Authors: Nicole Bungaro, Rossella Mattea Quinto, Lorenzo Scoma, Laura Dattilo, Claudio Imperatori, Marco Innamorati, Federica Genova, Michela Balsamo, Leonardo Carlucci
Categories: Review, Suicide risk, Ideation, Cognitive impairment, Older adults, Systematic
Source: Annals of General Psychiatry
Authors: Nicole Bungaro, Rossella Mattea Quinto, Lorenzo Scoma, Laura Dattilo, Claudio Imperatori, Marco Innamorati, Federica Genova, Michela Balsamo, Leonardo Carlucci
Late-life suicide represents a major public health issue, with older adults showing the highest lethality and completion rates. Mild Cognitive Impairment (MCI), a critical stage of increased psychological vulnerability, may represent an acute phase of suicide vulnerability. This systematic review aimed to synthesize existing evidence on suicide risk among older adults with MCI, examining the prevalence of suicide ideation and behaviors, as well as associated factors and therapeutic outcomes.
Following PRISMA guidelines, four databases (PubMed, Scopus, MEDLINE, PsycINFO) were searched. Eligible studies included participants aged ≥ 65 years, presence of MCI, and an assessment of suicide risk. Methodological quality was appraised using JBI Critical Appraisal Checklists.
Seventeen records published between 2002 and 2024 met inclusion criteria. Across studies, individuals with MCI consistently reported higher rates of suicide ideation compared both with cognitively intact controls and older adults showing more severe cognitive impairment. Prevalence estimates ranged from 0.5% in a large community sample to as high as 84% in a clinical cohort. Depressive symptoms, particularly late-life depression, emerged as the strongest correlate, along with social isolation, female gender, younger age, being single, and multimorbidity. Across three randomized controlled trials, psychotherapeutic interventions and supportive care conditions were associated with reductions in suicide ideation, likely mediated by improvements in emotional regulation and depression.
MCI may represent a particularly vulnerable stage of cognitive decline, marked by preserved insight, loss of autonomy, and heightened existential distress. Standardizing diagnostic criteria and suicidality measures, and implementing tailored psychotherapeutic approaches, are essential steps toward prevention and early interventions in this high-risk population.
The online version contains supplementary material available at 10.1186/s12991-026-00639-4.
Suicide risk in late life represents a major public health concern, presenting unique clinical and psychosocial challenges compared to other age groups [50]. Although suicide rates vary across regions, older adults consistently exhibit disproportionately high lethality and completion rates [8]. Distinctive features of late-life suicide risk include higher prevalence of medical comorbidities, greater social isolation, and less frequent communication of suicide intent compared with younger individuals or other vulnerable populations such as those with chronic medical illnesses [6, 18]. Moreover, recent reviews have highlighted that older adults tend to employ more lethal methods in suicide attempts, which contributes to higher fatality rates relative to adolescents and middle-aged adults (e.g., [3]). Understanding the epidemiology of suicide risk in late life requires careful consideration of age-specific vulnerabilities as well as their broader impact on healthcare and social systems [11].
Mild Cognitive Impairment (MCI) is a neurocognitive condition characterized by measurable cognitive decline that exceeds what is expected for the person’s age and education, yet it does not significantly interfere with daily functioning. According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5-TR), MCI corresponds to Mild Neurocognitive Disorder, defined by the evidence of modest cognitive decline in one or more domains (e.g., memory, executive function, attention, language) without substantial impairment in independence [1]. Similarly, the International Classification of Diseases (ICD-11) identifies Mild Neurocognitive Disorder as a cognitive decline that is not severe enough to meet the criteria for dementia [53]. Clinically, MCI is often diagnosed according to the Petersen et al.’s [40, 41] criteria, which include subjective cognitive complaints, objective impairment in one or more cognitive domains, largely preserved functional abilities, and absence of dementia. Prevalence estimates suggest that 10–20% of older adults are affected by MCI, with the incidence increasing with age [22].
Beyond its cognitive aspects, MCI has important implications for the psychological well-being and autonomy of older adults. Individuals with MCI frequently report loss of self-efficacy, reduced independence, and heightened emotional distress [30, 46]. Unlike those with dementia, individuals with MCI typically retain awareness of their cognitive decline, which may exacerbate feelings of helplessness, shame, and perceived burden on others [43]. Insight into one’s emerging deficit may therefore contribute to increased psychological suffering and vulnerability to suicide risk [47].
While a growing body of research has examined suicide risk in dementia and major neurocognitive disorders (e.g., [3, 10, 15]), this area remains relatively understudied. Even less attention, however, has been given to intermediate conditions such as MCI [47]. This gap in knowledge is clinically significant for several reasons. First, planning and decision-making strategies, and awareness of decline are preserved in individuals with MCI, possibly heightening existential distress and the risk for self-harm [43]. Second, early stages of cognitive impairment are more amenable to psychological and psychosocial interventions, making timely identification of suicidality crucial for prevention [49]. Finally, studying suicide risk at this stage may illuminate the trajectory of suicide risk across the neurocognitive continuum, from intact cognition to dementia [15].
In the context of MCI, specific risk factors should be considered. These include risk factors associated with cognitive deterioration, including heightened awareness of decline, loss of autonomy, the presence of functional limitations, and the internalization of stigma [15, 48], along with general factors like depression, social disconnection, medical comorbidities, and stressful life events [6, 9]. Conversely, protective factors (e.g., social support, resilience, and effective coping strategies) have been shown to mitigate suicide risk in late life, and they could play an even more important role for individuals with MCI [5, 14, 23].
Despite the growing recognition of this issue, the literature on suicide among older adults with MCI remains limited. While a recent systematic review examined suicidality across cognitive impairment [47], only a small number of studies have directly focused on suicide risk (i.e., suicide ideation and behavior, and pre-suicidal mental states) in older adults with MCI, and no synthesis has addressed this specific population. Therefore, the present review aims to address this gap by systematically summarizing the available evidence on suicidality in older adults with MCI, with particular attention to the prevalence of suicide risk and associated risk and protective factors (e.g., sociodemographic, clinical, psychological, psychosocial, and spiritual).
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA) guidelines [38, 39]. The PRISMA Checklist is provided in the Supplementary Material (Table S1). The protocol was registered with the International Register of Systematic Reviews (PROSPERO; registration CRD420251113550).
Four databases were selected (PubMed, Scopus, MEDLINE, and PsycINFO) and queried using a combination of terms, including MeSH terms and Boolean operators. For Mild Cognitive Impairment, terms identified in Mariani et al.’s review [28] were used, combined with synonyms and related terms for suicide risk and older adults (for full search strings, see the Supplementary Material, Table S2). The search was performed on 9th July 2025, without restrictions on publication date or language. A manual search was also performed using Google Scholar to identify additional relevant articles not retrieved through the biomedical and psychological databases.
Inclusion and exclusion criteria were identified according to the objectives of this review. The inclusion criteria Type of study = peer-reviewed studies reporting detailed results, such as original articles and brief reports; studies that were published and accessible in full text, with authors contacted if not readily available; studies published in any language and at any time.Study design = quantitative studies (i.e., cross-sectional and longitudinal studies), and mixed-method studies if statistical analyses for the variables of interest were performed.Population = studies including only older adults (≥ 65 years old) with a diagnosis of MCI.Intervention/exposure = studies investigating the prevalence of suicide risk in these populations; studies measuring the effect of at least one variable (e.g., sociodemographic, clinical, psychological, psychosocial, and/or spiritual variables) on suicide risk.Outcome = studies examining suicide risk (i.e., ideation and/or behavior) as an outcome.
Conversely, exclusion criteria (a) non-peer-reviewed materials (e.g., grey literature, case reports, study protocols, conference abstracts; master’s or doctoral dissertations); (b) qualitative studies and mixed-method studies that do not report separate statistical analyses for the variables of interest; (c) studies that included participants younger than 65 years old, individuals with a diagnosis other than MCI, and/or individuals with comorbid conditions (e.g., neurological disorders, psychosis or other severe psychiatric conditions); and (d) studies that did not investigate suicide risk as outcome and/or did not investigate suicide ideation or behavior (understood as a behavior based on a declared intention to take one’s own life).
One author (N.B.) performed the database search. Records from each database were imported into Rayyan software [21]. Rayyan identified possible duplicates that were verified and manually removed by comparing titles, authors, journals, and publication years. Two authors (R.M.Q. and L.S.) independently screened records against the eligibility criteria. First, records were screened based on title and abstract; second, full texts of selected records were examined. Possible discrepancies between the reviewers were resolved by a senior author (C.I.). Subsequently, another author (L.D.) conducted a manual search of the references in the retrieved papers to identify any additional relevant studies.
Data were extracted based on the aim of this review. Extracted information authors, year of publication, country, study design and type, sample size, participants’ details (e.g., mean age, clinical characteristics), instruments used to measure cognitive impairment, suicide risk, and relevant results (e.g., prevalence and/or variables significantly associated with suicide risk). Findings were synthesized using a planned narrative approach, according (a) cognitive status definition (formally diagnosed MCI vs. broader cognitive impairment); (b) type of suicide-related outcome assessed (suicide ideation vs. pre-suicidal mental states or suicide behavior/attempt); (c) study setting (community-dwelling, clinical, or institutional samples); and (d) study design, with particular attention to randomized controlled trials.
Risk of bias was assessed using design-specific Joanna Briggs Institute (JBI) Critical Appraisal Checklists [32], selected based on study design. Specifically, the JBI checklist for analytical cross-sectional studies was applied to cross-sectional studies, the JBI checklist for cohort studies to longitudinal observational studies, and the JBI checklist for randomized controlled trials to intervention studies. For longitudinal observational studies with a single cohort, items addressing between-group comparability and exposure assignment (i.e., items #1 and #2) were excluded from the appraisal.
Two authors (R.M.Q and L.S.) independently assessed each study, resolving discrepancies through discussion or consultation with a senior author (L.C.). Checklist items were rated by the authors as + (yes), − (no), ? (unclear), or NA (not applicable). To determine the overall Risk of Bias (RoB) for each study, the number of + responses was counted and classified according to commonly used studies with ≥ 75% of + were considered to have low RoB, studies with 50–74% of + were classified as moderate RoB, and studies with < 50% of + were considered to have high RoB.
Figure 1 illustrates the PRISMA flow diagram of the article selection process. According to the search strategy, a total of 1013 records were retrieved from the databases. After removing duplicates, 854 records were screened based on titles and abstracts. Of these, 65 full-text articles were assessed for eligibility. One full-text could not be retrieved, while 47 studies were excluded for the following qualitative design (n = 1), review article (n = 3), not peer-reviewed study (n = 4), book or letter (n = 2), no assessment of suicide risk (n = 2), absence of specific analysis for the MCI group and/or inclusion of participants with different diseases (n = 26), participants outside the target age range (n = 8), and duplicate (n = 1).Fig. 1Flow diagram of articles selection [38, 39]
The manual search did not identify additional eligible studies. Finally, 17 studies met the inclusion criteria and were included in the review.
Details of included studies (n = 17) are reported in Table 1. Most studies were conducted in the United States (n = 8), followed by Europe (n = 5) and Asia (n = 3). One study [48] analyzed data from the World Bank’s Study on Global Ageing and Adult Health (SAGE), which included participants from China, Ghana, India, Mexico, Russia, and South Africa. The publication years ranged from 2002 to 2024.Table 1Characteristics and main results of included studies (n = 17)Authors (year), countryStudy design and typeSample size and clinical characteristicsAgeM (SD) /Me (range)Suicide risk (measure)MCI definition & measureOther variables (measure)Main resultsRoBArslanoglou et al. (2019) [2]USALongitudinal RCT study (PATH vs. ST-CI groups; 12 weekly sessions)Participants (N = 63) were recruited through community agenciesAll aged ≥ 65 years old, with unipolar nonpsychotic MDD diagnosis;73% were female80.52 (7.7)Suicide**ideation(HDRS-item)Cognitive impairment defined as MMSE ≥ 17; cognitive severity assessed using DRSNo formal MCI diagnosis appliedNegative emotions(HDRS-subscale)Non-emotional symptoms(HDRS-subscale)Social support(DSS Index; MAI)Disability(MAI-IADL; WHODAS-II)Memory(HVLT-R)Dosage of antidepressants(CAD-R)41% of participants reported suicide ideation at baseline. Cognitive impairment ranged from mild to moderate deficits (DRS mean = 118.2 ± 12)MMSE scores did not differ between suicidal and non-suicidal participants (p > 0.05)ModerateBogner et al. (2009) [4]USAObservational cross-sectional studyParticipants (N = 305) were recruited from primary care officesAll aged ≥ 65 years;6.9% had depression;75% were female75.0 (5.8)Suicide ideation(items of CIDI and CES-D)Cognitive functioning assessed MMSE, FAS, HVLT, BTANo formal MCI diagnosis appliedAPOE-ε4**Functional status(SF-36)Depressive symptoms(CIDI; CES-D)37% of participants reported suicide ideationCognitively Impaired-Depressed Group had a 58.2% significantly higher probability of reporting suicide ideation, compared to 80.3% in the Depressed Group and 15% in the Non-Depressed Group (p < 0.05)ModerateFujiwara et al. (2013) [12]JapanObservational cross-sectional studyParticipants (N = 811) were recruited in 9 residential areasAll aged 65–84 years; 35.1% had depression and suicide risk;62% were females73.4 (4.8)Suicide ideation and behavior(5-item depression and suicide screen)MCI defined as MoCA-J < 26,with exclusion of dementia(MMSE < 24)Functional limitation(BADL)Functional capacity(TMIG-IC)Health habits(single items)Social activities(single items)Physical performance(PPT)Patients with MCI significantly reported more suicide ideation compared with those without cognitive impairments (22.6% vs. 12.5%;* p* < 0.001)LowGabryelewicz et al. (2004) [13]PolandCross-sectional observational studyParticipants (N = 102) were recruited through a Neurodegenerative Disorders Clinic70.6% were female; 46.1% had depression70.0 (7.5)Suicide ideation(MADRS-item)MCI defined on Petersen criteria, with GDS^1^ (stage 3) and CDR = 0.5Depression(MADRS)Participants were classified into two 1) MCI-Depressed Group,2) MCI-Non Depressed Group65.9% of MCI-Depressed Group reported suicide ideationHighGustavson et al. (2016) [16]USALongitudinal RCT study (PST vs. ST; 12 weekly sessions)Participants (N = 221) were recruited through clinicsAll aged ≥ 65 years, with nonpsychotic MDD diagnosis;65.5% were female73.0 (7.8)Suicide ideation(HDRS-item)Cognitive impairment defined as Mattis DRS ≤ 33 and SIT ≤ 25, with preserved global cognition(MMSE > 23)No formal MCI diagnosis appliedSociodemographic variables**Depression(SCID-R/DSM-IV)Cognitive burden(WCST)Disability(WHODAS-II)Executive functioning(HVLT)Dosage & frequency of antidepressants(CAD-R)Medical burner(CCI)61% of participants reported suicide ideation at baselineAfter 12 weeks, PST Group had significantly higher reduction in suicide ideation than ST Group (60.4% vs. 44.6%; OR = 0.50; p < 0.05)Patients who were single were less likely to show a reduction in suicide ideation; whereas disabled participants were more likely to show improvement (for both,OR = 1.05; p < 0.05)After 36 weeks, younger participants were less likely to report suicide ideation (OR = 0.57 per decade; p = 0.01)LowHeisel et al. (2002) [17]USACross-sectional observational studyParticipants (N = 90) were recruited from community centers, residents, and hospital wardsAll aged ≥ 65 years; 80% were female81.3 (7.6)Suicide ideation(GSIS; SSI)Cognitive functioning defined as MMSE > 20No formal MCI diagnosis appliedDepression(GDS^2^)Hopelessness(BHS)Social Hopelessness(SHQ)84% of participants reported suicide ideationSuicide ideation was associated with MMSE(r = −0.29; p < 0.01 for GSIS; r = −0.21; p < 0.05 for SSI), mediated by depression and hopelessness (path coefficient = 0.41 [SSI],0.82 [GSIS]; p < 0.0001)ModerateKiosses et al. (2017) [24]USALongitudinal RCT study (PATH vs. ST-CI; 12 weekly sessions)Participants (N = 74), were recruited from an Institute of Geriatric PsychiatryAll aged ≥ 65 years; with unipolar nonpsychotic MDD and cognitive impairment, and without active suicide thoughts withintent or plan;74.3% were female80.9 (7.48)Suicide ideation(MADRS-item)Cognitive impairment defined using Mattis DRS(score ≤ 7), with MMSE < 17 excludedNo formal MCI diagnosis appliedDepression(MADRS)Negative emotions(MADRS-Subscale)Disability(WHODAS-II)43% of participants reported suicide ideationIn both Groups, emotional improvement predicted lower suicide ideation at follow-up(OR = 0.68; 95%CI[0.53–0.86])LowLozupone et al. (2022) [27]ItalyCross-sectional observational studyParticipants (N = 1252) were recruited in HospitalsAll aged ≥ 65 years; 52,8% were female74.3 (6.52)Suicide ideation(C-SSRS Subscale)MCI diagnosis based on DSM-5 criteria, supported by MMSESociodemographic variablesClinical variablesBMI**Depression(SCID-I/DSM-IV-TR)Psychopathological symptoms(SCL-90-R)Social functioning(SDRS)2.3% of participants reported suicide ideationMachine learning approach identified MCI diagnosis among the major contributors to suicide ideationLowMattos et al. (2019) [29]USAEcological Momentary Assessment(14 days)Participants (N = 24) were recruited by telephone interviewThey were half female; half Amyloid-β positive; all had MCI75.4 (5.5)Suicide ideation(PHQ-item)MCI diagnosis based on the Alzheimer’s Disease Research Center (ADRS), supported by MMSEMood disorders(7-item EMA mood questionnaire)No suicidal ideation was observed following amyloid-β results disclosure in participants with MCIThere were not significant differences in mood symptoms between Aβ positive-Group and Aβ negative-Group (p > 0.05)HighMcAvay et al. (2005) [33]USALongitudinal observational study (1-month follow-up)Participants (N = 355) were recruited through medical home-careAll aged ≥ 65 years; 61,7% were female78 (65–100)Suicide ideation(SCID/DSM-IV)Cognitive impairment defined using MMSE orientation and recall items. No formal MCI diagnosis appliedSociodemographic variablesClinical variablesDepression(SCID/DSM-IV)6.5% of patients reported suicide ideationPatients without cognitive impairment were significantly more likely to report suicide ideation (p ≤ 0.10)ModerateMurri et al. (2019) [34]ItalyCross-sectional case–control studyParticipants (N = 86) were recruited from psychiatric unitsAll aged ≥ 65 years, with Late-Life Bipolar Disorders (LLBD), type 1 or 2;66,3% were female68.6 (6.4)Suicide ideation (MADRS-item)MCI diagnosis based on DSM-5 criteria, supported by MoCA and CDRSociodemographic variables**Psychiatric diagnosis(DSM-5 criteria; MINI interview)Mood symptoms(SIMD-R)Psychopathology symptoms severity(MADRS; YMRS)Aggressive behaviorClinical variablesPhysical comorbidities(CIRS)MCI and suicide ideation were not significantly associated (p > 0.05)LowOkamura et al. (2017) [35]JapanCross-sectional observational studyParticipants (N = 50) were recruited from homeless sheltersAll were elderly men (≥ 65 years old)72.8 (5.9)Pre-suicide mental states(MINI-interview)Cognitive impairment defined asMMSE < 24No formal MCI diagnosis appliedMental disorders(MINI-interview)8% of participants were at risk of suicideThere was no significant difference in the mean of MMSE between participants with and without a depressive episode (p > 0.05)HighOsvath et al. (2005) [36]HungaryCross-sectional case–control studyParticipants (N = 214) were recruited from psychiatric hospitalAll aged ≥ 65 years; 70,1% were female75.9 (6.7)Suicide attempt(structured interview)Cognitive impairment defined using MMSE(range: 28–25)No formal MCI diagnosis appliedSociodemographic variablesClinical variablesMental disorders(DSM-IV criteria)29% of participants were suicide attemptersThe Suicide-Group had a significantly lower cognitive impairment (Mean score = 18.6 vs. 22.15; p < 0.01), and reported higher rates of previous suicide attempt than the Control-Group (43.5% vs. 10%; χ^2^ = 31.68;* p* < 0.001)Increasing cognitive impairment was significantly associated with a lower risk of attempting suicide (OR = 0.52; p < 0.001)HighPalmer et al. (2007) [37]SwedenLongitudinal case–control study(3-years follow-up)Participants (N = 232) were recruited from hospitalsAll aged 75–95 yearsNASuicide ideation(CPRS-item)MCI diagnosis defined according to established subjective memory deficits, preserved ADL, no dementia, and objective cognitive impairment based on MMSEMood disorders(CPRS)Memory deterioration(psychiatric examination)Language fluency. Visuospatial functioning, Impairment(cognitive battery)At the baseline, suicide ideation was significantly more frequent in MCI-Group than in the Control-Group (17.6% vs 3.2%; p < 0.01), especially in the MCI- multidomain subgroup (16.7% vs 3.2%; p < 0.01)LowRymo et al. (2023) [45]SwedenCross-sectional case–control studyParticipants (N = 916) were recruited through private householdsAll aged ≥ 65 years old; 75,5% were female74 (5.1)Suicide ideation(Paykel questions)MCI diagnosis defined according to Winblad’s criteria, supported by IQCODE and MMSE. Dementia excluded using DSM-III criteriaDepression(MADRS)Neuropsychiatric symptoms(CPRS)MCI-Group reported more passive/active suicide ideation (16.0% vs. 1.1%; p < 0.001), and more lifetime suicide ideation (35.7% vs. 14.8%; p < 0.001) than the Control-GroupMCI was significantly associated with past-year and lifetime life- weariness (OR = 18.32, 95%CI [2.44–137.75]) and death wishes (OR = 5.30, 95%CI [1.19–23.64])LowSmith et al. (2024)[48]World Bank’s SAGECross-sectional observational studyParticipants (N = 13.623) were recruited through a world bank, included China, Ghana, India, Mexico, Russia, and South Africa countriesAll aged ≥ 65 years; the majority was femaleTo 71.2 (53.2)from 74 (13.8)Suicide ideation(WHOCIDI module)MCI diagnosis defined according to National Institute on Aging-Alzheimer’s Association (NIA-AA) guidelinesSociodemographic variables**Clinical variablesThe prevalence of suicide ideation ranged from 0.5% in China to 6% in IndiaThe prevalence of MCI ranged between 9.7% (Ghana) and 26.4% (China)In all countries, excluding India, the prevalence of suicide ideation was higher in the MCI-Group (95%CI [1.12–2.46])There was a low level of between-country heterogeneity (I^2^ = 0.0%)LowYuan et al. (2022) [51]USALongitudinal case–control study(90-days follow-up)Participants (N = 577.461) were recruited from nursing home residentsAll aged ≥ 65 years, and were newly admittedNASuicide ideation(PHQ-item)Cognitive impairment defined as BIMS and Cognitive Performance Scale. No formal MCI diagnosis appliedSociodemographic variablesClinical variablesDepressive symptoms(PHQ-9)*Palliative Care Index (PCI)*Pain severity and frequency(MDS 3.0 items)Frailty(MDS 3.0 items)At admission, 2.7% of participants reported suicide ideation. Among 38.9% experienced suicide ideation for ≥ 7 days during the last 2 weeks; 29% continued to report suicide ideation at 90 days. Four profiles were 1) Frail with all depressive symptoms (22.8%);2) Frail with some depressive symptoms (32.2%);3) Frail with severe cognitive impairment (22.9%);4) Pre-frail (22%)Profiles 1 and 2 were significantly more likely to report suicide ideation at 90 days than other profiles (aOR = 1.22; 95%CI [1.09–1.35]; aOR = 1.10; 95%CI [0.99–1.21], respectively)ModerateN = number; M = mean; SD = Standard Deviation; Me = median; NA = Not Available; RoB = Risk of Bias; p = p-value; r = correlation index; CI = Confidence Interval; OR = Odd Ratio; aOR = adjusted Odd Ratio
Study designs included cross-sectional observational studies (n = 10), longitudinal studies (n = 4), and randomized controlled trials (RCTs; n = 3). Sample sizes varied widely, ranging from 24 [29] to over 577,000 participants [51], with mean ages spanning from 68.6 [34] to 81.3 years [17]. Across all studies, participants were predominantly female, with proportions ranging from 52.8% [27] to 80% [17]. Exceptions included Mattos et al.’s [29] study, which had a more balanced gender distribution, and Okamura et al. [35], who investigated only men.
Most studies have assessed cognitive impairment using the Mini Mental State Examination (MMSE) or the Clinical Dementia Rating (CDR). Across studies, operationalization of cognitive status varied substantially, with MMSE cut-offs reported either as minimum thresholds (e.g., ≥ 17, > 20, or > 23) or as score ranges (e.g., 28–25). Other studies adopted alternative diagnostic approaches, including DSM-based criteria, the Montreal Cognitive Assessment (MoCA), or composite classifications that grouped cognitively intact individuals with those showing mild cognitive impairment. As summarized in Table 1, only a subset of studies applied formal diagnostic criteria for MCI, while others described broader or mixed levels of cognitive functioning.
Regarding suicide risk, all studies evaluated suicide ideation, typically through specific items of rating scales measuring depression (e.g., HDRS, MADRS, or PHQ), or via structured interviews (e.g., SCID, MINI, C-SSRS). Only one study [36] investigated suicide attempts, while another study examined pre-suicide mental states (i.e., loss of motivation, loss of future expectations, and negative feelings) through the MINI-Interview [35]. Detailed descriptions of suicide-related measures are reported in Table 1.
Overall, Risk of Bias (RoB) assessments were used to contextualize the strength and consistency of the evidence in the narrative synthesis rather than as exclusion criteria. Based on design-specific JBI Checklists, eight of the included studies (47.1%) were rated as having low RoB, six (35.3%) as moderate RoB, and three (17.6%) as high RoB (see Table 1). When stratified by study design, cross-sectional studies (n = 10) showed the greatest variability in methodological quality, with four rated as low RoB, three as moderate, and three as high RoB. Among longitudinal observational studies (n = 4), one study was rated as low RoB, two as moderate, and one as high risk. Randomized controlled trials (n = 3) were rated as having low or moderate RoB. Residual sources of bias in RCTs were primarily related to the impossibility of blinding participants and therapists in psychotherapeutic interventions. The detailed quality ratings are available in the Supplementary Material (Table S3).
As shown in Table 1, individuals with MCI reported a high risk for suicide ideation. Across the included studies, the prevalence of suicide ideation showed wide variability across studies, depending on the clinical setting, severity of cognitive impairment, and the assessment method used in the study. The lowest rates were observed in larger population-based or community samples (e.g., [27, 48, 51]), with suicide ideation prevalence ranging between 0.5 and 6%, whereas clinical and psychiatric samples presented markedly higher rates, particularly when comorbid depression was present [13, 16, 17]. In these settings, suicide ideation prevalence reached 60–84% of participants.
Across studies comparing individuals with and without cognitive impairment, MCI was generally associated with a higher likelihood of suicide risk than cognitively intact controls. Fujiwara et al. [12] found that participants with MCI reported more suicide ideation than those without cognitive impairment (22.6% vs. 12.5%). Similarly, Palmer et al. [37] observed a higher frequency of suicide ideation in the MCI group (17.6%) than in controls (3.2%), especially among those with multidomain MCI. Moreover, Rymo et al. [45] reported that older adults with MCI showed more frequent current (16%) and lifetime (35.7%) suicide ideation compared to who were cognitively intact. Only McAvay et al. [33] found that patients without cognitive impairment (compared to other patients) were significantly more likely to report suicide ideation, while another study suggested an inverse relationship between the presence of cognitive impairment and heightened suicide risk. Osvath et al. [36] reported that increasing cognitive impairment was associated with a lower probability of attempting suicide.
Only a limited number of studies systematically explored risk and protective factors associated with suicide risk beyond cognitive status. In these studies, depressive symptoms, especially late-life depression, consistently emerged as the strongest correlate of suicide ideation [13, 16, 17]. Other factors associated with elevated suicide risk included female gender, younger age, being single, smoking, higher body mass index, multimorbidity, greater psychopathology, APOE-ε4 allele, and age-related hearing loss [16, 27, 34]. In contrast, frailty combined with severe cognitive impairment was linked to lower odds of suicide ideation [51].
Three randomized controlled trials [2, 16, 24] investigated changes in suicide ideation following psychological interventions. Across these studies, the prevalence of suicide risk decreased over time, particularly following Problem Adaptation Therapy (PATH) or Problem-Solving Therapy (PST), with these interventions yielding greater reductions than supportive or standard treatments. Importantly, improvements in emotional regulation and depressive symptoms significantly predicted decreases in suicide ideation [2, 24].
Suicide in late life remains a pressing public health issue, with older adults exhibiting the highest lethality and completion rates among all age groups [8, 50]. Within this demographic, Mild Cognitive Impairment (MCI) represents a particularly vulnerable yet under-examined condition. Despite growing interest in the mental health consequences of neurocognitive decline, research on suicide risk in MCI has remained fragmented and inconsistent, often overshadowed by studies focusing on major neurocognitive disorders or dementia [3, 15].
The present systematic review sought to consolidate the existing evidence on suicide risk among older adults with MCI, examining the prevalence of suicide ideation and behaviors, as well as the clinical, psychological, and sociodemographic correlates of this phenomenon. By integrating data from 17 studies published across two decades, this review aimed to clarify whether MCI constitutes a critical window of vulnerability for suicidality, and to identify potential targets for intervention.
A key finding emerging from this synthesis is that the available literature focuses almost exclusively on suicide ideation, with rare studies investigating suicide behaviors in MCI patients. Only one study [36] reported data on suicide attempts, while another examined pre-suicide mental states, such as hopelessness and loss of motivation [35]. This imbalance highlights a significant research although ideation represents a key precursor to suicide attempts, understanding suicide behaviors and mental states prodromic of suicide risk is essential for effective prevention [19, 25].
Across studies, the prevalence of suicide ideation among individuals with MCI varied widely, ranging from 0.5% in large community samples [27, 48, 51] to over 60% in psychiatric or clinical populations, particularly when comorbid depression was present [13, 16, 17]. Despite this heterogeneity, a converging pattern compared with cognitively intact older adults, those with MCI consistently exhibited higher rates of suicide ideation [4, 12, 37, 45], whereas individuals with more advanced cognitive impairment or dementia often display lower rates of suicide ideation [2, 33], and attempts [36]. The main correlates of suicide risk included depressive symptoms (particularly late-life depression), female gender, being single, multimorbidity, smoking, and social isolation [16, 24, 27, 34]. In contrast, severe frailty and advanced cognitive deterioration appeared to attenuate ideation [51], suggesting a complex, non-linear association between cognitive decline and suicide vulnerability.
This inverse relationship between the degree of cognitive decline and suicide risk constitutes one of the most meaningful insights of this review. The evidence indicates that while intact cognition facilitates self-regulation and problem-solving, the partial preservation of insight in MCI may paradoxically amplify psychological suffering [15]. Individuals at this stage are often acutely aware of their cognitive losses and functional limitations, yet still capable of planning and reflecting on their situation [43, 47]. This combination of preserved awareness and declining autonomy may foster mental pain, helplessness, and perceived burdensomeness, mental states central for suicide risk [19, 50]. By contrast, as cognitive deterioration progresses toward dementia, the loss of self-awareness and executive control may limit both the capacity and intentionality required for suicidal acts [36, 51]. Consequently, MCI could represent a unique stage of elevated psychological vulnerability, one where prevention efforts are both most urgent and potentially most effective.
However, the interpretation of these findings is constrained by significant methodological limitations. First, the overall Risk of Bias (RoB) varied substantially across included studies, reflecting differences in study design, sample characteristics, and outcome assessment. Although design-specific quality appraisal tools were applied, fewer than half of the studies were classified as low RoB, and a non-negligible proportion was rated as high RoB. This variability limits the strength of causal inferences and underscores the need for cautious interpretation of the synthesized evidence.
In addition, significant methodological heterogeneity across studies was observed in the definition and operationalization of MCI. Only a subset of studies applied formal diagnostic criteria for MCI, Petersen [40, 41], Winblad et al. [52] or NIA-AA guidelines [13, 45, 48], DSM criteria [27, 34], or Alzheimer’s Disease Research Center operationalization of MCI [29]. Other studies did not diagnose MCI but included participants with broadly defined cognitive impairment, executive dysfunction, or mixed levels of cognitive functioning, ranging from intact cognition to moderate-severe impairment. As a result, part of the evidence synthesized in this review reflects suicide risk across a continuum of non-severe cognitive impairment rather than strictly defined MCI. This distinction is particularly relevant for studies conducted in institutional settings, where severe cognitive impairment was frequently represented and associated with lower levels of suicide ideation (e.g., [33, 51]). Conversely, studies focusing on clinically or epidemiologically defined MCI consistently reported a higher prevalence of suicide risk (e.g., [12, 45, 48]).
A closely related limitation concerns the widespread use of the Mini Mental State Examination (MMSE) to characterize cognitive status. Substantial evidence indicates that the MMSE has limited sensitivity for detecting MCI [7, 31]. Moreover, across studies, heterogeneous MMSE cut-offs were applied to define cognitive impairment or to exclude dementia, ranging from ≥ 17 [2, 24] to ≥ 23–24 [16, 35], while in other cases MMSE scores were reported descriptively without clear diagnostic thresholds (e.g., [17]). This variability increases the risk of misclassification, whereby individuals with intact cognition may be included in cognitively impaired samples, and individuals meeting criteria for MCI may remain undetected. In contrast, studies employing more sensitive screening tools were better positioned to capture early cognitive decline. For instance, the Montreal Cognitive Assessment (MoCA) has demonstrated greater sensitivity for identifying MCI and may be more appropriate for future research in this field [20, 42].
Accordingly, the study of suicidality in MCI would benefit from the use of standardized diagnostic frameworks (e.g., Petersen/Winblad guidelines, or DSM and ICD criteria), explicit reporting of cognitive score ranges, harmonized cut-offs across instruments, and clear differentiation between MCI, broader cognitive impairment, and dementia. Establishing uniform and culturally adjusted cognitive benchmarks is essential to improve comparability across studies and to isolate more precisely the risk for suicide attributable to MCI. A closely related limitation concerns the measurement of suicide risk across studies. Most investigations assessed suicidality indirectly, through single items embedded in depression scales (e.g., HDRS, MADRS, PHQ-9), rather than using dedicated suicide-specific instruments. Only one study utilized the Columbia–Suicide Severity Rating Scale (C-SSRS; [27]), a gold standard for assessing suicide risk [44]. The reliance on proxy measures introduces construct validity problems, as depression-related items may not capture the full spectrum of suicide risk, such as mental states prodromic of suicide or planning behaviors [19, 25]. These limitations are specifically relevant in cognitively impaired populations, in whom suicide thoughts may be underreported, poorly articulated, or conflated with depressive symptoms [15]. Consequently, variability in suicide measurement may contribute to inconsistent prevalence estimates and potentially attenuated associations between MCI and suicide risk across studies. Future research should prioritize the use of validated, multidimensional instruments specifically designed to assess suicidality, with attention to the frequency, severity, and temporality of suicide thoughts, as well as the presence of pre-suicidal mental states, particularly in older adults with cognitive impairment.
Beyond observational evidence, three randomized controlled trials (RCTs) provided valuable insights into potential therapeutic approaches [2, 16, 24]. These trials evaluated the effects of Problem-Solving Therapy (PST), Problem-Adaptation Therapy (PATH), and Supportive Therapy (ST) on suicide ideation among older adults with depression and MCI. Across studies, PATH and PST interventions yielded more significant reductions in suicide ideation over time than ST intervention, though their mechanisms appeared to differ. PST primarily targeted executive dysfunction and problem-solving deficits [16], whereas PATH emphasized emotional regulation and environmental adaptation [2, 24]. Moreover, improvements in negative emotions and depressive symptoms predicted reductions in suicide ideation, underscoring the role of affective dysregulation as a modifiable mechanism [17]. Although sample sizes were modest, these results demonstrate the feasibility of psychotherapeutic interventions tailored to the cognitive profiles of MCI individuals in line with previous findings (e.g., [26]).
Despite the methodological limitations, several strengths distinguish this review. First, design-specific JBI Checklists [32] were applied, and risk of bias assessment was explicitly reported across heterogeneous study designs, allowing findings to be interpreted in relation to methodological quality rather than treated as equivalent. Second, the majority of studies were published within the past decade, reflecting growing recognition of suicidality in MCI as a contemporary clinical concern rather than a peripheral topic. Third, this review adhered to PRISMA standards [38, 39], ensuring transparent and replicable methodology in study identification, selection, and synthesis.
In conclusion, this review provides the first systematic synthesis of suicide risk in MCI-older adults, revealing that mild cognitive decline is more associated with increased suicide ideation than a severe impairment. These findings reinforce the need to conceptualize MCI as a critical stage of increased psychological vulnerability, characterized by preserved insight, affective dysregulation, and existential distress. Future research should adopt standardized diagnostic and suicidality measures, employ longitudinal designs, and explore biopsychosocial mechanisms, including both risk and protective factors, underlying this vulnerability. Ultimately, improving the recognition and management of suicidality in MCI may bridge the gap between neurocognitive and mental health care in aging populations.
Additional file 1.