Study area and subjects
This cross-sectional study was performed in Northeastern China from May to September 2017. First, according to the information of pension institutions on the website of the Civil Affairs Bureau, the institutions with more than 500 beds were selected. There were 34 pension institutions with more than 500 beds. Then the pension institutions were divided according to the city level areas. Four cities were selected from them, and then they were selected in proportion to the number of pension institutions in each city. Finally, six public pension institutions were selected as the research objects: two in Shenyang, two in Anshan, one in Tieling, and one in Benxi. The inclusion criteria of the study population were: ① age≥60; ② stay time≥6 months; ③ not diagnosed with Alzheimer's disease; ④ voluntary participation in the study. Exclusion criteria of the study population:Elderly with acute diseases during the investigation. All the participants were interviewed using a questionnaire exploring cognitive function, demographic characteristics, and influencing factors. The questionnaire also made use of the Pittsburgh Sleep Quality Index (PSQI) and Geriatric Depression Scale (GDS). The selection of sample size is 5-10 times of the number of items of all scales. The sample size needs to consider 10-20% error . There were 52 items in this study. Therefore, the sample size should be 286 ~ 624. Considering the economic and human factors, the final sample size were 588.We contacted a total of 588 elderly people, of whom 553 were interviewed and completed questionnaires. After screening for the cognitive function using the Mini-Mental State Examination (MMSE) , 507 elderly adults without cognitive defects were chosen as participants.
Survey and quality control
The GDS  is a 30-item self-report assessment used to identify depression among the elderly. In the GDS, 20 questions are answered “yes” (scored 1) or “no” (scored 0), while the remain 10 questions are evaluated the opposite way (where “yes” is scored 0 and “no” is scored 1). The total score of the GDS ranges from 0 to 30, with a score of ≥11 indicating depression.
The PSQI  is a self-report questionnaire that assesses sleep quality over a one-month time interval. The measure scale consists of seven components, including subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, use of sleeping medication, and daytime dysfunction. The score for each component ranges from 0 to 3, where 0 indicates “no difficulty” and 3 indicates “very difficult.” The total score of the PSQI ranges from 0 to 21, with a score of >7 denoting poor sleep quality or having a sleep disorder.
The MMSE has been widely used to screen for cognitive defects among the elderly10. The total score of the MMSE ranges from 0 to 30. Combined with the educational level of the elderly, the diagnostic criteria for cognitive impairment among the elderly are as follows: illiteracy is ≤17 points, primary school is ≤20 points, secondary school (including technical secondary school) is ≤22 points, and university is ≤23 points.
All the scales we used were Chinese scales which were tested by reliability and validity in other studies. We also used a self-designed questionnaire for baseline information including age, sex, marital status, pension, economic status, chronic disease, self-care, body aches, chewing ability, ability to go out alone, exercise, regular diet, smoking, filial piety, and loneliness.
Thirty-five investigators were recruited for training. After unified training, the investigators in this study explained the contents of the questionnaire to the elderly one by one, and screened the cognitive function of the respondents to ensure the accuracy of the information source.
Items for which over 95% of participants had the same answers were excluded from the data analysis; these items included race (“Han” accounted for 96.8%) and medical insurance (“Have” accounted for 98.9%).
In addition, the proportion of missing items was <5% among elderly adults with poor or good sleep quality. No missing items were processed. We performed a statistical analysis of the data using SPSS Statistics 20.0. We conducted a univariate analysis using the chi-square test, and a multivariate analysis using binary unconditional logistic regression analysis. A P value less than 0.05 was considered statistically significant.