Comparative efficacy of group versus individual reminiscence therapy for depression in older adults: A network meta-analysis stratified by severity

Abstract

Objectives

To directly compare the efficacy of individual reminiscence therapy (IRT) and group reminiscence therapy (GRT) for alleviating depressive symptoms in older adults, and to evaluate their effects on secondary outcomes, including anxiety, cognitive function, life satisfaction, and quality of life.

Methods

A systematic review and network meta-analysis (NMA) was conducted. Eleven Chinese and English databases were searched from inception to December 1, 2025. Randomized controlled trials (RCTs) comparing IRT or GRT against a medication control group (MCG), a non-medication control group (NCG), or each other were included. Risk of bias (RoB) was assessed using the RoB 2 tool, and evidence certainty was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. Pairwise and network meta-analyses were performed, with comparisons further stratified by baseline depression severity (mild, moderate, severe). Treatment rankings were based on the surface under the cumulative ranking (SUCRA) statistic. Subgroup analyses were conducted to examine potential effect modifiers, including intervention format, baseline depression severity, control type, intervention duration, session frequency, total sessions, cumulative intervention time, geographical region, setting, and provider qualification.

Results

Twenty RCTs with 1,633 participants were included. For depression, both IRT and GRT were significantly superior to MCG and NCG, with no statistically significant difference between the two modalities overall (SMD = −0.03; 95 %CI: −0.49, 0.42). Severity-stratified analysis revealed that GRT had the highest probability of being optimal for mild depression (SUCRA = 92.5 %), while IRT ranked highest for moderate (SUCRA = 80.6 %) and severe depression (SUCRA = 84.5 %). For secondary outcomes, IRT was superior to GRT for reducing anxiety and improving cognitive function, whereas GRT was more effective for enhancing life satisfaction. Subgroup analyses showed no significant moderating effects for any of the examined variables (all P > 0.05).

Conclusions

Both IRT and GRT are effective for depression in older adults, with no statistically significant difference between them. GRT may be preferred for mild depression, while IRT may be more suitable for moderate-to-severe depression. IRT shows additional benefits for anxiety and cognition, whereas GRT excels in improving life satisfaction. These findings support a personalized, severity-informed approach to selecting RT delivery formats in clinical practice.

What is known?

• •Depression in older adults is a prevalent and serious mental health condition among older adults, associated with substantial personal and societal burden.
• •Reminiscence therapy (RT) is a structured psychosocial intervention widely used to improve psychological well-being in older populations.
• •Previous meta-analyses have largely evaluated individual reminiscence therapy (IRT) and group reminiscence therapy (GRT) as separate interventions, without direct comparative evidence of their relative efficacy.

What is new?

• •IRT and GRT are equally effective for reducing depressive symptoms in older adults, with no significant difference between the two formats.
• •GRT may be preferable for mild depression, while IRT appears more suitable for moderate-to-severe depression based on severity-stratified analysis.
• •IRT is superior at reducing anxiety and improving cognitive function, whereas GRT is more effective at enhancing life satisfaction.

1. Introduction

The accelerated aging of the global population represents a critical public health challenge [1]. Projections indicate that by 2050, older adults will number approximately 2 billion, constituting around 21 % of the global population [2]. Accompanying this shift is a growing burden of depression, a prevalent mental disorder in older adults that accounted for 10.6 % of disability-adjusted life years in 2021 [[3], [4], [5], [6]]. Beyond psychological distress, depression is robustly associated with reduced quality of life, accelerated cognitive and functional decline, worse outcomes for comorbid illnesses, and elevated mortality, incurring substantial personal and societal costs [[7], [8], [9]]. Consequently, identifying effective, scalable interventions for depressive symptoms in older adults is an imperative priority. Pharmacotherapy remains first-line [10,11]; however, age-related declines in hepatic metabolism and renal clearance increase the risk of adverse effects, compounded by high rates of multimorbidity and polypharmacy [12,13]. These limitations underscore the need for effective non-pharmacological alternatives, making psychotherapy an essential component of comprehensive treatment strategies [14,15].

Several psychotherapeutic approaches have proven effective for depressive symptoms in older adults, including cognitive behavioral therapy and problem-solving therapy [[15], [16], [17], [18]]. However, these often require specialized training and extended, structured delivery and may not fully accommodate age-related cognitive changes, sensory impairments, or a preference for emotionally meaningful, low-stress engagement [[19], [20], [21]]. In this context, reminiscence therapy (RT)—a structured psychosocial intervention conceptualized by Butler in the 1960s that improves well-being through guided recall and reappraisal of past experiences—has emerged as particularly promising for depressed older adults [[22], [23], [24]]. RT’s promise rests on several distinctive features: it is grounded in life-review theory, which links later-life reflection on past experiences to ego integrity and mitigation of depressive symptoms [24,25]; it capitalizes on the relative preservation of remote memory in aging, offering a low-burden, engaging, resource-efficient alternative to more demanding cognitive therapies [26,27]; and it possesses procedural flexibility and cross-cultural applicability, allowing adaptation across diverse settings, delivery formats, and provider skill levels, thereby enhancing scalability [28,29]. Through guided reexamination of significant life events, emotions, and relationships, RT facilitates meaning-making, strengthens self-continuity, and improves emotional states, directly targeting core mechanisms implicated in depression [[30], [31], [32]].

RT is typically delivered in two primary modalities: individual reminiscence therapy (IRT) and group reminiscence therapy (GRT), which differ in their underlying therapeutic mechanisms [[33], [34], [35]]. IRT relies on a strong, confidential therapeutic alliance, enabling therapists to guide patients in restructuring overgeneralized and negatively biased autobiographical memories in a private, focused setting [33,34]. GRT, in contrast, operates through shared narrative, peer validation, and group cohesion, utilizing collective reminiscence and social reinforcement to reduce isolation, foster universality, and enhance interpersonal support [33,35]. These distinct pathways suggest that modality choice may have important implications for treatment outcomes depending on patient characteristics and clinical context.

Although an expanding body of evidence [[36], [37], [38]] supports the efficacy of both IRT and GRT for depressive symptoms in older adults, their relative effectiveness remains unclear, as previous meta-analyses have largely treated them as separate interventions without direct head-to-head comparisons, leaving nursing practitioners without evidence-based guidance to support clinical decision-making regarding the choice of delivery format in geriatric nursing practice.

Baseline depression severity may profoundly influence RT efficacy [[39], [40], [41], [42]]. Individuals with severe depression often experience marked social withdrawal and cognitive impairment that can hinder active participation in group dynamics [39,40]. In contrast, those with mild depression may benefit more from the social normalization and peer support inherent to group-based approaches [41,42]. Despite these plausible differential effects, existing studies have not stratified results by severity, potentially obscuring crucial modality-by-severity interactions and limiting applicability across the full spectrum of depressive illness in older adults.

To address this gap, this network meta-analysis (NMA) directly compares GRT and IRT in older adults, stratified by mild, moderate, and severe depression. The analysis also evaluates secondary outcomes—such as anxiety, cognitive function, life satisfaction, and quality of life—to capture the broader therapeutic impact of each format. Our findings aim to provide nurses and other healthcare practitioners with nuanced, evidence-based guidance for selecting the most appropriate delivery format, supporting a personalized, severity-informed approach to RT in geriatric nursing practice.

2. Methods

This systematic review and NMA was conducted in strict accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. It was prospectively registered in the PROSPERO (CRD420250654250) [43,44].

2.1. Search strategy

A systematic search was conducted from database inception to December 1, 2025, across seven international databases (Web of Science, PubMed, Scopus, the Cochrane Library, APA PsycInfo, Embase, and CINAHL) and four Chinese databases (CNKI, VIP, Wanfang, and SinoMed). Grey literature was identified by reviewing reference lists of included studies and searching OpenGrey and ClinicalTrials.gov. Key search terms included “depression,” “reminiscence therapy,” and “aged.” The complete search strategy for each database is available in Appendix A.

2.2. Eligibility criteria

Inclusion criteria were as follows: 1) randomized controlled trials (RCTs); 2) older adults (age ≥60 years); 3) patients diagnosed with depression ranging from mild to severe, based on any validated standardized assessment scales, such as Geriatric Depression Scale (GDS), Hamilton Depression Rating Scale (HAMD), Self-Rating Depression Scale (SDS), the Beck Depression Inventory (BDI), and the Center for Epidemiologic Studies Depression Scale (CES-D); 4) intervention group receiving either IRT or GRT; 5) control group receiving pharmacological treatment defined as standard antidepressant therapy (medication control group, MCG), non-pharmacological usual care defined as routine care, waitlist, or health education (non-medication control group, NCG), or the alternative format of RT (e.g., GRT when the intervention group receives IRT, or IRT when the intervention group receives GRT); and 6) peer-reviewed journal articles.

Exclusion criteria were as follows: 1) RT was part of a multicomponent intervention were excluded unless the control group received all the same non-RT components; 2) no specific randomization method declared or quasi-random methods used; 3) Alzheimer’s disease, dementia, or severe cognitive impairment, as these may hinder engagement with RT or compromise outcome validity; 4) high risk of bias per Cochrane Risk of Bias tool 2 (RoB 2); 5) conference abstract and thesis; 6) retracted studies.

2.3. Study selection and data extraction

Two reviewers (S. Wang and S. Chen) independently removed duplicates using EndNote 20, screened titles/abstracts against eligibility criteria, and retrieved potentially relevant articles for full-text review. They then extracted data into a pre-designed Excel spreadsheet, including: publication details (author, year, country); baseline depression severity; demographics; group names and sample sizes; outcome measures (depression, anxiety, cognition, life satisfaction, quality of life); depression inclusion criteria (scale and cutoffs); intervention characteristics (modality, duration, session length/frequency, total sessions, cumulative time); setting; provider qualifications; and assessment time points. Discrepancies were resolved by consensus or by a third reviewer (C. Wu).

2.4. Risk of bias assessment

Two reviewers (S. Wang and S. Chen) independently assessed methodological quality using the RoB 2 [45], which evaluates five domains: randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of reported results. Overall risk of bias was classified as “low risk” (all domains low), “some concerns” (at least one domain with some concerns), or “high risk” (at least one domain high risk). Discrepancies were resolved through discussion or by consultation with a third reviewer (C. Wu).

2.5. Evaluation of the certainty of evidence

Evidence certainty for all outcomes was assessed using Grading of Recommendations Assessment, Development and Evaluation (GRADE) [46] across five domains (risk of bias, inconsistency, indirectness, imprecision, publication bias). Initial certainty was set too high as all included studies were RCTs. Downgrading was applied for considerable heterogeneity (I² ≥ 50 %) or imprecision (n ≤ 5); upgrading was applied for large effect sizes [standardized mean difference (SMD) or mean difference (MD) ≥ 0.8]. Certainty was graded as high, moderate, low, or very low. Two reviewers (S. Wang and S. Chen) independently conducted the assessment; disagreements were resolved through discussion or by consultation with a third reviewer (C. Wu).

2.6. Data analysis

The core analytical dataset was the mean change from baseline and its standard deviation (SD). When not reported, the mean change was calculated as post-intervention minus baseline, and the SD of the change score was imputed using the Cochrane-recommended formula [47] assuming a correlation coefficient of 0.5 [48], a conservative and widely accepted estimate.

Pairwise meta-analyses were performed using RevMan 5.4 with a random-effects model (if ≥ 5 studies) or a fixed-effect model (if < 5 studies), because a random-effects model requires sufficient studies to estimate the between-study variance reliably; with fewer than 5 studies, the estimation becomes unstable, making a fixed-effect model more appropriate. MD was used for outcomes assessed with the same scale; SMD otherwise. Analyses included only comparisons with at least two studies.

NMA was conducted in Stata 18.0 using the mvmeta package under a frequentist framework. Global inconsistency was evaluated using the design-by-treatment interaction model; local inconsistency was evaluated using node-splitting analysis in closed loops. Transitivity was validated by examining the distribution of key effect modifiers (baseline severity, duration, frequency, total sessions, cumulative time, region, setting, and provider qualification) across direct comparisons using Chi-square tests. A random-effects model estimated SMDs for each intervention, with treatments ranked by surface under the cumulative ranking (SUCRA) statistic (0–100 %). Effect estimates are presented as SMDs with 95 %CIs; statistical significance was defined as P < 0.05 (two-sided). Network structures were plotted with nodes indicating interventions, line thickness proportional to the number of studies, and node size corresponding to sample size. Result stability was tested using leave-one-out sensitivity analysis. Publication bias was assessed using funnel plots for comparisons with ≥10 studies.

Subgroup analyses examined potential effect modifiers: RT format, baseline depression severity, control type, intervention duration, session frequency, total sessions, cumulative intervention time, geographical region, intervention setting, and provider qualification.

3. Results

3.1. Study selection outcomes

The study selection process followed PRISMA guidelines (Appendix B). From 11 databases and grey literature sources, 3,417 records were identified. After removing 1,532 duplicates, 1,885 records underwent title and abstract screening, of which 1,734 were excluded. The remaining 151 full-text articles were assessed for eligibility; 131 were excluded for reasons detailed in Appendix C. Ultimately, 20 studies met all inclusion criteria and were included in the final analysis.

3.2. Characteristics of included studies

The NMA ultimately included 20 RCTs [[49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]] published between 2004 and 2025, comprising 1,633 participants. Study characteristics in Appendix D and E. Eleven studies used GRT [49,50,54,55,[58], [59], [60],63,65,66,68], eight used IRT [[51], [52], [53],56,57,61,64,67], and one compared both [62]. Seventeen studies [[49], [50], [51], [52], [53], [54], [55],57,[59], [60], [61],[63], [64], [65], [66], [67], [68]] employed NCG, while two studies [56,58] adopted MCG. Most studies were from China (n = 15) [[54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]], with others from Iran (n = 2) [49,50], Spain (n = 1) [53], Switzerland (n = 1) [52], and Singapore (n = 1) [51]. Intervention duration ranged 4–24 weeks; session frequency was once (n = 12) [[52], [53], [54], [55],[59], [60], [61], [62], [63],[65], [66], [67]], twice (n = 2) [50,68], or three times (n = 2) [56,64] weekly; total sessions ranged 4–24; cumulative time 188–900 min. Settings included hospitals (n = 9) [52,56,58,60,[63], [64], [65],67,68], nursing homes (n = 4) [50,54,57,62], community (n = 4) [49,55,59,66], and home (n = 2) [51,61]. Providers were qualified mental health professionals (n = 5) [50,53,58,60,66], trained personnel (n = 9) [52,55,57,59,[61], [62], [63],65,68], or unspecified (n = 6) [49,51,54,56,64,67]. Participants were classified as mild (n = 12) [[49], [50], [51], [52],54,55,57,[61], [62], [63],65,66], moderate (n = 4) [59,60,64,67], or severe (n = 4) [53,56,58,68] depression. Depression severity was classified based on the standardized cut-off scores of the assessment scales used in each trial (Appendix D) [[69], [70], [71], [72], [73], [74]], which were also applied to define study inclusion criteria: the GDS-30 (n = 7) [54,55,57,61,62,65,66], the GDS-15 (n = 3) [[49], [50], [51]], the HAMD-17 (n = 3) [56,58,68], the HAMD-24 (n = 3) [59,60,64], the SDS (n = 2) [63,67], the BDI-II (n = 1) [52], and the CES-D (n = 1) [53]. Follow-up assessments were conducted at 1 month (n = 3) [49,50,54] or 3 months (n = 2) [52,54].

3.3. Risk of bias assessment outcomes

All studies were rated as some concerns overall (Appendix F). Regarding the randomization process, all studies described random sequence generation and reported balanced baseline characteristics. Two studies explicitly described allocation concealment and were rated low risk [50,54]; the remainder raised some concerns due to insufficient information [49,[51], [52], [53],[55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]]. Given the psychological nature of the intervention, blinding of participants and personnel was not feasible; thus, all studies raised concerns about deviations from the intended interventions. For missing outcome data, 16 studies reported low dropout (<5 %) or used intention-to-treat analysis and were rated low risk [[49], [50], [51],[54], [55], [56],[58], [59], [60], [61], [62], [63], [64], [65],67,68]; four studies raised some concerns due to insufficient detail [52,53,57,66]. Outcome assessor blinding was implemented and rated low risk in four studies [53,54,57,60]; the others raised some concerns due to unclear blinding procedures [[49], [50], [51], [52],55,56,58,59,[61], [62], [63], [64], [65], [66], [67], [68]]. Two studies, which were prospectively registered and fully reported prespecified outcomes, were rated low risk for selective reporting [49,54]; the remaining studies raised some concerns due to lack of registration [[50], [51], [52], [53],[55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]].

3.4. Assessment of evidence certainty

The certainty of evidence was assessed using GRADE. As all evidence came from RCTs, the initial certainty was high. For depression, direct comparisons were of moderate certainty; indirect comparisons were of low certainty due to imprecision. For anxiety, most indirect comparisons had moderate certainty, except GRT vs NCG (low certainty). For cognition and life satisfaction, most comparisons had moderate certainty. For quality of life, indirect comparisons were of moderate certainty, while direct evidence (IRT vs NCG) was of low certainty. Detailed GRADE ratings are provided in Appendices G and H.

3.5. Consistency assessment

Consistency between direct and indirect evidence was examined across all closed loops in the network. Global inconsistency assessed via the design-by-treatment interaction model indicated no significant inconsistency for the primary outcome (χ² = 1.57, P = 0.457) or for the mild (χ² = 2.59, P = 0.108) and severe depression subgroups (χ² = 2.91, P = 0.088). Life Satisfaction also showed no significant inconsistency (χ² = 0.81, P = 0.367). Node-splitting analysis revealed no locally significant discrepancies between direct and indirect estimates for any treatment contrast in the full network or subgroup analyses (all P > 0.05, Appendix I).

3.6. Transitivity assessment

To validate the transitivity assumption, we assessed the distributional balance of eight potential effect modifiers (depression severity, intervention duration, session frequency, total sessions, cumulative intervention time, geographical region, setting, and provider qualification) across direct comparison pairs (GRT vs. NCG, IRT vs. NCG, GRT vs. MCG, IRT vs. MCG, GRT vs. IRT) using chi-square tests. No statistically significant differences were found for any modifier (all P > 0.05, Appendix J).

3.7. Primary outcome

3.7.1. Short-term efficacy on depressive symptoms

Short-term efficacy was measured immediately post-intervention. Pairwise meta-analysis showed that RT significantly reduced depressive symptoms compared with control conditions (SMD = −1.30; 95 %CI: −1.53, −1.08; Appendix K). Comparison network was shown in Fig. 1A. NMA confirmed that both IRT (SMD = −1.31; 95 %CI: −2.06, −0.56) and GRT (SMD = −1.28; 95 %CI: −2.03, −0.53) were superior to MCG, and similarly superior to NCG (IRT: SMD = −1.33; 95 %CI: −1.71, −0.94; GRT: SMD = −1.29; 95 %CI: −1.62, −0.97), with no significant difference between IRT and GRT (SMD = −0.03; 95 %CI: −0.49, 0.42), consistent with direct pairwise comparisons (Fig. 2). SUCRA rankings placed IRT first (85.2 %), followed by GRT (81.5 %), MCG (16.8 %), and NCG (16.6 %) (Appendix L). Leave-one-out sensitivity analysis confirmed robustness (SMDs range −1.34 to −1.21), and funnel plot inspection showed no substantial publication bias (Appendix M).

Fig. 1.

Network meta-analysis evidence network of depression across different severity levels.

Note: A: all levels of depression. B: mild depression. C: moderate depression. D: severe depression. GRT = group reminiscence therapy. IRT = individual reminiscence therapy. MCG = medication control group. NCG = non-medication control group.

Fig. 2.

League table of pairwise and network meta-analysis results for depressive symptoms.

Note: Results are presented in a matrix format. Below the diagonal (lower triangle) are effect sizes (SMD) with 95 %CIs from the network meta-analysis; above the diagonal (upper triangle) are effect sizes (SMD/MD) with 95 %CIs from the pairwise meta-analysis. GRT = group reminiscence therapy. IRT = individual reminiscence therapy. NCG = non-medication control group. MCG = medication control group.

3.7.2. Long-term efficacy on depressive symptoms

Long-term efficacy was defined as outcomes measured at one month post-intervention. The meta-analysis of three studies, all of which evaluated GRT, showed that GRT maintained a significant reduction in depressive symptoms compared with controls at one month (SMD = −1.36; 95 %CI: −1.79, −0.93; Appendix N). Leave-one-out sensitivity analysis supported robustness (SMDs range −1.64 to −1.02).

3.7.3. Efficacy according to depression severity

For mild depression, pairwise meta-analysis showed RT significantly reduced depressive symptoms versus controls (SMD = −1.10; 95 %CI: −1.28, −0.93; Appendix O). The comparison network was shown in Fig. 1B. NMA indicated both GRT (SMD = −1.16; 95 %CI: −1.39, −0.93) and IRT (SMD = −0.97; 95 %CI: −1.29, −0.65) were superior to NCG, with no significant difference between them (SMD = −0.19; 95 %CI: −0.54, 0.17), consistent with direct comparisons (Fig. 2). SUCRA rankings placed GRT first (92.5 %), followed by IRT (57.5 %) and NCG (0) (Appendix L). Sensitivity analysis confirmed robustness (SMDs range −1.33 to −1.22), and funnel plots showed no publication bias (Appendix P).

For moderate depression, RT significantly reduced depressive symptoms versus controls (SMD = −1.51; 95 %CI: −1.75, −1.28; Appendix O). The comparison network was shown in Fig. 1C. NMA showed IRT was superior to NCG (SMD = −1.81; 95 %CI: −3.29, −0.33), with no significant difference between IRT and GRT (SMD = −0.34; 95 %CI: −2.44, 1.75), inconsistent with direct comparisons (Fig. 2). SUCRA rankings favored IRT (80.6 %), followed by GRT (67.6 %) and NCG (1.8 %) (Appendix L). Sensitivity analysis supported stability (SMDs range −1.89 to −1.43).

For severe depression, RT significantly alleviated depressive symptoms versus controls (SMD = −1.43; 95 %CI: −1.65, −1.21; Appendix O). The comparison network was shown in Fig. 1D. NMA showed both IRT (SMD = −1.32; 95 %CI: −2.02, −0.62) and GRT (SMD = −1.28; 95 %CI: −1.98, −0.58) were superior to MCG, and similarly superior to NCG (IRT: SMD = −1.72; 95 %CI: −2.53, −0.91; GRT: SMD = −1.68; 95 %CI: −2.42, −0.94), with no significant difference between IRT and GRT (SMD = −0.04; 95 %CI: −0.90, 0.82) or between MCG and NCG (SMD = −0.40; 95 %CI: −1.25, 0.45; Fig. 2). SUCRA rankings placed IRT highest (84.5 %), followed by GRT (82.2 %), MCG (27.4 %), and NCG (6.0 %) (Appendix L). Sensitivity analysis confirmed robustness (SMDs range −1.60 to −1.32).

3.8. Secondary outcomes

3.8.1. Anxiety

Pairwise meta-analysis showed that RT significantly reduced anxiety compared with controls (SMD = −1.39; 95 %CI: −1.67, −1.11; Appendix Q). The comparison network is shown in Appendix R. Both IRT (SMD = −2.86; 95 %CI: –3.91, –1.81) and GRT (SMD = –0.82; 95 %CI: −1.18, −0.46) were superior to MCG. IRT also outperformed NCG (SMD = −2.82; 95 %CI: −3.34, −2.30) and was statistically superior to GRT (SMD = −2.04; 95 %CI: −3.02, −1.05; Appendix S). SUCRA rankings placed IRT highest (100.0 %), followed by GRT (65.6 %), NCG (19.1 %), and MCG (15.4 %) (Appendix T). Leave-one-out sensitivity analysis confirmed robustness (SMDs range −2.25 to −0.82).

3.8.2. Cognition

Pairwise meta-analysis showed that RT significantly improved cognitive function compared with controls (SMD = 1.90; 95 %CI: 1.54, 2.26; Appendix U). Comparison network was shown in Appendix R. Both IRT (SMD = 4.01; 95 %CI: 3.37, 4.66) and GRT (SMD = 0.92; 95 %CI: 0.48, 1.35) were superior to NCG, with IRT significantly outperforming GRT (SMD = 3.10; 95 %CI: 2.32, 3.87; Appendix S). SUCRA rankings placed IRT highest (100.0 %), followed by GRT (50.0 %) and NCG (0) (Appendix T). Leave-one-out sensitivity analysis confirmed robustness (SMDs range 0.92 to 4.01).

3.8.3. Life satisfaction

Pairwise meta-analysis showed that RT significantly improved life satisfaction compared with controls (MD = 5.26; 95 %CI: 3.48, 7.04; Appendix V). Comparison network was shown in Appendix R. Both GRT (SMD = 1.53; 95 %CI: 1.05, 2.02) and IRT (SMD = 0.58; 95 %CI: 0.11, 1.05) were superior to NCG, with GRT significantly outperforming IRT (SMD = 0.95; 95 %CI: 0.47, 1.43), consistent with direct comparisons (Appendix S). SUCRA rankings placed GRT highest (100.0 %), followed by IRT (49.7 %) and NCG (0.3 %) (Appendix T). Leave-one-out sensitivity analysis confirmed robustness (SMDs range 3.02 to 9.79).

3.8.4. Quality of life

Pairwise meta-analysis showed that RT significantly improved quality of life compared with controls (SMD = 1.48; 95 %CI: 1.26, 1.70; Appendix W). The comparison network is shown in Appendix R. The direct comparison confirmed that IRT was superior to NCG (SMD = 1.58; 95 %CI: 1.27, 1.90). Compared to MCG, IRT also led to a significant improvement (SMD = 1.63; 95 %CI: 0.54, 2.71; Appendix S). SUCRA rankings placed IRT highest (81.1 %), followed by GRT (78.6 %), MCG (24.9 %), and NCG (15.3 %) (Appendix T). Leave-one-out sensitivity analysis confirmed robustness (SMDs range 1.28 to 1.62).

3.9. Subgroup analyses

Overall, RT was significantly superior to the control conditions in alleviating depressive symptoms, with substantial heterogeneity (I² = 76 %). Subgroup analyses (Appendix O) revealed no significant differences in efficacy between GRT and IRT (P = 0.51). RT was effective across all depression severity levels (mild: SMD = −1.10; moderate: −1.51; severe: −1.43), with no significant between-subgroup difference (P = 0.11). Efficacy did not differ by control type (MCG vs. NCG, P = 0.97), intervention duration (P = 0.96), weekly frequency (P = 0.33), total sessions (P = 0.81), cumulative time (P = 0.72), geographic region (P = 0.93), intervention setting (P = 0.29), or provider qualification (P = 0.33).

4. Discussion

4.1. Comparative efficacy of group versus individual reminiscence therapy on depression

This NMA revealed no significant difference between IRT and GRT in ameliorating depressive symptoms in older adults. This equivalent efficacy may be attributed to their shared core therapeutic mechanisms. Both modalities operate through structured life review, emotional reprocessing, and narrative reconstruction, which collectively enhance self-continuity, promote meaning-making, and reduce negative cognitive biases [31,41]. While IRT facilitates deeper emotional exploration within a confidential one-on-one therapeutic alliance, GRT leverages peer validation, social comparison, and group cohesion to achieve similar antidepressant effects through distinct social pathways [[33], [34], [35]]. Thus, despite differences in delivery format, the two approaches converge on the same fundamental mechanisms of change, resulting in comparable clinical outcomes. This finding suggests that format selection may be less critical for overall antidepressant efficacy than previously assumed, shifting the focus toward patient-specific factors such as depression severity.

4.2. Moderating effect of depression severity on reminiscence therapy efficacy

For mild depression, GRT ranked highest (SUCRA = 92.5 %), likely because individuals with mild depression retain greater social engagement capacity, and social participation itself benefits mental well-being [75]. GRT provides structured social interaction, peer validation, and collective narrative sharing, reducing loneliness while enhancing happiness and self-efficacy through social connection and belonging [30,35,41,42].

Conversely, IRT ranked as more effective for moderate (SUCRA = 80.6 %) and severe (SUCRA = 84.5 %) depression. As depression severity increases, patients experience greater emotional distress, reduced energy, and social withdrawal, heightening the need for an intensive, private, supportive therapeutic relationship [39,40,76]. IRT’s individualized nature enables focused attention, emotional containment, and cognitive restructuring in a confidential setting, better suited to address the complex psychological state of moderate-to-severe depression [77]. Additionally, confidentiality provides a secure space for free expression without fear of judgment [78]. These findings establish baseline depression severity as a key moderator of relative efficacy between IRT and GRT. The two modalities should not be considered interchangeable; rather, clinical practice and future research should explore differentiated application pathways based on symptomatic profiles to achieve precise intervention matching.

4.3. Subgroup analyses on treatment moderators

Subgroup analyses showed no significant moderators of RT’s effect on depressive symptoms. For intervention parameters (format, duration, frequency, total sessions, cumulative time), the absence of significant moderation suggests that RT may have a low dose–response threshold; even brief or low-frequency protocols appear sufficient to achieve clinical benefits. For contextual factors (region, setting, provider qualification), the null findings support RT’s adaptability across diverse healthcare settings and resource levels. For clinical characteristics (baseline severity, control type), the lack of moderation indicates that RT’s superiority over controls is robust across severity levels. Our findings partially align with Lin et al. [37], who also reported no moderation by format or session number. Our study further extends their work by examining additional moderators (duration, frequency, cumulative time, region, setting, and provider qualification) and conducting a severity-stratified NMA.

4.4. Differential effects on secondary outcomes

IRT was significantly more effective than GRT in reducing anxiety and improving cognitive function. The private, one-on-one setting of IRT allows patients to express anxious thoughts without social pressure and enables therapists to tailor cognitive engagement strategies to individual needs [77,79]. Conversely, GRT outperformed IRT in enhancing life satisfaction, likely due to group-specific factors such as peer validation, social comparison, and group cohesion, which reduce loneliness and foster social belonging [80,81]. For quality of life, no significant difference was found between the two formats, suggesting that both similarly improve overall well-being through shared core mechanisms of life review and meaning-making. These differential benefits suggest that format selection should consider not only depression severity but also specific secondary treatment goals.

4.5. Implications for nursing practice

Our findings offer several implications for nursing practice. For mild depression, GRT may be preferred in community or nursing home settings due to peer support and resource efficiency. For moderate to severe depression, IRT is more suitable, offering a confidential therapeutic alliance for socially withdrawn patients. IRT is also preferable when anxiety or cognitive impairment is a concern, whereas GRT is better for improving life satisfaction and reducing loneliness. Importantly, trained nurses or paraprofessionals can effectively deliver RT, making it feasible in resource-limited or long-term care settings.

4.6. Strengths

This study offers several key contributions. It applies NMA to integrate direct and indirect evidence comparing GRT and IRT in older adults with depression, addressing a significant evidence gap. Systematic searches across 11 databases reduced publication and language bias. The severity-stratified framework delineates differential efficacy across clinically distinct presentations. Comprehensive subgroup analyses further characterized each intervention’s effects. Evaluation of secondary outcomes (anxiety, cognition, life satisfaction, quality of life) informs which intervention best suits different patient profiles.

4.7. Limitations

Several limitations warrant consideration. First, searching only English and Chinese databases may have introduced language bias. Second, only one study directly compared GRT and IRT, and only two included an active pharmacological control, limiting the precision of these comparisons. Third, due to limited studies, several clinically important comparisons (moderate depression, anxiety, cognition, quality of life) relied entirely on indirect evidence without closed loops. Finally, the analysis treated IRT and GRT as single nodes without examining intervention-intensity parameters. Thus, these findings reflect average effects across heterogeneous implementations and should be interpreted with caution.

4.8. Future research directions

A clear gap exists in high-quality head-to-head RCTs directly comparing IRT and GRT. Future work should prioritize methodologically rigorous trials with sufficient sample sizes and, where possible, an active comparator to generate stronger direct evidence. Additionally, research comparing dosage parameters (session frequency, duration, cumulative number) is needed to define optimal intervention protocols for each format.

5. Conclusions

Both IRT and GRT effectively reduce depressive symptoms in older adults, with no significant overall difference between them. However, GRT may be preferred for mild depression, while IRT is more suitable for moderate to severe cases. IRT shows additional benefits for anxiety and cognitive function, whereas GRT excels in improving life satisfaction. These findings support a personalized, severity-informed nursing approach to selecting reminiscence therapy formats in clinical practice.

Data availability statement

All data analyzed in this study are derived from previously published studies. The extracted datasets used for the analysis are available from the corresponding author upon reasonable request.

CRediT authorship contribution statement

Shan Wang: Conceptualization, Data curation, Methodology, Formal analysis, Validation, Visualization, Writing - original draft, Writing - review & editing. Shuang Chen: Data curation, Formal analysis, Methodology, Visualization, Writing - original draft, Writing - review & editing. Caiqin Wu: Conceptualization, Investigation, Methodology, Project administration, Writing - review & editing. Kunpeng Li: Conceptualization, Methodology, Resources, Software, Writing - review & editing. Libing Liang: Conceptualization, Methodology, Resources, Software, Writing - review & editing. Yonghong Shen: Conceptualization, Methodology, Project administration, Writing - review & editing. All authors commented on previous manuscript versions, and all authors read and endorsed the final manuscript. All authors meet the ICMJE authorship criteria. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Declaration of competing interest

The authors have declared no conflict of interest.

Appendices.

The following are the Supplementary data to this article: