009-sleep

Sleep monitoring dataset

Sleep monitoring includes sleep staging, body positioning, respiratory events, heart rate, oxygen saturation and snoring, measured concurrently over three nights.

Obstructive sleep apnea (OSA) is a sleep disorder in which a person’s breathing is interrupted during sleep due to obstruction of the upper airway as a result of relaxation of the throat muscles. This obstruction can lead to pauses in breathing for short periods of time, which can cause loud snoring, reduction in the blood oxygen levels, stress response, awakening and fragmented sleep.

A home sleep apnea test (HSAT) is a non-invasive test used to diagnose sleep apnea. During an HSAT, a patient wears a portable device overnight to monitor their breathing, heart rate, oxygen levels, snoring, and other sleep patterns. The device records this data and it is then analyzed by clinically-validated algorithms that compute apnea-related indices. The data can also be used to study sleep architecture and other aspects of sleep, such as heart rate variability (HRV) and arrhythmia.

The device used for sleep monitoring in the Human Phenotype Project is the FDA-approved WatchPAT-300 by Itamar Medical. This device contains 5 sensors: An actigraph worn on the wrist; a pulse oximeter and a Peripheral Arterial Tone (PAT) probe worn on the finger; and a microphone and accelerometer worn on the chest for respiratory effort, snoring and body position measurement.

The device’s algorithms provide 3 main indices for diagnosis of OSA: Apnea/Hypopnea Index (AHI), Respiratory Disturbance Index (RDI), Oxygen Desaturation Index (ODI). The device’s algoithms also provide indices for diagnosis of Atrial Fibrillation (not FDA-approved yet): Premature beats per minute, Atrial fibrillation longest event.

The WatchPAT device is given to participants during the visit to the clinical testing center who then record 3 nights of sleep at home during the following two weeks (Figure 4). The recorded tested nights should reflect the usual sleeping regimen / typical bedtime (for example, not after exceeding alcohol consumption).

Data availability:

The information is stored in 2 main statistics parquet files: sleep.parquet, hrv.parquet; In multiple events parquets and in multiple time series parquets / EDF files containing sensor channels for each night of logging.

from pheno_utils import PhenoLoader

pl = PhenoLoader('sleep')
pl

PhenoLoader for sleep with
473 fields
3 tables: ['sleep', 'hrv', 'age_sex']

Data dictionary

pl.dict

	field_string	description_string	folder_id	feature_set	field_type	strata	data_coding	array	pandas_dtype	units	...	stability	sexed	debut	completed	transformation	list_of_tags	min_plausible_value	max_plausible_value	dependency	parent_dataframe
tabular_field_name
collection_timestamp	Collection timestamp	Collection timestamp	9	sleep,sleep_raw	Datetime	Primary	NaN	Multiple	datetime64[ns, Asia/Jerusalem]	Time	...	Accruing	Both sexes	2020-01-15	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	NaN
collection_date	Collection date	Collection date	9	sleep,sleep_raw	Date	Primary	NaN	Multiple	datetime64[ns]	Date	...	Accruing	Both sexes	2020-01-15	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	NaN
night	Night	Night	9	sleep,sleep_raw	Integer	Primary	NaN	Multiple	int	NaN	...	Accruing	Both sexes	2020-01-15	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	NaN
warnings	Warnings	Analysis warnings	9	sleep,sleep_raw	Categorical (multiple)	Primary	009_04	Multiple	object	Text	...	Accruing	Both sexes	2020-01-15	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	NaN
timezone	Timezone	Timezone	9	sleep,sleep_raw	Categorical (single)	Auxiliary	001_03	Multiple	category	Text	...	Accruing	Both sexes	2020-01-15	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
REM	REM	REM	9	events_time_series	rows: event	NaN	NaN	NaN	string	Text	...	Accruing	NaN	NaN	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	events_time_series
REM Period	REM Period	REM Period	9	events_time_series	rows: event	NaN	NaN	NaN	string	Text	...	Accruing	NaN	NaN	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	events_time_series
Resp. Event	Resp. Event	Resp. Event	9	events_time_series	rows: event	NaN	NaN	NaN	string	Text	...	Accruing	NaN	NaN	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	events_time_series
Sleep	Sleep	Sleep	9	events_time_series	rows: event	NaN	NaN	NaN	string	Text	...	Accruing	NaN	NaN	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	events_time_series
Wake	Wak	Wak	9	events_time_series	rows: event	NaN	NaN	NaN	string	Text	...	Accruing	NaN	NaN	NaN	NaN	Sleep Monitoring	NaN	NaN	NaN	events_time_series

506 rows × 24 columns

Plot histogram and ecdf for ahi at baseline visit

from pheno_utils.basic_plots import hist_ecdf_plots

col = "ahi"
df = pl[[col] + ["age", "sex", "collection_date"]].loc[:,:,"00_00_visit",0,:]

# plot histogram and ecdf
hist_ecdf_plots(df.dropna(subset=[col,"sex", "age"]), col, gender_col="sex")

# stats
display(df[col].describe().to_frame().T)

	count	mean	std	min	25%	50%	75%	max
ahi	5562.0	10.593943	10.356211	0.0	3.2825	7.35	14.4625	89.74