Abstract
This study explored quality of life among stroke survivors.
According to the World Health Organization (WHO), each year approximately 15 million people globally experience a stroke, with 5 million dying and another 5 million experiencing permanent loss of physical function (Lindsay et al., 2019; Prokopiv et al., 2021; WHO, 2025). Poststroke symptoms vary depending on the brain region affected and the extent of damage; they include motor problems, such as spasticity or flaccidity; neurological dysfunctions; coordination problems; cognitive impairments; sensory–motor deficits; speech difficulties; and swallowing problems. These issues lead to a decline in stroke survivors’ quality of life (QoL), and the biopsychosocial impact can cause participation problems (Topçu & Bölüktaş, 2012).
The consequences of these impairments extend well beyond the physical domain, profoundly influencing psychological and social well-being. For instance, systematic reviews have documented a higher prevalence of poststroke depression and anxiety, which contribute to reduced social interaction and isolation. This cumulative burden not only hampers community reintegration but also affects long-term recovery. As a result, comprehensive rehabilitation strategies that address physical, cognitive, and emotional aspects are essential to improve QoL among stroke survivors and support their overall well-being (Lo Buono et al., 2017).
Given these multifaceted challenges, rehabilitation strategies must address both the physical and the psychosocial consequences of stroke to enhance recovery and QoL. In this context, occupational therapy emerges as a vital intervention, because it is specifically designed to promote health and well-being through engagement in meaningful activities (Christiansen et al., 2024). Occupational therapy interventions aim to help stroke survivors regain independence in daily activities, improve motor and cognitive functions, and facilitate social participation—objectives that are directly aligned with the identified needs of this population (Brown & Hollis, 2013; Kelley & Borazanci, 2009). Furthermore, understanding the latent profiles of stroke survivors—that is, subgroups characterized by varying degrees of physical, cognitive, and emotional impairments—can be crucial. Identifying these profiles beforehand allows clinicians to tailor rehabilitation strategies more precisely, ensuring that interventions are aligned with the individual’s specific needs and potential for recovery. This personalized approach not only enhances the effectiveness of occupational therapy but also supports comprehensive stroke rehabilitation by addressing the unique challenges faced by each survivor, ultimately facilitating community reintegration and long-term recovery.
Latent profile analysis (LPA) is a method used to group people who show similar patterns in their responses or traits and then examine the relationship of these different groups to other factors and outcomes (Collins & Lanza, 2009). It is ideal for research that examines how different combinations of variables influence outcomes, because it simplifies many continuous or categorical measures into a few distinct subgroups. This approach also helps practitioners understand why treatment effects may vary among individuals (Zyphur, 2009). LPA contributes to data-driven decision-making by allowing researchers and practitioners to uncover hidden subgroups in complex data sets. By grouping individuals on the basis of patterns across multiple indicators—such as clinical symptoms, cognitive measures, or emotional states—LPA transforms high-dimensional data into a manageable number of distinct profiles. These empirically derived profiles reveal nuanced patterns that might otherwise remain obscured in traditional analyses, thereby informing targeted interventions and resource allocation (Nylund-Gibson et al., 2023). In stroke studies, LPA is particularly valuable because of the heterogeneous nature of stroke symptoms and their varying severity depending on the brain region affected and the extent of damage. By revealing specific profiles and predictors, LPA enhances the development of targeted interventions, ultimately improving the QoL and rehabilitation outcomes of stroke survivors (Oberski, 2016). Many studies on stroke have conducted LPA to identify sleep pattern subtypes (Guo et al., 2023), anxiety subtypes (Zhang et al., 2023), cognitive status subtypes (Ma et al., 2025), and, last but not least, functional independence subtypes (Furuta et al., 2022). Together, these studies exemplify how LPA contributes to data-driven decision-making by generating actionable insights that lead to targeted, evidence-based interventions, ultimately improving the QoL and rehabilitation outcomes of stroke survivors (Oberski, 2016).
This study aimed to identify clusters among stroke survivors on the basis of their perceived level of QoL. Additionally, as a secondary outcome, the study sought to determine the predictors of the clusters obtained via LPA.
Method
Participants
This study included 696 adults age 18 yr or older who had experienced a stroke. Before their participation, all individuals were informed about the planned research. Each participant who agreed to join the study signed an informed consent form, confirming their willingness to participate. Patient charts and reports were thoroughly investigated to determine whether clients were eligible to participate in the study, and inclusion and exclusion criteria were used to determine eligibility for enrollment in the study. The inclusion criteria were (1) individuals age 18 years or older who have experienced a stroke and (2) literate individuals with a Mini-Mental State Examination score of 23 or higher (Cockrell & Folstein, 2002). The exclusion criteria were (1) having communication problems or another speech disorder (if the patient’s chart contained no information on speech pathologies, direct observation of communication methods was made); (2) presence of other chronic neurological, psychiatric, or orthopedic issues or serious organ dysfunction, respiratory failure, or malignant cancer; (3) chronic disease that might affect comprehension and cognitive functions, such as dementia, intellectual disability, and other psychiatric conditions; and (4) history of transient ischemic attack.
After determining which individuals met the inclusion criteria, participants underwent an evaluation and survey process lasting approximately 20–30 min, depending on the individual’s understanding and response rate. Responses were collected using a structured form.
Data Collection Instruments
All evaluations and surveys were conducted at the Physical Therapy and Rehabilitation Department of the T.C. Sağlık Bakanlığı Doç. Dr. Mustafa Kalemli State Hospital, Ministry of Health (Tavşanlı, Kütahya, Türkiye) and the Hacettepe University Faculty of Health Sciences Occupational Therapy Department (Ankara, Türkiye) where individuals had registered for an occupational therapy program. This study was approved by the Hacettepe University Health Sciences Research Ethics Committee (SBA 23/019) on September 5, 2023.
Demographic Form
We used a semistructured interview method to collect individuals’ demographic information. This information, which was collected via hospital charts and direct interviews, included age, gender, education level, occupation, marital status, affected hand, dominant hand, date of stroke, type of stroke, income level, employment status, living arrangements, and rehabilitation history.
Stroke Impact Scale
The Stroke Impact Scale (SIS; Duncan et al., 1999; Lai et al., 2002) is a comprehensive tool designed to evaluate the multidimensional effects of stroke on an individual’s health and QoL. The SIS provides a detailed assessment across multiple domains, making it a crucial component of this study. The SIS consists of 59 items divided into eight subscales: Strength, which assesses strength in the arms, legs, hands, and feet; Hand Function, which evaluates the ability to perform tasks requiring hand use, such as carrying heavy objects or turning a doorknob; Activities of Daily Living (ADLs) and Instrumental Activities of Daily Living (IADLs), which measure the ability to carry out daily activities independently, including feeding, bathing, dressing, and managing finances; Mobility, which assesses the ability to move around in different environments, such as walking and climbing stairs; Communication, which evaluates the ability to speak, understand, read, and write; Emotion, which assesses emotional well-being, including feelings of sadness, anger, and frustration; Memory and Thinking, which evaluates cognitive functions such as memory, attention, and problem-solving; and Participation/Role Function, which measures the impact of stroke on social participation and role fulfillment in the community.
Each SIS item is rated on a 5-point Likert scale, where higher scores indicate better function and less impact of stroke. The SIS provides a comprehensive picture of the physical, emotional, and social challenges faced by stroke survivors, and it is particularly useful for tracking changes over time and evaluating the effectiveness of interventions. The Turkish version of the SIS 3.0 was examined by Hantal et al. (2014) and had a Cronbach’s α of .70.
Barthel Index of Activities of Daily Living
Developed by Mahoney and Barthel (1965) and adapted into Turkish by Küçükdeveci et al. (2000), the Barthel Index (BI) assesses mobility and ADLs such as feeding, bathing, personal care, dressing, bowel control, bladder control, toilet use, transferring from bed to wheelchair, walking or wheelchair dependence, and climbing stairs. It consists of 10 items scored on a scale ranging from 0 to 10 or 15 points (scored in 5-point increments), depending on the question.
Impact on Participation and Autonomy Questionnaire
Originally developed by Cardol et al. (1999) in the Netherlands and adapted into Turkish by Kurt (2014), the Impact on Participation and Autonomy Questionnaire (IPA) measures the participation and autonomy of individuals with chronic disease in daily and social life. It is a simple, brief scale covering categories of activity and participation related to the International Classification of Functioning, Disability and Health (World Health Organization, 2001). The questionnaire assesses different dimensions of autonomy and participation, with individuals rating experienced limitations. It consists of 32 items divided into five subscales.
Statistical Analysis
LPA was applied to the data obtained from the SIS using Jamovi 3.24, an R programming-language-based software. LPA is used to identify a set of distinct and nonoverlapping latent classes by modeling patterns in responses to a set of profile indicators. Additionally, it involves model testing to discover the most suitable model for the research by comparing various data sets and models with different numbers and types of profiles.
In this study, we used several goodness-of-fit indices to determine the number of latent profiles, including Akaike’s information criterion (AIC), Bayesian information criterion (BIC), and sample-size-adjusted BIC (SABIC). A comprehensive evaluation of BIC, AIC, and SABIC was conducted to analyze and present the most appropriate model, with these indices collectively serving as a common fit index to identify the latent profiles being investigated.
Furthermore, the entropy index was estimated as an indicator of the precision and distinction between different profiles and classes. We expected a higher entropy value to better delineate profiles, with the highest value, that closest to 1, being preferred when deciding on the classes. To compare models using likelihood ratios, the bootstrap likelihood ratio test and Lo-Mendell-Rubin tests were also calculated.
After determination of the clusters, an inferential analysis was applied to compare the clusters’ participation and ADL dependency levels.
Additionally, as a secondary outcome, we aimed to determine the predictors of the clusters obtained through LPA. To achieve this, we used multiple logistic regression analysis. This method allowed for the identification of significant predictors that differentiate the latent clusters by assessing the relationship between the clusters and various independent variables (IPA and Barthel scores, demographic information). By doing so, it was possible to gain insights into the factors that contribute to the distinct profiles of stroke impact, thereby enhancing the understanding of the underlying characteristics and predictors of each cluster.
Sample Size
The necessary sample size for this study was determined on the basis of the requirements for conducting LPA with sufficient statistical power. To ensure reliable identification of latent classes and to achieve stable parameter estimates, a sample size of at least 500 participants was recommended, considering the complexity of the model and the anticipated number of latent classes. Power calculations were informed by previous studies using LPA with similar populations, which suggested that a minimum of 300–500 participants is typically needed to detect meaningful class distinctions with high confidence (Marsh et al., 2009; Tein et al., 2013). By recruiting 696 stroke survivors, we exceeded this threshold, thus enhancing the study’s power and ensuring robust, generalizable findings across the identified latent profiles. This sample size also allowed for accurate estimation in subsequent multinomial logistic regression analyses to identify significant predictors of class membership
Results
This study included 696 adult individuals who had experienced a stroke. The participants’ mean age was 64.4 yr (SD = 11.2), and they were deemed suitable for the study. Information on participants’ gender, hemiplegic side, time since stroke, education level, marital status, and employment status is provided in Table 1.
Sociodemographic Characteristics of the Participants (N = 696)
Upon examining the model fit indices, we determined that the three-class model demonstrates the best fit for this study, as evidenced by the lowest values of the BIC (50,674) and AIC (50,520). Although the four-class model shows a relatively high entropy value of .880, its BIC value does not reach statistical significance, suggesting a less-optimal fit. Therefore, the three-class model, with an entropy of .899, was chosen as the most appropriate representation of latent profiles among stroke participants, identifying three distinct classes.
The data indicate that the groups are distributed as follows: Class 1 (n = 322), Class 2 (n = 232), and Class 3 (n = 142). The mean ages for Classes 1, 2, and 3 are 64.9 yr (SD = 9.94), 61.1 yr (SD = 13.1), and 68.4 yr (SD = 8.42), respectively. Class 2 had the highest BI score, and Class 3 had the lowest. When examining IPA scores, Class 2 had the lowest scores, Class 3 had the highest scores, and Class 1 had average values. Overall, Class 2 tended to achieve the best scores across all scales, Class 3 to achieve the worst, and Class 1 to average out in the middle. On the basis of these values, Class 2 is identified as the high-QoL group, Class 1 as the moderate-QoL group, and Class 3 as the low-QoL group. The descriptive statistics for the classes are provided in Tables 2 and 3. Additionally, we investigated the differences between clusters of BI and IPA scores with one-way analyses of variance. The significant differences were, furthermore, analyzed with the Games-Howell post hoc test. All analyses showed significant differences between clusters and also between paired groups (p < .05).
Sociodemographic Characteristics by Classes (N = 696)
Descriptive Statistics for Classes (N = 696)
Note. ADLs = Activities of Daily Living; BI = Barthel Index; IPA = Impact on Participation and Autonomy Questionnaire; Min–Max = minimum and maximum values; SIS = Stroke Impact Scale.
aOne-way analysis of variance.
bSignificant between-groups differences (Games-Howell post hoc analysis).
To determine the factors influencing class distribution, we conducted a multinomial logistic regression model. The model included the parameters age, gender, BI score, IPA subscale scores (Indoors Autonomy, Family Role, Outdoors Autonomy, Social Life and Relationships, and Work and Education), stroke type, marital status, and education level, resulting in an R 2 of .646. According to these results, gender, IPA Social Life and Relationships, IPA Work and Education, and education level were found to be significant predictors for transitioning from one class to another (Table 4).
Summary of the Multinomial Logistic Regression Model
Note. IPA = Impact on Participation and Autonomy Questionnaire.
When examining the relationships between classes using the multinomial logistic regression model, we found that differences between the high-QoL and moderate-QoL groups were associated with BI scores and IPA Outdoors Autonomy, Social Life and Relationships, and Work and Education subscale scores, along with factors such as dominant hand and employment status. These were predictive of membership in the high-QoL group compared with the moderate-QoL group. In contrast, when comparing the moderate- and low-QoL groups, gender, education level, and employment status emerged as significant predictors. Finally, differences between the high-QoL and low-QoL groups were related to variations in BI scores, IPA Indoors Autonomy and Outdoors Autonomy scores, IPA Work and Education total scores, education level, affected hand, and employment status parameters. These findings indicate that scores on these measures are related to the likelihood of a stroke survivor being classified into a particular QoL group.
Discussion
In this study, latent classes were identified among stroke survivors on the basis of their QoL as determined with SIS outcomes. The analysis revealed three distinct classes: high QoL, moderate QoL, and low QoL. Individuals in the high-QoL class demonstrated minimal stroke-related impact, maintaining strong physical, emotional, and cognitive functioning, along with active social engagement. The moderate-QoL class experienced noticeable limitations but retained some level of independence in daily tasks. The low-QoL class, however, struggled significantly, facing profound difficulties in mobility, cognition, emotional well-being, and social participation. These findings provide health care professionals with valuable insights, helping them to identify individuals in greater need of targeted interventions and to tailor care plans accordingly to improve outcomes in populations with lower QoL.
This study investigated participants QoL by using the SIS 3.0. When comparing the SIS results of this study with existing literature, variations in outcomes across different domains are evident. The strength values in the low- and moderate-QoL classes in this study are lower than those observed in the Turkish and Italian assessments (Aran et al., 2023; Vellone et al., 2010), and the high-QoL class scores are comparatively higher. For hand function, the high-QoL class had higher results compared with the Brazilian (Carod-Artal et al., 2008), Korean (Choi et al., 2017), and Italian (Vellone et al., 2010) studies, with the low-QoL class results aligning with those of the Turkish study. In terms of mobility, the high-QoL class exhibited higher outcomes compared with the Brazilian and Italian studies, and the low-QoL class remained lower. ADLs and IADLs followed a similar trend, in which the high-QoL class surpassed previous findings and the low-QoL class remained comparable to those in the Korean and Italian studies. Memory and communication scores in the moderate- and high-QoL classes show elevated levels compared with the other studies, whereas the low-QoL class results demonstrate a notable decline, similar to the Brazilian and Italian assessments. For recovery and participation, the high-QoL scores in this study exceed those reported in the Brazilian, Italian, and Turkish groups, whereas the moderate-QoL scores align with the Korean findings.
In comparing the results of this study with other IPA assessments from different countries, we observed significant differences across various domains. Notably, the Indoors Autonomy scores in this study, particularly for the low-QoL class, are higher than those reported in the Swedish (Palstam et al., 2019), Thai (Suttiwong et al., 2018), and Chinese (Y. Li et al., 2020) studies, indicating greater limitations in indoor autonomy. Similarly, the Family Role scores for the low-QoL class are markedly higher than those in the Thai and Chinese populations, reflecting more pronounced challenges in family roles. In contrast, the moderate-QoL and high-QoL classes had scores closer to those found in other studies, suggesting fewer participation restrictions. In the Outdoors Autonomy domain, the low-QoL class scores were consistently higher than those reported in the Thai and Chinese studies, signifying greater difficulties in outdoor autonomy and participation. Additionally, the Social Life and Relationships scores for the low-QoL class are considerably elevated when compared with the Swedish and Thai studies, indicating more significant impairments in social participation, and they are similar to the results from the Chinese study. Finally, the Work and Education scores for the low-QoL class are notably higher than the Swedish results, highlighting greater limitations in work and educational activities.
Overall, across all domains, the low-QoL class in this study had the highest scores, which indicates worse participation and greater perceived problems, because higher scores on the IPA reflect more significant limitations (Cardol et al., 1999). This suggests variability in participation levels between different populations and regions, with the low-QoL class experiencing more substantial participation restrictions compared with counterparts in other countries. In comparing the BI scores from this study with existing literature (Arowoiya, 2014; Beltz et al., 2022; Gurková et al., 2023), the findings align with previous research indicating that higher BI scores correlate with better functional outcomes and shorter hospital stays. In particular, the high-QoL class in this study had BI scores similar to those in other studies, where scores of ≥60 predict better independence and recovery prospects. However, the low-QoL class, characterized by lower BI scores, is consistent with groups experiencing more severe disability and poorer long-term recovery, as reported in studies on stroke survivors (Q.-X. Li et al., 2020).
The regression results indicate that as age increases, stroke survivors are more likely to be classified in the low-QoL group, which is associated with poorer functional outcomes. This finding suggests that age plays a significant role in recovery, with older individuals experiencing more poststroke challenges. Our results are consistent with those of previous research that has demonstrated the impact of age on QoL (Bártlová et al., 2022; Jeon et al., 2017; Kim et al., 2005). In addition, men are at higher risk for these poor outcomes—meaning they have a greater probability of being classified into the low-QoL group—compared with women. This finding contradicts the literature, because most studies indicate that men had better QoL than women (Bártlová et al., 2022; Franco-Urbano et al., 2022; Ospel et al., 2023). Although our study was not primarily designed to comprehensively analyze all factors affecting QoL, these findings highlight the need for further research into gender differences in poststroke recovery. Future studies should explore the underlying mechanisms that may contribute to these unexpected gender disparities, considering factors such as social support, coping strategies, and cultural influences that might differentially affect QoL outcomes among male and female stroke survivors.
The findings indicate that social relationships and environmental interactions have a measurable impact on QoL group membership. Participants who scored better in the IPA Social Life and Relationships domain—meaning they experience fewer limitations in their interactions and social connections—tended to be classified in the high-function group. In contrast, those with poorer social interactions, reflected by higher IPA scores, were more frequently classified in the low-function group. This suggests that strong, supportive social networks and active engagement with the environment can positively influence rehabilitation outcomes. In other words, beyond physical and cognitive factors, the quality of a patient’s social life appears to be a crucial determinant of their overall functional status after stroke. Studies have consistently shown that robust social relationships and active engagement with one’s environment are associated with better functional outcomes after stroke. Carod-Artal (2012) demonstrated that stroke survivors with stronger social support networks and greater community participation tend to experience higher QoL and improved functional recovery. Similarly, the development and application of the IPA (Cardol et al., 1999) underscore the importance of social interactions in rehabilitation, because higher scores in social participation domains are linked to poorer functional status. Additionally, studies have shown that enhanced social integration is correlated with better recovery trajectories, supporting the view that quality social relationships are crucial for achieving higher levels of function poststroke (Schindel et al., 2021; Tiwari et al., 2021). The results of this study align with current literature. We believe that clinicians and researchers should consider planning interventions or detailed evaluations of social interactions after stroke.
Work and education status parameters significantly affect social participation, which in turn plays a critical role in transitions between functional classes by stroke survivors (Addo et al., 2012; Sun et al., 2023). Employed individuals and higher education level appear to have enhanced social engagement, which may facilitate better overall recovery. This observation is in line with recent studies that have shown that socioeconomic factors, such as employment and education, are closely linked to social participation and improved rehabilitation outcomes (Nguyen et al., 2024). Higher education may not only provide individuals with better access to health care resources but may also enhance their cognitive reserve, contributing to a more robust recovery process.
The multinomial logistic regression analysis offers key insights into factors influencing transitions between functional classes among stroke survivors. Age and BI score show significant effects, particularly in class transitions related to higher functional abilities. BI score consistently affects transitions from moderate to high function, with higher scores associated with better outcomes. Social participation measures, particularly the IPA Social Life and Relationships domain, significantly influence class transitions, highlighting the importance of social interaction in recovery. Similarly, work and education participation plays a role in determining transitions between functional levels, particularly in distinguishing high-QoL individuals from moderate- and low-QoL individuals. Additional significant predictors include education level, which affects transitions between moderate and low function, and employment status, which also plays a crucial role. Interestingly, factors such as gender, dominant hand, and affected hand have varying degrees of influence, indicating that personal factors and rehabilitation history contribute to functional outcomes. Overall, this analysis emphasizes the importance of functional independence, social participation, and personal characteristics in predicting recovery trajectories.
This study had some limitations: (1) The data were collected from two centers that may not represent the different cultures and settings of a country with a broad cultural background such as Türkiye and (2) the study relied on self-report measures (i.e., the SIS, BI, and IPA) to assess QoL and functional outcomes. Self-report instruments are subject to biases such as social desirability and recall bias, which might affect the accuracy of the responses.
Implications for Occupational Therapy Practice
This study provides insights into the varied QoL among stroke survivors, identifying distinct profiles that can guide occupational therapists in tailoring interventions. Understanding these profiles allows therapists to better address each individual’s specific needs, enhancing patient-centered care and improving functional outcomes. This study has the following implications for occupational therapy practice: ▪Tailored interventions for different QoL levels. Occupational therapists can use QoL classifications to design interventions that align with the patient’s functional status and specific limitations. For instance, those in the low-QoL group may benefit from intensive support in mobility and daily living skills, whereas those in the moderate- and high-QoL groups may need a focus on social participation and autonomy. ▪Focus on social participation and autonomy. The findings indicate that social relationships and environmental interactions significantly affect stroke survivors’ QoL. Therapists can prioritize interventions that foster social connection and facilitate community participation, particularly for those with low perceived QoL. ▪Incorporate education and advocacy for families. For individuals with substantial limitations, therapists can involve family members and caregivers in the therapeutic process to support the individual’s participation and autonomy at home and in the community. ▪Screening for key predictors. Screening for age, gender, social relationships, and education level may help therapists assess stroke survivors’ likely challenges and needs, enabling more precise intervention planning early in the rehabilitation process.
Conclusion
By determining latent classes among stroke survivors, this study offers a nuanced understanding of the varying impacts of stroke on QoL. Identifying these distinct subgroups—ranging from high to low QoL—provides a more personalized approach to stroke rehabilitation. It allows clinicians to focus interventions more precisely on the basis of the specific needs of each class. Moreover, this classification supports the development of targeted health care strategies, potentially improving long-term recovery outcomes and optimizing resource allocation for populations at greater risk of diminished QoL.