Diet and exercise get most of the public attention when people talk about a long life. A 2025 paper in SLEEP Advances asks whether sleep belongs much closer to the centre of that conversation.
The paper, led by Kathryn E. McAuliffe of Oregon Health & Science University with Andrew W. McHill as senior author, studied sleep insufficiency and life expectancy at the U.S. county level from 2019 to 2025. None of us are doctors, epidemiologists, or clinical researchers; what follows is reporting and interpretation of the study, not medical advice.
The result is worth taking seriously, but it should not be read as the final word. This is one study, not settled consensus, and it is especially important to say that because the finding is easy to turn into a slogan: sleep mattered more than exercise or food insecurity. The more accurate version is narrower. In this dataset, under the authors’ main mixed-effects model, county-level insufficient sleep was more strongly associated with lower life expectancy than physical inactivity or food insecurity, while smoking showed a stronger association.
What the study actually measured
The study did not follow individuals from childhood to death. It did not measure anyone’s sleep in a laboratory. It worked at the county level, using the Centers for Disease Control and Prevention’s Behavioral Risk Factor Surveillance System, a large telephone survey run with state health departments. The paper says the BRFSS data covered all U.S. counties and included risk behaviours and chronic conditions such as smoking, sleep duration, obesity and diabetes.
Sleep insufficiency was defined as less than seven hours of sleep in a 24-hour period. That threshold is not arbitrary. A 2015 joint consensus statement from the American Academy of Sleep Medicine and the Sleep Research Society, published in Sleep, recommended that adults sleep seven or more hours per night on a regular basis to support health.
In the OHSU paper, county-level sleep insufficiency was calculated from the survey question asking respondents how many hours of sleep they usually got in a 24-hour period. The researchers then compared the proportion of adults reporting insufficient sleep with life expectancy in each county, using correlations within states and mixed-effects models across multiple years.
This structure is useful for seeing broad public-health patterns. It is not designed to tell an individual reader how many years one habit will add or subtract from their life. A county can have more short sleep, more smoking, less insurance coverage, more unemployment, more chronic disease, different work schedules, different income patterns and different health-care access all at once. Statistical models can adjust for some of that. They cannot make county-level data behave like a personal medical forecast.
The main ranking belongs to one model
The headline finding comes from the study’s main model of county life expectancy from 2019 to 2025. In that model, the authors included age-adjusted insufficient sleep, physical inactivity, food insecurity, insurance, unemployment, high school completion, social connections and smoking as fixed factors, with state included as a random factor to account for counties being nested within states.
In that analysis, insufficient sleep was significantly associated with lower life expectancy. The paper reports an estimate of b = -0.17 for insufficient sleep, compared with b = -0.31 for smoking. Physical inactivity and food insecurity were included in the model, but their associations with life expectancy were smaller than the sleep estimate.
That is the ranking in the title. The scope matters. It is not saying sleep is always more important than exercise or diet for every person. It is saying that, across U.S. counties in this model and over this timeframe, the sleep insufficiency measure showed a stronger negative association with life expectancy than those particular county-level variables.
The paper is not an argument against exercise. Physical activity has extensive evidence behind it, including mortality research that the OHSU paper itself cites, such as a 2020 BMJ prospective cohort study on physical activity and all-cause and cause-specific mortality in U.S. adults. The point here is not that exercise matters less in general. It is that sleep looked unusually prominent in this particular county-level life-expectancy model.
The same care applies to diet. The model included food insecurity, not detailed individual diet quality. Food insecurity is a major social and health issue, but it is not the same thing as measuring what every person ate, how their food environment worked, or how diet interacted with sleep, work and stress.
The second model complicates the clean version
The cleanest version of the story is that only smoking outranked sleep. The paper’s second model complicates that. When the authors added obesity and diabetes, which they describe as possible mediators or bidirectional outcomes related to insufficient sleep rather than straightforward confounders, sleep insufficiency remained significant but its estimate became smaller.
In that second model, smoking and obesity both showed stronger relationships with life expectancy than insufficient sleep. The sleep estimate was still statistically significant, but the hierarchy changed. That is why the title’s phrase “main analysis” matters. Without it, the sentence would flatten the paper into a stronger claim than the authors’ full modelling supports.
This is also why the result should not be read as a competition between habits. Sleep, activity, diet, smoking, diabetes, obesity, insurance, employment and social connection do not sit in separate sealed boxes in real life. They shape one another. A person working night shifts, caring for family, living with noisy housing, lacking health insurance or managing chronic illness may not experience sleep as a simple personal choice.
Why the county level matters
The most interesting part of the OHSU paper may be the county scale. Earlier research has already connected short sleep with mortality risk. For example, a 2010 systematic review and meta-analysis in Sleep examined sleep duration and all-cause mortality across prospective studies. The OHSU authors were asking a different question: whether sleep insufficiency and life expectancy line up locally across the United States.
They report that in 2025, all but three states with more than 10 counties showed a significant negative correlation between county sleep insufficiency and county life expectancy. Similar relationships appeared from 2019 through 2024, with the proportion of states showing significant associations ranging from 84 percent in 2019 to 100 percent in 2024.
That local framing matters because sleep is often discussed as private discipline: go to bed earlier, stop looking at the phone, make better choices. Those may be relevant for some people. But county-level differences point toward larger conditions too: shift work, commute times, housing, caregiving demands, neighbourhood noise, economic strain, health-care access and untreated sleep disorders. The paper did not measure all of those mechanisms, and it says so. But the design makes it harder to pretend sleep is only an individual willpower issue.
What the paper cannot show
The main limitation is causation. The authors found associations between county-level sleep insufficiency and life expectancy. They did not show that insufficient sleep caused lower life expectancy in those counties. Unmeasured confounding remains possible, and the paper explicitly says it cannot exclude that risk.
The measurement also has limits. Sleep was self-reported as usual hours in a 24-hour period. That is practical for a national survey, but it is not the same as wearable or laboratory measurement. It also groups long sleepers with people who sleep seven hours or more. The authors note that this may matter because very long sleep has also been linked with poorer health in some earlier studies.
The timeframe is another complication. The years 2019 to 2025 include the COVID-19 pandemic and its aftereffects. The paper notes that pandemic changes in work, school and daily routines may have affected sleep behaviour, although similar sleep-life expectancy associations appeared across the years examined.
None of these caveats makes the study uninteresting. They make the finding more precise. The paper is strongest when read as evidence that sleep health may deserve more attention in local public-health thinking, not as proof that sleep is the master variable behind longevity.
The useful takeaway is quieter than the headline
The tempting overstatement is simple: sleep is more important than diet or exercise. The study does not establish that. The useful takeaway is quieter: in a large county-level analysis, insufficient sleep behaved like a serious population-health signal, strong enough to stand beside the behaviours that usually dominate longevity conversations.
That is worth noticing because sleep is often treated as leftover time. It is what remains after work, parenting, commuting, screens, stress, second jobs and late chores have taken their share. The OHSU paper suggests that when whole counties have higher levels of insufficient sleep, that pattern travels with shorter life expectancy, even after the model accounts for several familiar mortality-related behaviours.
If persistent sleep difficulty, heavy daytime sleepiness, loud snoring, breathing interruptions during sleep or health worries are part of someone’s life, a clinician is the right source of individual guidance. A county-level study can name a pattern. It cannot diagnose why one person is not sleeping, or what should be done about it.
What the paper changes is not the value of diet or exercise. It changes the balance of attention. A long life is still shaped by many things at once. But sleep, in this analysis, was not a soft extra around the edges. It was one of the strongest signals in the map.