We conducted literature searches across PubMed, Web of Science, the Cochrane Library, PsycInfo, CNKI, Wan Fang Data, VIP, and CBM databases from the inception of each database until February 20, 2025. The retrieval approach of the combination of free words and subject words was adopted and with the search strategy adjusted according to the search rules of each database. Terms related to the words “depressive disorders”, depression, “depressive disorder”, geriatric, senile, elderly, “late-life”, suicide risk, suicide-related behavior, suicide ideation, “suicide attempts”, impact factor, contributing factor, and risk factor were used to retrieve potentially relevant articles on the risk factors for suicide in LLD patients. The literature retrieval strategy takes PubMed as examples. The comprehensive PubMed search strategy is delineated in Table 1 . The detailed search strategy is provided in the Supplementary Material .

Inclusion criteria: (1) The study subjects were patients with depression as diagnosed by the Second Revision of the Chinese Classification Scheme and Diagnostic Criteria for Mental Diseases, the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV), the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-V), or the International Classification of Diseases-10 (ICD-10); (2) Age ≥60 years; (3) Type of study was case–control study, cohort study, and cross-sectional study; and (4) The study was on the risk factors for suicide in LLD patients, with a clear or calculable odds ratio (OR) and 95% confidence intervals (CIs) of each influencing factor of suicide risk.

Exclusion criteria: (1) review, case reports, or conference papers and other types of research, (2) non-Chinese and English literature, (3) literature with no relevant data, incomplete data that cannot be extracted, and (4) duplicate publication, the literature cannot be obtained in full text.

Literature search, checking, and screening were completed by two researchers independently using NoteExpress software, extracting data from the included literature and cross-checking them, and in cases of doubt or disagreement, the study was evaluated by a third reviewer until an interactive consensus was reached on the inclusion criteria. The first author, year of publication, where the study was conducted, diagnostic criteria for late-life depression, study design, sample size, influencing factors, definition (at least three relevant articles for each risk factor), odds ratio (OR), and 95% confidence interval (95% CI) were abstracted from the included studies to Microsoft Excel 2016. If the studies did not report OR equivalent measures, raw data were screened to determine whether ORs could be calculated.

The severity of depression was rated using the Hamilton Rating Scale for Depression (HAMD-17). Age of onset was defined as the age at first occurrence of depression meeting the diagnostic criteria. Suicidal behavior refers to the act of intentionally ending one’s own life, encompassing the processes of suicidal ideation, suicide planning, suicide attempt, and completed suicide. Suicide risk refers to the likelihood of an individual exhibiting a suicidal behavior based on the presence of suicidal ideation (18). Patients were at risk for suicide when they presented with any one of suicidal ideation, suicidal planning, or suicidal behavior (19).

Two systematically trained researchers evaluated the methodological quality of the included literature and cross-checked the results, asking a third researcher to assist in the judgment if disagreements were encountered. For cohort and case–control studies, the Newcastle–Ottawa Scale (NOS) (20) was used. The quality of the study was evaluated by eight items under the three categories of participant selection, comparability of study groups, and ascertainment of outcome or exposure. A score of ≥7 is classified as high-quality literature, 5 to 6 is classified as moderate-quality literature, and ≤4 is classified as low-quality literature. For the quality assessment of a cross-sectional study, the 11-item criteria recommended by the US Agency for Healthcare Quality and Research (AHRQ) were used (21). A score of 0 to 3 indicates low quality, 4 to 7 indicates medium quality, and 8 to 11 indicates high quality.

Statistical analysis was performed using Stata 18.0 software. Data expressed in the form of means were converted to ORs using the method recommended by the Cochrane Handbook of Systematic Reviews. Results were reported as OR and 95% CI. The χ ² test was used to evaluate the heterogeneity of the included studies (the test level was α = 0.1), and the size of the heterogeneity was assessed in combination with I ². When I ² <50% and P >0.05, it showed that there was less heterogeneity among the studies, the fixed-effects model would be used to combine the effect sizes. If I ² ≥50%, P < 0.05, it indicated high heterogeneity between studies. Effect sizes were combined using a random-effects model. When heterogeneity was large, sensitivity analyses were further conducted by comparing the consistency of the results of the random and fixed-effects models as well as by the leave-one-out analysis. Descriptive analysis was adopted for the influencing factors that were not suitable for meta-analysis. When three or more papers were included, publication bias was performed by quantitative analyses based on Begg’s and Egger’s test. P >0.05 indicated that there was no publication bias. P <0.05 was considered statistically significant.

From the total of 2,351 relevant articles retrieved, 534 duplicate literatures were eliminated, and 1,669 articles were removed after reviewing the titles and abstracts. These studies were excluded because their research subjects and designs did not meet the inclusion criteria and because they did not focus on suicide and suicidal-related factors. Then, the remaining 148 literatures were evaluated by reviewing the full text, and 12 literatures (13, 22–32) had finally remained. Figure 1 outlines the systematic literature screening process.

Of the 12 studies eventually included, eight were case–control studies, two were cross-sectional studies, and two were longitudinal studies. Seven studies were conducted in China, three studies were conducted in America, and one study each was conducted in Canada and Greece. The total cumulative sample size was 2,324 cases. The results of literature quality evaluation showed that eight studies (13, 22, 23, 25–27, 29, 30) were of high quality and four studies (24, 28, 31, 32) were of moderate quality. The basic characteristics of all 12 included literatures are illustrated in Table 2 .

Publication bias analysis of the influential factors using Begg’s and Egger’s test in Stata 18.0 software showed that no significant bias was seen in depression severity (t = 2.20, P = 0.079) and age of onset (t = -0.60, P = 0.579), as shown in Table 6 and Figures 4 and 5 .

Our study showed that depression severity is a risk factor for suicide in LLD patients. The greater the degree of depression in LLD patients, the greater the suicidal ideation in LLD patients due to the aggravation of pain, despair, negative emotions brought about by depression, and the impairment of cognitive functioning. Depression severity has been consistently identified as the most robust predictor of suicidal ideation in older adult populations across multiple studies (33). Moreover, the more severe the degree of depression and the more severe the clinical symptoms, the more prolonged and difficult it is to cure the disease, which will significantly increase the patient’s sense of despair and thus increase the patient’s risk of suicide. It has also been shown that LLD patients who have attempted suicide are more depressed (32). This is consistent with previous findings that depressive severity is strongly associated with a higher suicidal behavior (34, 35). The medical staff should focus on the people with higher levels of depression, pay attention to the changes in the level of depression, establish a trusting relationship with the patients, and assist them in actively cooperating with the treatment to reduce the level of depression and reduce the risk of suicide.

Previous studies have shown that the age of onset of depression in elderly patients with suicide attempts is younger than in elderly patients with depression without suicide attempt (32). Compared with late-onset depression patients, elderly individuals with early-onset depression who experienced their first episode before the age of 60 exhibit a higher recurrence rate, a greater number of prior episodes, and a higher prevalence of suicide attempt history (36). However, it has also been noted that patients with suicide attempt have a greater age of onset and show a positive correlation between age of onset and depression severity scores (13). In our study, the age of onset was not significantly associated with suicide risk in LLD patients. This may be due to the greater heterogeneity of the included studies. Current evidence remains inconclusive regarding this association. Future high-quality studies are still needed to elucidate the impact of age at first onset on suicide risk in late-life depression.

As all of the included studies were observational, the results of the meta-analysis of the influencing factors should be kept in research. The heterogeneous operationalization of suicide risk across studies introduces measurement bias that may limit the generalizability of our findings. The limited sample size and heterogeneity in measurement methodologies may have constrained the statistical power of our findings, and more high-quality studies and more sample sizes are still needed to further validate and supplement the results in the future. Concurrently, this study has not adequately parsed the potential moderating effects of cultural and regional disparities. Given that contextual factors such as cross-cultural attitudes toward suicide and regional variations in social support systems were not fully incorporated as covariates, interpretive biases may arise in the aggregated conclusions. Then, the subjects of this study were not differentiated into early-onset and late-onset types. In the future, a classification-based discussion can be conducted to explore the influencing factors of suicide risk in patients with early-onset depression and late-onset depression to implement targeted interventions.