Sleep Efficiency and Latency Calculator

Sleep efficiency from total sleep time and time in bed, with the wake after sleep onset the report may not print, recovered from the identity that time in bed is latency plus sleep plus wake.

Last updated 8 Oct 2026

Sleep efficiency, WASO and the latencies

TST over TIB, and the remainder

Total sleep time (minutes)Scored sleep across the whole recording. By definition it is the sum of the four stages, which is what the sleep stage percentage calculator uses to derive it — so if the stage minutes on your report do not add up to this number, one of the two is a transcription error.

Time in bed (minutes)Lights off to lights on. The NHLBI National Sleep Research Resource’s harmonised definition is the interval between lights-off or in-bed time and lights-on or out-of-bed time, and sleep efficiency is the ratio of total sleep time to it, expressed as a percentage. It cannot be smaller than the sleep inside it, and the page refuses if you enter it that way.

Sleep onset latency (minutes)Lights off to the first epoch of any sleep. This is the input that lets the page recover wake after sleep onset, because time in bed is latency plus sleep plus wake and nothing else. Latency plus sleep cannot exceed time in bed, and the page refuses if they do.

REM latency (minutes)Sleep onset to the first epoch of stage R. Measured from SLEEP ONSET rather than from lights off, which is the commonest way this number is misread; a report that measures it from lights off will be larger by the sleep onset latency. A short REM latency has a long differential and none of it is settled by the number.

77.3% of time in bedExample

Total sleep time 357 min, time in bed 462 min, sleep onset latency 22 min, REM latency 96 min

Formula

Sleep efficiency = 100 × total sleep time ÷ time in bed · wake after sleep onset = time in bed − latency − total sleep time

total sleep time: all scored sleep, which is by definition the sum of stages N1, N2, N3 and R. A report whose stage minutes do not sum to its total sleep time contains an arithmetic error somewhere
time in bed: the NSRR harmonised definition is the interval between lights-off or in-bed time and lights-on or out-of-bed time. It is not the recording time of a home device, which has no lights and no bed; a home study has no sleep efficiency at all, because it cannot measure either term
the identity: time in bed = sleep onset latency + total sleep time + wake after sleep onset. Everything in bed is one of those three, so the third follows from the other two. That is why this page asks for the latency: it buys the wake for nothing
the bound: efficiency lies in (0, 100]. Exactly 100% is attainable and means no wake at all in bed; 0% is not, because a study with no sleep has no efficiency rather than an efficiency of zero, and the page refuses it. A transposed numerator and denominator would print 200% on a real pair of inputs, so the refusal is also the guard against that slip
REM latency: sleep onset to the first epoch of stage R. Measured from sleep onset, not from lights off — the difference is the sleep onset latency, and a report that uses the other convention is larger by exactly that much

Worked example

Total sleep time 357 min, time in bed 462 min, sleep onset latency 22 min, REM latency 96 min
Sleep efficiency = 100 × 357 ÷ 462 = 77.3% of time in bed
Wake after sleep onset = 462 − 22 − 357 = 83 min, which the report may not have printed
Check the identity: 22 + 357 + 83 = 462, the time in bed. Everything in bed is latency, sleep or wake
Total wake in bed = 462 − 357 = 105 min, of which 22 min was before sleep began and 83 min after
REM latency 96 min is 26.9% of the night's sleep
Now hold the efficiency and move the wake about: total sleep time 357 with time in bed 462 and a latency of 100 min gives the same 77.3% with only 5 min of wake after sleep onset. One efficiency, two completely different nights, which is why the second row matters more than the headline
Enter a total sleep time of 480 with a time in bed of 240 and the page refuses rather than printing 200%: a transposed pair is the one slip this formula invites
Enter a total sleep time of 0 and it refuses too. A study that recorded no sleep has no efficiency, and 0% would read as a measurement rather than as a failed study

What the four numbers are measured from

Two of these five start at lights off and two start at sleep onset, and mixing the two origins is where the arithmetic goes wrong. REM latency in particular is measured from sleep onset; a report using lights off as its origin will be larger by the sleep onset latency, which on the worked night is 22 minutes.
Quantity	Measured from	To
Time in bed	lights off, or getting into bed	lights on, or getting out of bed
Sleep onset latency	lights off	the first epoch of any sleep
Total sleep time	—	the sum of all epochs scored N1, N2, N3 or R
Wake after sleep onset	the first epoch of sleep	lights on, counting only the wake epochs
REM latency	the first epoch of SLEEP	the first epoch of stage R

Why no normal range is printed here

The same decision the lung-function pages reached about predicted values: where the reference is an age-specific table, take it from the reader’s report instead of approximating it. Compare the approach on the percent predicted and z-score converter.
Source	What it offers	Why it cannot be built in
OHRI and Sunnybrook normal-PSG calculator	means with prediction intervals for efficiency, the latencies, the stage shares and the arousal index	it asks for sex, age and first-versus-later night, then reads two published tables. This engine has no table interpolation, and a fitted curve standing in for a published table is exactly the plausible, silent error worth avoiding
One hospital’s normative sheet	efficiency above 80% normal; below 80% reduced, not applying over age 70; sleep onset latency 10 to 20 min normal	no source is cited for any of them on the sheet itself, and they are one laboratory’s conventions rather than a published standard
Your own report	the laboratory’s own reference range beside the measurement	nothing — this is the one to use, and it is why the page computes rather than grades

One efficiency, two different nights

Sleep efficiency is the first number most readers look at and the least informative one on the page. It is total sleep time over time in bed, as a percentage — the NHLBI National Sleep Research Resource’s harmonised definition, and the one in general use — and it compresses two quite different problems into one figure. A person who takes ninety minutes to fall asleep and then sleeps soundly, and a person who falls asleep at once and then spends ninety minutes awake in the small hours, can have the same efficiency and nothing else in common.

The way out is an identity rather than a measurement. Everything that happens in bed is sleep onset latency, scored sleep, or wake after sleep onset, so the three sum to the time in bed and any one of them follows from the other two. Many reports print the efficiency and the latency and leave the wake after sleep onset out, and it is the number that distinguishes the two patients above. On the worked night, 105 minutes of the 462 in bed were awake: 22 before sleep began and 83 after.

REM latency is the other quantity readers routinely mis-transcribe, because its origin is different. It runs from the first epoch of sleep to the first epoch of stage R, not from lights off — so a report using lights off as the origin reports a figure larger by exactly the sleep onset latency. That is 22 minutes on this night, which is more than enough to move a value across a laboratory’s own cut-off.

What this page deliberately will not do is tell you whether any of these numbers is normal. The reference values are an age- and sex-specific distribution: the OHRI and Sunnybrook calculator built on a published meta-analysis needs sex, age and whether this is a first or a later night in the laboratory before it will return a mean, and reads the answer out of two tables. This engine has no table interpolation, and approximating one with a fitted curve is the kind of plausible, invisible error that is worse than no answer. Read this against the reference range your own laboratory printed beside the metric. Normal values here are age- and sex-specific distributions published as tables, and this page does not approximate one. Every number on a sleep study is scoring-rule-dependent and night-to-night variable: the same recording scored under two rules, or the same patient studied on two nights, gives different figures. This page does the arithmetic and says where every figure it quotes comes from. It renders no diagnosis and no clinical decision.

Frequently asked questions

How is sleep efficiency calculated?

Total sleep time divided by time in bed, times 100. The NHLBI National Sleep Research Resource’s harmonised definition is the ratio of total sleep duration to the in-bed period, expressed as a percentage, where the in-bed period runs from lights off or getting into bed to lights on or getting out of it. On the worked night, 357 minutes of sleep in 462 minutes in bed is 77.3%.

What is wake after sleep onset, and why is it not on my report?

It is the wake time between the first epoch of sleep and lights on, and many reports omit it. You can always recover it, because time in bed is sleep onset latency plus total sleep time plus wake after sleep onset and nothing else. On the worked night that gives 462 less 22 less 357 — 83 minutes. It is the number that distinguishes trouble falling asleep from trouble staying asleep, and the efficiency cannot.

Is REM latency measured from lights off or from sleep onset?

From sleep onset, to the first epoch of stage R. A report that measures it from lights off gives a figure larger by exactly the sleep onset latency — 22 minutes on the worked night. If a value looks unexpectedly long, check which origin the laboratory used before reading anything into it.

What sleep efficiency is normal?

That depends on age, on sex and on whether it is a first night in the laboratory, which is why this page prints no band. The reference calculator built on the published meta-analysis asks for all three before returning a mean with a prediction interval, and reads it from a table this calculator engine cannot interpolate. Use the range your own laboratory printed beside the measurement. One hospital sheet read while preparing this page uses 80% and notes it does not apply above age 70 — and cites no source for it.

Can a home sleep study give a sleep efficiency?

Not a real one. Sleep efficiency needs both total sleep time and time in bed, and a home device measures neither: it has no EEG, so it cannot score sleep, and no lights-off marker. That is the same limitation that forces a home study to report a respiratory event index over recording time rather than an AHI over sleep time.

Related calculators

References

National Sleep Research Resource (NHLBI). Harmonized variable: sleep efficiency (nsrr_effsp_f1).
Ottawa Hospital Research Institute / Sunnybrook. Normal Polysomnography Parameters in Healthy Adults — a sex-, age- and night-specific reference calculator built on a published meta-analysis.
Trillium Health Partners. Normative Values for the Sleep Study Report — cited as an exhibit of one laboratory’s own reference sheet, not as a source of thresholds.
American Academy of Sleep Medicine. Summary of Updates in Version 3 of The AASM Manual for the Scoring of Sleep and Associated Events. Darien, IL: AASM; February 2023.
Kapur VK, Auckley DH, Chowdhuri S, et al. Clinical practice guideline for diagnostic testing for adult obstructive sleep apnea: an American Academy of Sleep Medicine clinical practice guideline. J Clin Sleep Med. 2017;13(3):479–504.

Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/