This article provides the first comprehensive systematic review concerning epidemiological features of SI among the elderly in China. To our knowledge, this was the first meta-analysis to estimate the odds ratios of factors associated with SI among the elderly in China. Based on the study by Dong et al.[13], this study also provided a substantial update and complementary to the prevalence of SI among Chinese elderly with a more precise method. According to this meta-analysis, the pooled prevalence of SI among the elderly in China was 11.47% (95% CI 7.82–15.71%) calculated by a total of 79,861 subjects in 31 cross-sectional studies; Sixteen statistically significant factors associated with SI among Chinese elderly were identified. In the context of continuous deepening aging, the data above was certain to provide a credible reference for the development of effective SI prevention strategies for Chinese elderly.
Different cultural backgrounds and economic levels can result in varying prevalence of SI among the elderly across countries[13]. The prevalence of SI among the elderly in China from this study is relative high among the global elderly population. Taking countries with different levels of economic development as examples, a nationwide cross-sectional survey in South Korea, including 58,590 elderly, showed 15.72% prevalence of SI[53]; In Mexico, belonging to developing countries, a cross-sectional survey on elderly aged 65 years and older found 13.5% lifetime prevalence of SI[54]. However, one national survey that included 5,191 older Black Americans showed that the lifetime prevalence of SI was only 6.1%[55]. The reasons for the high prevalence of SI among the Chinese elderly may be that, firstly, rapid urbanization has, to some extent, dissolved traditional Chinese clan ties, exacerbating the loneliness and even depression of the elderly, especially those left behind, which may further lead to SI; secondly, China has a large elderly population, but mental health services have difficulty in meeting such a large demand; thirdly, in the unique Chinese culture, the elderly are often reluctant to trouble their young children, and SI has a certain stigmatizing effect, which may lead to the elderly not seeking help in time when they have psychological problems. Studies have shown that suicide mortality rates in China, including among the elderly, have declined significantly over the past few decades[56, 57]. Possible reasons for the high prevalence of SI and low mortality among the elderly are that suicide is a more sensitive public health issue with more adequate prevention and interventions at the national level than SI, and that individuals with SI also need to have factors such as frustration of belonging, hopelessness, perceived burdensomeness, and pain tolerance[6, 58–60]. This further warns us to be cautious about inferring suicide from SI.
To explore the sources of heterogeneity, this study conducted comprehensive subgroup analyses in domains including demographics, physical, mental, economic, spatial and temporal distribution, and measurement tools with time points. Several important findings concerning distribution characteristics in the prevalence need much attention. Firstly, inequalities in SI rates exist between urban and rural elderly, consistent with urban-rural disparities in suicide rates among the elderly in China found in Li and Katikireddi's meta-analysis[61]. The underlying mechanisms for these inequalities are still unclear[61]. The economic, educational, and lifestyle differences hidden behind the urban-rural dichotomy in China, along with China's urbanization process, may have contributed to the inequalities. Secondly, the elderly with poor physical conditions, especially chronic disease and ADL disability, tend to have a higher prevalence of SI. Former studies have found a positive association between chronic disease and SI, such as cardiovascular disease, stroke, ischemic heart disease, cancer, diabetes, and renal failure[62, 63], which was similar to the findings of our study. Functional disability, such as ADL disability, was regarded as a determinant of “high” psychological distress closely related to SI[64]. Thirdly, older groups with depression symptoms and low life satisfaction had a rather high prevalence of SI (41.78% and 35.36%), further pointing out the significance of regular depression screening and psychological interventions for the elderly. Fourthly, the elderly in the lowest economic-level regions (only China mainland included) had the lowest prevalence of SI. The low economic level means poor health care support and inadequate mental health education for the elderly, which may impact SI. Last but not least, a dramatic and significant difference in the prevalence of SI among Chinese elderly between 2001–2010 and 2011–2020 needs much attention. It is still unclear what caused the difference. Since 1999, China has undergone various sweeping changes, including rapid urbanization and ongoing aging. According to China Development Report 2020: Trends in Population Aging in China Trends and Policies on Population Aging in China, the rise in China's 65-year-old population was 30 million from 2000 to 2010, while 60 million from 2010 to 2020. The dramatic rise of the elderly population has changed the demographic structure of China, which undoubtedly had a significant impact on the senior care system and the health care system. Multiple studies have demonstrated that the elderly have the highest risk of suicide of any age group[2, 3]. A larger elderly population base and deeper urbanization between 2011 and 2020 would directly impact on healthcare, nursing, and health, which were closely related to the elderly population. Based on the above subgroup analysis of health and mental domains, we speculated that this may explain, to some extent, the higher prevalence of SI among the elderly from 2011 to 2020. Considering that the aging population will further increase, the prevalence of SI among the elderly in China is likely to remain at a relatively high level in the future, and precise prevention and intervention measures should be carried out as early as possible.
In addition, this study also explored associated factors with SI in the elderly from a variable perspective using meta-analysis. Sixteen factors were significantly associated with SI in the elderly, most of which were consistent with previous conclusions. There exist several factors for extra discussion or emphasis. In demographic domains, advanced age was a potential risk factor for SI, partly because those were in worse physical condition, less mobile, and more psychologically isolated. In the physical domain, comorbidity has received increasing attention recently as a risk factor for SI in the elderly. Two studies in Korea and America have demonstrated a significant association between comorbidity and SI in the elderly[65, 66]. The elderly with comorbidity were more likely to have a disability, poor physical conditions, and poor mental conditions than healthy ones[67], thus increasing the likelihood of SI[50]. In the mental domain, the association of pressure and religiosity with SI in the elderly was not significant, inconsistent with previous studies [24, 68]. The relatively small amount of included literature may likely cause inconsistent results. Above all, marriage and employment were significantly and negatively associated with SI in the elderly, suggesting that, to some extent, companionship and work engagement may reduce SI in the elderly[22, 24, 68].
Lastly, this study also highlights the significant impact of statistical methods and measuring tools with time points on the results of meta-analysis, further confirming the conclusions of previous related studies[14, 21]. For the statistical method used in the meta-analysis, the majority of previous studies did not consider the actual distribution of the prevalence when doing the meta-analysis on the prevalence but defaulted to the binomial or normal distribution. According to Barendregt et al., the prevalence does not always follow the standard binomial distribution[14]. When the prevalence of a disease is around 0.5, disregarding the actual distribution of the prevalence does not have a large impact[14]. However, when the prevalence of a disease is too large or too small, there is a large variance variability in the study results if the prevalence of each original study is not transformed to correct for its distribution[14]. Among the two commonly used methods for prevalence transformation, the double arcsine transformations method has better accuracy than the logit method[14]. Given the low prevalence of SI among the elderly, this study corrected the distribution of the prevalence using the double arcsine transformations method recommended by the researchers[14, 15, 21], and the pooled prevalence of SI among the elderly in China was 11.47% (95% CI 7.82–15.71%). However, the prevalence obtained by the direct method without transformation and the Logit method were calculated as 12.84% (95% CI 10.78–14.89%) and 9.45% (95% CI 6.39–13.95%), respectively. The results transformed by different methods show much difference, reinforcing the necessity of reporting the statistical transformation methods in the meta-analysis of the prevalence. The reasons are also one of the key reasons why we updated to the meta-analysis of Dong et al.[13]. Besides, this study found statistically significant differences regarding the prevalence in measuring tools and time points used in these studies. In comparing the effects of different time points on the prevalence of SI, we divided all time points into two groups (past ≤ 12 months, and past>12 months) in order to ensure that as many studies as possible were available at the time points used for comparison. We found that longer time points did not imply higher prevalence, consistent with the conclusions in Li et al. and Xiao et al.[21, 22]. Retrospective bias and the proximate effect of event occurrence may be responsible for this inconsistency. Therefore, a narrower time frame in future studies is necessary for the elderly due to possible memory loss and cognitive impairments[22]. Similarly, the prevalence with different measuring tools varied considerably. More than half of the included studies used single-item questions to assess SI. A former study showed that the validity of single-item tools for SI is poorer than multi-item scales[69]. Nevertheless, for the sake of improving the response rate, large national epidemiological surveys still continue to use single-item question. In general, more standardized tools with narrow time points need to be promoted in the future.
4.1 Strengths and limitations
Firstly, this study was the first meta-analysis concerning factors of SI among Chinese elderly. Secondly, this study used a more precise method to fill a nearly 8-year gap in the prevalence of SI among Chinese elderly. Finally, this review identified 16 factors that significantly associate with SI in the elderly by pooling the effect sizes.
There also existed some limitations in this review. Firstly, heterogeneity remained high and could not be avoided after subgroup analysis, which might contribute to the publication bias; Secondly, the definition of SI was not entirely consistent in the included studies, which may have biased the findings to some extent. Finally, more than half of the studies used a single item from well-known or self-designed scales to rate SI for convenience, whose validity may be lower than the multi-item standardized tools.
4.2 Implication
The findings of this review have provided valuable insight into SI among Chinese elderly. Given the high lethality of suicide in the elderly and China’s context of accelerating population aging, developing more targeted treatment or intervention approaches for preventing SI is necessary. Additionally, this review highlighted the importance of choosing the proper method to transform the prevalence of original studies when calculating the pooled prevalence of some diseases. Finally, considering that longer time points may introduce retrospective bias and proximate effect, developing credible measuring tools with narrow time points for SI is equally important.