This study uses data from the China Family Panel Studies 2018 (CFPS) (14), which were funded by the 985 Program of Peking University and carried out by the Institute of Social Science Survey (ISSS) of Peking University. The CFPS is a nationally representative, the annual longitudinal project reviewed and approved by the ISSS of Peking University. Taking into account regional differences and survey costs, CFPS implemented probability proportional to size sampling with implicit stratification. Each subsample in CFPS went through three stages of extraction (districts/counties–villages/communities-households) (14). The 2018 CFPS surveyed about 15,000 households and collected nearly 44,000 copies of questionnaires. The CFPS mainly adopted face-to-face interviews to collect data and 22% were aided by phone interviews.

The total sample size included 37,354 individuals. We firstly excluded 2,620 participants aged<16 years old. Subsequently, we excluded the individuals whose index values were unknown, rejected, not applicable and missing: vocabulary and mathematics scores (10,950), physical activity (32), sleep duration (41), smoking (1), drinking (2), self-reported height (277), self-reported weight (52), self-reported health status (3), residency (31), employment (1,203), and marital status (1,014). Finally, our analytic sample included 21,128 participants with complete information. The flow diagram of subject recruitment is shown in Figure 1.

The project was retrieved from the CFPS website for public access to secondary data,¹ which excludes all identifiable information about individual participants.

This study was conducted in accordance with the Declaration of Helsinki, and all procedures involving human subjects were approved by the Ethics Committee of Peking University (No. IRB00001052–14010). All participants were asked to provide written informed consent before completing the survey.

In the 2018 CFPS, three items on sleep duration were constructed from relevant survey questions (15), including “On average, how many hours do you sleep each day?,” “On a typical workday, how many hours do you sleep?,” and “On a typical free day, how many hours do you sleep?.” If individuals reported the first question, the latter two questions would be missing; conversely, individuals would report the latter two questions. According to their reply, sleep duration was calculated by (hours on school days*5 + hours on weekends*2)/7 and was categorized as <7 h, 7–10 h, and ≥ 10 h (16).

The duration of physical activity was assessed by the question: “How much time in total do you exercise each week?” (16). Physical activity was subsequently classified as: reporting no activity per week; reporting activity from 1 to 149 min/week; reporting activity ≥150 min/week.

The American psychologist Cattell has divided cognitive ability into two components: fluid intelligence and crystalized intelligence (17). In the CFPS 2018, all individuals performed the cognitive ability tests (including vocabulary and mathematics tests) by self-answer questionnaire (14, 16, 18). The theoretical basis of the CFPS cognitive ability tests was the design of Guttman Scale in psychometry, which showed good reliability and validity (19).

The vocabulary test was consisted of 34 Chinese characters drawn from the language textbooks used in primary and secondary schools and sorted in ascending order of difficulty. This test seeks to measure one’s vocabulary by how difficult he or she can recognize a character and the final score is the rank of the hardest question that the respondent answers correctly. Respondents did not know the rules before taking the test, so they would not fail on purpose. To make the test more efficient, all survey respondents were assigned to one of three entry points, based on their self-reported highest level of education. The respondents were asked to recognize the increasingly difficult characters one by one until they failed to recognize three consecutive characters. The mathematics test was consisted of 24 mathematical questions. The procedures of mathematics test were similar to vocabulary test. The mathematics test scores were categorized based on the same rank-order rule as that in the vocabulary test (18).

To provide more compelling evidence by reducing the amount of confounding influence, a set of variables which were previously revealed to associate with cognitive function (i.e., gender, age, educational level, residency, employment, obesity, smoking, drinking, health status, marital status and depression) were adjusted for. Gender, residency status and marital status were defined as dichotomous variables (gender: male or female; residency: rural or urban; marital status: unmarried, married or divorced/widowed/ cohabitation). Age was measured by years, and educational level was measured by asking the respondent to indicate the highest level of education completed: low (completed primary school or lower), middle (completed junior middle school but did not undergo the tertiary entrance exam) and high (took a tertiary entrance exam or higher) (17). The smoking and drinking status were assessed by the question: “Have you used at least one cigarette/drink during the recent month?” The responses were coded as follows: 1 = yes; 2 = no. BMI was calculated through the self-reported height and weight. The obesity was defined as a BMI ≥ 28 kg/m². Health status was measured according to respondents’ self-rated physical health status: good/average and poor. Depressive symptoms were measured by the brief self-report 8-item version of The Center for Epidemiologic Studies Depression Scale (CES-D). The total score of CES-D was from 0 to 24, with a higher score indicating a higher level of depressive symptoms. A cut-off of greater than or equal to 9 has been suggested for this scale as indicating significant depressive symptoms (20).

For descriptive purposes, continuous data was grouped into categorical data, and categorical data are presented as numbers and percentages. Linear regression was used to determine the associations among sleep duration, physical activity and cognitive ability. In the model 1, age and sex were adjusted for. In the model 2, education, residency and employment were additionally adjusted for. In the model 3, apart from the variables included in model 2, obesity, smoking, drinking, health status, marital status and depression were additionally adjusted for. In addition, physical activity and sleep duration were adjusted for mutually in the models. Further analysis would be performed to evaluate the association between sleep duration and physical activity with cognitive performance in subgroups of age (16–50 years old or > 50 years old), sex (male or female) and employment (employed or unemployed), respectively. All analyzes were weighted to account for the complex, multi-stage. All analyzes were carried out using STATA 17.0 (Stata Corp, College Station, TX, United States). Statistical significance was set at 0.05.

Compared with individuals reporting 150 min/week or more of activity, we observed a 116% higher risk of getting lower vocabulary scores (coefficient: -1.16, 95% CI: −1.55 ~ −0.78) and a 61% higher risk of getting lower mathematics scores (coefficient: −0.61, 95% CI: −0.78 ~ −0.44) among those reporting no physical activity. Compared with those who slept for 7–10 h/day, the adjusted coefficient for vocabulary score was −1.34 (95%CI, −1.86 ~ −0.83), and the coefficient for mathematics scores was −0.68 (95% CI, −0.94 ~ −0.42) for individuals reporting sleeping more than 10 h/ day in fully adjusted model. All the results were detailed in Tables 2, 3.

The joint associations between physical activity and sleep duration with cognitive performance are presented in Table 4. Compared with those who slept for 7–10 h/day reporting more than 150 min/week, the adjusted coefficient for vocabulary scores were −2.58 (95% CI, −3.33 ~ −1.82) for reporting no physical activity with a sleep duration≥10 h/day, and −1.00 (95% CI, −1.88 ~ −0.12) for reporting more than 150 min/week with a sleep duration≥10 h/day. Similarly, those who reporting no physical activity with a sleep duration≥10 h/day had a 130% higher risk of getting lower mathematics scores (95% CI: −1.67 ~ −0.93), 82% for reporting 1 to 149 min/week with a sleep duration≥10 h/day (95% CI: −1.58 ~ −0.05), 53% for reporting more than 150 min/week with a sleep duration≥10 h/day (95% CI, −0.96 ~ −0.09), respectively. Those shorter sleepers (<7 h/day) reporting no activity had lower vocabulary scores (coefficient: -0.83, 95% CI: −1.34 ~ −0.31) and mathematics scores (coefficient: -0.69, 95% CI: −0.95 ~ −0.43).

In the subgroups of age, sex and employment, compared with those reporting a sleep duration for 7-10 h/day, short sleeper (<7 h/day) who are younger than 50 years old had a 35% higher risk of getting lower mathematics scores (coefficient: −0.35, 95% CI = −0.64 ~ −0.07). Those individuals who were employed with a sleep duration ≥10 h/day tend to get lower vocabulary scores (coefficient: −1.47, 95% CI: −2.03 ~ −0.91) and mathematics scores (coefficient: −0.81, 95% CI: −1.11 ~ −0.51). And other results observed were consistent with our primary findings. All the results were detailed in Table 5.