Suicidal Ideation, Suicide Attempts, and Suicide Deaths in Adolescents and Young Adults With Type 1 Diabetes: A Systematic Review and Meta-analysis

Adolescence (10–19 years of age) and young adulthood (20–24 years of age) are unique developmental periods during which youth establish their autonomy and personal identity (1,2). For youth living with a chronic condition, these stages are further complicated by the daily demands of the condition. In adolescents and young adults (AYA) with type 1 diabetes (T1D), concurrent physiologic and psychological changes are occurring, including a deterioration in glycemic stability (glycated hemoglobin [HbA_1c]) (3), decreased adherence to self-care management (4), and increased risk of psychiatric disorders and involvement in risk-taking behaviors (5).

Suicide is the third-leading cause of death among youth 15–24 years old in the U.S., accounting for 6,062 deaths among this population in 2022 (6). Suicide also represents important costs to society, with the average cost of suicide (i.e., economic value of lost economic productivity) estimated at $835,288 (expressed in 2014 international dollars) per youth death in the U.S. (7).

Previous reviews have been conducted to investigate suicide risk in T1D (8,9). However, to our knowledge no systematic review or meta-analysis has been conducted to estimate the prevalence of suicide-related behaviors (i.e., suicidal ideation, suicide attempt, and suicide death) in AYA with T1D. Nevertheless, it has been shown that depression, suicidal ideation, and suicide attempts are highly prevalent in adolescents with a chronic condition and may have negative effects on self-management and health-related outcomes (10). Thus, we hypothesized that AYA with T1D are at increased risk of suicide-related behaviors; in turn, such behaviors may affect their T1D self-management. The aim of this systematic review and meta-analysis was to synthesize the evidence reported on the prevalence of and factors associated with suicide-related behaviors in AYA with T1D.

We registered the details of our systematic review with the International prospective register of systematic reviews (PROSPERO) (CRD42022358303). Methods adhere to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (11).

In collaboration with a medical librarian, we developed a comprehensive search strategy to identify relevant studies. A systematic search was conducted on MEDLINE, Embase, and PsycInfo for studies published from inception to 3 September 2023. We used terms related to T1D, suicide-related behaviors, adolescence, and young adulthood (Supplementary Appendix 1). Two reviewers (O.R.-C. and A.S.G.) screened the titles and abstracts for relevance. Following the initial screen, full-text articles were assessed for eligibility. Consensus was reached through discussion. Additionally, reference lists of included articles and identified reviews were manually reviewed.

We considered for inclusion full-text articles published in English or French where participants were aged 10–24 years with a diagnosis of T1D and where the prevalence of one or more suicide-related behaviors (i.e., suicidal ideation, suicide attempt, suicide death) assessed with any tool was reported. We also included studies reporting on younger and older individuals if the mean age of the participants was within our defined age range or if the age of individuals at the time of performing suicide-related behaviors fell within that range. Cross-sectional studies, case-control studies, and cohort studies were eligible for inclusion. We excluded case reports and studies where the majority of participants were youth with type 2 diabetes or were followed for a depressive episode. Where suicide-related behaviors were reported from studies in overlapping cohorts of patients, only the most relevant and recent study with the largest sample of participants was included.

Two reviewers (O.R.-C. and A.S.G.) extracted data from included studies using a standard form for the following study characteristics: 1) first author, 2) year of publication, 3) location, 4) study design, 5) sample size, 6) sex or gender distribution, 7) age distribution, 8) mean or median HbA_1c, 9) method of assessment for measures of suicide-related behaviors, 10) prevalence of suicide-related behaviors, 11) factors associated with suicide-related behaviors, and 12) other significant findings (e.g., suicide methods).

The methodological quality of the articles was evaluated with the Joanna Briggs Institute critical appraisal checklist for studies reporting prevalence data (12). The Joanna Briggs Institute checklist comprises nine criteria rated as yes (1 point), no (0 points), not applicable (0 points), or unclear (0 points), with a global score for each study ranging from 0 to 9. Higher scores represent higher-quality studies. We adapted the tool to our study (Supplementary Appendix 2). Two reviewers (O.R.-C. and A.S.G.) assessed the quality of the included articles independently. Disagreements were resolved through discussion.

To calculate the pooled prevalence of suicidal ideation and suicide attempts across all available nonoverlapping studies, we used a random-effects model with the restricted maximum likelihood procedures (13). Proportions were transformed (Freeman-Tukey double arcsine transformation) before pooling of study estimates (14). Heterogeneity was evaluated with use of the I² statistic, with classification as mild (0–30%), moderate (31%–50%), substantial (51%–80%), and considerable (81%–100%).

The prevalence of suicidal ideation and prevalence of suicide attempts were calculated for all timeframes combined, and the prevalence of suicidal ideation was also calculated for timeframe subgroups (i.e., past 12 months vs. lifetime) (15). When values were reported for more than one timeframe, the value for the longest timeframe was included in the analysis. To investigate heterogeneity, we used meta-regression and subgroup analysis. Subgroup analysis was used for the categorical variables timeframe and sex. Meta-regression was used for the following continuous variables: mean age, mean HbA_1c, year of publication, and study risk of bias score. We only performed subgroup analyses when there were four or more studies included in the respective subgroups and meta-regression when there were six or more studies with information on the variable of interest (16). All reported CIs represent 95% CIs, and a P value <0.05 was considered significant. Statistical analyses were conducted with use of the meta and metafor packages in R, version 4.3.0.

Synthesis was conducted for the outcome of suicide death in a narrative format due to heterogeneity in study designs. Prevalence of suicide deaths was reported as a range to represent the variation across the included studies. We also used a narrative approach to synthesize factors associated with suicide-related behaviors (17).

All data generated or analyzed during this study are included in the published article (and Supplementary Material).

The initial search yielded a total of 463 articles (Fig. 1). After removal of 108 duplicates, 355 titles and abstracts were screened for eligibility; 83 studies met criteria for full-text assessment, and 271 studies were excluded. One additional study identified from the reference lists of relevant articles met inclusion criteria. A total of 31 studies were included in the review (18–48), of which 20 were included in the quantitative analysis (18–20,22,23,25,27–29,32,35,37–39,42–45,47,48).

Of the 31 included studies, 14 included only participants within the target age range of 10–24 years (18,23,26,28,35–40,42,45,46,48). For 23 studies investigators used random probabilistic sampling techniques or incorporated the inclusion of all consecutive individuals within the sampling frame (18–24,27–33,35,36,38–41,44,46,47). A sample size calculation was performed in only one study (26). Investigators of 16 studies described the study participants and setting in detail including age, sex, and HbA_1c (18,20,21,23,25,28,32,35,36,38–40,42,43,45,46). In 21 studies there was sufficient coverage of the identified sample, with no subgroup of participants responding at a lower rate (19,22–24,26,28,30–33,35,38–47). In all but two studies investigators reported the use of validated self-reported questionnaires, semistructured interviews, or official registers to assess suicide-related behaviors (34,37). All but one study included measurement of suicide-related behaviors in a standard, reliable way (37). Twenty studies provided all the data needed to determine pooled prevalence for any of the suicide-related outcomes and thus were included in the quantitative analysis (18–20,22,23,25,27–29,32,35,37–39,42–45,47,48). Twenty studies had a response rate considered adequate (19,21–24,26,28,30–33,35,38,40,41,43–47) and four did not but investigators conducted an analysis to compare the characteristics of responders and nonresponders (18,25,27,42). Three studies had a low response rate without appropriate explanation (20,36,37), and for four the response rate was not clearly reported (29,34,39,48). The results of our assessment for the risk of bias are presented in Supplementary Appendix 3.

Main characteristics of included studies are presented in Table 1. Of the 31 studies included in this review, 20 were cross-sectional studies (18–21,23,25,26,28,29,32,35–37,39,42,43,45–48), 8 were retrospective cohort studies (22,30,31,33,34,40,41,44), 2 were prospective cohort studies (27,38), and 1 was a case-control study (24). The years of publication ranged from 1977 to 2023, with 20 studies published since 2011 (18,20–23,25,26,32,35–40,43–48). Overall, 16 studies included analysis of data from North America (21,23,27–29,32,35–39,43–47), 12 from European countries (18,20,22,24,25,30,31,33,34,40–42), and 3 from the African continent (19,26,48). Sample sizes of youth with T1D ranged from 6 to 28,887. Suicidal ideation was most frequently assessed using the nine-item Patient Health Questionnaire (PHQ-9). In assessment of suicide attempts in studies investigators used either hospital records and registers or questionnaires. Factors associated with suicide-related behaviors are presented in Supplementary Appendix 4.

In the 18 studies (n = 5,198 participants) in which prevalence of suicidal ideation was reported and that allowed for quantitative analysis, the pooled estimate from the random-effects meta-analysis was 15.4% (95% CI 10.0–21.7; I² = 94.1%) (Fig. 2). Four of these studies included a control group (n = 51,499 individuals), which allowed for the calculation of a pooled prevalence of suicidal ideation, 11.5% (0.4–33.3; I² = 99.5%), in individuals without T1D. For subgroup analyses based on timeframe, the prevalence of suicidal ideation in youth with T1D was lower with reporting for the past 12 months (9.6% [95% CI 6.8–12.6]) than with reporting for lifetime (32.2% [19.0–46.9]; P < 0.001). For subgroup analysis by sex, data were available for five studies (n = 1,484 participants) and showed a prevalence estimate of 17.6% (95% CI 6.9–31.8) in females vs. 7.0% (5.2–9.0) in males (P = 0.059). For meta-regression analysis, an association was shown for year of publication, with lower prevalence in more recent studies (β-coefficient = −0.01 [SD 0.00], P < 0.01), but heterogeneity remained high (I² = 94.5%). Of note, four of the five studies reporting lifetime suicidal ideation data were published before 2011, while the majority of studies reporting on suicidal ideation for the last 12 months were published after 2011. Mean age, mean HbA_1c, and risk of bias score did not account significantly for heterogeneity in prevalence estimates (Supplementary Appendix 5).

The prevalence of suicide attempts was reported in eight studies (n = 21,407 participants). The pooled estimate from the random-effects meta-analysis was 3.5% (95% CI 1.3–6.7; I² = 91.5%) (Fig. 3). Five of these studies included a control group (n = 3,135,236 individuals), which allowed for the calculation of a pooled prevalence of suicide attempts, 2.0% (0.0–6.4; I² = 99.9%), among individuals without T1D. Only two studies (n = 221 participants) reported data on suicide attempts by sex, and for that reason we did not perform a subgroup analysis. In both studies, there were more suicide attempts in females, with a total of 14 suicide attempts out of 119 female patients in comparison with 3 suicide attempts out of 102 male patients. In their study, Radobuljac et al. (42) found a statistically significant higher proportion of suicide attempts in females (P = 0.026). In meta-regression analysis, mean age, year of publication, and risk of bias score did not significantly account for heterogeneity in prevalence estimates (Supplementary Appendix 6).

Six studies reported a prevalence of suicide deaths that ranged from 0.04% to 4.4% among individuals with T1D. Across the six studies, suicide deaths represented between 7.8% and 28.6% of the total number of deaths. In the four studies with a comparison of the prevalence of suicide deaths between males and females (n = 12,276), there were twice as many suicide deaths in males (9 vs. 4 in females).

Overall, 11 of 31 studies included assessment of factors potentially associated with increased prevalence of suicide-related behaviors. Four studies looked at the relationship between T1D self-management and suicide-related behaviors; three reported an association between more T1D self-management difficulties and higher prevalence of suicide-related behaviors (21,27,28), while one did not (43). Seven studies included assessment of the relationship between HbA_1c and suicide-related behaviors, with mixed findings (18,20,21,26,42,43,46). Among the findings, in four studies there was significant association (20,21,26,43), indicating that higher HbA_1c levels in AYA were linked to suicidal ideation and/or suicide attempts. Of these, in one study an association was identified solely with suicidal ideation—not suicide attempts (20). In contrast, in three studies no significant association was detected between HbA_1c levels and suicide-related behaviors (18,42,46). It is noteworthy that only one of the included studies included adjustment for potential confounding variables (18). This analysis did not reveal a significant association between HbA_1c levels and the presence of suicidal ideation. In an additional study Matlock et al. (38) looked at changes in the 12 months after mental health intervention in patients initially shown to have suicidal ideation. While patients showed improvement in depressive symptoms and endorsement of suicidal thoughts, no significant change in mean HbA_1c was reported.

In five studies investigators looked at the association between suicidal ideation and depressive symptoms. Findings of two studies showed an association between suicidal ideation and depressive symptoms (21,38), one an association at baseline but not over follow-up (27), and another an association between suicidal ideation and hopelessness scores only in univariate models (28). There was only one study with findings that did not show an association (48).

Additionally, three studies included assessment of methods used in previous suicide attempts (27,35,45). Substance or medication overdose (including insulin) was the most common method, used in 50% to 73% of all attempts. Of note, Goldston et al. (27) and Majidi et al. (35) both showed that most attempts (55% and 50%, respectively) involved methods related to aspects of diabetes care (e.g., insulin overdose, omitting insulin, refusing to eat, and self-harm with syringes).

In our systematic review and meta-analysis we aimed to synthesize the literature on the prevalence of and factors associated with suicide-related behaviors in AYA with T1D. In the meta-analysis, we estimated a prevalence of suicidal ideation of 15.4% (95% CI 10.0–21.7) and a prevalence of suicide attempts of 3.5% (1.3–6.7) in youth with T1D. In comparison, our prevalence estimates in those without T1D were 11.5% (0.4–33.3) for suicidal ideation and 2.0% (0.0–6.4) for suicide attempts. Prevalence of suicide deaths ranged from 0.04% to 4.4% among AYA with T1D. Subgroup analysis by sex did not reveal any differences in prevalence of suicidal ideation in individuals with T1D. In our narrative synthesis, suicide attempts were more frequent in females, while suicide deaths were more frequent in males. Furthermore, depressive symptoms and difficulties with T1D self-management were frequently reported to be associated with higher rates of suicide-related behaviors. However, findings on the association of glycemic levels and suicide-related behaviors were inconsistent. Additionally, drug overdose (including insulin) was the most common method reported in suicide attempts, and most attempts involved methods associated with diabetes care.

Suicidal ideation and suicide attempts were prevalent in AYA with T1D. Many aspects of T1D self-management (e.g., daily blood glucose monitoring) can be a source of significant psychological distress in youth. This is referred to as diabetes distress, i.e., the negative emotions that result from the challenge of living with and managing T1D (49). A systematic review in adolescents aged 10–20 years with T1D (50) has shown high rates of diabetes distress, which may contribute to the prevalence of suicide-related behaviors in this population. In addition to diabetes distress, high rates of symptoms of depression and anxiety have been reported in youth with T1D (51). Hill et al. (52) recently developed a clinical model and recommendations for better understanding and ultimately prevention of suicide-related behaviors in youth with T1D. Based on ideation-to-action models of suicide, they proposed risk factors for suicide-related behaviors that are particular to T1D in youth, including loneliness, feelings of burdensomeness related to the demands of the condition, as well as access to insulin, a potentially lethal medication. Correspondingly, we identified in our systematic review that methods related to aspects of diabetes care are commonly used in suicide attempts in youth with T1D, which highlights the need for suicide prevention–focused interventions with the specifics of this population taken into consideration (52).

We did not observe evidence of differences in prevalence of suicidal ideation and suicide attempts in comparisons between AYA with and AYA without T1D. Results remain inconclusive as to whether a diagnosis of T1D is associated with higher rates for experiencing suicidal thoughts or engaging in suicide attempts among AYA. Arguably, those with T1D might be more reluctant to disclose suicide-related behaviors, leading to underreporting in studies. Indeed, many of these youth often express feelings of being a burden to their family and might prefer not to add to their health-related problems (53). Furthermore, AYA with T1D are known to experience high rates of stigma notably due to the visibility of their medical equipment and hypoglycemic symptoms, which may in turn lead to blame and discrimination (54). As psychiatric disorders are also heavily stigmatized (55), AYA who already endure stigma due to their T1D diagnosis may fear additional discrimination from peers in seeking help for mental health concerns. In any case, the prevalence of suicidal ideation and suicide attempts was not negligible in all AYA, as previous studies have shown (56). This reaffirms the imperative for mental health support for AYA both with and without T1D.

There were limited data on suicide deaths in AYA with T1D. Nevertheless, findings of a systematic review and meta-analysis on suicide deaths in adults of all ages with T1D showed an increased risk of suicide among those with T1D in comparisons with those without (8). Interestingly, the relative risk of suicide in those with type 2 diabetes compared with those without was not significantly higher. This reinforces the idea that T1D, as a chronic condition emerging in childhood and requiring insulin, presents unique challenges. Of note, youth with type 2 diabetes have high rates of comorbid mental illness, suicide attempts, and suicide deaths (57). When presenting in childhood, type 2 diabetes shares many similarities with T1D in terms of challenges related to self-management (58).

Our subgroup analysis did not reveal a statistically significant difference between sexes in terms of suicidal ideation. This was unexpected given the well-established sex differences in suicide-related behaviors among youth, with females typically reporting suicidal thoughts more frequently than males (56). It is possible that the experience of T1D mitigates sex differences in regard to suicidal ideation, given that diabetes distress appears to be present in both sexes fairly equally (50). In our narrative synthesis, suicide attempts were more frequent in females, while suicide deaths were more frequent in males, in line with previous research on suicide-related behaviors in youth without T1D (59). This serves as a reminder that while females might present with more suicide-related behaviors in clinic, males are known to use more lethal methods and are less likely to seek help for mental health issues (60).

Difficulties with T1D self-management were found to be associated with suicide-related behaviors. This association can be explained bidirectionally: suicidal thoughts may limit the ability of an individual to effectively complete self-management tasks, while difficulty with self-management behaviors may represent a significant stressor that could contribute to the development of suicidal ideation (28). However, the association between glycemic levels and suicide-related behaviors was inconsistent among studies. This demonstrates the clinical importance of not using glycemic levels as a marker for those who may be at risk for suicide-related behaviors. Of note, social determinants of health, known to be important drivers of suicide-related behaviors (61), were not always accounted for in reported studies and may act as confounders in examining the association between suicide-related behaviors and HbA_1c.

This systematic review and meta-analysis has several limitations. First, there was a considerable level of heterogeneity in meta-analysis of both suicidal ideation and suicide attempts, which remained largely unexplained by variables examined in subgroup and meta-regression analyses. Heterogeneity may be partially attributable to the various self-reported questionnaires and semistructured interviews used across studies to assess suicide-related behaviors. Second, most of the included studies lacked a control group, hindering the possibility of comparing suicide-related behaviors between AYA with and AYA without T1D. This limitation notwithstanding, studies with inclusion of a control group had a large number of control participants. Third, subgroup and meta-regression analyses were limited by the number of included studies with information provided on variables of interest. Notably, only two studies reported the prevalence of suicide attempts by sex. Fourth, while our goal was to determine prevalence for adolescence and young adulthood combined, there are likely differences between those stages of development that might have been overlooked in combining these two populations. Fifth, the narrative approach used to examine factors associated with suicide-related behaviors limits the possibility of quantifying the strength of evidence. There is also a risk of introducing subjectivity due to the challenge of comparing findings across studies with different methodologies.

The current literature on the prevalence of and factors associated with suicide-related behaviors, especially suicide deaths, in AYA with T1D is limited. Assessments of past suicidal ideation and suicide attempts for research should be performed with use of validated clinician-rated scales complemented with validated self-report questionnaires. This would reduce the risk of overestimating or underestimating the true prevalence of suicidal ideation and suicide attempts. Taken together, there is a need to conduct longitudinal studies to describe the evolution of suicide-related behaviors from T1D diagnosis to the end of young adulthood, incorporating key covariates such as sex, HbA_1c, and factors related to structural and social determinants of health.

In conclusion, in the current systematic review and meta-analysis we found that suicidal ideation and suicide attempts are prevalent in AYA with T1D. Yet, there is no evidence that the rates are higher in youth with T1D than in those without. More research is needed to better quantify the prevalence of suicide deaths in this population, especially in light of the frequent use of insulin and other methods related to diabetes care in suicide attempts. Furthermore, for those with T1D a more comprehensive examination is needed of the evolution in time of suicide-related behaviors within the context of other social determinants of health, during the adolescent and young adulthood developmental period. Such insight will help with better tailoring prevention and intervention efforts for this population in the future.

Acknowledgments. The authors are grateful for the contribution of Geneviève Gore, MLIS, medical science librarian at the Schulich Library of Physical Sciences, Life Sciences, and Engineering, McGill University, Montreal, Canada, who worked with the authors to develop and implement the search strategy.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. O.R.-C., E.M., M.N., and P.L. designed the protocol of the study. O.R. and A.S.G. designed the search strategy and performed the search. O.R.-C. and A.S.G. performed screening, quality assessment, data collection, and data synthesis. O.R.-C., E.M., J.K., M.N., and P.L. contributed to interpretation of data in the manuscript. All authors contributed to writing of the manuscript.

Prior Presentation. Parts of this study were presented in abstract form at the Diabetes Canada/Canadian Society of Endocrinology and Metabolism Professional Conference during Vascular 2023, Montreal, Quebec, Canada, 25–29 October 2023.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Stephanie L. Fitzpatrick.