Introduction

Burden of stroke remains high at global and national levels as being the fourth leading cause of disability-adjusted life years in 20211. Stroke can be categorized into 3 primary subtypes on the basis of its pathophysiological mechanism: ischemic stroke (IS), spontaneous intracerebral hemorrhage (ICH), and spontaneous subarachnoid hemorrhage (SAH)1. Globally, 65.3% of new strokes are IS, 28.8% are ICH, and 5.8% are SAH. Of the 3 stroke subtypes, patients with SAH experience highest in-hospital mortality rates and extended hospital stays, whereas patients with IS experience lowest mortality rates and shortest hospital stays2. Moreover, patients with SAH had lower activity function at admission as well as at discharge from the inpatient rehabilitation than those with other subtypes3. The early differences in functional status theoretically play a critical role in shaping long-term outcomes4,5,6,7. Functional recovery continues after discharge and varies by stroke pathology8,9,10; it is hypothesized that SAH is associated with worst long-term prognosis due to being at a higher risk of delayed cerebral ischemia11. However, the long-term trajectory of these functional limitations across these subtypes and their impact on participation restrictions—challenges in active involvement across various life domains—remain largely unexplored4.

Research on this topic has reported that level of functional recovery in patients with hemorrhagic stroke were lower or closed to patients with IS, using motor or cognitive impairment9,12,13, global function (modified Rankin Scale, mRS)14, activity function8,9, and health-related quality of life (HRQoL)15,16 as outcomes of interest. Despite these findings, previous studies have not fully captured the multifaceted nature of participation restrictions across all 3 stroke subtypes, particularly in the long term. Further investigation is required to determine whether the subtype of stroke affects participation restrictions after hospital discharge, which are a primary focus during the late phase of stroke rehabilitation17,18. Understanding the recovery patterns and prognoses of different subtypes of stroke is essential for optimizing postdischarge health care and advancing therapeutic efforts toward addressing different types of stroke.

Accordingly, the objectives of this study were, first, to characterize the 1-year recovery trajectories for participation, activity function, HRQoL, and global cognition since hospital discharge among patients with IS, ICH, and SAH, and second, to compare differences in these outcomes between the subtypes. Finally, this study aims to explore whether factors readily available at hospital discharge, such as stroke severity (National Institute of Health Stroke Scale, NIHSS), global function (mRS), and potentially person-environmental or clinical characteristics, can predict long-term participation outcomes for each stroke subtype.

Methods

Study design and data source

In this retrospective study, data were collected from 2 independent, prospective observational studies4,5, which included adult patients with stroke admitted to 1 of 3 hospitals located in Taipei metropolitan area between August 2016 and September 2018. Functional outcomes and personal and environmental variables were retrieved at 3 time points: at hospital discharge (T1), at 3-month follow-up (T2), and at 12-month follow-up (T3). Personally identifiable information, such as chart numbers, had been removed to ensure patient anonymity. This study was approved by the Ethics Committee of Taipei Medical University—Joint Institutional Review Board (approval no. N202406014), including a waiver of informed consent. We adhered to the ethical principles set forth in the Helsinki Declaration for research involving human participants.

Participants

Patients meeting the following criteria were included in the study: being 20 years of age or older, having stroke confirmed by a physician (International Classification of Diseases, Ninth Revision codes 430–438), and having at least 1 outcome of interest at any time point. Specifically, patients with IS were identified by codes 433, 434, or 435; those with ICH by code 431; and those with SAH by code 430. The outcomes included participation (measured using the Participation Measure-3 Domains, 4 Dimensions [PM-3D4D])19, activity function (measured using the outpatient short-form Activity Measure for Post-Acute Care [AM-PAC])20, HRQoL (measured using the EuroQol 5-Dimension 3-Level [EQ-5D-3L])21, and cognitive function (measured using the Montreal Cognitive Assessment [MoCA])22. Patients meeting the following criteria were excluded from the analysis: having an unspecified stroke subtype or a diagnosis unrelated to IS, ICH, or SAH (International Classification of Diseases, Ninth Revision codes 432.9, 436, and 438) and having a hospital stay longer than 6 months or shorter than 2 months. According to Taiwan’s National Health Insurance, patients can receive acute care for up to 28 days poststroke and inpatient rehabilitation for up to 6 months poststroke, depending on their need. The inclusion timeframe was selected to exclude patients with either unusually short or prolonged hospitalizations and to ensure consistent rehabilitation exposure. A flowchart of the patient selection process is depicted in Fig. 1.

Fig. 1
figure 1

Flowchart indicating the subject selection.

Full size image

Outcome measures

Participation as primary outcome was evaluated using the PM-3D4D, which covers 3 distinct domains: (1) Social participation (6 items), including activities that contain the degree of interaction between the respondent and others; (2) Community participation (12 items), including shopping, exercise, and leisure activities that have to go out of the house; (3) and Productivity (6 items), including paid work, unpaid work, education, and housekeeping19,23. Each item was assessed in 2 dimensions: participation frequency (objective) and perceived difficulty (subjective). A frequency score was determined for each domain by averaging the summed ratings within each domain (range: 0 to 6 for social and community participation, 0 to 4 for productivity). Difficulty scores were transformed into a Rasch score (range: 0–100).

Activity function was measured using the AM-PAC outpatient short form, which is used to evaluate basic mobility (18 items), daily activities (15 items), and applied cognition (19 items)20,24. Standardized t-scores were calculated for each domain (range: 29.41–80.3 for basic mobility, 28.09 to 76.03 for daily activities, and − 5.3 to 64.97 for applied cognition). HRQoL was evaluated using the EQ-5D-3L, which covers 5 domains: mobility, self-care, routine activities, pain or discomfort, and anxiety or depression21. For quantifying population level, health states from 5 domains were converted into weighted time trade-off values (index score) by using the Taiwan model (range: − 0.484 to 1)25. For indexing individual level, a vertical visual analog scale (VAS), with 0 representing the “worst imaginable health state,” was used to evaluated self-rated health (range: 0 to 100). Global cognitive function was measured using the MoCA (range of total score: 0 to 30), which is used to evaluate visuospatial and executive skills (5 items), naming (3 items), attention (6 items), language (3 items), abstraction (2 items), delayed memory (5 items), and orientation (6 items)22,26. Mandarin versions of each outcome measure were used, with a higher score representing more favorable function or health status.

Personal and environmental factors

Various personal and environmental factors related to poststroke participation recovery were extracted and used as potential predictors or covariates4,7,27. These variables included (1) demographic and personal variables, such as age, sex (male as reference), educational level (< 6 years as reference), relationship status (single/widowed/separated/divorced status as reference), income (< NT$50,000 as reference), and pre-stroke employment status (no job as reference); (2) clinical variables, such as stroke severity (NIHSS), pre-existing comorbidities, and global activity function (mRS); and (3) environmental variable was the level of social or family support (Multidimensional Scale of Perceived Social Support, MSPSS)28.

Statistical analysis

The IBM SPSS Statistics, version 24 (IBM, Armonk, NY, USA) was used for all data analysis. Descriptive statistics were used to summarize demographics, clinical and environmental characteristics, and outcome scores, with stratification by stroke subtype. The categorical variables were expressed as number (%), while numerical variables were reported as median (interquartile range), mean (standard deviation), and estimated marginal mean (95% confidence interval [CI]). Moreover, the recovery rates of each outcome after 12 months stroke, i.e. percentages of the maximum amount possible of each outcome at T3, were calculated29. Baseline between-group comparisons were conducted using a χ2 test for categorical variables and a Kruskal–Wallis test for numerical variables. Post-hoc multiple χ2 tests and Mann–Whitney U tests were conducted, with statistical significance set at P < 0.017.

Generalized estimating equations (GEEs) were used to investigate the effects of time, group (stroke subtype), and time-by-group interactions on outcomes of interest. Proper GEE modeling depended on the goodness of fit testing summarized in Supplementary Material (Table S1). GEE was selected because it can be used for population-level estimations and can be used when some longitudinal data are missing. In cases involving significant effects (P < 0.05), post-hoc pairwise comparisons were conducted and the type I error was corrected with Bonferroni’s method aligning with P < 0.006 for a significant between-group difference (0.05/9) and P < 0.025 for a significant within-group difference (T1 vs. T2, T2 vs. T3). Baseline variables with significant between-group differences were included as covariates. Full- adjusted and crude GEE models were estimated separately, and the later modeling is considered as sensitivity analysis.

Hierarchical multivariable linear regression analyses that used age and sex as fixed predictors were conducted for each stroke subtype to explore the factors associated with each PM-3D4D outcomes at 12-month follow-up. To prevent overfitting, only significant variables from the univariable linear analyses were included in stepwise multivariable models, with a P value of less than 0.05 as an independent predictor30. A predictor was considered redundant due to multicollinearity when the variance inflation factor exceeded 5 or the condition index was greater than 3031. The performance of the final models was assessed based on goodness-of-fit metrics (R2 and adjusted R2) and the F statistic, with P < 0.05 indicating significance. Within the models, predictors were reported with corresponding regression coefficient (B), standard error, and standardized coefficient (β).

Results

Table 1 presents a summary of the demographic characteristics and baseline variables included in this study. Of a total of 646 patients (254 women [39.3%]; median age = 57 years, interquartile range [IQR] = 50–62 years; median duration from stroke to discharge = 4 months), 335 had IS (132 women [39.4%]; median age = 58 years, IQR = 50–63 years), 288 had ICH (109 women [37.8%]; median age = 56 years, IQR = 50–62 years), and 23 had SAH (13 women [56.5%]; median age = 55 years, IQR = 48–63 years). Patients with ICH had a significantly higher prevalence of hypertension compared with those with IS (168 patients [58.3%] vs. 149 patients [44.5%]). Patients with IS were significantly more likely to have diabetes mellitus compared with those with SAH (75 patients [22.4%] vs. 0 patients). Patients with IS were also significantly more likely to have partners (272 patients [81.2%] vs. 208 patients [72.2%]) and to receive social support (median MSPSS score [IQR]: 55 [48, 60] vs. 52 [47, 60]) compared with those with ICH. Stroke severity was greater in patients with SAH (median NIHSS score [IQR]: 13 [9.5, 14.5]) than in those with IS (median NIHSS score: 9 [7,11]) and ICH (median NIHSS score: 9 [7,13]). These 5 unequal factors were included in GEE modeling as covariates. The descriptive statistics of all outcomes are summarized in Table 2, and the number of missing data for each outcome depending on subtype and time points are listed in Table S2.

Table 1 Baseline characteristics of stroke patients with specified subtypes.
Full size table
Table 2 The mean (standard deviation) of the outcomes by time point and subtype.
Full size table

Participation outcomes over 12 months postdischarge

Table 3 summarizes the effects of time, stroke subtype, and their interaction on all outcomes. Overall, the stroke survivors demonstrated positive trends in productivity, social, and community participation frequency and perceived difficulty over the 12-month follow-up period (all P < 0.001 for time effects), particularly between T2 and T3 (Fig. 2). Notably, the SAH group demonstrated non-significant changes over time in productivity frequency or social difficulty. Among the measured participation outcomes, the recovery rate was lower for objective participation (frequency scores: 1–37.8%, Fig. 2A) compared to subjective participation (difficulty scores: 48.7–72.1%, Fig. 2B). Moreover, results showed that recovery rate was higher for social domain compared to community and productivity domains. The highest level of 1-year outcome was social difficulty for all groups. Both the IS and ICH groups achieved 72.1% of the maximum possible score (100), while SAH group reached 57.9% of the maximum score. Conversely, patients with IS and ICH showed the lowest recovery rate in community frequency (IS: 8.8%, ICH: 9.3%) and the lowest, 1% of recovery rate for the SAH group was observed in productivity difficulty score (Fig. 2).

Table 3 Model effects on outcomes via general estimating equation.
Full size table
Fig. 2
figure 2

Recovery patterns of PM-3D4D stratified by stroke subtype. One-year trajectory of poststroke participation frequency (a) and perceived difficulty (b) across the domains of social participation, community participation, and productivity. A higher difficulty score indicates lower perceived difficulty. Percentages of the maximum possible amount of each score are showed at T3. The dots and error bars indicate estimated means and standard errors with National Institute of Health Stroke Scale, MSPSS, hypertension, diabetes, and partnership scores adjusted for. ICH, intracerebral hemorrhage; IS, ischemic stroke; MSPSS, Multidimentional Scale of Perceived Social Support score; PM-3D4D, Participation Measure-3 Domains, 4 Dimensions; SAH, subarachnoid hemorrhage; T1, at hospital discharge; T2, at 3-month follow-up; T3, at 12-month follow-up. #Significant difference with previous time point (P < 0.025). aIS > SAH (P < 0.006). bICH > SAH (P < 0.006).

Full size image

Significant subtype effects were observed in participation frequency across all domains (all P < 0.001), with the IS and ICH groups having 1.8 to 40 times higher scores in comparison with the SAH group (P < 0.006 for all, Fig. 2A). The estimated mean (95% CI) for social participation frequency was as follows: IS group—2.02 (1.9, 2.14) at T1, 2.10 (1.99, 2.21) at T2, and 2.18 (2.01, 2.29) at T3; ICH group—2.10 (1.98, 2.22) at T1, 2.19 (2.07, 2.31) at T2, and 2.27 (2.14, 2.40) at T3; SAH group—1.08 (0.84, 1.33) at T1, 1.19 (0.95, 1.44) at T2, and 1.23 (0.98, 1.49) at T3. For community participation frequency, the estimated mean (95% CI) scores were: IS group—0.47 (0.41, 0.53) at T1, 0.48 (0.43, 0.54) at T2, and 0.53 (0.46, 0.59) at T3; ICH group—0.49 (0.43, 0.55) at T1, 0.52 (0.45, 0.59) at T2, and 0.56 (0.49, 0.63) at T3; SAH group—0.10 (0, 0.24) at T1, 0.11 (0, 0.25) at T2, and 0.13 (0, 0.27) at T3. In productivity frequency, the estimated mean (95% CI) scores were: IS group—0.32 (0.26, 0.37) at T1, 0.36 (0.30, 0.42) at T2, and 0.39 (0.33, 0.45) at T3; ICH group—0.39 (0.32, 0.46) at T1, 0.42 (0.35, 0.49) at T2, and 0.44 (0.37, 0.51) at T3; SAH group—0 (0, 0) at T1, 0 (0, 0.02) at T2, and 0.04 (0, 0.09) at T3.

In terms of participation difficulty, a significant subtype effect was found only within social domain (P = 0.023), with the IS and ICH groups having 1.2 times higher social difficulty scores than the SAH group (Fig. 2B). Specifically, the estimated mean (95% CI) scores were: IS group—68.96 (66.92, 71.07) at T1, 70.96 (68.92, 73.06) at T2, and 72.10 (70.07, 74.19) at T3; ICH group—68.16 (66.02, 70.38) at T1, 70.69 (68.50, 72.95) at T2, and 72.11 (69.93, 74.35) at T3; SAH group—56.78 (48.83, 66.03) at T1, 57.26 (49.40, 66.38) at T2, and 57.85 (50.08, 66.81) at T3. A significant time-by-subtype interaction was observed within the productivity domain (P = 0.012). The ICH group exhibited consistent improvements over time: 55.34 (53.23, 57.45) at T1, 57.21 (55.14, 59.27) at T2, and 58.01 (55.99, 60.02) at T3. In contrast, the IS and SAH groups had stable estimated means between T1 (IS: 54.51 [52.50, 56.51], SAH: 48.03 [40.18, 55.89]) and T2 (IS: 55.19 [53.32, 57.07], SAH: 47.71 [39.67, 55.75]); however, they exhibited delayed improvements, with gains occurring primarily between T2 and T3 (IS: 55.92 [54.03, 57.81], SAH: 48.73 [40.88, 56.58]). Neither subtype effect nor time-by-subtype interaction was significant within the community domain (P = 0.176 for stroke subtype, P = 0.286 for interaction). These results regarding time, subtype, and their interaction remained consistent after covariates removal (Table S3).

Functional outcomes over 12 months postdischarge

GEE analyses revealed significant effects of time on AM-PAC, EQ-5D-3L, and MoCA scores (all P ≤ 0.007 for time), indicating overall functional recovery across all stroke subtypes following hospital discharge (Fig. 3). However, the SAH group showed non-significant changes over time in EQ-5D-3L index score. In general, the recovery rates of these functional outcomes ranged from 43.4% to 72.2% (Fig. 3). The highest level of functional outcome achieved at 1 year was observed in AM-PAC basic mobility scores, where the estimated mean (95% CI) scores were 58.8 (58.1, 59.6) for the IS group, accounting for 73.2% of the maximum value of 80.3; 58.7 (57.9, 59.6) for the ICH group, representing 73.1% of the maximum; and 54.2 (51.5, 57.2) for the SAH group, corresponding to 67.5% of the maximum. Subtype effects were significant for AM-PAC (P = 0.027 for both basic mobility and daily activity; P < 0.001 for applied cognition), EQ-5D-3L (P < 0.001 for both index and VAS scores), and MoCA (P < 0.001, Table 3). The IS and ICH groups had similar functional outcomes from discharge until 12 months after discharge. These outcomes were more favorable than those of the SAH group, particularly at 3 months after discharge (i.e., T2, Fig. 3). For example, in terms of AM-PAC, the estimated mean (95% CI) for basic mobility were 57.47 (56.48, 58.48) for the IS group, 57.49 (56.42, 58.58) for the ICH group, and 51.01 (47.04, 55.31) for the SAH group (Fig. 3A). The estimated mean for daily activity were 50.82 (50.02, 51.64) for the IS group, 50.84 (49.99, 51.70) for the ICH group, and 45.89 (42.98, 49.01) for the SAH group (Fig. 3B). The estimated mean for applied cognition were 40.55 (39.37, 41.72) for the IS group, 39.67 (38.37, 40.98) for the ICH group, and 33.51 (30.14, 36.87) for the SAH group (Fig. 3C). Regarding EQ-5D-3L, the estimated mean (95% CI) index scores were 0.54 (0.52, 0.56), 0.53 (0.51, 0.56), and 0.42 (0.34, 0.50) for the IS, ICH, and SAH group, respectively (Fig. 3D). The estimated mean VAS scores were 61.7 (59.3, 64.1) for the IS group, 60.2 (57.5, 62.8) for the ICH group, and 39.3 (29.0, 49.5) for the SAH group (Fig. 3E).

Fig. 3
figure 3

Functional recovery patterns stratified by stroke subtype. One-year trajectory of poststroke activity function (a–c), HRQoL (d, e), and global cognitive function (f). Percentages of the maximum possible amount of each score are showed at T3. The dots and error bars indicate estimated means and standard errors with N National Institute of Health Stroke Scale, MSPSS, hypertension, diabetes, and partnership scores adjusted for. AM-PAC, Activity Measure for Post-Acute Care; HRQoL, health-related quality of life; ICH, intracerebral hemorrhage; IS, ischemic stroke; MoCA, Montreal Cognitive Assessment; MSPSS, Multidimentional Scale of Perceived Social Support score; PM-3D4D, Participation Measure-3 Domains, 4 Dimensions; SAH, subarachnoid hemorrhage; T1, at hospital discharge; T2, at 3-month follow-up; T3, at 12-month follow-up; VAS, Visual Analog Scale. #Significant difference with previous time point (P < 0.025). aIS > SAH (P < 0.006). bICH > SAH (P < 0.006).

Full size image

Notably, the recovery patterns for MoCA scores significantly differed between stroke subtypes (interaction P < 0.001, Fig. 3F). The IS and ICH groups exhibited early improvements in the estimated means between T1 (IS: 19.4 [18.7, 20.1], ICH: 19.4 [18.6, 20.2]) and T2 (IS: 20.1 [19.5, 20.7], ICH: 20.1 [19.5, 20.7]). Conversely, the SAH group exhibited delayed improvements, with gains occurring from T2 (15.8 [14.1, 17.4]) to T3 (17.1 [15.5, 18.6]). Additionally, the IS group exhibited a decline in MoCA scores between T2 and T3 (19.3 [18.9, 19.8]), suggesting possible cognitive challenges at later stages. All GEE findings remained consistent after covariates were removed, with the exception of the effect of stroke subtype on AM-PAC basic mobility, which became nonsignificant (Table S3).

Factors at discharge influencing participation outcomes at 12 months after stroke, stratified by stroke subtype

In patients with IS (Table 4), multivariable linear regression analyses revealed that younger age, better global activity function (mRS), higher cognitive function (as indicated by MoCA or AM-PAC applied cognition scores), greater social support, and lower income were associated with higher social (adjusted R2 = 0.34) and productivity (adjusted R2 = 0.24) participation frequency after 12 months poststroke. Additionally, younger age, better global activity function (mRS), and greater social support were associated with higher community participation frequency (adjusted R2 = 0.29). For social difficulty, better global activity function (mRS), higher cognitive function (AM-PAC applied cognition), and greater social support were associated with lower perceived difficulty (adjusted R2 = 0.42). Lower perceived difficulty in community participation was associated with female sex, better physical activity function (AM-PAC basic mobility), and greater social support (adjusted R2 = 0.32). Finally, better physical activity function (AM-PAC basic mobility), higher cognitive function (AM-PAC applied cognition), and greater social support were associated with lower productivity difficulty (adjusted R2 = 0.27). Simple linear regressions for IS patients were summarized in Table S4.

Table 4 Multivariable linear regression of PM-3D4D over 12 months after ischemic stroke (n = 250).
Full size table

In patients with ICH (Table 5), multivariable linear regression analyses revealed that younger age, better global activity function (mRS), higher cognitive function (AM-PAC applied cognition), greater social support, and lower income were associated with higher social participation frequency (adjusted R2 = 0.40) after 12 months poststroke. These factors, along with having a partner, were also associated with higher community participation frequency (adjusted R2 = 0.33). Younger age, better global activity function (mRS), and lower income were associated with higher productivity participation frequency (adjusted R2 = 0.33). For social and community participation difficulty, younger age, better activity function (mRS or AM-PAC basic mobility), higher cognitive function (AM-PAC applied cognition), and greater social support were associated with lower perceived difficulty (social participation: adjusted R2 = 0.32; community participation: adjusted R2 = 0.24). Finally, younger age, better activity function (mRS), higher cognitive function (AM-PAC applied cognition), and lower income were associated with lower productivity difficulty (adjusted R2 = 0.24). Simple linear regressions for ICH patients were presented in Table S5.

Table 5 Multivariable linear regression of PM-3D4D over 12 months after intracerebral hemorrhage (n = 225).
Full size table

In patients with SAH, significant models were found in predicting social participation frequency (adjusted R2 = 0.75, P < 0.001), community participation frequency (adjusted R2 = 0.38, P = 0.014), and community difficulty scores (adjusted R2 = 0.44, P = 0.011). Specifically, younger age, male, and higher AM-PAC daily activity scores were associated with higher social participation frequency. Higher AM-PAC daily activity scores was associated with higher community frequency, while higher AM-PAC daily activity along with greater social support were associated with lower community participation difficulty. The multivariate models for social participation difficulty, as well as productivity frequency and difficulty, were not significant. Table S6 presents the simple linear regression results, and Table S7 shows the multivariate regression results for SAH patients.

Discussion

It is undoubtable that stroke negatively impacts multifaceted functioning and health status32. This study provides an in-depth comparison of postdischarge outcomes, including participation, activity, HRQoL, and global cognition, in patients with 3 common subtypes of stroke. We discovered significant variation in participation and functional outcomes by stroke subtype over a follow-up period of 1 year, thereby showing subtype-specific recovery profile. Social engagement usually regained a higher level in comparison with participation in community and productivity activities after any type of stroke. Patients with SAH consistently had lower scores compared with those of patients with IS and ICH, even after baseline variables that differed significantly between groups, such as stroke severity, social support, partnership, and comorbidities were adjusted for. After age and sex were controlled for, activity function emerged as an early predictor of poststroke participation outcomes across all stroke subtypes, which indicates it should be considered in postdischarge care planning.

Our results demonstrated that patients after stroke, regardless of their pathology, were able to re-participate in part of their everyday life activities in different domains and to regain the capability of engaging the activities during 12 months after discharge. The positive trends observed are partially consistent with previous studies using the Utrecht Scale for Evaluation of Rehabilitation-Participation7,33 or the Activity Card Sort34 as participation measures. However, the extent of participation frequency and perceived difficulty remained considerably lower than the general healthy level, particularly concerning severe challenges in life activities that occur outdoors or are related to productivity34,35,36. This domain-specific pattern suggests that social connection may have more readily adaptable approaches for engagement, such as using videophones instead of face-to-face visits. In contrast, such activities involving outdoor leisure, return to work, or caring for family often require specific physical and cognitive capabilities.

To our knowledge, this is the first study to compare longitudinal participation and functional outcomes following different stroke subtypes. The majority of studies on poststroke participation have either not employed subtype differentiation4,7,34,37 or focused on recovery profiles for specific subtypes, such as SAH33. Multiple studies have compared the outcomes of IS, ICH, and SAH in cross-sectional cohorts at different time points after discharge2,38. Katzan et al.38 indicated that patient-reported outcomes were similar across all 3 stroke subtypes in terms of physical and psychosocial function. Our study adds threefold evidence: first, patients with IS and ICH exhibit similar participation, activity, HRQoL, and cognitive outcomes and their recovery trajectories between discharge and 12 months follow-up; second, patients with SAH experienced poorer outcomes across these domain during the same period in comparison with those with IS and ICH; and third, highest recovery rate at 12 months postdischarge, among 3 stroke subtypes, was approximately 70% in mobility function (AM-PAC basic mobility). The Korean Stroke Cohort for functioning and rehabilitation (KOSCO) study recently identified a subgroup of SAH patients who demonstrated nearly full functional independence from 3 months to 5 years poststroke39. These patients had good global function (mRS = 0–3) at acute stage and an averaged hospitalization of less than 1 month. Our patients, by contrast, were individuals who had been receiving inpatient care for months after they experienced stroke. Consequently, our findings underscore the importance of rehabilitation during and after inpatient care, particularly for SAH patients with worse inital function.

Multiple factors can contribute to severe outcomes following SAH. For instance, SAH can lead to several types of pathophysiological phenomena, such as parenchymal bleeding (which also occurs in ICH), increased intracranial pressure, and acute global ischemia, all of which may lead to early neurological deterioration11,40. During the first 3 weeks after the onset of stroke, patients with SAH are at a risk of delayed cerebral ischemia or even cerebral infarction, which can result in neuronal damage and lead to persistent functional and cognitive deficits11,41,42. Although guidelines for management of SAH have been published, studies of specifically focusing on rehabilitation in chronic patients are scarce18,43. This study’s findings of poorer long-term outcomes in our SAH cohort highlight the need for more targeted rehabilitation strategies.

In contrast to previous studies8,9,14,15, we observed similar recovery trajectories for patients with IS and ICH over time. This discrepancy may be due to our focus on estimated means of outcomes rather than their dichotomization into positive and negative categories8,14,15. Additionally, we examined the trajectories of recovery from hospital discharge, whereas many studies that have investigated such trajectories did so start at the point of stroke onset8,9. Although patients with ICH had lower activity function at admission, they demonstrated potential for improvement during the inpatient care period, resulting in discharge function similar to that of patients with IS44. Importantly, our finding that ICH and SAH have distinct recovery patterns contrasts with studies pooling them as a hemorrhagic cohort9,45, reinforcing that they should be considered separate conditions in research and clinical care.

In this study, we disclosed that discharge factors, including physical activity function, cognitive function, and social support, can significantly predict 1-year participations after controlling for age and sex. Physical activity function had the largest effect and was associated with most participation outcomes across stroke subtypes. These findings align with previous studies4,7,33,37,46,47, which have identified similar predictors and provided support for the development of rehabilitation programs targeting modifiable factors to enhance poststroke participation outcomes48. However, the small sample size of SAH patients limited our power to identify specific predictors in this group. Future research with larger cohorts is needed to further explore these associations.

Limitations

This study has 4 main limitations. First, the low incidence of SAH in Taiwan (2.8%) resulted in a small sample size for this type of stroke49. This limited our statistical power to detect certain recovery patterns and predictors. Second, the KOSCO studies suggest that functional outcomes may continue to improve or decline beyond 12 months poststroke9,39. Therefore, further research with longer follow-up periods and larger multicenter cohorts is required to validate and expand on our findings and to ensure adequate representation of all stroke subtypes. Third, although lesion characteristics—such as lesion location and lesion volume—are known to influence functional prognosis50, the present study did not include detailed brain image data. Finally, the external validity of this study is limited by the extended duration of inpatient rehabilitation available under Taiwan’s National Health Insurance system.

Conclusions

This study provides valuable insights into the differences between IS, ICH, and SAH as major stroke subtypes. Compared with patients with IS and ICH, those with SAH typically have poorer outcomes across multiple domains, with particularly pronounced barriers to re-engaging in productivity-related activities. These findings underscore the need for targeted rehabilitation—especially home- and community-based approaches—to support reintegration in SAH patients. In addition, activity function evaluated shortly after discharge is a robust predictor of long-term participation outcomes across all stroke subtypes, indicating it is pivotal in poststroke recovery. Furthermore, factors such as cognitive function and social support independently predict the recovery trajectories of patients with IS and ICH, indicating these factors can be considered potential targets for personalized poststroke rehabilitation. Future research should focus on larger, multicenter studies with longer follow-up to confirm these findings and explore effective interventions tailored to each stroke subtype.