Post-stroke depression management in the Israeli primary care setting: a retrospective cohort study Authors: Olga Spivak, Liat Lev-Shalem, Shlomit Yaari, Limor Adler Published: 2026 ABSTRACT AbstractObjectivesTo characterise post-stroke depression (PSD) diagnostic and treatment patterns in an outpatient primary care setting, including timing, screening methods, therapeutic interventions and associations with mortality.DesignA retrospective cohort study.SettingMaccabi Healthcare Services, the second largest Healthcare Maintenance Organisation in Israel, covering more than 2.7 million citizens between 2016 and 2022.ParticipantsParticipants were adult patients with a new stroke diagnosis between 2016 and 2022 and a subsequent diagnosis of depression according to International Classification of Diseases clinical criteria or antidepressant medication initiation. Patients with a diagnosis of depression or antidepressant treatment prior to stroke were excluded from the study.Primary and secondary outcomesPrimary outcomes included PSD diagnosis rates, time to diagnosis and treatment, use of screening questionnaires, specialty of the physician making the diagnosis and all-cause mortality. Secondary outcomes included referral rates to mental health services, rehabilitation participation and its impact on mortality.ResultsAmong 11 499 patients, PSD occurred in 4620 (40.2%) patients. Primary care physicians diagnosed 53.1% of cases based on clinical assessment; only 4.5% of patients underwent Patient Health Questionnaire-2 screening. Most diagnoses occurred in the first year (53.4%). Antidepressants, predominantly selective serotonin reuptake inhibitors (58.3%), were initiated within 30 days in 65.1% of diagnosed patients. Patients with PSD demonstrated higher rehabilitation participation (69.5% vs 48.5%, p<0.001) and paradoxically lower mortality rates (22.1% vs 27.9%, p<0.001). Patients without PSD were older with a greater comorbidity burden. Cox regression identified physiotherapy visits (HR=0.625, p<0.001) and mental health consultations (HR=0.642, p<0.001) as protective factors.ConclusionsIn this cohort, in an outpatient primary care setting, primary care physicians diagnosed the majority of PSD cases, predominantly relying on clinical criteria, without using screening tools. The majority of PSD diagnoses occurred during the first year after stroke (53.1%), suggesting that repeated evaluation might be appropriate during this follow-up period. Early antidepressant treatment and enhanced rehabilitation engagement may positively influence mortality outcomes in patients with PSD. Low utilisation of standard screening tools and mental health referrals represents an area for potential improvement in PSD management. STRENGTHS AND LIMITATIONS OF THIS STUDY The study has a large sample, encompassing 11 499 patients with stroke. The study has a large sample, encompassing 11 499 patients with stroke. The population-based design spanning 6 years of follow-up captures real-world practices in the primary care setting. The population-based design spanning 6 years of follow-up captures real-world practices in the primary care setting. Specifically addressing the chronic phase after stroke in an outpatient setting fills a gap in the literature. Specifically addressing the chronic phase after stroke in an outpatient setting fills a gap in the literature. The study relies on medical records, which may contain information bias and incomplete data. The study relies on medical records, which may contain information bias and incomplete data. The study could not assess cerebrovascular accident-related disability severity, which could influence rehabilitation patterns and outcomes. The study could not assess cerebrovascular accident-related disability severity, which could influence rehabilitation patterns and outcomes. Introduction Post-stroke depression (PSD) is the most common neuropsychiatric complication of cerebrovascular accident (CVA), with approximately one-third of CVA survivors suffering from depression.1 2 According to a recent literature review, PSD remains underdiagnosed as depressive symptoms are often attributed to normal reactions to severe illness1 3 and stroke-related neurological symptoms may hinder practitioners’ identification of PSD.4 PSD has a multifactorial pathophysiology, involving a combination of various ischaemia-induced neurobiological dysfunctions in the context of psychosocial distress.5 Risk factors for developing depression include female sex, personal history of psychiatric disease, CVA severity, multiple CVA, lack of social support and significant disability caused by the CVA.14 PSD is associated with poor rehabilitation outcomes, decreased cognitive abilities and low quality of life.46 More importantly, it is associated with high mortality rates.13 These outcomes place a significant burden on the patient, family and society; hence, early diagnosis and treatment may prove beneficial.5 7 8 Much of the research on PSD was conducted in hospitals, rehabilitation facilities or specialised neurologic clinics using one of the available screening tools, including the patient health questionnaire (PHQ).9 10 These studies demonstrate that periodic systematic screening increases depression diagnosis rates, improves appropriate treatment, reduces depressive symptoms and improves outcomes.5 11 12 International guidelines differ in their recommendations on optimal timing, frequency and setting for screening for depression.1315 Many patients without depressive symptoms during the acute post-stroke phase are discharged to primary care without ongoing mental health evaluation. This gap is particularly important because primary care physicians (PCPs) maintain longitudinal relationships with patients and are ideally positioned for ongoing surveillance. The aim of this study was to evaluate the approach to diagnosing PSD in an outpatient primary care setting—specifically, the timing of diagnosis, treatment initiation, referrals to multidisciplinary evaluation and impact on all-cause mortality. Methods Standard protocol approvals, registrations and patient consents The study was approved by the institutional review board of Maccabi Health Services (0006–23 MHS, 05/03/2023). All methods were carried out in accordance with ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. Informed consent requirement was waived owing to the study design. Study design and setting In this retrospective cohort study, all adult patients with a new CVA diagnosis between 2016 and 2022 were followed using data from MHS database, Israel’s second largest Healthcare Maintenance Organisation covering more than 2.7 million Israeli citizens nationwide. The database contains information on diagnoses, use of screening questionnaires, antidepressant prescriptions, including treatment initiation and duration, referral to multidisciplinary teams and data on sociodemographic characteristics of MHS members. Participants The study included all adult patients with a CVA diagnosed between 2016 and 2022 and a new diagnosis of depression or antidepressant treatment initiated after stroke diagnosis. Patients diagnosed with depression or antidepressant treatment prior to stroke diagnosis were excluded from the study (figure 1). Patient selection flowchart summarising the process of study population identification. MHS, Maccabi Health Services Variables All variables defining stroke and depression were extracted from the MHS registries, which included disease diagnosis codes and dates. The definition of depression conformed to one of the following criteria: a formal International Classification of Diseases (ICD) diagnosis based on physicians’ evaluation or prescribing antidepressant initiated post-stroke. Data on the specialty of the diagnosing physician, time gap between stroke and depression diagnosis and use of PHQ-2/PHQ-9 screening questionnaires were collected. Data on total annual prescriptions, dosages and pack sales of antidepressant medications were evaluated for 2016–2022, comprising the study period. The time elapsed between stroke diagnosis and antidepressant prescription was evaluated. Antidepressant class, dosage and timing of treatment initiation and duration were captured. Mental health referrals, psychiatric visits and physiotherapy participation were assessed. Demographic variables included age, sex, socioeconomic status (SES levels categorised as low (1–4), medium (5–6) or high (≥7)) and geographic region (peripheral vs non-peripheral). Comorbidities included hypertension, ischaemic heart disease, congestive heart failure, chronic kidney disease, diabetes, dementia and malignancy. All-cause mortality was derived from the MHS registry. Statistical methods Patients with and without PSD were compared. Categorical data were reported as numbers and percentages in brackets. Continuous variables were reported as mean±SD. To compare categorical data, the χ2 test was used; for continuous data, the t-test/Mann–Whitney test was used (according to the distribution of the data). P-values<0.05 indicated significance. The survival of patients with and without PSD was compared using Kaplan–Meier survival analysis. This method estimated the time from stroke to death for both groups and the probability of survival over time. The differences in survival distributions were statistically evaluated using the log-rank test. A Cox proportional hazards regression model was employed to adjust for potential confounding variables and evaluate the impact of various covariates on time to death. The model included age, sex, SES, district of residence, number of physiotherapy treatments, referral to psychological treatments and comorbid conditions. Comorbid conditions included hypertension, malignancy, congestive heart failure, chronic kidney disease, diabetes mellitus, dementia and cardiovascular disease. The assumption of proportional hazards was checked for each explanatory variable. This model provided HRs, quantifying the relative risk of death associated with PSD after adjusting for other covariates. SPSS V.28 (IBM Corp.) was used. Data availability policy and statement Anonymised data not published as part of the present article will be considered for release on request from qualified investigators, subject to the restrictions of the Helsinki committee. Patient and public involvement None Results Participants Among 11 499 patients with stroke included in the study between 2016 and 2022, 4620 (40.2%) met the criteria for PSD through ICD diagnosis (37.2%) or antidepressant treatment (62.8%). The remaining 6879 (59.8%) did not receive a depression diagnosis or treatment with antidepressant medications. Patients without PSD were older, with a higher proportion of patients aged ≥80 years (22.1% vs 18.7%, p<0.001) and a lower proportion of females compared with PSD patients (38.1% vs 45.8%, p<0.001). Patients without PSD had higher prevalence of chronic conditions, specifically ischaemic heart disease (64.2% vs 58.8%, p<0.001), congestive heart failure (94.8% vs 93.1%, p<0.001), dementia (93.1% vs 81.5%, p<0.001) and chronic kidney disease (86.8% vs 83.5%, p<0.001). Patient characteristics are presented in table 1. No significant differences were observed in SES levels (p=0.076) or district of residence (periphery with 30.1% vs 31.4%, and non-periphery with 69.9% vs 68.6%, p=0.142) between the two patient groups. Table 1Socio-demographic and chronic illness characteristics of patients with post-stroke depression and those without depresssionVariableTotal out 11499CVA patients with depression, n=4620CVA patients without depression, n=6879P valueSex (%) Males6764 (58.8%)2503 (54.2%)4261 (61.9%)<0.001 Females4735 (41.2%)2117 (45.8%)2618 (38.1%)Age group ≤50906 (7.9%)353 (7.6%)553 (8%)<0.001 51–601745 (15.2%)707 (15.3%)1038 (15.1%) 61–703111 (27.1%)1282 (27.7%)1829 (26.6%) 71–803352 (29.2%)1414 (30.6%)1938 (28.2%) >802385 (20.7%)864 (18.7%)1521 (22.1%)SES Low2356 (20.5%)901 (19.5%)1455 (21.2%)0.076 Medium3722 (32.4%)1497 (32.4%)2225 (32.3%) High5421 (47.1%)2222 (48.1%)3199 (46.5%)District of residence Periphery3551 (30.9%)1391 (30.1%)2160 (31.4%)0.142+ Non-periphery7948 (69.1%)3229 (69.9%)4719 (68.6%)Ischaemic heart disease2715 (58.8%)4414 (64.2%)<0.001Congestive heart failure4300 (93.1%)6522 (94.8%)<0.001Chronic kidney disease (GFR< 60 mL/min)3856 (83.5%)5974 (86.8%)<0.001Dementia3763 (81.5%)6405 (93.1%)<0.001Diabetes4392 (95.1%)6568 (95.5%)<0.303Hypertension4245 (91.9%)6366 (92.5%)<0.194Malignancy4277 (92.6%)6368 (92.8%)<0.602CVA, cerebrosvacular accident; GFR, glomerular filtration rate. Socio-demographic and chronic illness characteristics of patients with post-stroke depression and those without depresssion CVA, cerebrosvacular accident; GFR, glomerular filtration rate. Outcome data Diagnosis: PCPs diagnosed a large proportion of patients, accounting for 53.1% of the depression diagnoses, whereas psychiatrists (15.3%), geriatrics (10.1%), neurologists (5.8%) and patients diagnosed in hospital (12.1%) accounted for the rest. Use of standard screening tools leading to further investigation was limited, with only 520 out of 11 499 (4.5%) patients undergoing PHQ-2 assessment. Among those who tested positive on the PHQ-2, only 94 patients completed the PHQ-9 questionnaire. Thus, most diagnoses relied primarily on clinical assessment without preceding utilisation of standard screening tools to aid the diagnosis Timing of diagnosis and treatment: the rate of depression diagnoses was evaluated at defined intervals after stroke: 1, 2 and 3 months, 4–6 months, 7–12 months, 2 years and ≥2 years post-event. The majority of PSD diagnoses (53.4%) occurred within the first year post-stroke, with 37.2% occurring during the first 4–6 months after stroke. Among patients with a formal ICD diagnosis, 92.9% received antidepressant therapy, with 65.1% initiating treatment within 30 days of diagnosis. SSRIs comprised 58.3% of prescriptions, followed by mirtazapine (12%), serotonin–norepinephrine reuptake inhibitors (10.9%), tricyclic antidepressants (9%) and trazodone (3%). Referral patterns: out of 11 499 patients with stroke, only 1105 (9.6%) received mental health evaluation, nearly all in the PSD group (16.9% vs 3.4% respectively, p<0.001). Virtual consultations comprised 5.3% of mental health visits. No statistically significant difference was found when comparing mental health consultations according to SES (9.1% vs 10.3%% vs 9.4%, in low, intermediate and high SES, respectively, p=0.213), nor according to district of residence (peripheral vs non-peripheral, 39.1% vs 40.6%, p=0.142). Rehabilitation patterns: we examined patients’ participation in different rehabilitation options, specifically the number of physiotherapy visits. Patients with PSD had significantly higher physiotherapy participation (69.5% vs 48.5%, p<0.001). Mortality: all-cause mortality was lower in patients with PSD than in those without depression (22.1% vs 27.9%, p<0.001). Kaplan–Meier survival analysis confirmed this survival advantage, with the log-rank test (χ² = 280.3, p<0.001) (figure 2). Mortality rates were similar between patients with a formal ICD diagnosis and those receiving antidepressants without diagnostic codes (22.2% vs 21.2%, p=0.472). Kaplan–Meier plot comparing survival rates of patients with and without post-stroke depression. In a Cox regression analysis, age was a significant predictor of mortality, with an HR of 1.081 (p<0.001). Several comorbid conditions, including hypertension, cancer, ischaemic heart disease, diabetes and dementia, were major predictors of mortality, each associated with a HR >1 (table 2). Specifically, the presence of each condition significantly elevated the risk of mortality, underscoring their impact on survival outcomes. Both physiotherapy visits (HR=0.625, p<0.001) and mental health consultations (HR=0.642, p<0.001) emerged as protective factors. Table 2Variables in the Cox analysisVariablesp-valueHR95.0% CILowerUpperAge<0.0011.0811.0771.085Sex (male)0.0901.0670.9901.151District of residence0.2430.9510.8751.034Socio-economic status<0.001Low–medium0.0090.8760.7930.968High<0.0010.6810.6150.755Hypertension<0.0012.6272.0433.378Malignancy0.1241.1170.9701.285Congestive heart failure<0.0010.7580.6710.856Chronic kidney disease<0.0011.8981.6612.170Diabetes<0.0012.0181.5202.679Cerebrovascular disease0.0041.1811.0541.323Dementia<0.0011.7171.5161.945Coronary artery disease<0.0011.5411.3531.754Physiotherapy visits<0.0010.6250.5800.674Mental health visits<0.0010.6420.5330.773Depression after stroke<0.0010.8370.7710.909 Variables in the Cox analysis The overall model fit was assessed using the omnibus test of model coefficients, which yielded a −2 log-likelihood χ2 of 2998.6 (p<0.001), indicating that the model significantly explained the variability in survival times. The Cox-adjusted survival curves (figure 3) for patients with and without PSD mirrored the Kaplan–Meier curves (figure 2) but demonstrated a slightly narrower gap between the two groups, particularly after adjusting for covariates Cox model showing survival function at mean of covariates of patients with and without post-stroke depression. Discussion Main findings This study of 11 499 patients with stroke reveals that PCPs diagnose most PSD cases based on clinical evaluation without prior formal screening, with most diagnoses occurring during the first year after stroke. SSRIs were the main treatment of choice for depression. In this study, patients with PSD were younger, with lower rates of chronic diseases. They received prompt antidepressant treatment and had a significantly higher number of rehabilitation visits and mental health evaluations, factors that could potentially contribute to lower mortality rates in this patient group. Interpretation In this study, 40.2% of patients with stroke were diagnosed with PSD, slightly exceeding published estimates of approximately 30%,13 likely reflecting our inclusive definition that captured both ICD-based diagnoses and antidepressant treatment. In our view, this approach better represents real-world practice, where PCPs may empirically treat suspected depression without assigning a formal diagnosis, particularly given the diagnostic significance of treatment response in borderline and uncertain cases. Both Israeli and international guidelines for CVA management emphasise the importance of PSD screening during the acute post-stroke period,1316 which is usually conducted in institutionalised and specialised clinics.12 The guidelines differ regarding the optimal frequency, setting and timing for repeated evaluation.1311 12 PCPs are in a pivotal place to diagnose PSD owing to their long therapeutic alliance with their patients. In this study, PCPs diagnosed 53.1% of PSD cases, underscoring their central role in post-stroke mental health surveillance. Despite guideline recommendations for structured screening with one of the available screening tools, only 4.5% of patients underwent PHQ-2 assessment, and the majority were diagnosed based on clinical criteria. Reliance on clinical evaluation alone might have missed atypical presentations, particularly in older and frailer adults with multiple comorbidities leading to underdiagnosis of the condition. In this study, most patients were diagnosed with depression during the first year after stroke (53.4%), with 37.2% diagnosed during the first 4–6 months post-stroke, suggesting that repeated screening at 3, 6 and 12 months of follow-up might be appropriate. Nearly universal treatment of formally ICD-diagnosed patients (92.9%), starting mainly within the first month after diagnosis, aligns with evidence demonstrating the therapeutic benefits of early medical intervention.1719 These practices may also account for the higher rehabilitation participation among patients with PSD compared to a non-PSD patient group, as untreated PSD is highly correlated with the inability to resume pre-stroke social and physical activities.6 20 21 According to current literature, patients with PSD have worse mortality outcomes.1 3 17 21 22 However, our study found opposite results. Several factors may account for this discrepancy. First, patients without PSD were older, with substantially higher comorbidity burden, potentially leading to higher mortality. In addition, PSD symptomatology differs from typical depression, presenting with emotional dysregulation and less anhedonia.6 23 24 Combined with physical disability and cognitive impairment, these factors may have led to misattribution of depressive symptoms to medical comorbidities, particularly in older and frailer adults, leading to underdiagnosis of the condition. Second, our ambulatory stroke population likely represents less severe stroke with better functional status. Although some studies have shown that the extent of physical damage does not always correlate with the severity of PSD and depends on stroke location,2325 others demonstrate an association between stroke severity, related disability and reduced survival.137 20 21 26 27 Third, near-universal antidepressant treatment, which has a positive effect on dependence, disability, neurological impairment, anxiety and depression in patients with stroke,171928 and rehabilitation participation may have mitigated depression’s negative impact.17 Finally, both physiotherapy (HR=0.625) and mental health consultations (HR=0.642) emerged as protective factors in adjusted analyses, suggesting that an active multidisciplinary care approach may also contribute to reduced mortality rate. Patients diagnosed with PSD may represent a subgroup with more regular clinical follow-up, and this greater engagement with healthcare services may confer a survival advantage, introducing a form of surveillance bias into the observed mortality outcomes. Strengths and limitations This study’s strengths include its large sample (11 499 patients; 4620 with PSD), 6 years of follow-up and comprehensive capture of real-world primary care practices beyond acute hospitalisation. The large health system database provided detailed longitudinal data on diagnoses, treatments, healthcare utilisation and mortality. Several limitations of this study warrant consideration. First, the retrospective design is associated with potential information bias and incomplete documentation. Second, our depression definition included patients receiving antidepressants without formal diagnostic code, thereby potentially including individuals treated for other psychiatric conditions such as anxiety, adjustment disorder, insomnia and appetite disturbances. However, these conditions frequently overlap with and may constitute components of depressive disorder as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, criteria. Furthermore, no significant difference in mortality was observed between the subgroups identified by ICD diagnostic codes and the subgroup initiated on antidepressant therapy following stroke, lending some validity to this broader definition. Finally, the study lacked stroke severity data, which may have influenced both depression risk and mortality outcomes, and introduced residual confounding into our analysis. The approach described in this study reflects the characteristics of the study sample and the healthcare setting in which it was conducted and may not be directly generalisable to other healthcare systems or populations. That being said, reporting these findings remains valuable, as it provides insight into real-world clinical practice within our setting and may serve as a reference point for comparison with other outpatient healthcare models. Conclusions In this cohort of ambulatory stroke survivors, in primary care setting, PSD diagnosis relies predominantly on clinical judgement, with most cases identified within the first year after stroke. The mortality benefit among patients with PSD likely reflects prompt antidepressant treatment, enhanced rehabilitation engagement and lower baseline comorbidity. Substantial underdiagnosis in older adults with multiple comorbidities remains likely. Systematic screening questionnaires implementation and repeated testing during the first year after stroke represent an opportunity for improvement in PSD detection in primary care.