How to Fix Bad Sleep: 10 Habits With Real Evidence

Across 60,977 UK Biobank participants and more than 10 million hours of accelerometer data, sleep regularity predicted all-cause mortality more strongly than sleep duration did. The four most regular fifths of the cohort had 20% to 48% lower all-cause mortality risk than the least regular fifth. The most useful sleep variable turned out to be the one almost nobody tracks.

Why does sleep timing matter more than sleep length?

The Sleep Regularity Index scores how similar your sleep and wake times are from one day to the next. In the 2024 analysis published in Sleep, higher regularity was associated with 16% to 39% lower cancer mortality and 22% to 57% lower cardiometabolic mortality across the top four quintiles, adjusted for age, sex, ethnicity, and sociodemographic, lifestyle and health factors. It is observational, so it cannot establish cause. But the comparison is internal to one cohort, which is what makes the regularity-beats-duration result interesting.

Caffeine has unusually precise numbers attached. A 2023 meta-analysis in Sleep Medicine Reviews pooled 24 crossover studies and found it reduced total sleep time by 45 minutes and cut deep sleep by 11.4 minutes, with a modelled cutoff for a 107 mg coffee of 8.8 hours before bed. A 2024 randomized crossover trial in Sleep refined this by dose: 100 mg was tolerated up to 4 hours before bed, while 400 mg disrupted sleep architecture even 12 hours out.

Which 10 habits does the evidence support?

1. A fixed wake time, including weekends, the single input the regularity index rewards most
2. Caffeine cutoff around 9 hours before bed for a coffee, earlier for pre-workout doses
3. Treating a large afternoon coffee as a 12-hour input, not a 5-hour one
4. A dark sleeping environment, since night light is independently associated with worse cardiometabolic outcomes
5. Bright light within the first hours of waking, which stabilizes circadian phase
6. A bedtime routine, valuable mainly because it protects the fixed wake time
7. Alcohol moved earlier or reduced, given its measured effect on sleep architecture (see how alcohol affects overnight recovery)
8. Screens dimmed rather than banished, intensity and timing matter more than device type
9. Getting out of bed when awake for long stretches, the core stimulus-control step in CBT-I
10. Measuring rather than guessing, because the caffeine trial found people judged their own sleep quality poorly

That last point is worth sitting with. In the 2024 trial, objective disruption and perceived sleep quality came apart: participants did not notice what caffeine was doing to them.

The bottom line

• Sleep regularity predicted mortality risk more strongly than sleep duration across 60,977 people, with 20% to 48% lower all-cause mortality in the most regular quintiles
• Caffeine cost 45 minutes of total sleep time on average, with a modelled cutoff of 8.8 hours before bed for one coffee
• People misjudge their own sleep quality, which is why the fix usually starts with measurement rather than willpower