Table 2 The top 15 most important features influencing CKD detection across models

From: Ensemble learning approaches for early prediction of chronic kidney disease based on polysomnographic phenotype analysis

Rank

Features

Description

1

difbak

Difficulty falling back asleep after waking during the night.

2

am930cig

Time since the last cigarette, recorded at 9:30 AM.

3

candur

Duration (in hours) of usable nasal cannula signal during polysomnography.

4

arremop

Number of times a person wakes up during REM sleep while sleeping in a non-supine position (side or stomach).

5

am7drk

Time since the last caffeinated drink as of 7:00 AM.

6

remlaiip

Time to enter REM sleep after initially falling asleep, excluding periods of wakefulness.

7

lowexclude

A measure used to exclude non-significant apnea events.

8

aprtdurop

Total duration of apnea episodes in REM sleep while sleeping in a non-supine position.

9

carop5

The average duration of central apnea episodes in REM sleep, with oxygen desaturation of 5% or more.

10

quabdo

Quality of abdominal effort signal recorded during polysomnography.

11

oindex

An overall index measuring sleep-disordered breathing severity.

12

am930drk

Time since the last caffeinated drink as of 9:30 AM

13

pctsa90hge2

Percentage of sleep time when oxygen levels drop below 90%.

14

ortdurop

Total duration of obstructive apnea episodes in REM sleep while sleeping in a non-supine position.

15

alpdel

Brain activity (alpha waves) during non-REM sleep indicates disturbed sleep.