Prevalence of depression in post-stroke aphasia: systematic review and meta-analysis Authors: Amy Vogel-Eyny, Ye-Jee Jung, JungMoon Hyun, Eleanor Siegle, Aisha Haroun, Samsun Nahar, Sarah Sarkas, Stewart A Shankman, Samantha Slattery, Sameer Ashaie Published: 2025 ABSTRACT AbstractBackgroundDepression is a significant consequence of post-stroke aphasia, adversely affecting quality of life; however, its frequency remains unclear. The objective of this systematic review and meta-analysis is to quantify depression prevalence rates among individuals with post-stroke aphasia.MethodsWe systematically reviewed studies on depression in post-stroke aphasia, conducting a meta-analysis that focused on the frequency of depression using diagnostic tools. We conducted a systematic search to identify studies reporting original data on depression frequency in individuals with post-stroke aphasia. From each included study, we extracted data on patient characteristics, diagnostic methods, and aphasia-friendliness. The quality of each study was assessed to ensure reliability and accuracy. We estimated the pooled prevalence using random effect models.ResultsSixteen articles, conducted between 1993 and 2023, met our inclusion criteria. The pooled prevalence of post-stroke depression in people with aphasia was 31.7% (95% CI = 0.2354, 0.4044). Self-reports were the most common method for assessing depression, with participants’ average time post-stroke onset being 1.75 years (range = 0.02–6.76 years). Our meta-analysis found no significant difference in depression prevalence between self-reported and proxy-reported measures, nor any effect of time post-stroke onset on depression.ConclusionsDepression is a common concern in post-stroke aphasia. The lack of observed effects from assessment method or time since stroke may be due to methodological inconsistencies in reporting across studies. Psychometrically sound assessment tools validated for people with aphasia, along with more consistent reporting of patient characteristics, are essential for improving depression screening and mental health care in this population.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12883-025-04570-1.Keywords: Depression, Aphasia, Stroke, Prevalence Introduction Each year, more than 795,000 people in the United States experience a stroke [1]. Aphasia, an acquired neurogenic language impairment, affects approximately one-third of these cases, and an estimated 2.5 million people nationwide are currently living with aphasia [2, 3]. The communication difficulties and activity limitations associated with post-stroke aphasia lead to higher healthcare costs, reduced social participation, and negatively impact quality of life [4–6]. An important factor contributing to quality of life and functional outcomes is mental health, specifically post-stroke depression [7, 8]. Understanding the prevalence of depression in people with aphasia is critical for planning appropriate mental health services, ensuring that assessment practices account for language impairment, and designing clinical research that accurately supports this population. While current estimates suggest that depression affects upwards of one-third of the stroke population [9], its prevalence among individuals with aphasia remains less clear. This connection between post-stroke depression and aphasia has been a long-standing area of discussion in aphasiology [10–12]. Indeed, depression may be more prevalent among stroke survivors with aphasia than those without; however, methodological limitations hinder accurate estimates, with reported prevalence rates ranging from 30% to 60% [8, 12, 13]. In stroke research, eligibility criteria are often underspecified, making it unclear whether individuals with aphasia were included [14]. Further, the wide range of reported prevalence rates for depression in people with aphasia may be due, in part, to inconsistencies in how depression is identified across studies. Many studies use language-based diagnostic tools, such as structured interviews or self-report measures, which rely on verbal comprehension and expression. These tools impose a high linguistic burden (e.g., PHQ)—an issue noted in prior research [15]—and may not be appropriate for individuals with aphasia. As a result, the linguistic and communicative demands of these tools, and their reliance on self-report, often lead to the exclusion of individuals with aphasia or to inaccurate identification of depressive symptoms [12, 13, 15, 16]. In response to these linguistic demands, researchers have employed a range of assessment approaches, including proxy- or informant-based tools, clinical diagnoses, and modified self-report measures. Of note, a review found that fewer than half of studies used diagnostic tools appropriately adapted for aphasia, highlighting a widespread gap in how depression is measured in this population [16]. Although DSM-5–based structured clinical interviews are considered the gold standard for diagnosing depression in the general population [11], their reliance on language makes them less suitable for individuals with aphasia. Alternative measures validated for this population include self-assessments designed to reduce linguistic demands (e.g., visual analog scales) as well as proxy measures [15, 17]. However, proxy ratings should be interpreted with caution, as caregivers’ own emotional states can influence their evaluations, potentially biasing depression scores [17]. Taken together, these methodological differences—particularly the type of assessment and its associated linguistic burden—represent a critical limiting feature of the current literature and likely contribute to variability in reported prevalence. Accordingly, we recorded the type of depression assessment used in each included study to describe the range of tools applied and to support accurate interpretation of the synthesized prevalence estimate. In addition to assessment type, another underexplored source of variability is the time point at which depression is measured post stroke. For example, researchers found that more individuals with aphasia were diagnosed with major depression at 12 months than at 3 months post onset, suggesting that the diagnostic profile of depression may evolve over time [12]. These limitations raise the possibility that methodological factors—such as depression assessment type and time post onset—may influence the proportion of individuals with post-stroke aphasia identified as depressed across studies. In this review, we formally evaluated whether these variables moderated the pooled prevalence of depression reported in our meta-analysis. While individual studies have examined depression in people with post-stroke aphasia, there has been no systematic review or meta-analysis focused exclusively on this population. For example, one widely-cited meta-analysis [18] synthesized data on post-stroke depression across 51 observational studies but explicitly excluded individuals with aphasia and other communication impairments due to the challenges of mood assessment in this group. Similarly, Mitchell and colleagues conducted a structured interview–based meta-analysis of post-stroke mood disorders and included only five analyses that reported prevalence estimates for individuals with aphasia—representing a small subset of the total sample [19]. More recently, Liu et al. conducted a systematic review and meta-analysis of post-stroke depression and attempted to include studies that enrolled individuals with aphasia; however, they noted that most of these studies included only individuals with mild-to-moderate aphasia, due to the linguistic demands of depression assessments [20]. As a result, their pooled prevalence estimates likely underrepresent individuals with more severe aphasia, thereby limiting what can be known about depression in this subgroup. To date, only a few individual studies, such as Kauhanen et al. —which reported separate depression prevalence rates for people with and without aphasia [12]—and Ashaie and colleagues —which studied only people with post-stroke aphasia [17]—have directly addressed this population. Yet no prior study has synthesized depression prevalence estimates across the range of aphasia severity. This systematic review and meta-analysis had two primary aims. First, we sought to estimate the pooled estimate of depression prevalence among individuals with post-stroke aphasia. Second, we examined whether selected methodological features of the included studies were associated with variability in reported prevalence estimates. For this second aim, we focused a priori on two factors: (1) time post-stroke onset and (2) the reporter of depression (self-report vs. proxy). We also documented the specific depression assessment tools used and whether they incorporated aphasia-friendly adaptations (e.g., simplified response formats, large print, read-aloud administration). These latter variables were summarized descriptively to contextualize the pooled prevalence estimates. Methods The study protocol was registered in advance of data extraction with the Open Science Framework (https://osf.io/bfz9q). Study identification Identification and retrieval of studies were conducted in accordance with the PRISMA 2020 statement: an updated guideline for reporting systematic reviews [21]. A comprehensive search strategy, which employed both subject headings and keywords, was run in MEDLINE (Ovid), PsycINFO (EBSCO), Web of Science Core Collection, CINAHL (EBSCO), and Embase (Elsevier) on November 15, 2023, with no date restrictions. Searches were designed to yield only peer-reviewed articles, and to exclude conference proceedings, dissertations, and theses. The full search queries are available in Appendix A. Search results were exported and de-duplicated into the Covidence software [22]. Eligibility criteria & study selection Article eligibility for inclusion was based on criteria from Baker and colleagues’ review [10] that investigated interventions to treat or prevent depression in individuals with post-stroke aphasia. We applied two modifications tailored to our study aims. First, while Baker et al. included intervention studies, the present study focused exclusively on observational research (i.e., non-intervention designs such as cross-sectional, cohort, or medical-record based studies) reporting the frequency of depression using a diagnostic tool (e.g., Center for Epidemiologic Studies Depression Scale [CESD]) [23] or clinician-administered assessment procedures (e.g., structured interviews or clinician-completed rating scales). Formal clinical diagnoses based on medical records or ICD codes were also eligible. Intervention studies were excluded because their samples are recruited based on treatment eligibility rather than to estimate the distribution of depression in the broader post-stroke aphasia population; therefore, they do not meet the aims of descriptive prevalence research [24]. Second, whereas Baker et al. included studies with mixed clinical populations when individuals with aphasia comprised at least 25% of the sample and data were reported separately, our review included only studies that explicitly focused on individuals with post-stroke aphasia. Included studies met the following criteria:(1) participants had post-stroke aphasia, without other neurological diseases or traumatic brain injury, at any time post-onset; (2) the study reported original, study-level data on depression frequency (i.e., the number or percentage of participants with post-stroke aphasia identified as having depression); (3) reported the number of individuals with post-stroke aphasia and the number classified as having depression, allowing us to manually calculate the prevalence of depression in the sample; (4) included diagnostic tools for depression; and (5) was peer-reviewed and published in English. If multiple articles reported outcomes from the same study population, data were extracted from the most recent publication. After de-duplication, the search yielded 990 articles. Titles and abstracts were independently screened by two reviewers per article using Covidence. Full-text screening followed the same procedure, with two reviewers independently assessing each article; S.A. joined the team during this phase. A record advanced to the next stage only when both reviewers voted yes. When reviewers disagreed (for example, one yes and one no), the record was flagged for consensus discussion rather than being moved forward automatically. Consistency between reviewers was ensured through pilot screening rounds held before each stage of screening, during which reviewers screened a pilot set of studies to norm decision making and finalize predetermined eligibility criteria. Discrepancies at either stage were resolved through group discussion among all reviewers. A PRISMA flowchart of study identification with reasons for exclusion is depicted in Fig. 1. Screening of studies Quality assessment and data extraction Study quality was assessed prior to data extraction to ensure that only methodologically sound studies informed the synthesis. Studies were assessed for bias using the Joanna Briggs Institute (JBI) critical appraisal tool for prevalence studies, which assessed sampling, sample size, measurement reliability, and statistical reporting across studies [25]. Two authors (A.V.E. and S.A.) rated each study using the tool’s nine questions, assigning one of four possible ratings: “Yes,” “No,” “Unclear,” or “Not applicable.” Discrepancies in ratings were resolved through discussion. The full set of questions and rating criteria are included in the Supplemental Materials Table 1. After quality assessment, data from the included studies were extracted and summarized. Two reviewers (Y.J.J. & J.H.) independently extracted data, each reviewing 50% of studies. A second team member (A.V.E. or J.H.) cross-checked all entries for accuracy. Discrepancies were resolved through group discussion; while no preset reliability threshold was used, this double-extraction and adjudication process increased confidence in the accuracy and completeness of extracted data. For each study, we extracted participant characteristics (age, gender, time post-onset, and aphasia severity), the depression diagnostic tool used (e.g., CESD, ICD), method of administration (i.e., self, proxy, or mixed), and whether the assessment included aphasia-friendly adaptations. For Time Post Onset (TPO), we recorded study-level summary values (most often means, or medians when specified); no study provided participant-level TPO data. Where available, ranges or standard deviations were also extracted. From each diagnostic tool, we recorded (1) the prevalence of depression reported (2), the type of reporter, and (3) whether the tool incorporated aphasia-friendly features. We defined aphasia-friendly assessments as those incorporating communication supports such as large print, read-aloud formats, simplified response formats, or structured cueing hierarchies or modifications to reduce linguistic demands. Aspects such as reading level and sentence length were considered within this broader definition of aphasia friendliness, as described in prior work (e.g., Pompon et al.; Pompon et al.) [26, 27]. We also recorded the reported depression outcome (i.e., the number or percentage of individuals with aphasia classified as having depression). All variables, including aphasia-friendliness, were independently extracted by two reviewers, with discrepancies resolved through discussion. These data are presented in Tables 1 and 2. Variables were selected to support the primary aim of estimating pooled prevalence and the secondary aim of evaluating two moderators, time post-stroke onset and reporter of depression (self vs. proxy), while also describing additional assessment-related features (specific assessment tools and whether they incorporated aphasia-friendly adaptations). Table 1Participant demographicsFirst Author, YearSampleSizeMean AgeMean Time PostOnsetGender(M; F)AphasiaSeverity Ashaie, 2019 [17]14456.5 ± 12.330.6 mos.± 33.184 M; 60 FModeratea Cocchini, 2013 [34]2660.9 ± 15.9 (19–84)41.2 wks. ±9.3 (1–306)20 M; 10 FMild: 3bModerate: 8Severe: 15Cruice,2011 [41]3070.7 ± 8.4(57–88)41.1 mos. ±25.6(10–108)14 M;16 FMild-to-moderateaDamecour, 1991[35]5464 ± 12.8511.6 mos. ±5.63N/ASevere: 43%cModerate: 30%Mild: 28% El Hachioui, 2014[39]14767 ± 15Baseline: within 2 wks. post-stroke78 F; 69 M3 mos.: Milda12 mos.: Mild Herrmann, 1993[42]42Acute: 62(26–81)Chronic: 62(27–79)1 mos.(0.5–2.5)35 mos.(7–132)26 M; 16 Fn.s. Hunting Pompon, 2018[27]7264.5 ± 10.91Chronic: 81.1 mos. ±5429 F; 43 MNot severea Kauhanen, 2000[12]3 mos.: 2712 mos.: 2165.83 mos.12 mos.n.s.Mild: 9bModerate: 10Severe: 17 Kemper, 2011[43]145N/Awithin 1 year post-stroken.s.n.s. Laures-Gore, 2017[36]2548.7 ± 13.75 (19.95–70.57)2.7 yrs. ±3.00 (0.06–11.61)16 M; 9 FVery severe: 1bSevere: 4Moderate: 5Mild: 13Beyond cut-off: 2 Lee, 2020[40]2 wks.: 683 mos.: 202 wks. post − 66.2 ± 11.373 mos. post − 66.2 ± 12.362 wks.3 mos.2 wks. post: 36 M; 32 F3 mos. post: 14 M; 6 F2 wks. post:a severe3 mos. post: moderate Lin, 2023[33]26,75466.3 ± 13.5n.s.16,673 M; 10,081 Fn.s. Mazaux, 2013[37]10065 ± 13.512–18 mos.51 M; 49 FBaseline: 62%c had severe aphasia Pompon, 2022[26]12163.4 years ± 11.30(26–85)79.4 mos. ±54.19 (4–216) 82 M; 39 FNot severea Wang, 2018[44]148 (baseline)57 ± 14.214 ± 2 days (baseline)n.s.n.s. Zanella, 2023[38]40666 ± 15.141 wk.206 M; 200 FMild-to-moderate = 111bSevere = 107Mute = 188Where available, standard deviations (SD) and/or ranges are included; n.s. refers to “not specified.” Severity was determined as reported in each respective study and may reflect different approaches. Mean age and mean TPO reflect study-level summary values (mean or median), not individual participant dataaCategorical description onlybCounts reportedcPercentages reported Participant demographics Mild: 3b Moderate: 8 Severe: 15 Cruice, 2011 [41] 70.7 ± 8.4 (57–88) 41.1 mos. ±25.6 (10–108) Severe: 43%c Moderate: 30% Mild: 28% 3 mos.: Milda 12 mos.: Mild Acute: 62 (26–81) Chronic: 62(27–79) 1 mos.(0.5–2.5) 35 mos. (7–132) 3 mos.: 27 12 mos.: 21 3 mos. 12 mos. Mild: 9b Moderate: 10 Severe: 17 Very severe: 1b Severe: 4 Moderate: 5 Mild: 13 Beyond cut-off: 2 2 wks.: 68 3 mos.: 20 2 wks. post − 66.2 ± 11.37 3 mos. post − 66.2 ± 12.36 2 wks. 3 mos. 2 wks. post: 36 M; 32 F 3 mos. post: 14 M; 6 F 2 wks. post:a severe 3 mos. post: moderate 63.4 years ± 11.30 (26–85) 79.4 mos. ±54.19 (4–216) Mild-to-moderate = 111b Severe = 107 Mute = 188 Where available, standard deviations (SD) and/or ranges are included; n.s. refers to “not specified.” Severity was determined as reported in each respective study and may reflect different approaches. Mean age and mean TPO reflect study-level summary values (mean or median), not individual participant data aCategorical description only bCounts reported cPercentages reported Table 2Characteristics of included studies measuring aphasia and depressionFirst Author, YearDX Aphasia ToolPrevalence (%)DX Depression ToolReporter TypeAphasia Friendly Ashaie, 2019 [17]WAB-R41.6%CESDSselfY Cocchini, 2013 [34]AAT42%SADBD-ItalianmixedN Cruice, 2011 [41]WAB30%GDSselfNDamecour, 1991 [35]BDAE15%Zung-SORSproxyn.s. El Hachioui, 2014 [39]SL/TT/ASRS11% (1 year.)Medical records, interviewing patients/relatives, antidepressant usemixedN Herrmann, 1993 [42]AATAcute: 19%Chronic: 10%CDS-GermanmixedN Hunting Pompon, 2018 [27]CAT17%PHQ-8selfY Kauhanen, 2000 [12]WAB3 mos. − 70% 12 mos. − 62%DSM-III-RselfN Kemper, 2011 [43]ICD codes32%ICD-10selfN Laures-Gore, 2017 [36]WAB-R40%SADQ-10; ADRSproxyn.s. Lee, 2020 [40]K-WAB2 wks. post: 43%3 mos. post: 30%BDI-Korean, GDS-Korean, psychiatrist observationmixedN Lin, 2023 [33]ICD codes6.30%ICD-9 & 10selfN Mazaux, 2013 [37]ASRS25%ADRS-Frenchproxyn.s. Pompon, 2022 [26]CAT, CRTT50%PHQ-8selfY Wang, 2018 [44]WAB47.5%SADQ-Hproxyn.s. Zanella, 2023 [38]Item 9 (Best Language) NIH Stroke Scale43%PHQ-2proxyn.s.n.s. refers to “not specified.” The column ‘Prevalence (%)’ indicates the proportion of participants with aphasia identified as having depression in each study. Where studies reported percentages directly, we entered those values; where only raw numbers were provided, we calculated the percentage Characteristics of included studies measuring aphasia and depression Acute: 19% Chronic: 10% 3 mos. − 70% 12 mos. − 62% 2 wks. post: 43% 3 mos. post: 30% n.s. refers to “not specified.” The column ‘Prevalence (%)’ indicates the proportion of participants with aphasia identified as having depression in each study. Where studies reported percentages directly, we entered those values; where only raw numbers were provided, we calculated the percentage Statistical analysis A random-effects model was computed to estimate the pooled prevalence with 95% CI of depression following post-stroke aphasia. Of the methodological factors extracted (assessment tool, aphasia-friendly adaptations, time post onset, and reporter of depression), two were evaluated as moderators in the meta-analysis. Meta-regression and subgroup analyses were performed to investigate the impact of time post onset and reporter of depression (i.e., self vs. proxy) on the prevalence of post-stroke depression following aphasia. All raw proportions were transformed using the double arcsine approach [28]. For subgroup analyses (e.g., self vs. proxy), prevalence estimates were arcsine-transformed, and between-group differences were tested using the Q statistic for subgroup differences. Assessment of publication bias was not performed due to it not being necessary for prevalence data [28]. Meta-analyses were performed using both the metafor [29] and meta [30] packages in R statistical software [31]. Van Dijk and colleagues [32] identified influential studies using a combination of Baujat plot analysis, externally studentized residuals, and influence diagnostics [28, 29], and was excluded from the meta-analysis. In this systematic review and meta-analysis, “pooled prevalence” refers to the meta-analytic estimate of depression prevalence across all included studies generated using random-effects models. Results Sixteen studies, including a total of 28,288 unique participants (median sample size per study = 70; range = 20–26,754), met our inclusion criteria (see Table 1). When a study reported more than one time point from the same cohort, the cohort was counted once toward the total number of unique participants, but each time point contributed a separate prevalence estimate to the meta-analysis. Although Fig. 1 displays 17 included studies, one study [32] was removed following outlier diagnostics prior to meta-analysis due to its disproportionate influence on pooled estimates. It remains in the PRISMA diagram to accurately reflect the outcome of the full-text screening process. Study quality was assessed using the JBI Critical Appraisal Checklist for Prevalence Studies. Overall, methodological quality was high. Seven studies met all nine JBI criteria (100%), eight studies met eight out of nine criteria (88%), and one study met seven out of nine criteria (77%). Across studies, the strongest domains were appropriate sampling frames, clear descriptions of participants and settings, and use of standardized or reliable methods for identifying depression. The domains with the most “No” or “Unclear” responses were sample-size adequacy (Q3) and response-rate reporting/management (Q9). Full JBI ratings for each study are provided in Supplemental Table 1. Prevalence For the meta-analysis, studies that reported prevalence at multiple time points were divided into separate entries, as each time point provided an independent prevalence estimate. For descriptive purposes, however, these studies were counted once, consistent with the 16 unique studies included in the systematic review (see Table 1). Accordingly, the total number of participants contributing to pooled prevalence estimates was 28,350. The pooled prevalence of depression in post-stroke aphasia was 31.70% (95% CI = 23.54%, 40.44%) based on data from 28,350 participants (see Fig. 2). There was significant between-study heterogeneity, as indicated by τ2 = 0.0339 (95% CI = 0.0173, 0.0795), I2 = 95.91% (95% CI = 92.2620, 98.2117), and the Q-statistic = 1043.63 (p < 0.001). Meta-analysis on the prevalence of depression in post-stroke aphasia. A study with multiple time points was divided into separate studies, with the letter “a” referring to an earlier time point. Each square represents the prevalence estimate from an individual study, with the size of the square proportional to the study’s weight in the analysis. Horizontal lines indicate 95% confidence intervals. The diamond represents the pooled prevalence estimate across studies. Cases = number of participants classified as having depression; Total = number of participants with aphasia in that study Given that the overall sample was dominated by one very large study [33] (n = 26,754), we conducted a supplemental (sensitivity) analysis excluding this study to determine whether it disproportionately influenced the pooled estimate. After removing that study, the recalculated pooled prevalence—based on the remaining 1,596 participants in 15 studies—was 33.83% (95% CI: 26.09%–42.01%), which closely aligns with the original estimate of 31.70% (see Supplemental Fig. 1). Characteristics of participants with aphasia1 The participants with aphasia had a mean age of 63.35 years (range = 48.7–70.7 years) across studies. Because individual-level age data were not available, the mean age reported here reflects a mean of study-level means. Specifically, we calculated the average of the mean ages reported in each study. The range shown reflects the lowest and highest mean ages reported across studies, not individual participants’ minimum or maximum ages.The gender distribution across studies indicates that there was a total of 17,334 males (62.0%) and 10,625 females (38.0%). Because individual-level TPO data were not available, the average time post-onset (1.75 years) reflects a mean of the study-level means or medians. The reported range (0.02–6.76 years) represents the lowest and highest study-level TPO values, not individual-level values. Although many studies commented on the overall aphasia severity of their participant populations, only a portion of studies provided numerical distributions of participants within distinct severity categories. Six studies [12, 34–38] reported detailed severity data. Across these, authors classified approximately 40 participants as mild [12, 34–36], 111 as mild-to-moderate [38], 39 as moderate [12, 34–36], and 229 [12, 34–38] as severe (including very severe). These classifications reflect what was reported directly by the study authors. In two cases [35, 37], percentages were converted into counts based on the reported sample size; no reclassification by the study authors was applied. The remaining studies did not provide categorical distributions of aphasia severity. Instead, Ashaie [17] described participants as moderate overall [17]; El Hachioui [39] described the sample as mild at both 3 and 12 months [39]; Hunting Pompon [27] and Pompon [26] both described participants as not severe [26, 27]; Lee [40] described participants as severe at 2 weeks and moderate at 3 months [40]; and Cruice [41] reported a mean severity score in the mild-to-moderate range [41]. Four studies—Herrmann [42], Kemper [43], Lin [33], and Wang [44]—did not report severity data [33, 42–44]. Depression diagnostic tools Twelve diagnostic methods or test types were identified: (1) the Aphasia Depression Rating Scale (ADRS, ADRS-French); (2) the Beck Depression Inventory (BDI-Korean); (3) the Center for Epidemiologic Studies Depression Scale (CESDS); (4) the Cornell Depression Scale (CDS-German); (5) the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R); (6) the Geriatric Depression Scale (GDS, GDS-Korean); (7) the International Classification of Diseases (ICD, ICD-9, ICD-10); (8) the Patient Health Questionnaire (PHQ-2, PHQ-8); (9) the Stroke Aphasic Depression Questionnaire (SADQ, SADQ-10, SADQ-H); (10) the Structured Assessment of Depression in Brain Damaged Individuals (SADBD-Italian); (11) the Zung Significant Other Rating Depression Scale (Zung-SORS); and (12) one diagnostic method used in a single study [39] in which depression was identified using medical record review, patient or relative interviews, and antidepressant use. Although less standardized than psychometric measures, the use of medical records, diagnostic codes, and antidepressant prescriptions as indicators of depression has precedent in epidemiological research. Among the standardized instruments noted above, the PHQ (PHQ-2 and PHQ-8) was the most frequently used tool, although it was only used three times, while the ICD codes, SADQ, and ADRS were each used twice (see Table 2). In examining the administration methods for depression assessments, we categorized tools based on whether they were self-administered, proxy-administered, or used a mixed approach. Self-administered depression tools refer to diagnostic assessments completed directly by the individual with aphasia or instances where a diagnostic billing code (e.g., ICD) was assigned, indicating that a healthcare professional conducted an interview with the individual with aphasia to formally diagnose depression based on established criteria. Proxy measures, by comparison, refer to assessments completed by family members, caregivers, or nursing staff without the direct participation of the individual with aphasia. Mixed administration approaches generally combined information from both the individual with aphasia and external informants (e.g., clinicians, caregivers, or nurses), integrating patient self-report measures with observer ratings of depressive symptoms using structured and often validated assessment tools. However, one study in this category did not fully specify its depression assessment method; instead, depression was determined based on medical record review, patient and relative interviews, and antidepressant use, without clearly identifying a diagnostic tool [39]. Of the 16 studies, 7 used self-administered measures, 5 were administered by proxy, and 4 used a mixed administration method. Among the proxy-administered assessments, respondents varied across studies. In Damecour [35], the Zung-SORS was completed by a “concerned individual,” though the authors did not specify who this referred to. Laures-Gore [36] used the SADQ-10 and ADRS, which were completed by caregivers or other informants familiar with the participant. Similarly, Wang [44] and Zanella [38] relied on nursing staff, in some cases supplemented by family members or caregivers, while Mazaux [37] administered the ADRS-French but did not identify the respondent. As defined in the Methods, studies were coded as using aphasia-friendly assessments if they incorporated communication supports such as large print, read-aloud formats, or structured cueing hierarchies. Of the 11 assessments that included self-administration (either exclusively or as part of a mixed approach; [25–34]), only three incorporated aphasia-friendly modifications [17, 26, 27]. Specifically, one assessment used a large font and read-aloud format [17]; another employed an external support hierarchy [27]; and a third applied supportive communication strategies [22]. In one study of depression in people with post-stroke aphasia [35], other aphasia-friendly, self-administered measures of depression were included; however, because the depression frequency data were derived from a proxy measure, this study was not logged as including an aphasia-friendly assessment. See Supplemental Fig. 2 for analysis of aphasia-friendly assessment tools on depression prevalence. Subgroup analysis Impact of time post onset stroke For this subgroup analysis, three studies [33, 37, 43] were excluded as they did not report an exact time post onset stroke. As a result, 13 studies were included in this meta-regression analysis, with TPO treated as a continuous moderator. We note that no study provided participant-level exact TPO data; instead, study-level summary values (most often means, or medians when specified) were analyzed. Results revealed that the regression coefficient was not significant (−0.0002; Z [14] = −0.4272, p = 0.6692) (Fig. 3). A scatter plot including participants’ time post onset stroke as the moderator. Time post-stroke onset (in months); Percentage = frequency of depression Impact of reporter of depression (proxy versus self-report) Among the 16 studies included in the meta-analysis, 12 employed exclusively self-administered (e.g., PHQ‐8) or proxy‐administered (e.g., caregiver, nurse, etc.) measures, whereas 4 used a mixed‐administration approach. Because our analysis focused solely on single‐mode administration studies, we ultimately included only those 12 studies. Mixed-administration studies were excluded from subgroup comparisons because their combined data could not be separated into distinct self- versus proxy-reported estimates, which would have obscured interpretation. The subgroup effect size difference between self-reported (n = 7) and proxy-reported assessments (n = 5) was not significant (QM (1) = 0.041, p-value = 0.839). Although the between-group difference was not statistically significant, we report subgroup prevalences descriptively to provide transparency regarding the range of estimates across self- and proxy-reported studies. For this sub-group analysis, a common between-study variance between the two subgroups was not assumed, and data were transformed using the double arcsine approach [23]. The overall estimate was 34.44% (95% CI = 24.66%, 44.90%, see Fig. 4). For studies that used proxy assessment tools, the subgroup prevalence was 33.86% (95% CI = 22.87%, 45.79%), while for studies that used self-reported assessment tools, the prevalence was 36.39% (95% CI = 16.84%, 58.52%). Subgroup meta-analysis on the prevalence of depression in post-stroke aphasia by the characteristic of depression assessment tool (proxy vs. self). Each square represents the prevalence estimate from an individual study, with the size of the square proportional to the study’s weight in the analysis. Horizontal lines indicate 95% confidence intervals. The diamond represents the pooled prevalence estimate across studies for each subgroup. Cases = number of participants classified as having depression; Total = number of participants with aphasia in that study We also explored whether age moderated depression prevalence. Results showed that the regression coefficient was not significant (−0.0087; Z [16] = −0.879, p = 0.380). See Supplemental Fig. 3. Discussion Our systematic review and meta-analysis indicate that approximately 32% of individuals with post-stroke aphasia experience depression following stroke onset, as reported across 16 research studies involving 28,350 participants. Although one study contributed the majority of participants, a sensitivity analysis excluding it yielded a nearly identical pooled prevalence, supporting the robustness of the estimate [33]. Notably, our review identified that variations in depression prevalence could not be fully explained by differences in stroke chronicity or the reporter of depression (self vs. proxy). This suggests that depression is a consistent concern regardless of stages of recovery and modes of assessment. We identified substantial variability in the diagnostic tools used to detect depression across studies, with 12 different tools employed. Among these, variants of the PHQ emerged as the most commonly used. Some tools were self-administered, others were proxy-reported, and a few used mixed methods. Only a limited number of studies incorporated aphasia-friendly adaptations, such as simplified language or supportive communication strategies, to accommodate the language impairments in individuals with aphasia. This inconsistency in diagnostic methods, particularly the limited use of aphasia-friendly modifications, suggests a need for accessible tools that ensure accurate depression screening across the spectrum of aphasia severity. Our findings underscore the importance of refining diagnostic practices to improve mental health care for this population. In particular, the variability in depression prevalence appears to stem more from differences in measurement, such as the choice of tool, mode of administration, and degree of aphasia-friendliness, than from time since stroke or reporter type. Developing accessible assessment protocols that minimize linguistic demands and offer coordinated self- and proxy-report options may help produce more consistent estimates and better support clinical decision-making. The pooled prevalence of 32% found in this study aligns with depression rates reported in the broader stroke literature [18]. For example, a widely cited meta-analysis, not including individuals with aphasia, estimated a pooled prevalence of post-stroke depression [18]. Although our pooled prevalence estimate integrates findings across studies, individual investigations of people with post-stroke aphasia have reported widely varying prevalence rates, with some studies reporting higher rates [12, 19] and others reporting lower rates [17]. These discrepancies likely reflect differences in study design, diagnostic tools, and thresholds for detecting depression, highlighting the value of a synthesized estimate. Our finding that time post-stroke did not influence depression prevalence among individuals with aphasia aligns with several previous studies. A recent retrospective analysis that included 121 individuals with chronic aphasia found no association between time post onset of stroke and post stroke depression [26]. Additionally, studies that included stroke patients both with and without aphasia [19], or excluded aphasia altogether [18], have reported similar outcomes in their broader stroke populations. For example, a meta-analysis including individuals with post-stroke depression found no statistically significant effect of time on the prevalence of any depressive disorder, whether major depression or subsyndromal conditions like minor depression [19]. Similarly, in a systematic review of observational studies of general post-stroke depression, the researchers found an average rate of depression of 33% that remained constant across various intervals from 1 month to over 1 year post-onset [18]. Most of the studies in our systematic review were conducted in the acute phase, with relatively fewer representing later stages of recovery, which may have limited our ability to detect patterns beyond the early months post-stroke. Our results nevertheless contrast with prior research [7, 45] that noted significantly higher rates of post-stroke depression during the first year, peaking in the initial months and then gradually decreasing after 12 months. In our moderator analysis of TPO, we did not observe this downward pattern. This discrepancy may reflect both the limited variability in TPO across included studies and the reliance on study-level summary values rather than participant-level data. Additionally, our lack of an observed effect of time since stroke on depression prevalence may be due to post-stroke depression measures not distinguishing nonchronic depressive disorders, which may resolve more easily, from chronic depressive conditions like persistent depressive disorder (aka dysthymia) [15, 46]. Future studies should assess depression at multiple time points post-stroke to clarify when and how depression changes across the recovery course. Our analysis found no difference in the prevalence of post-stroke depression in people with aphasia when comparing proxy and self-report measures. The PHQ was the most frequently used tool, followed by the ICD codes, SADQ, and ADRS, with each of these latter three appearing twice. Two of these measures are self-report and two are proxy, yet none have been specifically validated for use in people with aphasia following stroke [17]. The PHQ is widely used for assessing post-stroke depression and is psychometrically valid for this population [47, 48]. Although it has not been specifically validated for people with post-stroke aphasia [13], it has been used effectively in this group and aligns with the reading level of other patient-reported measures commonly used with individuals with aphasia [26, 27]. Previous research suggests that, even with a communicative support hierarchy—as applied in two studies in this review [26, 27]—limited understanding of the PHQ items may still compromise its validity. Similarly, ICD codes for depression rely on clinical assessment, which generally involves self-report. This approach, however, provides a standardized measure for identifying depression but may miss nuances in individuals with post-stroke aphasia due to communication barriers that affect self-reporting accuracy. While self-report measures have limitations, using informant reports presents its own challenges, as caregivers may over- or underestimate depression based on their own mood [49]. Indeed, prior work has shown that caregivers often rate individuals with stroke as more depressed than those individuals rate themselves [50]. The SADQ-10 and ADRS have been recommended as complementary tools for assessing depression in individuals with post-stroke aphasia, either together [36] or in combination with self-report measures such as the CESD-R [13]. Even among proxy-rated measures, responses can vary depending on whether they are completed by someone close to the person with post-stroke aphasia or by an inpatient medical team, potentially affecting reliability. This variability is compounded by the fact that no gold-standard measure of depression currently exists for individuals with post-stroke aphasia [11], and this heterogeneity likely contributes to the observed lack of effect between proxy and self-report outcomes, limiting comparability across assessment tools. Clinical implications In our review of the post-stroke depression literature, only 16 studies qualified for inclusion based on their focus on people with post-stroke aphasia, yet they exhibited extensive variability in reporting on aphasia-related factors. This inconsistency aligns with issues noted in a recent study by Shiggins et al. that highlighted the frequent exclusion or inadequate inclusion of people with aphasia in stroke research [14]. Shiggins et al. [14] argue that this lack of consistent, detailed reporting and appropriate modifications limits the generalizability and validity of findings for the broader stroke population. The need for deliberate inclusion, with protocols that incorporate validated and aphasia-friendly measures, is essential to ensure accurate representation and better address the mental health needs of people with post-stroke aphasia. Given the heterogeneity in language profiles among people with aphasia, diagnostic guidance may be more effective when structured as a flexible protocol that supports multiple response modalities and tailored adaptations, rather than reliance on a single uniform screening tool. Such protocols may incorporate supported communication techniques, simplified formatting, visual or gestural response options, or opportunities for corroborating proxy input when appropriate. Emerging approaches, including speech- or acoustics-based markers as complementary indicators of mood [51], also warrant further investigation.Our findings reinforce those of prior studies that call for such practices to improve the quality and applicability of stroke research [13–15]. Limitations In this systematic review and meta-analysis, findings are limited by differences in stroke features, clinical characteristics, timing of assessments, and types of depression assessments used. One example is TPO, which was reported only as study-level summary values (typically means or, when specified, medians), not participant-level data. As a result, the TPO meta-regression relied on study-level averages and should be interpreted cautiously; studies without a usable TPO summary were excluded from that analysis. Additionally, future research could provide a more comprehensive understanding by exploring the levels of depression severity (e.g., mild, moderate, severe) and the persistence of depressive symptoms over time in individuals with post-stroke aphasia. The current studies included in the review demonstrated significant variability in how both aphasia and depression severity were reported. Some studies provided detailed severity categories, whereas others offered only broad classifications or omitted severity information altogether. These inconsistencies limit the ability to compare findings across studies and to draw nuanced conclusions about how aphasia severity may relate to depression prevalence. To strengthen future research, studies should report severity information consistently, document whether communication supports were used during depression assessments, and clearly describe the administration procedures (e.g., self vs. proxy). Additionally, accounting for potential confounding factors would help clarify the complex relationship between aphasia severity, communication profiles, and depression. Conclusion This systematic review and meta-analysis of 16 studies found a pooled prevalence of depression in post-stroke aphasia of 32%. No significant differences were observed based on assessment method (self vs. proxy) or time post-onset. These findings provide the first pooled estimate of depression prevalence specifically in individuals with aphasia, highlighting depression as a common concern in this population. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Acknowledgements We would like to thank research assistant, Mary-Grace Zwilling, for her assistance in the initial conceptualization and review process. Appendix Complete search strategies Embase (Elsevier Platform):706 results on 11-1-23 (due to availability, this search was run two weeks earlier than the other databases’). ('cerebrovascular accident'/exp OR 'cerebrovascular accident' OR stroke:kw,ti,ab OR ((cerebrovascular NEAR/2 (disease OR accident* OR attack* OR ischemia OR infarction)):ab,kw,ti) OR ((cerebral NEAR/2 (disease OR accident* OR attack* OR ischemia OR infarction)):ab,kw,ti) OR (transient:ab,kw,ti AND ischemic:ab,kw,ti AND attack*:ab,kw,ti)) AND ('depression'/exp OR depression:ab,kw,ti OR depressed:ab,kw,ti OR depressive:ab,kw,ti) AND ('aphasia'/exp OR'dysphasia'/exp OR aphasia:ab,kw,ti) NOT (('cerebrovascular accident'/exp OR'cerebrovascular accident' OR stroke:kw,ti,ab OR ((cerebrovascular NEAR/2 (disease OR accident* OR attack* OR ischemia OR infarction)):ab,kw,ti) OR ((cerebral NEAR/2 (disease OR accident* OR attack* OR ischemia OR infarction)):ab,kw,ti) OR (transient:ab,kw,ti AND ischemic:ab,kw,ti AND attack*:ab,kw,ti)) AND ('depression'/exp OR depression:ab,kw,ti OR depressed:ab,kw,ti OR depressive:ab,kw,ti) AND ('aphasia'/exp OR'dysphasia'/exp OR aphasia:ab,kw,ti) AND ('conference abstract'/it OR 'conference paper'/it OR 'conference review'/it)). Ovid MEDLINE(R) ALL <1946 to November 14, 2023>: search run on 11/15/23 1exp Stroke/1760082stroke.mp3800553(Cerebrovascular adj2 (disease or accident* or attack* or ischemia or infarction)).mp307544(Cerebral adj2 (disease or accident* or attack* or ischemia or infarction)).mp. 757575Transient Ischemic Attack*.mp1437961 or 2 or 3 or 4 or 5 4499747exp depression/or exp mood disorders/3043158(depression or depressed or depressive).mp62154697 or 8 654368106 and 91558611exp Aphasia/1334112aphasia.mp206901311 or 12 206901410 and 13355 CINAHL (EBSCO): search run on 11/15/23 ((MH"Stroke+") OR (TI stroke OR AB stroke) OR (TI ((Cerebrovascular N2 (disease or accident* or attack* or ischemia or infarction))) OR AB ((Cerebrovascular N2 (disease or accident* or attack* or ischemia or infarction)))) OR (TI ((Cerebral N2 (disease or accident* or attack* or ischemia or infarction))) OR AB ((Cerebral N2 (disease or accident* or attack* or ischemia or infarction)))) OR (TI "Transient Ischemic Attack*" OR AB "Transient Ischemic Attack*")). AND ((MH"Affective Disorders+") OR (TI (depression or depressed or depressive) OR AB (depression or depressed or depressive))) AND ((MH"Aphasia+") OR (TI (MH "Aphasia+") OR AB (MH "Aphasia+"))) 129 results, limit to Academic Journals ->118 results PsycINFO (EBSCO): search run on 11/15/23 ((DE"Cerebrovascular Accidents") OR (TI stroke OR AB stroke) OR (TI ((Cerebrovascular N2 (disease or accident* or attack* or ischemia or infarction))) OR AB ((Cerebrovascular N2 (disease or accident* or attack* or ischemia or infarction)))) OR (TI ((Cerebral N2 (disease or accident* or attack* or ischemia or infarction))) OR AB ((Cerebral N2 (disease or accident* or attack* or ischemia or infarction)))) OR (TI "Transient Ischemic Attack*" OR AB "Transient Ischemic Attack*")) AND ((DE"Depression (Emotion)" OR DE "Affective Disorders" OR DE"Disruptive Mood Dysregulation Disorder" OR DE "Major Depression" OR DE "Persistent Depressive Disorder" OR DE "Premenstrual Dysphoric Disorder" OR DE "Seasonal Affective Disorder") OR (TI (depression or depressed or depressive) OR AB (depression or depressed or depressive))). AND ((DE"Aphasia" OR DE "Dysphasia") OR (TI (MH"Aphasia+") OR AB (MH "Aphasia+"))) 121 results, limit to academic articles ->109 results Web of science core collection: search run on 11/15/23 TS=(stroke) OR TS=(Cerebrovascular NEAR/2 (disease or accident* or attack* or ischemia or infarction)) OR TS=(Cerebral NEAR/2 (disease or accident* or attack* or ischemia or infarction)) OR TS=Transient Ischemic Attack* AND TS=(depression or depressed or depressive) AND TS=(aphasia) 476 results, limit to Article, Review Article, Early Access ->461 results Authors’ contributions AVE and SA conceptualized the study, developed the review protocol, and contributed to manuscript preparation and revision. YJJ and JH contributed to the study design, assisted with data analysis, and participated manuscript preparation and revision. ES, AH, SN, and SS2 were involved in literature screening, data extraction, and contributed to drafting and revising the manuscript. SAS provided oversight of statistical analysis and contributed to the interpretation of findings and critical manuscript revision. SS1 contributed to the development of the review protocol and supported methodological decisions and data organization. All authors read and approved the final manuscript. Funding AVE was supported by National Institute on Aging under award number R21AG073974. SAA was supported by the National Institute on Deafness and Other Communication Disorders of the National Institutes of Health under award number K23DC020757 and by National Eunice Kennedy Shriver National Institute of Child Health and Human Development under award number R03HD111008. SAS was supported by National Institute of Mental Health award number R01 MH134369 and R01 MH118741. Data availability The study protocol and corresponding search terms was pre-registered with the Open Science Framework https://osf.io/bfz9q. Extracted data is summarized in the tables provided in this manuscript. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes The information presented is based on baseline data. We used baseline values to allow consistent comparison across studies, as many did not report these participant characteristics (e.g., aphasia severity, age, gender) at later time points Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Data Availability Statement The study protocol and corresponding search terms was pre-registered with the Open Science Framework https://osf.io/bfz9q. Extracted data is summarized in the tables provided in this manuscript.