Participants and procedures
A two-stage stratified cluster sampling method was used to select research sites. Three provinces were selected based on the GDP ranking in mainland China: Shandong (top 10), Hunan (11–20), and Guangxi (21–31). A total of 12 counties were randomly selected from three provinces based on the rank of an annual income of each county within each province. A case-control psychological autopsy study was devised and implemented to collect participants’ information from June 2014 to September 2015[12, 13].
In each of the selected counties, suicide cases aged 60 years and older were collected consecutively based on the death certification system. The controls were community members matched by age (± 3 years), gender, and location. For every participant, two informants were selected. Generally, the first informant was a family member, and the second informant was a friend, a neighbor, or other important people. A total of 968 interviews were completed for 242 pair of suicides and controls with 4 interviews in one pair. The survey was conducted through face-to-face interview with an average interview time of 90 minutes.
The investigators consisted of teachers and graduate students from Shandong University, Central South University and Guangxi Medical University. All investigators were trained intensively for 10 days in a standardized way on the instruments and the skills of the interview.
The study was approved by the Institutional Review Board of the Shandong University, Central South University, and Guangxi Medical University. The aim and procedure of the research were explained to all participants. Written informed consent was obtained before interviews were conducted.
Measurements
Demographic characteristics
Demographic variables included age (in years), gender (male/female), marital status (stable/unstable), family income (yuan), if having a physical disease (yes/no), if having a mental disorder (yes/no), if living alone (yes/no) and being left behind (yes/no). In the study, people who were married and living with a spouse, or cohabitation with a partner were classified as “stable”, while people with other marital status were classified as “unstable”.
Life Event Scale for the Elderly
Life events were measured by the Life Events Scale for the Elderly (LESE), which was developed specifically for older Chinese adults [14]. It is a valid and reliable scale used among the elderly generally, and elderly people who died by suicide in particular [14, 15]. A total of 46 life events were categorized into three separate categories: (1) Health/Hospital (16 items); (2) Family/Home (18 items); (3) Friends/Relationships (12 items). Each life event was assessed by five questions: (1) The date it happened (never occur, one month, one year, and more than one year); (2) Whether it was positive or negative for the target person; (3) The effect on participants’ mental health, measured by a five-point Likert scale from 1=no impact to 5=very severe impact; (4) The duration of the event (three months, six months, one year, and more than one year); (5) The number of times it happened.
In this study, life events were recognized as a recent life event if happened within one month. Life events were recognized as long- term life events if happened one month before death and interview for living control.
Statistical analysis
The information provided by two informants was combined as proxy data for the suicides and controls. The demographic characteristics based on the information provided by the first informant was relied on. Answers which were associated with an increased risk of suicide were used when two informants reported differently for each item of LESE. The rationale for this practice is that a targeting behavior may exist as long as one of the two informants has observed it.
All participants were divided into two groups using the median number of life events types. The t-test and chi-square test were used to compare different demographic characteristics between the elderly who experienced fewer life events and more life events both among suicides and living controls. χ2 was used to analyze the incidence of 46 life events between suicides and controls. χ2 was also used to compare the frequencies of the top 10 life events between males and females among suicides. Logistic regression analysis was used to explore the probability-changing factors for suicides.
All analyses were performed using SPSS 24.0 for Windows (SPSS Inc, Chicago, IL). In the study, statistical significance was determined if P less than 0.05.