Abstract
Objective
This study aimed to evaluate postdischarge class-complete secondary prevention medication nonadherence after ischemic stroke and identify overall and comorbidity-specific factors associated with 180-day nonadherence.
Methods
We conducted a retrospective cohort study using prospectively collected follow-up data from a stroke registry, including consecutive patients hospitalized with acute ischemic stroke. Medication use was assessed at discharge, 90 days, and 180 days after discharge. Class-complete discontinuation—that is, discontinuation of at least one of the secondary prevention medication classes prescribed at discharge, including antiplatelet agents, anticoagulants, lipid-lowering medications, antihypertensive medications, and glucose-lowering medications—was the primary outcome. Any-use adherence, defined as continued use of at least one secondary prevention medication class, was used for comparison. Multivariable logistic regression was performed to identify factors associated with 180-day class-complete nonadherence, with subgroup analyses among patients with hypertension and diabetes mellitus.
Results
A total of 2549 patients with ischemic stroke were included. At 180 days, 1408 patients (55.2%) had discontinued at least one prescribed secondary prevention medication class, whereas only 738 (29.0%) had discontinued all prescribed classes; 1141 (44.8%) continued every class prescribed at discharge. Class-complete adherence declined from 59.4% at 90 days to 44.8% at 180 days, whereas any-use adherence was 85.9% and 71.0%, respectively. Each additional discharge medication class was associated with higher odds of 180-day class-complete nonadherence in a prespecified multivariable model (adjusted odds ratio [OR], 1.50; 95% CI, 1.27–1.79). Cardioembolism (OR, 2.11; 95% CI, 1.37–3.26), dyslipidemia (OR, 1.50; 95% CI, 1.09–2.07), and large-artery atherosclerosis (OR, 1.24; 95% CI, 1.03–1.49) were also associated with nonadherence; hypertension was of borderline significance (OR, 1.23; 95% CI, 0.99–1.54; P = 0.067). In subgroup analyses, the association with number of discharge medication classes was similar across the hypertension and diabetes mellitus subgroups, with no evidence of effect modification (P for interaction = 0.32 and 0.67), whereas the association with large-artery atherosclerosis was stronger among patients with hypertension (OR, 1.37; 95% CI, 1.08–1.73) or diabetes mellitus (OR, 1.43; 95% CI, 1.10–1.87) than among those without (P for interaction = 0.036 and 0.009, respectively).
Conclusion
Class-complete medication nonadherence was common within 180 days after ischemic stroke discharge and often reflected partial discontinuation rather than complete cessation of secondary prevention therapy. Medication-class–specific follow-up may help identify incomplete regimen continuation and guide targeted support for patients with greater treatment complexity, atherosclerotic stroke mechanisms, or chronic vascular comorbidities.
Keywords: ischemic stroke, secondary prevention, medication adherence, medication nonadherence, class-complete adherence, postdischarge follow-up, hypertension, diabetes mellitus
Introduction
Ischemic stroke remains a leading cause of death and disability worldwide, and survivors continue to face a substantial risk of recurrent vascular events after hospital discharge.1 Guideline-recommended secondary prevention depends on sustained use of multiple medication classes, including antithrombotic therapy, lipid-lowering therapy, antihypertensive therapy, and glucose-lowering therapy when clinically indicated, with recent registry data indicating a 90-day recurrence risk of approximately 5%, and long-term recurrence rates reaching up to 12.3%.2–4 Although discharge prescriptions establish the intended prevention regimen, the effectiveness of secondary prevention depends on whether patients continue these medications during the transition from hospital-based care to long-term community management.
Existing studies of poststroke medication adherence have often focused on longitudinal trajectories of individual medication classes, such as antithrombotics or statin, highlighting suboptimal adherence and significant discontinuation rates within the first year.5,6 These approaches may not capture whether patients maintain the complete set of secondary prevention medication classes prescribed at discharge. For example, a patient who continues antiplatelet therapy but discontinues lipid-lowering or antihypertensive therapy may still have some ongoing secondary prevention treatment but has not fully continued the intended regimen. These comorbidities are highly prevalent among stroke survivors and significantly amplify the risk of recurrence; for instance, diabetes has been shown to independently increase the risk of recurrent disabling or fatal stroke by more than twofold (hazard ratio, 2.23),6 and is associated with an elevated 5-year recurrence risk of 12.7%.7 However, real-world evidence describing class-complete medication nonadherence and its comorbidity-specific patterns after ischemic stroke remains limited.
Established adherence measures quantify different aspects of medication-taking behavior. Pharmacy-based measures such as the proportion of days covered and the medication possession ratio quantify the supply of a single medication over time; self-report scales quantify reported medication-taking behavior; and persistence measures quantify the time until a given medication is stopped. Each of these is applied to one medication or one medication class at a time, and none indicates whether the complete set of classes prescribed at discharge is still being taken. A class-complete measure addresses this specific question and is therefore complementary to, rather than a replacement for, established measures.
Therefore, we conducted a retrospective analysis of prospectively collected follow-up registry data from patients hospitalized with ischemic stroke. We assessed secondary prevention medication use at 90 and 180 days after discharge and defined class-complete medication nonadherence as discontinuation of at least one secondary prevention medication class prescribed at discharge. We aimed to characterize postdischarge adherence trajectories, identify factors associated with 180-day class-complete medication nonadherence, and describe adherence patterns among patients with hypertension or diabetes mellitus. We hypothesized that class-complete nonadherence would be substantially more frequent than complete cessation of all secondary prevention classes, and that patients discharged with more medication classes would more often discontinue at least one class while remaining on some form of secondary prevention.
Methods
Study Design and Data Source
This was a retrospective cohort study based on a prospectively enrolled stroke follow-up cohort. Patient information was obtained from the stroke follow-up platform of the Shenzhen Clinical Research Center for Neurological Diseases and the clinical information system of the People’s Hospital of Longhua, Shenzhen. Patients with ischemic stroke onset between February 2021 and May 2022 were included.
The primary outcome was 180-day class-complete medication nonadherence. Class-complete adherence was defined as continued use of all secondary prevention medication classes prescribed at discharge through the 180-day follow-up period. Class-complete nonadherence was defined as discontinuation of at least one secondary prevention medication class prescribed at discharge during follow-up, regardless of the total number of medication classes prescribed at discharge.
This study was approved by the Medical Ethics Committee of the People’s Hospital of Longhua, Shenzhen, China (approval number [2025] 081). Patient data were analyzed retrospectively and were de-identified before analysis. The committee waived the requirement for informed consent because of the retrospective design of the study.
Study Population
Patients were eligible for inclusion if they were aged 18 years or older and had ischemic stroke confirmed by computed tomography or magnetic resonance imaging. Patients whose diagnosis was changed during hospitalization to hemorrhagic cerebrovascular disease or another nonischemic condition were not entered into the registry cohort and are therefore not represented in the numbers below. Among the patients screened, those for whom medication adherence outcome information was unavailable or whose 180-day follow-up assessment was not completed were excluded. A total of 2896 patients were initially screened. Of these, 201 were excluded because they were lost to follow-up or because outcome information was unavailable, leaving 2695 eligible patients. A further 146 were excluded because no secondary prevention medication was prescribed at discharge (n = 126) or discharge medication data were missing (n = 20), for whom the outcome is undefined; these patients had more severe strokes than those included (mean NIHSS 8.0 vs 3.6) and were older (mean 63.0 vs 60.5 years). The final analytic cohort comprised 2549 patients, among whom 180-day medication status was available for all; no further loss to follow-up occurred. Baseline data were not available for the 201 patients excluded before cohort assembly, so these patients could not be compared with those retained. A comparison of the 2549 included and 146 excluded patients is provided in Supplementary Table S1. Among the 2549 patients in the final analytic cohort, 1141 were class-complete adherent and 1408 were class-complete nonadherent at 180 days. For multivariable logistic regression analyses, 2545 patients were included because 4 patients with missing or unclassifiable Trial of Org 10172 in Acute Stroke Treatment (TOAST) subtype were excluded from models involving stroke etiology (Figure 1).
Figure 1.
Flowchart of study population selection. Patients hospitalized with acute ischemic stroke and enrolled in the stroke follow-up registry of the People’s Hospital of Longhua, Shenzhen, with onset between February 2021 and May 2022, were assessed for eligibility (N = 2896). Of these, 201 were excluded because they were lost to follow-up or because outcome information was unavailable, and a further 146 were excluded because no secondary prevention medication was prescribed at discharge (n = 126) or discharge medication data were missing (n = 20), for whom the outcome is undefined. The final analytic cohort comprised 2549 patients, of whom 2545 were included in the multivariable models (4 patients with missing or unclassifiable TOAST subtype were excluded from models involving stroke etiology). Class-complete nonadherence was defined as discontinuation of at least one secondary prevention medication class prescribed at discharge within 180 days; class-complete adherence was defined as continued use of all prescribed classes. Percentages are of the final analytic cohort.
Assessment of Medication Adherence
Secondary prevention medication classes included antiplatelet agents, anticoagulants, lipid-lowering medications, antihypertensive medications, and glucose-lowering medications. Discharge medication classes were determined from physician-issued discharge prescriptions. Medication use at 90 and 180 days after discharge was ascertained by neurologists and dedicated follow-up nurses through structured assessments conducted either face-to-face in the outpatient clinic or by telephone. For patients assessed in the outpatient clinic, reported medication use was cross-checked against hospital prescription records. For patients assessed by telephone, the use of each prescribed medication class was ascertained item by item, using the discharge medication list and the medication record from the preceding follow-up assessment as the reference framework. Medication adherence was assessed at the medication-class level. Continued use of any medication within the same class was considered continuation of that class. Changes in dose were not considered nonadherence if the medication class was continued. If a medication class prescribed at discharge was not being used at the follow-up time point, that class was recorded as discontinued. Patients with missing follow-up medication information were excluded from the analysis.
Outcomes
The primary outcome was 180-day class-complete medication nonadherence after discharge. Class-complete adherence was defined as continued use of all secondary prevention medication classes prescribed at discharge through the follow-up time point. Class-complete nonadherence was defined as discontinuation of at least one secondary prevention medication class prescribed at discharge by the follow-up time point. A secondary adherence definition, any-use adherence, was used to describe overall continuation of secondary prevention therapy. Any-use adherence was defined as continued use of at least one secondary prevention medication class at the follow-up time point. Under this definition, patients were considered nonadherent only if all secondary prevention medication classes were discontinued. The class-complete adherence definition was used to capture partial discontinuation of prescribed secondary prevention medication classes, whereas the any-use adherence definition was used to distinguish complete cessation of secondary prevention therapy from partial discontinuation. Medication adherence was assessed at both 90 days and 180 days after discharge. The multivariable regression analysis used 180-day class-complete medication nonadherence as the primary outcome. For patients discharged with only one secondary prevention medication class, class-complete adherence and any-use adherence were equivalent and were determined according to whether that medication class was continued at follow-up. Consistent with the concept of persistence used in previous secondary prevention studies, discontinuation was defined without reference to the reason for stopping. Registry follow-up did not record whether a medication class was stopped by the patient or on the advice of a clinician; therefore, clinically appropriate discontinuation—for example, withdrawal of a glucose-lowering agent after glycemic normalization or of an anticoagulant because of bleeding risk—would also be counted as discontinuation under this definition. The outcome name (class-complete nonadherence) and the manuscript title are unchanged; this statement clarifies what the existing definition counts, rather than introducing a new outcome.
Covariates and Subgroups
Baseline covariates were obtained from the hospital information system, laboratory information system, medical record system, and the neurological disease follow-up registry platform. Demographic and socioeconomic variables included age, sex, ethnicity, living alone, health insurance status, education level, current smoking, and current drinking. Medical history and vascular risk factors included hypertension, diabetes mellitus, dyslipidemia, coronary heart disease, atrial fibrillation, prior cerebral infarction, prior transient ischemic attack, heart failure, peripheral artery disease, and stroke-associated pneumonia. Stroke severity and functional status variables included baseline National Institutes of Health Stroke Scale score, baseline Glasgow Coma Scale score, prestroke modified Rankin Scale score, discharge National Institutes of Health Stroke Scale score, and discharge modified Rankin Scale score. Laboratory variables included high-sensitivity C-reactive protein, low-density lipoprotein cholesterol, and serum creatinine. Stroke etiology was classified according to the TOAST criteria. TOAST categories included large-artery atherosclerosis, small-vessel occlusion, cardioembolism, other determined etiology, and undetermined etiology. In regression models, small-vessel occlusion was used as the reference category. In-hospital treatment and discharge-related variables included endovascular therapy, intravenous thrombolysis, difficult oral intake, nasogastric tube placement, discharge within 7 days after onset, antiinfective treatment, and the number of secondary prevention medication classes prescribed at discharge. Subgroup analyses were performed among patients with hypertension and among patients with diabetes mellitus. Hypertension and diabetes mellitus were defined according to documented medical history or discharge diagnosis. These subgroups were not mutually exclusive; therefore, patients with both conditions could be included in both subgroup analyses.
Statistical Analysis
Continuous variables were summarized as medians with interquartile ranges, and categorical variables were summarized as numbers and percentages. Baseline characteristics were compared between patients who were class-complete adherent and those who were class-complete nonadherent at 180 days. Continuous variables were compared using the Mann–Whitney U-test. Categorical variables were compared using the χ2 test or Fisher exact test, as appropriate. Medication adherence rates were calculated at discharge, 90 days, and 180 days. Adherence trajectories were described separately for the overall ischemic stroke cohort, patients with hypertension, and patients with diabetes mellitus. Adherence was calculated using both the class-complete adherence definition and the any-use adherence definition. Covariates for the primary multivariable model were prespecified on clinical grounds and prior literature rather than selected by statistical screening, and comprised age, hypertension, diabetes mellitus, dyslipidemia, coronary heart disease, atrial fibrillation, peripheral artery disease, baseline NIHSS score, prestroke modified Rankin Scale score, serum creatinine, TOAST subtype (small-vessel occlusion as reference), endovascular therapy, antiinfective treatment, and the number of secondary prevention medication classes prescribed at discharge. All prespecified covariates were entered simultaneously; no data-driven selection was applied. The previously used approach—univariable screening at P < 0.10 followed by backward elimination—was retained as a sensitivity analysis (Supplementary Table S2). Adjusted odds ratios with 95% confidence intervals were reported for all covariates included in the prespecified model. Statistical significance in the final model was defined as a 2-sided P < 0.05. For subgroup analyses, logistic regression models were fitted separately among patients with hypertension and among patients with diabetes mellitus. The same covariates included in the overall prespecified model (omitting the stratifying comorbidity itself) were used in subgroup models without subgroup-specific variable re-selection. Subgroup analyses were exploratory. Because the subgroup analysis in the original submission involved 24 comparisons (8 factors in each of 3 cohorts), the Benjamini–Hochberg procedure was applied to the prespecified subgroup-related comparisons, and adjusted q values are reported alongside unadjusted P values. No adjustment was applied to the primary model. If a subgroup-defining variable was constant within a subgroup, it was omitted from that subgroup model. Missing covariate data were handled using multiple imputation by chained equations. All variables included in the analysis had less than 10% missingness before imputation. Patients with missing outcome information or incomplete 180-day follow-up were excluded from the analytic cohort. All analyses were performed using Python 3.14 in Jupyter Notebook. A 2-sided P value less than 0.05 was considered statistically significant for final multivariable analyses. The registry recorded whether recurrent stroke occurred within 180 days but did not record dates of death or of recurrent events. Time-to-event and competing-risk models therefore could not be fitted. As the analytic cohort was by definition restricted to patients alive and assessed at 180 days, and as the objective was to identify factors associated with discontinuation rather than to estimate its incidence, discontinuation was modeled with logistic regression at fixed follow-up time points. A sensitivity analysis excluding patients with recurrent stroke within 180 days was performed. No a priori sample size calculation was performed because the study included all eligible patients in the registry during the enrolment period. In the primary model, 1405 events were available for 17 candidate covariates, corresponding to 82.6 events per variable, well above the commonly recommended minimum of 10 and indicating that the model was not at risk of overfitting. Model performance of the prespecified primary model was assessed using the C statistic and a calibration plot with bootstrap correction for optimism (1000 resamples).
Results
Study Population and Baseline Characteristics
Among 2549 patients with ischemic stroke included in the analytic cohort, 1141 patients were class-complete adherent and 1408 patients were class-complete nonadherent at 180 days after discharge, corresponding to a 180-day class-complete nonadherence rate of 55.2% (Figure 1 and Table 1).
Table 1.
Baseline Characteristics According to 180-Day Class-Complete Medication Adherence Status
| Characteristic | Overall (N = 2549) |
Adherent (n = 1141) |
Nonadherent (n = 1408) |
P Value |
|---|---|---|---|---|
| Age, years, median (IQR) | 59.00 (51.00–70.00) | 59.00 (51.00–69.00) | 60.00 (51.75–70.00) | 0.073 |
| Male sex, No. (%) | 1821 (71.4) | 826 (72.4) | 995 (70.7) | 0.360 |
| Han ethnicity, No. (%) | 2523 (99.0) | 1125 (98.6) | 1398 (99.3) | 0.126 |
| Living alone, No. (%) | 99 (3.9) | 40 (3.5) | 59 (4.2) | 0.432 |
| Health insurance, No. (%) | 1764 (69.2) | 774 (67.8) | 990 (70.3) | 0.192 |
| Education level | 0.549 | |||
| Primary or below | 75 (2.9) | 37 (3.2) | 38 (2.7) | |
| Middle school | 1404 (55.1) | 620 (54.3) | 784 (55.7) | |
| High school | 792 (31.1) | 366 (32.1) | 426 (30.3) | |
| College or above | 278 (10.9) | 118 (10.3) | 160 (11.4) | |
| Current smoking, No. (%) | 895 (35.1) | 413 (36.2) | 482 (34.2) | 0.322 |
| Current drinking, No. (%) | 541 (21.2) | 244 (21.4) | 297 (21.1) | 0.897 |
| Hypertension, No. (%) | 1563 (61.3) | 609 (53.4) | 954 (67.8) | <0.001 |
| Diabetes mellitus, No. (%) | 1243 (48.8) | 501 (43.9) | 742 (52.7) | <0.001 |
| Dyslipidemia, No. (%) | 189 (7.4) | 67 (5.9) | 122 (8.7) | 0.009 |
| Coronary heart disease, No. (%) | 354 (13.9) | 124 (10.9) | 230 (16.3) | <0.001 |
| Atrial fibrillation, No. (%) | 107 (4.2) | 35 (3.1) | 72 (5.1) | 0.014 |
| Prior cerebral infarction, No. (%) | 2337 (91.7) | 1049 (91.9) | 1288 (91.5) | 0.730 |
| Prior TIA, No. (%) | 227 (8.9) | 97 (8.5) | 130 (9.2) | 0.565 |
| Heart failure, No. (%) | 2 (0.1) | 0 (0.0) | 2 (0.1) | 0.505 |
| Peripheral artery disease, No. (%) | 124 (4.9) | 45 (3.9) | 79 (5.6) | 0.064 |
| Stroke-associated pneumonia, No. (%) | 36 (1.4) | 11 (1.0) | 25 (1.8) | 0.119 |
| Baseline NIHSS score, median (IQR) | 2.00 (1.00–4.00) | 2.00 (1.00–4.00) | 2.00 (1.00–4.00) | 0.081 |
| Baseline GCS score, median (IQR) | 15.00 (15.00–15.00) | 15.00 (15.00–15.00) | 15.00 (15.00–15.00) | 0.253 |
| Prestroke mRS score, median (IQR) | 0.00 (0.00–1.00) | 0.00 (0.00–1.00) | 0.00 (0.00–1.00) | 0.042 |
| Discharge NIHSS score, median (IQR) | 1.00 (0.00–3.00) | 1.00 (0.00–3.00) | 1.00 (0.00–3.00) | 0.871 |
| Discharge mRS score, median (IQR) | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) | 0.456 |
| High-sensitivity C-reactive protein, mg/L, median (IQR) | 1.50 (0.60–3.20) | 1.50 (0.63–3.20) | 1.50 (0.57–3.20) | 0.465 |
| LDL cholesterol, mmol/L, median (IQR) | 2.91 (2.30–3.54) | 2.91 (2.30–3.59) | 2.91 (2.31–3.49) | 0.284 |
| Serum creatinine, μmol/L, median (IQR) | 75.60 (64.00–89.00) | 75.30 (63.90–87.20) | 75.60 (64.00–90.05) | 0.074 |
| Stroke etiology (TOAST classification) | 0.001 | |||
| Large-artery atherosclerosis | 975 (38.3) | 412 (36.1) | 559 (39.8) | |
| Small-vessel occlusion (reference) | 1047 (41.1) | 505 (44.3) | 542 (38.6) | |
| Cardioembolism | 132 (5.2) | 43 (3.8) | 89 (6.3) | |
| Other determined etiology | 241 (9.5) | 102 (8.9) | 139 (9.9) | |
| Undetermined etiology | 154 (6.0) | 78 (6.8) | 76 (5.4) | |
| Endovascular therapy, No. (%) | 145 (5.7) | 50 (4.4) | 95 (6.7) | 0.013 |
| Intravenous thrombolysis, No. (%) | 421 (16.5) | 186 (16.3) | 235 (16.7) | 0.834 |
| Difficult oral intake, No. (%) | 121 (4.7) | 55 (4.8) | 66 (4.7) | 0.950 |
| Nasogastric tube placement, No. (%) | 81 (3.2) | 38 (3.3) | 43 (3.1) | 0.778 |
| Discharged within 7 days, No. (%) | 1064 (41.7) | 488 (42.8) | 576 (40.9) | 0.365 |
| Antiinfective treatment, No. (%) | 116 (4.6) | 37 (3.2) | 79 (5.6) | 0.006 |
| No. of secondary prevention medication classes at discharge, median (IQR) | 3.00 (2.00–3.00) | 3.00 (2.00–3.00) | 3.00 (2.00–3.00) | <0.001 |
Notes: Data are No. (%) or median (IQR); Mann–Whitney U, χ2, or Fisher exact test.
Abbreviations: GCS, Glasgow Coma Scale; IQR, interquartile range; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; TIA, transient ischemic attack; TOAST, Trial of Org 10172 in Acute Stroke Treatment.
Baseline characteristics according to 180-day class-complete medication adherence status are shown in Table 1. Compared with adherent patients, nonadherent patients had higher proportions of hypertension [954 of 1408 (67.8%) vs 609 of 1141 (53.4%); P < 0.001], diabetes mellitus [742 of 1408 (52.7%) vs 501 of 1141 (43.9%); P < 0.001], dyslipidemia [122 of 1408 (8.7%) vs 67 of 1141 (5.9%); P = 0.009], coronary heart disease [230 of 1408 (16.3%) vs 124 of 1141 (10.9%); P < 0.001], and atrial fibrillation [72 of 1408 (5.1%) vs 35 of 1141 (3.1%); P = 0.014]. Peripheral artery disease was more frequent in the nonadherent group than in the adherent group [79 of 1408 (5.6%) vs 45 of 1141 (3.9%)], although the difference did not reach statistical significance (P = 0.064). Baseline NIHSS score and serum creatinine also differed numerically between groups but did not reach the prespecified P < 0.05 threshold. The distribution of stroke etiology differed between the adherent and nonadherent groups (P = 0.001). Large-artery atherosclerosis accounted for 559 of 1408 nonadherent patients (39.8%) and 412 of 1141 adherent patients (36.1%), whereas cardioembolism accounted for 89 nonadherent patients (6.3%) and 43 adherent patients (3.8%). Endovascular therapy was more common in the nonadherent group than in the adherent group [95 of 1408 (6.7%) vs 50 of 1141 (4.4%); P = 0.013], as was antiinfective treatment [79 of 1408 (5.6%) vs 37 of 1141 (3.2%); P = 0.006]. The number of secondary prevention medication classes prescribed at discharge also differed between groups (P < 0.001) (Table 1).
Medication Adherence Trajectories from Discharge to 180 Days
Medication adherence trajectories from discharge to 180 days are shown in Figure 2. Under the class-complete adherence definition, adherence in the overall ischemic stroke cohort was 100.0% at discharge, 59.4% at 90 days, and 44.8% at 180 days. Among patients with hypertension, class-complete adherence was 100.0% at discharge, 52.9% at 90 days, and 39.0% at 180 days. Among patients with diabetes mellitus, class-complete adherence was 100.0% at discharge, 53.8% at 90 days, and 40.3% at 180 days.
Figure 2.
Trajectories of secondary prevention medication adherence from discharge to 180 days after ischemic stroke. (A) Class-complete adherence, defined as continued use of all secondary prevention medication classes prescribed at discharge. (B) Any-use adherence, defined as continued use of at least one prescribed secondary prevention medication class. Each panel shows the overall cohort and the hypertension and diabetes mellitus subgroups at discharge, 90 days, and 180 days.
Under the any-use adherence definition, adherence in the overall cohort was 100.0% at discharge, 85.9% at 90 days, and 71.0% at 180 days. Among patients with hypertension, any-use adherence was 100.0% at discharge, 85.5% at 90 days, and 72.0% at 180 days. Among patients with diabetes mellitus, any-use adherence was 100.0% at discharge, 84.7% at 90 days, and 73.2% at 180 days. At 180 days, the difference between any-use adherence and class-complete adherence was 26.2 percentage points in the overall cohort, 33.0 percentage points among patients with hypertension, and 32.9 percentage points among patients with diabetes mellitus.
Factors Associated with 180-Day Class-Complete Medication Nonadherence
The multivariable logistic regression model included 2545 patients, of whom 1405 patients (55.2%) had 180-day class-complete medication nonadherence. Results are shown in Table 2. In the prespecified model, each additional secondary prevention medication class prescribed at discharge was associated with higher odds of 180-day class-complete medication nonadherence (adjusted OR, 1.50; 95% CI, 1.27–1.79; P < 0.001). Compared with small-vessel occlusion, cardioembolism was associated with higher odds of nonadherence (adjusted OR, 2.11; 95% CI, 1.37–3.26; P < 0.001), as was large-artery atherosclerosis (adjusted OR, 1.24; 95% CI, 1.03–1.49; P = 0.025). Dyslipidemia was also associated with higher odds of nonadherence (adjusted OR, 1.50; 95% CI, 1.09–2.07; P = 0.014). Hypertension was of borderline significance (adjusted OR, 1.23; 95% CI, 0.99–1.54; P = 0.067). The apparent C statistic for this model was 0.627 (95% CI, 0.605–0.649); after bootstrap correction for optimism (1000 resamples), the C statistic was 0.613 (95% CI, 0.593–0.635), the calibration slope was 0.883 (95% CI, 0.713–1.092), and the calibration intercept was 0.022 (95% CI, −0.078–0.117) (Supplementary Figure S1). In a sensitivity analysis using the previous approach of univariable screening at P < 0.10 followed by backward elimination, estimates were closely similar; the only material difference was hypertension, which reached significance in the backward model (adjusted OR, 1.24; 95% CI, 1.00–1.54; P = 0.045) (Supplementary Table S2). In a sensitivity analysis excluding the 20 patients with recurrent stroke within 180 days (n = 2525; 1395 events), the findings were unchanged (number of discharge medication classes: adjusted OR, 1.51; 95% CI, 1.27–1.80) (Supplementary Table S2).
Table 2.
Factors Associated with 180-Day Class-Complete Medication Nonadherence: Prespecified Multivariable Logistic Regression (Primary Model)
| Factor | With characteristic, No./total (%) | Without characteristic, No./total (%) | Adjusted OR (95% CI) | P Value |
|---|---|---|---|---|
| Demographic and clinical characteristics | ||||
| Age (per 10-year increase) | 562/966 (58.2) | 843/1579 (53.4) | 1.01 (0.95–1.07) | 0.82 |
| Hypertension | 953/1561 (61.1) | 452/984 (45.9) | 1.23 (0.99–1.54) | 0.07 |
| Diabetes mellitus | 741/1242 (59.7) | 664/1303 (51.0) | 0.94 (0.77–1.14) | 0.52 |
| Dyslipidemia | 121/188 (64.4) | 1284/2357 (54.5) | 1.50 (1.09–2.07) | 0.01 |
| Coronary heart disease | 230/354 (65.0) | 1175/2191 (53.6) | 1.18 (0.90–1.53) | 0.23 |
| Atrial fibrillation | 72/107 (67.3) | 1333/2438 (54.7) | 1.18 (0.75–1.88) | 0.48 |
| Peripheral artery disease | 78/123 (63.4) | 1327/2422 (54.8) | 1.22 (0.83–1.82) | 0.32 |
| Stroke severity and functional status | ||||
| Baseline NIHSS (per 5-point increase) | 341/607 (56.2) | 1064/1938 (54.9) | 1.02 (0.92–1.13) | 0.73 |
| Prestroke mRS score | 180/373 (48.3) | 1225/2172 (56.4) | 0.96 (0.89–1.04) | 0.33 |
| Serum creatinine (per 10 μmol/L) | 194/328 (59.1) | 1211/2217 (54.6) | 1.01 (1.00–1.03) | 0.16 |
| Stroke etiology (reference, small-vessel occlusion) | ||||
| Large-artery atherosclerosis | 559/971 (57.6) | 846/1574 (53.7) | 1.24 (1.03–1.49) | 0.02 |
| Cardioembolism | 89/132 (67.4) | 1316/2413 (54.5) | 2.11 (1.37–3.26) | <0.001 |
| Other determined etiology | 139/241 (57.7) | 1266/2304 (54.9) | 1.31 (0.98–1.76) | 0.07 |
| Undetermined etiology | 76/154 (49.4) | 1329/2391 (55.6) | 1.00 (0.70–1.41) | 0.99 |
| In-hospital treatment | ||||
| Endovascular therapy | 94/144 (65.3) | 1311/2401 (54.6) | 1.36 (0.91–2.06) | 0.14 |
| Antiinfective treatment | 79/116 (68.1) | 1326/2429 (54.6) | 1.47 (0.98–2.26) | 0.07 |
| Discharge regimen | ||||
| No. of discharge prevention medication classes | 1007/1628 (61.9) | 398/917 (43.4) | 1.50 (1.27–1.79) | <0.001 |
Notes: Prespecified logistic model, 17 covariates entered simultaneously (N = 2545; 1405 events). Rates, class-complete discontinuation with vs without each characteristic; for continuous variables, with vs without age ≥65 years, NIHSS ≥5, prestroke mRS ≥2, creatinine >104 μmol/L, or ≥3 medication classes (odds ratios remain per unit as labeled).
Abbreviations: NIHSS, National Institutes of Health Stroke Scale; mRS, modified Rankin Scale.
The proportion of patients who discontinued at least one class increased with the number of classes prescribed at discharge (43.4% for 2 classes, 60.0% for 3, 66.4% for 4, and 88.9% for 5, although only 9 patients were prescribed 5 classes). Patients discharged with a single medication class (n = 80) were not included in this trend because the class-complete and any-use definitions are equivalent for these patients. However, the mean proportion of prescribed classes that were discontinued varied only modestly across these strata (37.3%, 41.3%, 45.4%, and 42.2%, respectively), and discontinuation of all prescribed classes became less frequent as the number of prescribed classes increased (31.2%, 26.7%, 28.1%, and 11.1%) (Supplementary Table S3). Under a null model in which each prescribed class was discontinued independently at its observed class-specific rate and the number of prescribed classes had no effect, the simulated odds ratio per additional class was 2.44 (2.5th-97.5th percentile, 2.13–2.78; 2000 simulations), higher than the observed univariable odds ratio of 1.60 (95% CI, 1.44–1.78) and the adjusted odds ratio of 1.50 (95% CI, 1.27–1.79) (Supplementary Table S3). The null model assumes independent discontinuation across classes; within-patient correlation would change the magnitude, not the direction, of this comparison.
Subgroup Analyses Among Patients with Hypertension or Diabetes Mellitus
Subgroup models used the same prespecified covariates as the overall model, omitting the stratifying comorbidity (Table 3). The subgroup models included 1561 patients with hypertension, of whom 953 (61.0%) were nonadherent, and 1242 patients with diabetes mellitus, of whom 741 (59.7%) were nonadherent. The association between the number of discharge medication classes and 180-day discontinuation was similar across subgroups, with no evidence of effect modification (hypertension subgroup: OR, 1.42; 95% CI, 1.16–1.74; diabetes subgroup: OR, 1.41; 95% CI, 1.14–1.75; P for interaction = 0.32 for hypertension and 0.67 for diabetes mellitus). In contrast, the association with large-artery atherosclerosis differed by comorbidity status: it was stronger among patients with hypertension (OR, 1.37; 95% CI, 1.08–1.73) or diabetes mellitus (OR, 1.43; 95% CI, 1.10–1.87) than among those without (P for interaction = 0.036 and 0.009, respectively). Dyslipidemia was associated with nonadherence in the hypertension subgroup (OR, 1.48; 95% CI, 1.02–2.19; P = 0.045) but not in the diabetes subgroup (OR, 1.46; 95% CI, 0.91–2.37; P = 0.122). Cardioembolism, which was associated with nonadherence in the overall cohort, was not statistically significant in either subgroup (hypertension: OR, 1.48; 95% CI, 0.79–2.85; P = 0.228; diabetes: OR, 1.34; 95% CI, 0.68–2.70; P = 0.408). Other covariates in the model were not statistically significant in subgroup analyses. Subgroup analyses were exploratory; Benjamini-Hochberg-adjusted q values for the interaction tests and subgroup-specific estimates are reported in Table 3.
Table 3.
Subgroup Analysis: Factors Associated with 180-Day Class-Complete Medication Nonadherence, Prespecified Model, Overall Cohort and by Comorbidity Subgroup
| Factor | Overall cohort (N = 2545; 1405 events), OR (95% CI) | Hypertension subgroup (n = 1561; 953 events), OR (95% CI) | Diabetes subgroup (n = 1242; 741 events), OR (95% CI) | P for interaction (Benjamini-Hochberg q) |
|---|---|---|---|---|
| No. of discharge prevention medication classes | 1.50 (1.27–1.79) P < 0.001 |
1.42 (1.16–1.74) P < 0.001 |
1.41 (1.14–1.75) P = 0.002 |
Hypertension: 0.322 (q = 0.368) Diabetes: 0.671 (q = 0.671) |
| Large-artery atherosclerosis | 1.24 (1.03–1.49) P = 0.02 |
1.37 (1.08–1.73) P = 0.009 |
1.43 (1.10–1.87) P = 0.008 |
Hypertension: 0.036 (q = 0.048) Diabetes: 0.009 (q = 0.015) |
| Cardioembolism | 2.11 (1.37–3.26) P < 0.001 |
1.48 (0.79–2.85) P = 0.23 |
1.34 (0.68–2.70) P = 0.41 |
— |
| Dyslipidemia | 1.50 (1.09–2.07) P = 0.01 |
1.48 (1.02–2.19) P = 0.04 |
1.46 (0.91–2.37) P = 0.12 |
— |
| Hypertension | 1.23 (0.99–1.54) P = 0.07 |
Stratifying variable | 1.37 (1.00–1.88) P = 0.048 |
— |
Notes: Factors associated with the outcome in Table 2, plus hypertension. Same 17 covariates, omitting the stratifying comorbidity. Interaction tests prespecified for the first two factors; q, Benjamini-Hochberg adjusted.
Discussion
In this retrospective analysis of prospectively collected follow-up registry data, 55.2% of patients with ischemic stroke had 180-day class-complete medication nonadherence after discharge, and only 44.8% maintained all secondary prevention medication classes prescribed at discharge. Although most patients continued at least one secondary prevention medication class, fewer than half maintained the complete discharge regimen. This gap suggests that partial discontinuation of the intended secondary prevention regimen is a common pattern of postdischarge nonadherence. The number of prescribed medication classes, cardioembolism, large-artery atherosclerosis, and dyslipidemia were associated with class-complete nonadherence in the overall cohort, and hypertension was of borderline significance. In patients with hypertension or diabetes mellitus, the association with the number of prescribed classes was similar to that in the overall cohort, with no evidence of effect modification, whereas the association with large-artery atherosclerosis was stronger than among patients without these comorbidities.
The separation between any-use adherence and class-complete adherence has important clinical implications. Any-use adherence identifies patients who remain on at least one component of secondary prevention, whereas class-complete adherence evaluates whether the full discharge prevention regimen is maintained. In this study, approximately 7 in 10 patients continued at least one medication class at 180 days, but fewer than half maintained all prescribed classes. This finding indicates that postdischarge nonadherence often occurs as incomplete regimen continuation rather than complete cessation of therapy. Because guideline-directed secondary prevention targets multiple vascular pathways and modifiable risk factors, continuation of one medication class does not necessarily compensate for discontinuation of another indicated class.8 Reliance on any-use adherence alone may therefore overestimate the adequacy of postdischarge secondary prevention, particularly among patients with hypertension or diabetes mellitus.
These findings are consistent with previous studies showing that persistence with secondary prevention medications declines after ischemic stroke or transient ischemic attack.9,10 Prior studies have commonly evaluated adherence to individual medication classes, such as antiplatelet agents, statins, or antihypertensive medications, or used pharmacy refill–based or self-reported measures of medication use.11 For example, recent meta-analyses have reported 1-year adherence estimates for individual therapeutic classes, such as statins or antithrombotic agents, in the range of approximately 60% to 75%, although estimates vary by medication class, data source, population, and adherence definition.12,13 In the present study, the 180-day class-complete adherence estimate was 44.8%, suggesting that single-class assessments may provide a more optimistic view of secondary prevention implementation than regimen-level assessment. These approaches are important for assessing specific components of secondary prevention but may provide a limited view of whether patients continue the complete regimen recommended at discharge. By shifting the unit of assessment from individual medications or single medication classes to the full set of prescribed secondary prevention medication classes, the present study provides a more stringent measure of postdischarge regimen implementation. This medication-class–based definition extends previous evidence by quantifying incomplete regimen continuation and showing that broader adherence measures may underestimate clinically relevant nonadherence.
The association between the number of prescribed classes and class-complete discontinuation is in part structural: a patient prescribed more classes has more opportunities to meet the outcome. Our simulation quantifies this. Under a null model with no effect of regimen size, the expected odds ratio per additional class was 2.44, whereas the observed adjusted odds ratio was 1.50. The observed association is therefore weaker than would be expected from the construction of the outcome alone, and the mean per-class discontinuation rate varied only modestly across strata (37.3% to 45.4%), without a consistent gradient. These findings do not support an interpretation in which greater treatment burden causes higher class-specific discontinuation. What they do show is that discontinuation in patients prescribed more classes was usually partial rather than complete: discontinuation of at least one class became more common as regimen size increased, whereas discontinuation of every prescribed class did not (31.2% for 2 classes and 26.7% to 28.1% for 3 to 4 classes). Both measures depend on the number of classes prescribed, which is why regimen-level continuation cannot be captured by a single any-use question and why follow-up should be class-specific, particularly for patients with the most complex regimens. Mechanisms such as pill burden, dosing complexity, concerns about adverse effects, practical barriers to obtaining medications, and uncertainty about the purpose of each medication have been described as contributors to treatment burden in qualitative studies of stroke survivors.14,15 None of these was measured in the present study, and our data neither support nor exclude them as explanations for the class-specific discontinuation observed here. Patients may also be more likely to discontinue medications that do not provide immediate symptomatic benefit, particularly therapies for lipid, blood pressure, or glucose control.15,16 Large-artery atherosclerosis was associated with nonadherence in the overall cohort and in both hypertension and diabetes subgroups, suggesting that this etiologic subtype may identify patients with a chronic atherosclerotic risk phenotype who require sustained multidrug prevention and long-term medication support.17 The association between cardioembolism and nonadherence in the overall cohort may reflect the complexity of anticoagulation management, perceived bleeding risk, monitoring requirements for some agents, or cost considerations for others—factors identified in real-world studies of oral anticoagulant discontinuation.18
The comorbidity-specific findings highlight the need to integrate poststroke secondary prevention with chronic disease management.19,20 Patients with hypertension or diabetes mellitus often require long-term multidrug therapy, and many may continue some medications while discontinuing others.21 For these patients, asking whether they are taking any secondary prevention medication may miss clinically important discontinuation of specific medication classes. Medication-class–specific review should therefore be combined with monitoring of blood pressure, glucose, and lipid control.22 Potential strategies include early postdischarge contact, pharmacist-led medication reconciliation, medication-class–specific checklists, refill reminders, and regimen simplification, including fixed-dose combinations when clinically appropriate.23 Evidence from cardiovascular prevention and pharmacist-led care studies suggests that structured medication management and regimen simplification can improve adherence in selected settings, although the optimal approach for class-complete secondary prevention after ischemic stroke requires further evaluation.24,25 Coordination between stroke services, primary care, and chronic disease management programs may be needed to sustain complete, medication-class–specific secondary prevention after discharge.26 In practice, follow-up should ask about each prescribed class rather than about medication use in general. The classes most often stopped in this cohort—anticoagulant (59.9% by 180 days), glucose-lowering (52.7%), and antihypertensive therapy (48.7%)—warrant specific review, particularly in patients discharged on three or four classes, of whom 60.0% to 66.4% had stopped at least one class by 180 days.
Taken together, these findings indicate that class-complete assessment identifies a pattern of incomplete regimen continuation that any-use assessment does not capture. Whether this measure predicts clinical outcomes better than established adherence measures was not tested here and requires validation against pharmacy-based measures and against recurrent vascular events. Incorporating medication-class–based assessment into routine follow-up may help detect partial discontinuation earlier and guide targeted interventions for patients with higher treatment complexity, atherosclerotic stroke mechanisms, or chronic vascular comorbidities. These findings support a shift in poststroke follow-up from documenting any medication use toward assessing continuation of the complete prescribed regimen.
Limitations
This study has several limitations. First, it was a retrospective analysis conducted at a single center in one city in southern China, and the findings may not generalize to other health systems, regions, or ethnic groups. Second, the registry did not record the reason for stopping a medication class, so patient-driven and clinician-directed discontinuation could not be distinguished; clinically appropriate discontinuation would also be counted as discontinuation under our definition. Third, medication use in patients assessed by telephone was self-reported and therefore subject to recall bias, although structured item-by-item questioning against the discharge medication list and the preceding follow-up record was used to reduce under-reporting. Fourth, the registry did not record dates of death or of recurrent events, and rehospitalization was not captured, so time-to-event, recurrent-event, and competing-risk analyses could not be performed; a sensitivity analysis excluding patients with recurrent stroke within 180 days gave unchanged results. Fifth, the outcome counts a patient as having discontinued if any one prescribed class is stopped, so patients prescribed more classes have more opportunities to meet the outcome; this structural relationship is quantified in Supplementary Table S3. Sixth, 201 patients were excluded before cohort assembly because of loss to follow-up or unavailable outcome data, and baseline information for these patients was not available for comparison; the analytic cohort was further restricted by discharge prescribing rather than by follow-up completion, with 180-day medication status available for all included patients.
Conclusion
In this cohort of patients with ischemic stroke, 180-day class-complete medication nonadherence was common, affecting more than half of patients after discharge and often reflecting partial discontinuation rather than complete cessation of secondary prevention therapy. A greater number of medication classes prescribed at discharge was associated with nonadherence, with a similar association across the hypertension and diabetes mellitus subgroups, and large-artery atherosclerosis was more strongly associated with nonadherence among patients with hypertension or diabetes mellitus than among those without. Medication-class–specific assessment during poststroke follow-up may help identify incomplete continuation of prescribed secondary prevention regimens and guide targeted support for patients with higher treatment complexity, atherosclerotic stroke mechanisms, or chronic vascular comorbidities. These findings are based on a single-center cohort, and the class-complete measure requires further validation against established adherence measures and clinical outcomes before it is adopted more widely.
Acknowledgments
We thank the healthcare professionals who participated in the follow-up work for this study.
Funding Statement
This work was supported by the Construction Funds of Key Clinical Medical Disciplines in Longhua District, Shenzhen (MKD202508210309).
Disclosure
The author(s) report no conflicts of interest in this work.
References
- 1.GBD 2019 Australia Collaborators . The burden and trend of diseases and their risk factors in Australia, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Public Health. 2023;8(8):e585–13. doi: 10.1016/s2468-2667(23)00123-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hejazian SS, Shahjouei S, Sadr AV, et al. Risk of ischemic stroke after transient ischemic attack in patients receiving comprehensive versus noncomprehensive diagnostic workup: a systematic review and meta-analysis. J Am Heart Assoc. 2025;14(21):I–XLI. doi: 10.1161/JAHA.125.042175 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Zhang L, Delord M, Marshall I, Wolfe C, Wang Y, O’connell MDL. Long-term risk of stroke recurrence: a competing risk analysis based on the South London stroke register. J Am Heart Assoc. 2026;15(6):e043626. doi: 10.1161/jaha.125.043626 [DOI] [PubMed] [Google Scholar]
- 4.Wang SW, Lin HS, Hsu CW, Chen CY, Saito H, Chen FS. Longitudinal adherence trajectories to statins and antithrombotic therapy after ischemic stroke and their impact on low-density lipoprotein control and cardiovascular outcomes. J Am Heart Assoc. 2026;15(10):e048456. doi: 10.1161/jaha.125.048456 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Chen C, Reeves MJ, Farris KB, Morgenstern LB, Lisabeth LD. Sex differences in nonadherence to secondary stroke prevention medications among patients with first-ever ischemic stroke. J Am Heart Assoc. 2024;13(24):e036409. doi: 10.1161/jaha.124.036409 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hobeanu C, Lavallée PC, Charles H, et al. Risk of subsequent disabling or fatal stroke in patients with transient ischaemic attack or minor ischaemic stroke: an international, prospective cohort study. Lancet Neurol. 2022;21(10):889–898. doi: 10.1016/s1474-4422(22)00302-7 [DOI] [PubMed] [Google Scholar]
- 7.Kristensen FPB, Svane HML, Laugesen K, et al. Risk of mortality and recurrence after first-time stroke among patients with type 2 diabetes: a Danish nationwide cohort study. Eur Stroke J. 2025;10(1):190–197. doi: 10.1177/23969873241260956 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Pan Y, Li Z, Li J, et al. Residual risk and its risk factors for ischemic stroke with adherence to guideline-based secondary stroke prevention. J Stroke. 2021;23(1):51–60. doi: 10.5853/jos.2020.03391 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Dalli LL, Kim J, Thrift AG, et al. Patterns of use and discontinuation of secondary prevention medications after stroke. Neurology. 2021;96(1):e30–e41. doi: 10.1212/wnl.0000000000011083 [DOI] [PubMed] [Google Scholar]
- 10.Bushnell CD, Olson DM, Zhao X, et al. Secondary preventive medication persistence and adherence 1 year after stroke. Neurology. 2011;77(12):1182–1190. doi: 10.1212/WNL.0b013e31822f0423 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Dalli LL, Olaiya MT, Kim J, et al. Antihypertensive medication adherence and the risk of vascular events and falls after stroke: a real-world Effectiveness study using linked registry data. Hypertension. 2023;80(1):182–191. doi: 10.1161/hypertensionaha.122.19883 [DOI] [PubMed] [Google Scholar]
- 12.Yang M, Cheng H, Wang X, et al. Antithrombotics prescription and adherence among stroke survivors: a systematic review and meta-analysis. Brain Behav. 2022;12(10):e2752. doi: 10.1002/brb3.2752 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Al AlShaikh S, Quinn T, Dunn W, Walters M, Dawson J. Predictive factors of non-adherence to secondary preventative medication after stroke or transient ischaemic attack: a systematic review and meta-analyses. Eur Stroke J. 2016;1(2):65–75. doi: 10.1177/2396987316647187 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gallacher K, Morrison D, Jani B, et al. Uncovering treatment burden as a key concept for stroke care: a systematic review of qualitative research. PLoS Med. 2013;10(6):e1001473. doi: 10.1371/journal.pmed.1001473 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Jamison J, Graffy J, Mullis R, Mant J, Sutton S. Barriers to medication adherence for the secondary prevention of stroke: a qualitative interview study in primary care. Br J Gen Pract. 2016;66(649):e568–76. doi: 10.3399/bjgp16X685609 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Souter C, Kinnear A, Kinnear M, Mead G. Optimisation of secondary prevention of stroke: a qualitative study of stroke patients’ beliefs, concerns and difficulties with their medicines. Int J Pharm Pract. 2014;22(6):424–432. doi: 10.1111/ijpp.12104 [DOI] [PubMed] [Google Scholar]
- 17.Dawson J, Béjot Y, Christensen LM, et al. European stroke organisation (ESO) guideline on pharmacological interventions for long-term secondary prevention after ischaemic stroke or transient ischaemic attack. Eur Stroke J. 2022;7(3):I–XLI. doi: 10.1177/23969873221100032 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Buck J, Fromings Hill J, Martin A, et al. Reasons for discontinuing oral anticoagulation therapy for atrial fibrillation: a systematic review. Age Ageing. 2021;50(4):1108–1117. doi: 10.1093/ageing/afab024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Liu L, Li Z, Zhou H, et al. Chinese stroke association guidelines for clinical management of ischaemic cerebrovascular diseases: executive summary and 2023 update. Stroke Vasc Neurol. 2023;8(6):e3. doi: 10.1136/svn-2023-002998 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yan LL, Gong E, Gu W, et al. Effectiveness of a primary care-based integrated mobile health intervention for stroke management in rural China (SINEMA): a cluster-randomized controlled trial. PLoS Med. 2021;18(4):e1003582. doi: 10.1371/journal.pmed.1003582 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Osasu YM, Cooper R, Mitchell C. Patients’ and clinicians’ perceptions of oral anticoagulants in atrial fibrillation: a systematic narrative review and meta-analysis. BMC Fam Pract. 2021;22(1):254. doi: 10.1186/s12875-021-01590-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gynnild MN, Aakerøy R, Spigset O, et al. Vascular risk factor control and adherence to secondary preventive medication after ischaemic stroke. J Intern Med. 2021;289(3):355–368. doi: 10.1111/joim.13161 [DOI] [PubMed] [Google Scholar]
- 23.Turner GM, Jones R, Collis P, et al. Structured follow-up pathway to support people after transient ischaemic attack and minor stroke (SUPPORT TIA): protocol for a feasibility study and process evaluation. BMJ Open. 2022;12(6):e060280. doi: 10.1136/bmjopen-2021-060280 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sarfo FS, Voeks J, Adamu S, et al. A cardiovascular polypill for secondary stroke prevention in a tertiary center in Ghana (SMAART): a phase 2 randomised clinical trial. Lancet Glob Health. 2023;11(10):e1619–e1628. doi: 10.1016/s2214-109x(23)00347-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Cengiz KN, Midi I, Sancar M. The effect of clinical pharmacist-led pharmaceutical care services on medication adherence, clinical outcomes and quality of life in patients with stroke: a randomised controlled trial. Int J Clin Pharm. 2025;47(1):99–106. doi: 10.1007/s11096-024-01811-0 [DOI] [PubMed] [Google Scholar]
- 26.Dalli LL, Kilkenny MF, Olaiya MT, et al. Chronic disease management to enhance medication adherence trajectories in long-term survivors of stroke: a population-based cohort study. Pharmacoepidemiol Drug Saf. 2025;34(5):e70148. doi: 10.1002/pds.70148 [DOI] [PMC free article] [PubMed] [Google Scholar]