Data was collected from April to July 2018 in Wuhu city, Anhui province, China. We adopted the cluster sampling method to select the participants. The sampling process consists of four steps: (a) According to the administrative divisions of Wuhu city, it was divided into four districts: Jinghu, Jiujiang, Sanshan, and Yijiang. (b) A district was selected from the city, which was Jinghu district. (c) All communities in the Jinghu district were selected. (d) To survey all the parents who lost their only child in the communities. The inclusion criteria were: a) Being older than 49(Considering that the health bureau had brought the parents who lost their only child aged 49 or older into the systematic management, and the mothers who lost their only child aged 49 or older were no longer in the fertile period), b) Having normal cognitive functions, willing and able to cooperate throughout the survey. The exclusion criteria were: a) The parents who lost their only child had refused to accept the government's or others' condolence or survey, b) Moving to another place from the district.
20 family planning officials in the local communities and two professors from Zhejiang Chinese Medical University were recruited and trained as investigators. All participants were clearly informed of the purposes of the study and were asked to sign in the consent form. All participants were ensured of their rights to decline to participate or to withdraw from the study at any time. Privacy and confidentiality of the participants were ensured. Each interview lasted for about 30 minutes. The criteria for identifying a valid questionnaire were: The information of the two scales used was complete, and there were very few missing even no missing sociodemographic information or other data, and no contradictory content. 350 participants were surveyed. 306 valid questionnaires were recovered, and the effective recovery rate was 87.4%.
The Social Support Rating Scale (SSRS) designed by Xiao, which is one of the most commonly used instruments for measuring social support in China. It consists of 10 items measuring three dimensions: subjective support (4 items), objective support (3 items), and utilization of support (3 items) . The total score consists of the sum of the scores of the three dimensions. A higher score shows a higher level of social support. A total score of less than 20 signifies low social support, a total score of 20-30 signifies moderate social support, and a total score of more than 30 signifies high social support. The SSRS total score among the parents who lost their only child was regarded as the dependent variable.
2.2.2. EQ-5D scale
The European quality of life-5 dimensions (EQ-5D) was introduced by EuroQol Group in 1990. There are five dimensions in the scale: mobility, self-care, daily activities, pain or discomfort, anxiety or depression. There are three levels of "no difficulty", "some difficulty" and "total difficulty" in each dimension, which are used to evaluate the quality of life of the respondents. Due to the lack of calculation method of China's EQ-5D index score at present, the integral conversion table of Japan for calculation was chosen to use, which borders on China and also located in the east of Asia. The scale index score is -0.11~1.00 points. A higher EQ-5D total score shows better health. The last part of the scale is the European quality of life visual analog scale (EQ-VAS), which is a 20-centimeter visual scale. Participants were asked to rate their own health for the day, with a score of 100 representing their “best health” and a score of 0 representing their “worst health” .
2.2.3. Independent variables
The independent variables included the following: gender(1=male, 2=female, male as the reference value), marital status(1=in marriage, 2=not in marriage, in marriage as the reference value), the grandchildren(1=have, 2=no, have as the reference value). Age, monthly income, number of chronic diseases and EQ-5D score were continuous variables. Educational level (1= middle school or less, 2= high school, 3= technical college or more) and self-rated health status（1=good, 2=moderate,3=poor）were regarded as continuous variables in the generalized linear regression model.
A multiple-choice question was used to survey the number of chronic diseases among respondents, "How many chronic diseases do you have?" Sixteen chronic diseases were listed for selection, including diabetes, hypertension, the malignant tumor, hyperlipidemia, cerebrovascular disease, coronary heart disease, Cerebral infarction (stroke), senile dementia, gynaecology disease, chronic liver disease, arthritis, osteoporosis, gout, asthma, hematopathy, Tuberculosis (TB), chronic low back pain, cataract. A higher score means that the participants have more chronic diseases.
2.3. Quality control
The ages of the respondents were confirmed using the household registration system. During the face-to-face field investigation, these trained family planning officials and professors explained how to fill in the questionnaires and helped respondents complete them in their homes or community residents committee office. The EpiData3.1 software was used to set up the database, and double input was conducted to ensure accuracy.
2.4. Data analysis
Data were analyzed using the SAS version 9.1 software. Sociodemographic variables of the participants were expressed in terms of frequencies. Considering the non-normal distribution of the SSRS score, the rank sum test was adopted to compare the SSRS scores among the parents who lost their only child with different socio-demographic characteristics. The generalized linear regression models were performed to analyze factors influencing all dimensions’ scores of the scale.