Data sources
CHARLS (China Health and Retirement Longitudinal Study) is a nationally representative follow-up survey, which is designed to investigate the economic and health of the populations aged 45 years old and above. The baseline national wave of CHARLS(W1) was being fielded in late summer 2011-March 2012 and includes about 10,000 households in 150 counties/districts and 450 villages/resident committees. The individuals will be followed up every two years, Wave 2 (W2) was fielded in 2013 and Wave 3 (W3) in 2015. A special life history wave was fielded in 2014 (W4) [25]. CHARLS questionnaires include information about self-reported and objective measures of health among middle-aged and elderly in China, these include health status and functioning, general health, physician-diagnosed chronic illnesses, lifestyle and health-related behaviors, subjective expectation of mortality, activities of daily living (ADL), cognition testing, depression and so on. The Biomedical Ethics Review Committee of Peking University (IRB00001052-11015) gave the ethic approval and allowed the CHARLS research group to collect data. Requiring all interviewees to sign informed consent was the first step of the study.
The study sample was derived from the baseline data W1 and follow-up data W2, W3 and W4 and specially was those who could be followed-up in the 4-year period, thus, a total of 4824 subjects were included.
Definition Of Successful Aging
Our concept of successful aging took the definition of Rowe and Kahn[6], including the following 5 components: 1) absence of major diseases, 2) freedom from disability, 3) high cognitive function, 4) no depressive symptoms, 5) active social engagement in life.
1 Absence of major diseases: To judge the status of chronic diseases, respondents were asked by the following question: “Have you been diagnosed with conditions listed below by a doctor?” The conditions include cancer, chronic lung disease, diabetes, heart disease and stroke. The research indicated that those diseases mentioned above may cause major disease burden for the elder[26], the respondents were classified as having no major diseases if they reported have no any of the above five chronic diseases.
2 Freedom from disability: The activities of daily living (ADL) scale was used to assess the ADLs[27], according to the following questions: “Because of a physical, mental, emotional or memory problem, do you have any difficulty with one type of everyday activity, excluding any that you expect to last less than three months?” The everyday activities include dressing, bathing, or showering, eating, getting into or out of bed, using the toilet, and controlling urination and defecation. Respondents were classified as having no disability if they reported that they had no difficulty with the everyday activities of six items mentioned above.
3 High cognitive function[14] : Cognitive function was assessed with the Telephone Interview of Cognitive Status (TICS).This includes both immediate and delayed recall of ten words on a list, serial subtraction of seven from 100 (up to five times), and naming the day of the week, month, day, year, and season, and drawing the picture. The score of cognitive function ranged from 0 to 21. Respondents were considered to have high cognitive functioning if they achieved a median or above score, and the median score was 11.
4 No depressive symptoms: Depressive symptoms were assessed using the CES-D 10 (10-item Center for Epidemiological Studies Depression Scale). The cut-off value is less than 10 points, which was used to identify no depressive symptoms.
5 Active social engagement in life: Respondents were defined as being actively social engaged if they participate in any of the following types of social groups: voluntary or charity work, provided help to family, friends, or neighbors, gone to a sport, social, or other kind of club in the month preceding the interview. The participant who met all five indicator criteria mentioned above was defined as “successful aging”, otherwise as “non-successful aging”.
Other Variables
Control variables includes following factors: age (60–74 years/75 years or older), gender (Male/Female), education level (Primary school and below/Junior high school or above), income ( low: <650yuan,medium: 650-100580yuan,high:>10058yuan) marital status (Married/Cohabitating/Divorced/Separated/Widowed/Never married), community type (Rural/Urban), smoking (Yes/No/Quit), and drinking (Drink more than once a month/Drink but less than once a month/Do not drink).
Mortality
Participants enrolled in W1 were followed up in W2, W3 and W4. We collected the all-cause mortality and survival information of the respondents during the four wave surveys (2011–2015). W2 recorded of the respondents’ both status information (dead or alive) and death time, while W3 and W4 only provided the interview status information (dead or alive). We recorded and calculated the survival time of those who had the accurate all-cause death time by the interval between the interview time of W1 and the death time in W2. If respondents’ accurate death time was not available, we calculated the specific value by the interval between the interview time of W1 and the specific wave with death information and then defined the median of this value as the survival time. The survival time of those who were alive during whole follow-up interview was the interval between W1 and W4.
Statistics Analysis
All descriptive statistical analyses were performed by gender. We first used chi-square test to compare individual characteristics (including dichotomous or categorical variables) with and without successful aging. Next, the survival analysis was used to examine the association between successful aging and all-cause mortality, the Cox proportional hazards regression models were used to estimate the unadjusted and adjusted hazard rations (HRs) and 95% confidence intervals (CLs) of successful aging. At last, whether the education could modify the effect of successful aging on the all-cause mortality was assessed, in this stage, we also used the Cox proportional hazards regression models to examine the educational mediating effect on all-cause mortality among successful aging and non-successful aging populations. All statistical analyses were performed by SAS 9.3. The significance level was set at 0.05.