Irregular sleep is sticky: after six irregular months, it continued 80% of the time

Sleep advice is overwhelmingly about duration. Regularity — how much sleep timing and duration vary from one night to the next — is a different dimension, and one that has been studied in short windows, because the persistence and long-term effects of irregular sleep have remained unclear largely due to short monitoring periods in past work [s1].

A study published in the Journal of Sleep Research extended the observation window to roughly 20 months, which changes what can be asked [s1].

The dataset

The authors analysed approximately 20 months of under-mattress sleep sensor data from 95,819 global participants [s1].

Each person-month was classified as irregular if sleep duration variability was at least 90 minutes and sleep midpoint variability at least 60 minutes, measured as intra-individual standard deviations — thresholds previously found to correlate with poor cardiovascular health [s1].

Approximately 20% of person-months met those criteria [s1].

An under-mattress sensor is a meaningful methodological choice. It requires no charging, no wearing and no daily action from the participant, which removes the adherence decay that shortens wearable studies. What it cannot do is verify who is in the bed.

The three findings

Irregularity is common. One person-month in five was irregular by the study's criteria [s1].

It is persistent. After six consecutive irregular months, there was over an 80% likelihood of continuation [s1]. This is the finding that a two-week study could not have produced. Irregular sleep is not, for most people, a fluctuating state — it is a stable feature of how their life is arranged.

It tracks with worse sleep and higher blood pressure. Participants with chronic irregular sleep, defined as at least 75% of months irregular, showed 8% lower sleep efficiency, 40 minutes more wake after sleep onset, and 46% higher odds of hypertension, compared with those in the regular group (0% to 25% of months irregular) [s1].

What the design cannot establish

This is observational. The people whose sleep timing varies by more than an hour night to night are disproportionately shift workers, parents of young children, people with irregular or multiple jobs, and people with untreated sleep disorders. Every one of those is independently associated with hypertension risk.

The 46% figure describes an association within a self-selected group of people who bought and installed a sleep sensor [s1] — a population that skews toward higher income and higher health interest than the general public, and in a direction that would, if anything, tend to attenuate the association rather than inflate it.

The persistence finding is the more robust of the two, because it is descriptive rather than causal. It says something reliable about the natural history of irregular sleep: it does not resolve on its own.

The measurement problem underneath

If regularity is worth acting on, someone has to measure it. Objective indices from actigraphy or wearables are robust but resource-intensive, and the practical alternative is a questionnaire.

A validation study in the Journal of Circadian Rhythms tested how well the Sleep Regularity Questionnaire corresponds to what devices and diaries record [s2].

In the first part, 31 adults wore a smart ring continuously for 21 nights, generating device-derived regularity metrics including the Sleep Regularity Index, interdaily stability, social jetlag, composite phase deviation, and standard deviations of sleep onset and wake time [s2]. In the second, 52 adults completed a one-week sleep diary [s2].

Associations between questionnaire scores and the device-derived measures were small, with absolute correlations of 0.36 or below [s2]. Questionnaire scores related more strongly to subjective sleep quality: higher Global and Sleep Continuity scores were moderately associated with better scores on the Brief Pittsburgh Sleep Quality Index, r between −0.37 and −0.44 [s2].

In the diary part, the questionnaire's Global and Circadian Regularity scores showed small-to- moderate associations with higher diary-rated sleep quality and lower bedtime variability, r around 0.40 and between −0.32 and −0.34, while correlations with other diary metrics and the sleep quality index were weak, at 0.25 or below [s2].

Both samples are small — 31 and 52 adults [s2] — which limits how firmly this can be stated. But the pattern is coherent: the questionnaire appears to capture how regular people feel their sleep is, which is not the same quantity as how regular a device records it to be.

What to watch

Regularity is arriving in sleep research faster than the tools to measure it in ordinary care. The open questions are whether a clinically feasible measure can be validated against device-derived indices, and — the one no observational dataset can answer — whether deliberately stabilising sleep timing changes any downstream outcome.

This article is informational and is not medical advice.