Health-related quality of life and its associated factors among Chinese seasonal retired migrants in Hainan

Study setting and design

With local residents of approximately 111.2 thousand people, Wuzhishan city is located in the central area of Hainan Island at an altitude of approximately 300 m, covering 1,144 km². Owing to the low latitude, the ambient temperature during the winter in Wuzhishan city is much higher than that in high-latitude regions. In addition, approximately 75% of the areas in Wuzhishan city are covered by forests, providing fresh and clean air. Owing to the superior environment, a substantial population of seasonal retired migrants, most of whom reside in urban areas, move to Wuzhishan city in winter. According to a report from the Hainan Health Commission (2019), the population of seasonal retired migrants in winter can reach approximately one-third of the local resident population in Wuzhishan city. Hence, this city is an ideal location for studying seasonal retired migrants. We conducted this community-based cross-sectional survey in the urban areas of Wuzhishan city from October to November 2022.

Study setting and participants

The current study focused on seasonal retired migrants aged 50 years and older who were recruited from communities in the urban areas of Wuzhishan city, Hainan. The seasonal retired migrants were snowbirds who spend winter in Wuzhishan city, with a length of stay ranging from one month to eleven months per year. Sunbirds that escaped summer in their original places and did not overwinter in Wuzhishan city were excluded from this study because they represented a small proportion of the seasonal migrants in Wuzhishan city and were different from snowbirds in terms of their characteristics.

The inclusion criteria of this study were as follows: (1) aged 50 years and older; (2) owned or rented residences in Wuzhishan city; (3) had a stay of at least 30 consecutive days, which is consistent with previous studies on seasonal retired migrants (Happel & Hogan, 2002; Smith & House, 2006); and (4) had experienced at least two rounds of seasonal migration.

The exclusion criteria were as follows: (1) living in Wuzhishan city for more than eleven months per year to distinguish our participants from local permanent residents; (2) arriving in Wuzhishan city between March and August and left between September and December to exclude sunbirds who did not spend winter in Wuzhishan city; (3) refusal to complete the questionnaire; and (4) unable to complete the questionnaire independently or with help from the interviewers.

We used the following formula (Desu & Raghavarao, 1990) to calculate the sample size: n = (Z_α/2∗σ/d)², where n = the desired sample size, Z_α/2 = the upper 100(α/2) percentile point of the standard normal distribution, σ = the assumed population standard deviation, and d = the desired precision. In this study, the confidence level was set to 0.95; hence, Z_α/2 = 1.96. The assumed population standard deviation was 1.53 according to the standard deviation of EuroQol-visual analogue scale (EQ-VAS) scores in the norms in the urban Chinese population (Yang et al., 2018), resulting in σ = 1.53. The desired precision was set to one tenth of σ; hence, d = 0.15. The minimum sample size for the current study was 400.

To select a representative sample, we utilized a cluster sampling method to randomly select participants for our study. We randomly selected 10 communities in the urban areas of Wuzhishan city. Within each community, the local neighborhood committee, an organization that is responsible for residential management at the grassroots level, identified potential seasonal retired migrants based on their seasonal movements. The potential seasonal retired migrants were further screened on the basis of the predefined inclusion and exclusion criteria by trained interviewers.

Eligible participants were then invited to the local study site, where trained interviewers first instructed them about the study’s aim. The participants were then required to complete self-administered questionnaires. Those who encountered problems reading or understanding the questionnaires could receive help from interviewers. All completed questionnaires were assessed onsite by interviewers to ensure that there were no missing values. If any missing items were identified, the participant was required to complete the missing items instantly. As a result, there were no missing values in the current study.

Variable measurements

Statistical analysis

We provided frequencies and percentages for participants’ demographic characteristics, presence of chronic diseases, categorical BMI, and behavioral factors to describe the participants’ characteristics. In addition, we also provided the means and standard deviations for age and BMI. To describe participants’ HRQoL, we provided frequencies and percentages for participants’ responses to each dimension of the EQ-5D-5L descriptive system. The EQ-VAS scores were summarized using means and standard deviations and medians and interquartile ranges (IQRs).

When analyzing the factors associated with HRQoL, as described in previous studies (Kim et al., 2018; Wong et al., 2021), we first dichotomized the responses in each dimension of the EQ-5D-5L descriptive system into two levels: (1) reporting no problems and (2) reporting any problem (slight, moderate, severe or extreme problems). We then conducted separate logistic regression modeling of the probability of reporting any problem in each dimension with the same regressors, including selected demographic and behavioral characteristics, BMI, and the presence of chronic diseases. Because participants who were unmarried reported no problems in the dimension of self-care, leading to quasicomplete separation problems, the marital status variable was not included in the self-care model. Similarly, because participants with an education level of middle school or lower and those who were current smokers reported no problems in the self-care dimension, these participants were excluded from the self-care model. Odds ratios (ORs) were computed from the five logistic regression models. The five logistic regression models can be expressed as follows: ${\displaystyle logit\left(P\left(Y=1\right)\right)={\beta }_0+{\beta }_{1}Sex+{\beta }_{2}Age+{\beta }_{3}Marital+{\beta }_{4}Education+{\beta }_{5}Income+{\beta }_{6}ChronicDiseases+{\beta }_{7}BMI+{\beta }_{8}Smoking+{\beta }_{9}Alcohol+{\beta }_{10}Sleep+{\beta }_{11}PhysicalActivity}$

Here,

• P(Y = 1) is the probability of the participant reporting any problem in the corresponding dimension of the EQ-5D-5L descriptive system.

• β₀ is the intercept, and β₁ to β₁₁ are the coefficients representing the effect of each independent variable on the log-odds of the outcome.

We also conducted a multiple linear regression model using the EQ-VAS score as the dependent variable and the same regressors used in the logistic regression models. The multiple linear regression model can be expressed as follows: ${\displaystyle EQ-VASscore={\beta }_0+{\beta }_{1}Sex+{\beta }_{2}Age+{\beta }_{3}Marital+{\beta }_{4}Education+{\beta }_{5}Income+{\beta }_{6}ChronicDiseases+{\beta }_{7}BMI+{\beta }_{8}Smoking+{\beta }_{9}Alcohol+{\beta }_{10}Sleep+{\beta }_{11}PhysicalActivity}$

Here,

• β₀ is the intercept, and β₁ to β₁₁ are the coefficients representing the effect of each independent variable on the EQ-VAS score.

All the statistical analyses were performed by using SAS software version 9.4 (SAS Institute, Cary, NC, USA). A two-sided p value of less than 0.05 was considered statistically significant.

Characteristics of participants

A total of 1069 eligible participants were invited to participate in the current study, 992 of whom completed the survey, yielding a response rate of 92.8%.The mean age of the participants was 67.87 ± 6.98 years, and approximately half of them (56.5%) were female. The majority of the participants (93.6%) were married. University or above was the most common level of education (36.7%), followed by high school (30.3%). Slightly more than half of the seasonal retired migrants (54.7%) had an income level of less than 5000 CNY per month. A total of 41.5% had chronic diseases. The mean BMI of the participants was 24.11 ± 3.03. The majority of the seasonal retired migrants had never smoked (89.4%) or drunk alcohol (62.5%). Details regarding the characteristics of the study population are presented in Table 1.

Health-related quality of life

Table 2 shows the distribution of the participants’ responses to each dimension of the EQ-5D-5L descriptive system and their EQ-VAS scores. Almost all of them reported no problems in the self-care dimension (98.8%) or usual activities dimension (98.0%), followed by the mobility dimension (94.6%) and anxiety/depression dimension (91.7%). The participants most commonly reported problems (slight to extreme problems) in the dimension of pain/discomfort (16.7%). In addition, the median EQ-VAS score was 84 (IQR: 80–89).

Table 3 presents the results of logistic models for each dimension of the EQ-5D-5L descriptive system. Sleep quality was the most important factor and was negatively associated with all five dimensions of the EQ-5D-5L descriptive system. In addition, the presence of chronic diseases was negatively associated with three dimensions of the EQ-5D-5L descriptive system (mobility, pain/discomfort, and anxiety/depression). Compared with those with low-level physical activity, participants with moderate-level physical activity were less likely to report problems in the pain/comfort and anxiety/depression dimensions. In addition to the pain/comfort and anxiety/depression dimensions, high-level physical activity was additionally associated with reporting fewer problems in the usual activities dimension. The responses to the EQ-5D-5L descriptive system were also associated with other factors, including education level, income level, smoking status, and alcohol consumption.

Table 4 presents the results of the multiple linear regression for the EQ-VAS score. Lower EQ-VAS scores were significantly associated with age (β = −0.15, p = 0.001), the presence of chronic diseases (β = −2.24, p < 0.001), and poor sleep quality (β = −6.90, p < 0.001). Compared with low-level physical activity, moderate-level physical activity (β = 5.01, p < 0.001) and high-level physical activity (β = 5.35, p < 0.001) were associated with higher EQ-VAS scores.

Sex, marital status, and BMI were not significantly associated with any dimension of the EQ-5D-5L descriptive system or the EQ-VAS score.