of sleep EEG-predicted risk. The outcomes include ischemic stroke, intracranial hemorrhage, mild cognitive impairment, dementia, atrial fibrillation, myocardial infarction, type 2 diabetes, hypertension, bipolar disorder, depression, and mortality. Diagnoses were based on diagnosis codes, brain imaging reports, medications, cognitive scores, and hospital records. We used the Cox survival model with death as the competing risk.

Results

There were 8673 participants from MGH and 5650 from SHHS. For all outcomes, the model-predicted 10-year risk was within the 95% confidence interval of the ground truth, indicating good prediction performance. When comparing participants with poor, average, and good sleep, except for atrial fibrillation, all other 10-year risk ratios were significant. The model-predicted 10-year risk ratio closely matched the observed event rate in the external validation cohort.

Discussion

The incidence of health outcomes can be predicted by brain activity during sleep. The findings strengthen the concept of sleep as an accessible biological window into unfavorable brain and general health outcomes.

Multi-dimensional sleep and mortality: The Multi-Ethnic Study of Atherosclerosis

STUDY OBJECTIVES

Multiple sleep characteristics are informative of health, sleep characteristics cluster, and sleep health can be described as a composite of positive sleep attributes. We assessed the association between a sleep score reflecting multiple sleep dimensions, and mortality. We tested the hypothesis that more favorable sleep (higher sleep scores) is associated with lower mortality.

METHODS

The Multi-Ethnic Study of Atherosclerosis (MESA) is a racially and ethnically-diverse multi-site, prospective cohort study of US adults. Sleep was measured using unattended polysomnography, 7-day wrist actigraphy, and validated questionnaires (2010-2013). 1726 participants were followed for a median of 6.9 years (Q1-Q3, 6.4-7.4 years) until death (171 deaths) or last contact. Survival models were used to estimate the association between the exposure of sleep scores and the outcome of all-cause mortality, adjusting for socio-demographics, lifestyle, and medical comorbidities; follow-up analyses examined associations between individual metrics and mortality. The exposure, a sleep score, was constructed by an empirically-based Principal Components Analysis on 13 sleep metrics, selected a priori.

RESULTS

After adjusting for multiple confounders, a 1 standard deviation (sd) higher sleep score was associated with 25% lower hazard of mortality (Hazard Ratio [HR]: 0.75; 95% Confidence interval: [0.65, 0.87]). The largest drivers of this association were: night-to-night sleep regularity, total sleep time, and the Apnea-Hypopnea Index.

CONCLUSION

More favorable sleep across multiple characteristics, operationalized by a sleep score, is associated with lower risk of death in a diverse US cohort of adults. Results suggest that interventions that address multiple dimensions may provide novel approaches for improving health.

Use and misuse of random forest variable importance metrics in medicine: demonstrations through incident stroke prediction

BACKGROUND

Machine learning tools such as random forests provide important opportunities for modeling large, complex modern data generated in medicine. Unfortunately, when it comes to understanding why machine learning models are predictive, applied research continues to rely on 'out of bag' (OOB) variable importance metrics (VIMPs) that are known to have considerable shortcomings within the statistics community. After explaining the limitations of OOB VIMPs - including bias towards correlated features and limited interpretability - we describe a modern approach called 'knockoff VIMPs' and explain its advantages.

METHODS

We first evaluate current VIMP practices through an in-depth literature review of 50 recent random forest manuscripts. Next, we recommend organized and interpretable strategies for analysis with knockoff VIMPs, including computing them for groups of features and considering multiple model performance metrics. To demonstrate methods, we develop a random forest to predict 5-year incident stroke in the Sleep Heart Health Study and compare results based on OOB and knockoff VIMPs.

RESULTS

Nearly all papers in the literature review contained substantial limitations in their use of VIMPs. In our demonstration, using OOB VIMPs for individual variables suggested two highly correlated lung function variables (forced expiratory volume, forced vital capacity) as the best predictors of incident stroke, followed by age and height. Using an organized analytic approach that considered knockoff VIMPs of both groups of features and individual features, the largest contributions to model sensitivity were medications (especially cardiovascular) and measured medical risk factors, while the largest contributions to model specificity were age, diastolic blood pressure, self-reported medical risk factors, polysomnography features, and pack-years of smoking. Thus, we reach very different conclusions about stroke risk factors using OOB VIMPs versus knockoff VIMPs.

CONCLUSIONS

The near-ubiquitous reliance on OOB VIMPs may provide misleading results for researchers who use such methods to guide their research. Given the rapid pace of scientific inquiry using machine learning, it is essential to bring modern knockoff VIMPs that are interpretable and unbiased into widespread applied practice to steer researchers using random forest machine learning toward more meaningful results.

Hour-to-hour variability of respiratory sleep indices and ambulatory blood pressure

STUDY OBJECTIVES

Respiratory sleep indices are traditionally reported on the basis of the average total sleep time. The relationship between the hour-to-hour variability of these parameters and blood pressure (BP) has not been reported.

METHODS

We evaluated the associations of the hour-to-hour variability of the apnea-hypopnea index (AHI), oxygen desaturation index (ODI), and lowest oxygen saturation with the 24-h ambulatory BP in patients with hypertension and newly diagnosed obstructive sleep apnea. A total of 147 patients underwent polysomnography, based on which obstructive sleep apnea was diagnosed in 106 patients; these patients underwent 24-h ambulatory BP monitoring within the next 30 days. Each polysomnogram was divided into hourly reports to calculate the variability of the respiratory sleep indices. Variability independent of the mean was considered the primary measure of variability.

RESULTS

The median number of hourly polysomnogram reports was 7 (range, 4-8). The hour-to-hour variability of both AHI and ODI, but not of the lowest oxygen saturation, was correlated with the 24-h pulse pressure, 24-h systolic BP, and awake systolic BP (p < 0.05 for all). The fully adjusted linear regression analysis indicated that the hour-to-hour variability of AHI and ODI remained associated with the 24-h pulse pressure (AHI: β coefficient, 0.264 [95% CI = 0.033-0.495], p = 0.026; ODI: β coefficient, 0.450 [95% CI = 0.174-0.726], p = 0.002).

CONCLUSIONS

The hour-to-hour variability of AHI and ODI is independently associated with the 24-h pulse pressure. Further investigations are warranted to evaluate the clinical relevance of this new-found association.

Age estimation from sleep studies using deep learning predicts life expectancy

Sleep disturbances increase with age and are predictors of mortality. Here, we present deep neural networks that estimate age and mortality risk through polysomnograms (PSGs). Aging was modeled using 2500 PSGs and tested in 10,699 PSGs from men and women in seven different cohorts aged between 20 and 90. Ages were estimated with a mean absolute error of 5.8 ± 1.6 years, while basic sleep scoring measures had an error of 14.9 ± 6.29 years. After controlling for demographics, sleep, and health covariates, each 10-year increment in age estimate error (AEE) was associated with increased all-cause mortality rate of 29% (95% confidence interval: 20-39%). An increase from -10 to +10 years in AEE translates to an estimated decreased life expectancy of 8.7 years (95% confidence interval: 6.1-11.4 years). Greater AEE was mostly reflected in increased sleep fragmentation, suggesting this is an important biomarker of future health independent of sleep apnea.

A Novel EEG Derived Measure of Disrupted Delta Wave Activity during Sleep Predicts All-Cause Mortality Risk

RATIONALE

Conventional markers of sleep disturbance, based on manual electroencephalography scoring, may not adequately capture important features of more fundamental electroencephalography-related sleep disturbance.

OBJECTIVES

This study aimed to determine if more comprehensive power-spectral measures of delta wave activity during sleep are stronger independent predictors of mortality than conventional sleep quality and disturbance metrics.

METHODS

Power spectral analysis of the delta frequency band and spectral entropy-based markers to quantify disruption of electroencephalography delta power during sleep were performed to examine potential associations with mortality risk in the Sleep Heart Health Study cohort (N = 5804). Adjusted Cox proportional hazard models were used to determine the association between disrupted delta wave activity at baseline and all-cause mortality over an ~11y follow-up period.

RESULTS

Disrupted delta electroencephalography power during sleep was associated with a 32% increased risk of all-cause mortality compared with no fragmentation (hazard ratios 1.32 [95% confidence interval 1.14, 1.50], after adjusting for total sleep time and other clinical and life-style related covariates including sleep apnea. The association was of similar magnitude to a reduction in total sleep time from 6.5h to 4.25h. Conventional measures of sleep quality, including wake after sleep onset and arousal index were not predictive of all-cause mortality.

CONCLUSIONS

Delta wave activity disruption during sleep is strongly associated with all-cause mortality risk, independent of traditional potential confounders. Future investigation into the potential role of delta sleep disruption on other specific adverse health consequences such as cardiometabolic, mental health and safety outcomes has considerable potential to provide unique neurophysiological insight.