Basic information of the participants
Table 1 shows the basic information collected from the 1658 participants. The average age of the respondents was 70.41 years old (SD 7.63). The loneliness scores in the survey ranged between 21 and 80. A total of 31 (18.4%), 924 (55.7 %), and 429 (25.9%) of the respondents were assessed as having low, moderate, and high levels of loneliness, respectively, and the mean overall score of the participants was 43.17±9.46. Of all the participants, 50.7% were female, and 15.2% were over 80 years of age. Individuals who were in a couple relationship accounted for 69.8% of the respondents, and 31.1% of the participants reported less than 1000 yuan in economic income. A total of 62.1% of the participants reported that their income came from their children; 69.1% had a primary school education or below, and 63.0% worked in manual labor. In total, 30.9% reported smoking, and 46.1% reported consuming alcohol. In addition, 94.7% had good relationships with their family members, 30.2% reported being empty nesters, and 17.3% reported having either 1 child or 0 children. A total of 32.8% reported having a chronic disease.
Comparison of loneliness scores
There were significant differences in loneliness scores among older adults with different characteristics, mainly including age (p<0.001), marital status (p=0.001), occupation (p<0.001), education (p<0.001), source of economic income (p<0.001), quantity of economic income (p=0.002), smoking behavior (p<0.001), alcohol consumption (p<0.001), relationships with family members (p=0.045), empty nester status (p<0.001), number of children (p<0.001) and NCD experience (p<0.001), as shown in Table 1. The average loneliness score of elderly individuals with poor sleep quality was 45.04±8.86, which was significantly higher than that of elderly individuals with good sleep quality (41.66± 9.65) (t=7.742, p< 0.001).
Comparison of PSQI total and dimension scores
The mean sleep quality score was 6.673.42. Among the 1658 rural older people in Shandong Province with global scores greater than 7, which accounted for 44.9% of the sample, the prevalence rate of poor sleep quality was 44.9%. As shown in Table 2, the average scores for sleep quality, subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction were 6.67±3.42, 1.14±0.76, 1.24±0.90, 0.61±0.84, 0.84±1.04, 1.34±0.67, 0.36±0.76 and 1.15±0.78, respectively. There were statistically significant differences in the total sleep quality scores and 5 of the PSQI dimensions scores (all except subjective sleep quality and habitual sleep efficiency) among participants with different levels of loneliness (p<0.05).
Contour analysis of the average PSQI scores of participants with different levels of loneliness
A contour analysis of the average PSQI total scores of rural older people with different levels of loneliness was conducted. The profiles of older people in rural areas who experienced low, moderate, and high levels of loneliness were not parallel to each other (F=12.000, p=0.000), and the contours of the high group were higher than those of the moderate and low groups. In the horizontal profile analysis, the average scores for the 7 dimensions, namely, subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication and daytime dysfunction, were different (F= 38.103, p=0.000), as shown in Figure 1.
Association between sleep quality and loneliness
The level of loneliness in the rural older population was used as the dependent variable (Y=0, low level of loneliness; Y=1, moderate level of loneliness; and Y=2, high level of loneliness). The sleep quality score was used as the independent variable, and age, marital status, occupation, source and quantity of economic income, relationships with family members, empty nester status, number of children, smoking behavior, alcohol consumption, BMI, chronic disease experience and quality of life were used as the control variables. The relationship between the sleep quality and loneliness of older people was analyzed by the three ordinal regression models. Even after marital status, alcohol consumption, relationships with family members, occupation, source and quantity of economic income, chronic disease experience and quality of life were controlled, an increase in the odds of loneliness was associated with an increase in the sleep quality score (adjusted odds ratio [aOR] = 1.111, 95% confidence interval [95% CI] = 1.078-1.145). The worse the quality of sleep was, the higher the degree of loneliness in the older adult sample. Scores for subjective sleep quality (aOR = 0.765, 95% CI = 0.649-0.902), sleep latency (aOR = 1.346, 95% CI = 1.178-1.537), sleep duration (aOR = 1.316, 95% CI = 1.139-1.522), use of sleeping medication (aOR = 1.175, 95% CI = 1.005-1.372), and daytime dysfunction (aOR = 1.267, 95% CI = 1.079-1.487) were significantly different between older adults with high levels of loneliness and those with moderate and low levels of loneliness. Older adults with higher PSQI scores for these 5 dimensions had a higher risk of loneliness. Although older adults with higher scores for habitual sleep efficiency and sleep disturbances had a higher risk of loneliness, these differences were not statistically significant, as shown in Table 3.
Sensitivity analysis
We conducted a sensitivity analysis to further assess the robustness of the model and better explain the reliability of the results. A sensitivity analysis that excluded participants with either one child or no children (n = 287) yielded a similar result to that of Model 2 (OR 1.102; 95% CI 1.064–1.141; p = 0.000). The second sensitivity analysis that excluded participants with a BMI ≥ 24 or a BMI<18.5 (n = 882) also yielded a result similar to that of Model 2 (OR 1.113; 95% CI 1.063–1.164; p = 0.000). The third sensitivity analysis that excluded participants with poor quality of life (n = 160) also yielded a result similar to that of Model 2 (OR 1.105; 95% CI 1.070–1.141; p = 0.000).