Brain Wave Sleep Data Can Predict Future Health Outcomes

Summary: Brain wave data collected during sleep predicts the future incidences of 11 health-related outcomes, including dementia, cardiovascular health, psychological disorders, and mortality.

Source: American Academy of Sleep Medicine

The wealth of brain wave data collected during overnight sleep studies can predict future health outcomes, strengthening the concept of sleep as a “window” into health, according to a study from researchers at Massachusetts General Hospital and Beth Israel Deaconess Medical Center.

Results show that a model based on quantitative analyses of sleep data was able to predict the 10-year risk of all 11 health outcomes selected for analysis. The three outcomes with the highest risk ratios associated with poor-to-average sleep were dementia (RR = 6.2), mortality (RR = 5.7), and mild cognitive impairment or dementia (RR = 4.0).

“It is surprising that sleep can predict the future incidence of 11 outcomes covering neurological, psychiatric, cardiovascular conditions, and mortality even before the actual diagnosis, using data where the outcomes occur about three years later from the baseline sleep recording,” said lead author Haoqi Sun, who has a doctorate in machine learning and computational neuroscience and is an instructor in neurology at Massachusetts General Hospital.

“We already know sleep is informative about health, but it is still quite a surprise that the brain activity during sleep can reflect so much information, let alone other important signals during sleep such as respiration and heart rates.”

The study involved quantitative analyses of the sleep microstructure in 8,673 adults who had a diagnostic sleep study in a sleep center using polysomnography. The participants had an average age of 51 years; 51% were female.

The researchers extracted 86 features from the data generated by overnight EEG, a recording that measures brain waves. They sorted participants into three groups — poor, average and good sleep —and used a statistical model to assess health risk.

Health outcomes were determined using medical codes, brain imaging reports, medications, and/or cognition scores. Results were controlled for potential confounders including age, sex, body mass index, and certain prescription medication use.

While sleep medicine physicians routinely use sleep studies for the diagnosis of sleep disorders such as obstructive sleep apnea, Sun said the findings suggest that decoding sleep data could play a more significant role in health care.

“The ability to use noninvasive physiologic measurements of sleep to predict future incident health outcomes would be significant, because it might allow early interventions to prevent unfavorable outcomes,” he said.

Funding: The study was supported by the AASM Foundation 2019 Strategic Research Award. Principal investigators of the study were Dr. M. Brandon Westover, director of data sciences for the McCance Center for Brain Health at Massachusetts General Hospital, and Dr. Robert Thomas, pulmonary, critical care and sleep medicine specialist at Beth Israel Deaconess Medical Center.