Study design and data collection
Data for this study was derived from the fifth National Health Services Survey 2013 conducted in Jiangsu province(J-NHSS) that is located in eastern coastal area of China, and has a population of 79.38 million (2013) with top level the socio-economic indicators among all provinces in China.
The 2013 J-NHSS adopted a multistage stratified random sampling strategy. First, the counties (rural areas)/ districts (urban areas) were chosen randomly. Second, towns were drawn from each county/district, and then villages/communities were drawn randomly from each town. Finally, considering diversities in geographic location and economic status, the number of households drawn from each village/community was determined according to the proportion of the number of households (details of the NHSS design have been published ).The J-NHSS covered 18 counties (rural areas) or districts (urban areas) of Jiangsu. A total of 12,600 households from 180 villages/communities that were from 90 towns were selected to participate in the survey.
A questionnaire was administered to the selected participants by interviewers recruited from local health workers. The investigations were conducted by professionals after written consent forms (consent to participate under the ‘Ethics, Consent and Permissions’ heading and to publish their data in this study in any form) were explained and accepted. Ethical approval was obtained from the Ethics Committee of Medical Faculty, Nanjing Medical University.
A total of 36,381 individuals (from 12,600 households) in Jiangsu province completed the survey in 2013. For the purpose of this study, a final sample of 10,257 individuals aged ≥60 years were enrolled. Exclusion criteria for individuals were cognitive or awareness problems, and the inability to understand or answer the questions.
Measurement of HRQoL
Eq-5d-3 L is preselected for the measurement of an individual’s self-perceived HRQoL in the NHSS. It is a five-item questionnaire encompassing the domains of mobility, self-care, usual activities, pain/discomfort, and anxiety/depression, as well as overall health rated on a Visual Analog Scale (VAS) [13, 14]. For the above five dimensions, each has 3 response levels (i.e., 1 = no problems, 2 = some/moderate problems, and 3 = extreme problems). This allows for 243 possible combinations of health conditions. HRQoL results measured by the Eq-5d-3 L were converted to an index score (a proxy health utility score) using the Japan Time-Trade Off (TTO) value set which ranges from − 0.11 to 1.00 , in which a score close to 1 indicates better the health. In contrast, for each domain, a higher score indicates a more serious problem. The Eq-VAS scores ranged from 0 (worst health) to 100 (best health). The equivalence between the Chinese version and the English version of EQ-5d-3 L has been proved . Also, the Chinese version has passed the test of reliability and validity both in previous studies  and this study. The Cronbach’s α of EQ-5d index in this analysis was 0.852. The Correlation coefficient between VAS and index score was 0.540 (P < 0.05); the Correlation coefficients for VAS and the five domains were − 0.472,-0.422,-0.470,-0.477 and − 0.378 (P < 0.05), respectively.
Body mass index
All respondents were asked to report their height and weight. BMI was calculated as the weight (in kilograms) divided by the square of the height (in meters) (kg/m2). The study subjects were classified into “underweight” (< 18.5), “normal weight” (18.5–24.0), “overweight” (24.0–28.0) and “obese” (≥28.0), based on the BMI quartiles which were formed according to the “Criteria of Weight for Chinese Adults” .
Three categories of control variables were included in the analyses. The first category was demographic characteristics, including age, marriage, education, registered residence, and income. The second category was health related behaviors, including smoking (at present or in the past), alcohol consumption in the last 12 months, and physical exercise (in the last 6 months, when the average weekly exercise exceeded once per week). The third category inquired about two kinds of health conditions: whether there was a diagnosis with any chronic disease or not and whether there was a hospitalization or not in the last 12 months.
All statistical analyses were performed using SPSS for windows (version 18.0, SPSS, Inc., Chicago, IL). Chi-square tests were used to compare the frequencies of Eq-5d responses among the different BMI groups. Then, One-way ANOVA analyses were selected to compare the scores among BMI groups. Additionally, Post Hoc Multiple Comparisons were also conducted using the Dunnett’s T3 methods, since equal variances were not assumed.
Binary Logistic regressions were used to examine the influence of BMI on Eq-5d scores. The dependent variables were divided into two categories depending on whether the Eq-5d score (Eq-5d index in model I-1 and I-2, Eq-5d VAS in model II-1 and II-2) was lower than the Chinese elderly norm. A previous study reported the norm value of EQ-5d-3 L index/VAS among Chinese population ≥ 60 years. The norm value of index/VAS for males was 0.82/76.95, and for females was 0.73/73.55 . According to the research objective, BMI groups were the core independent variable in these regression models. In addition, the control variables were adapted including demographic characteristics, health related behaviors and health conditions.