Is Sleep Consistency More Important Than Sleep Duration? In 61,000 People, Regularity Predicted Mortality Better

Most sleep advice, and most sleep features, start with how long. Seven to nine hours is the number on every poster, and duration is the first thing a sleep tracker reports.

The strongest evidence of the past few years points somewhere else. In a study of 60,977 UK Biobank adults who wore an accelerometer for a week, the most regular sleepers had 20% to 48% lower risk of dying over the following six years than the least regular, and regularity predicted mortality better than sleep duration did [1]. Adding duration to a model that already contained regularity did not significantly improve it.

That is an observational finding, and it does not prove that changing someone’s schedule extends their life. But it changes what a sleep product should measure and show. This is what regularity is, what its numbers mean in minutes, and how to compute it from the data a wearable already collects.

Is sleep consistency more important than sleep duration?

For predicting mortality, the largest study to compare them says yes. It used the Sleep Regularity Index, which scores how often a person is in the same state, asleep or awake, at the same time on consecutive days [1][2]. On the study’s scale, 100 means identical sleep and wake times every day, and 0 means no relationship from one day to the next.

Associations held for cancer and cardiometabolic deaths too, though in the fully adjusted models only the two most regular fifths reached significance [1].

What is a good sleep regularity score?

There is no clinical cut-off, but the UK Biobank distribution gives useful reference points: a median of 81, and below 71.6 for the least regular fifth [1]. To see what those mean, it helps to translate them into minutes. The index is defined minute by minute, but for someone who sleeps once a night it simplifies to a formula anyone can reason about. Each minute that bedtime moves from one night to the next, and each minute wake time moves, is a minute in which the two days disagree. The index becomes:

Sleep Regularity Index = 100 × (1 − (shift in bedtime + shift in wake time, in minutes) ÷ 720)

That turns the UK Biobank thresholds into something concrete:

• The median of 81 corresponds to bedtime and wake time shifting by a combined about 137 minutes from one night to the next.
• The least regular fifth, below 71.6, corresponds to a combined shift of about 204 minutes or more.

In practice, accelerometer data also captures naps and long night-time awakenings, which lower the index, so treat these as approximate. But they show the scale: the people at highest risk were not slightly irregular. Their sleep moved by more than three hours a day.

Are weekend lie-ins bad for sleep regularity?

Less than a nightly swing. Duration and regularity can disagree completely, so we computed the index for four schedules, each over four weeks:

Two results stand out. First, the schedule with the most sleep is not the most regular: weekend lie-ins add sleep and cost regularity, but still leave the index above the UK Biobank median. Second, a person who sleeps a steady 7.5 hours a night can still land in the least regular fifth, simply by moving bedtime two hours back and forth. A duration score would rate that person perfectly. A regularity score would flag them.

Do Apple, WHOOP and Oura measure sleep consistency?

Some trackers already score consistency, each in a different way:

• Apple’s Sleep Score gives 50 points to duration, 30 to bedtime consistency over the last 13 nights, and 20 to interruptions [3]. It considers when you fell asleep, not when you woke up.
• WHOOP reports a sleep consistency percentage, which is part of its Sleep Performance and available through its API [4].
• Oura’s Sleep Score includes a timing contributor among seven, which concerns when you sleep relative to your body clock rather than how much that changes from night to night.

We compared the full constructions in what is a good sleep score. The gap the UK Biobank result exposes is narrower than “nobody measures consistency”. It is that most products fold consistency into a single sleep number dominated by duration, where a user cannot see it, and most measure only bedtime, when the index weighs wake time equally.

How do you calculate sleep regularity from wearable data?

For one main sleep a night, the formula above is all you need:

def minutes_of_day(t):
    return t.hour * 60 + t.minute

def clock_shift(a, b):
    """Shortest difference between two clock times, in minutes."""
    d = abs(minutes_of_day(a) - minutes_of_day(b)) % 1440
    return min(d, 1440 - d)

def sleep_regularity_index(nights):
    """nights: consecutive nights as (bedtime, wake_time) in local time,
    one main sleep each. Returns the Sleep Regularity Index, 100 = identical."""
    pairs = [
        100 * (1 - (clock_shift(b1, b2) + clock_shift(w1, w2)) / 720)
        for (b1, w1), (b2, w2) in zip(nights, nights[1:])
    ]
    return sum(pairs) / len(pairs)

Five details decide whether the number means anything:

1. Use consecutive nights only. Compare a night with the next calendar night, and skip any pair where a night is missing. Missing nights are rarely random, since people take devices off when they are ill, travelling or out late, and every regularity metric is highly sensitive to data missing for a reason [5]. Report how many nights the number is based on.
2. Use enough nights. The UK Biobank study required at least five days of valid data out of seven [1]. A week is the practical minimum; more is better.
3. Handle naps and split sleep. The shortcut assumes one sleep a night. For people who nap or sleep in blocks, compute the full minute-by-minute version: mark each minute asleep or awake, and compare every minute with the same minute 24 hours later [2].
4. Work in local time, and watch the clock change. Regularity is about the clock people live by, so use local time, and exclude or flag the night clocks change, which we covered in daylight saving time and sleep data.
5. Treat shift workers separately. A rotating schedule produces a low index by design. Showing those users a “highly irregular” warning they cannot act on helps no one.

How should a sleep app show regularity?

Regularity is easier to act on than duration. “Sleep longer” is vague; “your wake time moved by two hours on Tuesday” is not. Three practical patterns:

• Show regularity beside duration, not folded into one score, so a steady 7.5-hour sleeper with a two-hour swing sees why their number is low.
• Point to the biggest swing. The largest night-to-night change in bedtime or wake time is usually the one worth fixing.
• Treat wake time as half the score. Most consistency features look at bedtime; the index weighs both equally, and a fixed alarm is often the easiest change to make.

How long people sleep still matters. But on the best evidence available, when they sleep, and how much that changes from one night to the next, matters at least as much, and it is the part most sleep products show least clearly. It is also the part a wearable measures well, and that users can change tomorrow.