The Inuential factors of Adherence to Physical Activity and Exercise among Community-Dwelling Stroke Survivors: A Path Analysis

Background: Regular physical activity and exercise (PAE) after stroke is essential, but the adherence to PAE and its inuential factors is rarely studied and unclear in developed countries. The aim of this study was to investigate the status of PAE adherence, and to identify its inuential factors among Chinese community-dwelling stroke survivors. Methods: A cross-sectional study was conducted, 208 stroke survivors were randomly selected from three communities (60.10% men, mean age 70.25 ± 9.08 years). Physical Activity and Exercise Adherence Scale (PAEAS) was used as a measure of compliance, knowledge, attitude, self-ecacy and social support were measured using relevant scales. Inuential factors of PAE adherence and associations between those variables were explored using multiple linear regression and path analysis. Results: The mean adherence rate was moderate (62.0%), stroke survivors tended to be more adherent to do PAE than monitoring and seeking advices (70.30% > 53.50% > 48.30%). Regression results revealed that seven factors signicantly associated with PAE adherence, including knowledge, attitude, self-ecacy, social support, self-care ability, community rehabilitation experience, and times since stroke (R 2 = 75.10%, F = 45.593, p < 0.05). Furthermore, path analysis showed that knowledge had a signicant indirect positive inuence with self-ecacy as a mediator on adherence (β= 0.16, p < 0.05); while attitude and social support impacted both directly and indirectly on adherence with self-ecacy as the mediator; what’s more, self-ecacy was an important predictor that performed the strongest direct effect on community-dwelling stroke survivors’ PAE adherence (β= 0.44, p < 0.01 ). These four variables accounted for a total of 67.00% of the variance of PAE adherence among community-dwelling stroke survivors. Conclusions: Physical activity and exercise adherence of community stroke survivors is needed to be improved. Attitude and social support presented both direct and indirect impacts on PAE adherence through self-ecacy, while knowledge only exhibited an indirect inuencing effect on adherence. Health care professionals should a more comprehensive and systematic assessment to seek to improve self-ecacy

However, stroke survivors may cease physical activity or reduce the frequency and duration, studies have shown that low adherence to recommended physical activity and exercise is a major problem among stroke survivors (5,11,12). Even continuous monitoring during an exercise intervention with stroke survivors living in a community resulted in a mean exercise class attendance of only 60%, and after one year's follow-up, only 44% of participants completed the recommended exercise of walking 10,000 steps per day (13). In a study conducted by Gunnes (12) in Norway, the rate of adherence to long-term physical activity and exercise was 51.2% at the beginning and increased by 2.6% per month with individualized monthly coaching by a physical therapist. Another investigation conducted in USA also revealed that only 65.3% of stroke patients who completed 6 months of rehabilitation with a professional organization reported adhering to at least part of the home exercise program (11). Stroke survivors who discharged from hospital spent only 13% to 45% (14) of their days engaging in therapeutic activities. Research from China also demonstrated that 70.6% of the stroke survivors spent less than 8.5 hours per week on moderate activity (15). So, it is very important to explore the reasons for adherence or non-adherence in order to design more effective and long-lasting interventions.
Social Cognitive Theory (SCT) has been widely used to explore and explain physical activity and exercise behavior, evidences recommend that researchers should report direct, indirect, and total effects from all constructs to behavior (16). According to SCT, stroke survivors' adherence behavior may be determined by personal, social factors, and environmental factors (17,18). Personal correlates, demographic and health variables, have been related to exercise (19). Among social factors, social support has been shown to correlate with physical activity in general populations (20) and older patients (21). Environmental factors of physical activity have recently received attention (22), while rehabilitation service in community have not been rigorously studied. The SCT postulates that psychosocial, physical and behavioral factors potentially in uence each other both theoretically and structurally, and there are several studies that have examined and identi ed the pathways used by various psychosocial factors to in uence physical activity and exercise (23,24). However, none have examined these interrelationships with respect to physical activity and exercise adherence behavior in a cohort of community-dwelling stroke survivors.
In this study, we aimed to investigate the physical activity and exercise adherence of community-dwellers after stroke, explore its association with personal, social, and environmental factors in an agentic and systematic perspective, nally to identify the pathways in which factors, such as knowledge, attitude, selfe cacy and social support. The hypothesized model based on a SCT and previous research (19) (20) (21) was evaluated through a path analysis which tested the contributions of various paths in the model; the hypotheses and model are illustrated in Figure 1.

Methods Participants
A cross-sectional survey study was conducted with 208 community-dwelling stroke survivors. By convenience sampling, three communities (Nan-guan, Qin-ling, and Chang-jiang) with a total of eleven heath service stations in Zhengzhou city were selected as the sites of the study. Participants were randomly selected from the health service stations according to their health record number. Of the 468 standard stroke patients in these communities, 50.6% (236) met the study's eligibility criteria: stroke diagnosed by MRI or CT, lucid enough to cooperate with the research interviewers, able to voluntarily sign the information and consent form, able to participate in a face-to-face interview, live in their home rather than any institutional care centers. Among the eligible participants, 16 refused to be interviewed after being informed of the study purpose, and 12 could not be reached via the contact information recorded in their les. Finally, 208 quali ed and consenting participants were included.

Data collection
Three trained investigators conducted the investigation, and informed consent was obtained from all participants. Participants were interviewed face-to-face by the three investigators at the patient's home or the community center. To guarantee the accuracy of the data, all participants were asked to carefully review and con rm their answers to the questionnaires immediately after completing them. A copy of the Manuals of Daily Life for Stroke Patients designed by our research group was given to each participant as a gift at the end of the investigation.

Independent variables
The demographic and disease-speci c data questionnaire was designed and used to collect the Responses are divided into three levels: "important", "not sure", and "not important", scored as 3 points, 2 points, and 1 point, respectively. The sum of the points re ects the respondent's overall attitude toward physical activity and exercise. The Cronbach's α and Content Validity Index (CVI) of the knowledge section are 0.85 and 0.92, respectively, the Cronbach's α and CVI of the attitude section are 0.88 and 0.87, respectively. Overall, the Cronbach's α and CVI for the reliability-validity of all questionnaire items are 0.90 and 0.95, respectively(26).
Perceived Social Support Scale (PSSS): The PSSS was described by McDowell and Zimet et al. (27,28). It measures the perceived support that patients receive from their family, friends or communities. This scale contains 10 items divided into three sections: objective support (OS), with 3 items; subjective support (SS), with 4 items; and utilization of social support (USS), with 3 items. The nal score is the sum of each of the 10 item scores. The Cronbach's alpha for PSSS in present study is 0.92, and the item-total correlations range from 0.89 to 0.94.
Self-E cacy for Managing Chronic Disease 6-Item Scale (SES6C): The SES6C was designed by Lorig K et al. (29) and is widely used to measure the con dence of chronic disease patients in doing certain activities (30,31). The measure consists of six items. It uses a 10-point Likert scale a psychometric scale commonly used in questionnaire-based research, with scores ranging from 1 ("not con dent at all") to 10 ("completely con dent"). The scale is interpreted by calculating a mean score over at least four of the six items thus allowing a maximum of two missing item responses. The scores re ect the self-e cacy level of the respondent, with higher scores indicating greater con dence. Cronbach's alpha for the SES6C is 0.88 and the split-half is 0.80 (32).

Dependent variable
Physical Activity Adherence and Exercise Scale (PAAES): The PAAES was previously tested (33) and widely used (34) to determine the adherence to the recommended rehabilitation physical activity and exercise regimen of community-dwelling stroke survivors (10). PAAES contains 14 items organized into three sections: adherence to PAE participation (PAEP), with 8 items; adherence to PAE monitoring (PAEM), with 3 items; and adherence to PAE advice-seeking (PAEA), with 3 items. The scores for each question range from 1 ("never") to 4 ("always"), and higher scores indicate better adherence. Adherence rate M can be calculated using the formula "M=actual score/max", max is obtained by multiplying the number of entries by 4. The adherence rate is de ned as high at ≧75%, moderate at 50% -75% and low at ≦50% (35, 36). Cronbach's alpha value of PAAS is 0.900, in relation to each domain are 0.705 -0.897, the Scale Content Validity Index (S-CVI) is computed as 0.950 (37).

Statistical analysis
We investigated participants' demographic characteristics and path analysis assumptions using IBM SPSS (version 21). Descriptive statistics (means, standard deviations and percentages) were calculated to represent the demographic characteristics, along with the scores of PAAES, SKAQ, SES6C and PSSS. Two unrelated groups' mean scores were compared using a two-independent-sample t-test, differences of three or more groups' scores were tested by means of one-way analysis of variance (ANOVA). Multiple twogroup comparisons were conducted using a least signi cant difference (LSD) t-test. The relationships between predictive factors and adherence were analyzed using Pearson correlation coe cients. Stepwise linear regression was performed to identify the predictors of adherence. The results were checked for accuracy, and two tailed p values which were signi cant at <0.05.
The hypothesized model was tested using IBM SPSS AMOS 21.0 software. There were seven hypothesized paths in this study: adherence regressed on knowledge, attitude, social support and selfe cacy; self-e cacy regressed on knowledge, attitude and social support. Path analysis is an extension of the regression model, a special case of Structural Equation Modeling (SEM) where all variables included in the model are observed (non-latent) variables, it allows for the simultaneous estimation of these equations (24,38), this path analysis was de ned as path analysis with observed variables (PA-OV) (39).
Additionally, path analysis can involve modeling indirect and direct effects as well as hypothesis tests on the parameters of the effects. The path of the model is shown by squares and arrows. A single-headed arrow points from cause to effect. A double-headed, curved arrow indicates that variables are merely correlated. Regression weight is predicated by the model, and the residual error terms re ect the unexplained variance and measurement error (39). Then the goodness-of-t statistic is calculated in order to test the consistent with the hypotheses that went into the model. Because these criteria were very sensitive, alternative incremental t indices were determined such as root-mean-square error approximation (RMSEA) <0.080, goodness of t index (GIF) >0.900, and comparative t index (CFI) > 0.900 (40).

Characteristics of the Study Sample
Among the 208 stroke survivors, mean age was 70.25 ± 9.08 years (range: 50-90 years); 60.10% of the participants were male, and 75.48% were married. All the data were transformed into categorical variables so that it would be easier to test the differences of physical activity and exercise adherence between groups with different characters (Table 1).  Table 2 illustrates the physical activity and exercise adherence rate, scores of SKAQ, PSSS, SES6C, and relevant proportion or value of conversion. The mean rate of adherence was moderate, at 62.00%, stroke survivors tended to be more adherent to do physical activity than monitoring and seeking advices (70.30%>53.50%>48.30%). The constituent ratio of adherence levels is shown in Figure 2.

Regression results
Pearson correlation coe cients demonstrated that the physical activity and exercise adherence rates had signi cant statistical correlations with scores of knowledge, attitude, social support and self-e cacy; these were presented in Table 3. The regression analysis was constructed using the factors with a statistically signi cant contribution shown in Tables 1 and 3. Finally, after analysis, seven independent variables entered into the regression model, with a R 2 of 75.1%, the F-value was 45.593 and p-value was 0.001 (Table 4).

Path analysis results
The path coe cients are regression weights obtained from the regression analysis. As shown in Figure 3

Discussion
Although previous studies had con rmed, adherence to physical activity and exercise is very important for stroke survivors' long-term outcomes (12,36,42), it is a worldwide concern as well. The results of the present study showed that the mean adherence rate was moderate (62.00%), only 20.19% of the participants had a high level of adherence rate. The results were similar to but slightly lower than those in Duncan's study (36); in that study, the average adherence rate was 68.2% in the acute phase and 69.5% in the post-acute phase. Another study by Ovbiagele et al. (43) found that nearly 70% of 144 stroke survivors maintained regular exercise upon being discharged from hospital, which was similar to but slightly higher than our ndings. Despite differences in research design, cultural background, and economical level, lack of conventional management (44), not ideal health care seeking behavior (45) have been main barriers for adherence behavior, and it is likely to present an even greater in uence on Chinese community-dwelling stroke survivors' long-term outcomes over the next few decades.
To further understand the physical activity and exercise adherence status, we analyzed the adherence rate of each subdimension. Results showed that the rate of adherence to participation was the highest, approximately 70.30%, while the rate of adherence to monitoring was 53.49% and that of adherence to advice-seeking was only 48.32%. This illustrates that most of the stroke survivors preferred to do regular physical activity and exercise but did not pay much attention to self-monitoring when performing exercise, and only approximately 15.38% of them had ratings consistent with a high level of adherence, as shown in Figure 1. Moreover, the results also indicated that community-dwelling stroke survivors rarely received positive advices when they ran into barriers because only 14.42% were rated as having high adherence to PAE advice seeking. Possible explanations are that consciousness, language, movement, and other disabilities may affect stroke patients' intention to communicate with others (46). In addition, according to the National norms for basic public health services in China (47), no regular home visits or phone communication service will be provided to community-dwelling stroke survivors, and there is no routine mandatory follow-up service from hospital as well. Therefore, this may be helpful in explain the reason that why most of stroke survivors are unaware of the importance of asking for advice, for they do not know how to get help.
Regarding to the possible in uential factors. The hemiplegic patients with lower self-care ability performed worse physical activity and exercise adherence, however it was revealed that hemiplegia was not a signi cant predictor in the multivariate linear regression analysis. All in all, the more dependent patients were, the lower their adherence rates were, and these results are reasonably similar to other investigations (48). In addition, patients are willing to receive assistance from caregivers (49), however, the inconsistent perceived needs between stroke survivors and their caregivers should be considered (50). What's more, the interpersonal relationship challenges among stroke survivors and family caregivers should be considered as well (51). Regarding the correlation between marital status and adherence, there was no signi cant difference in adherence rates between stroke survivors who had a spouse as a caregiver and those who did not (0.624 vs. 0.607, p >0.05). We were surprised to nd that the adherence rates of the patients who had caregivers were lower than those of the patients who looked after themselves. Perhaps it is necessary to conduct a systematic review regarding the association of marital status with adherence for further understanding.
Our ndings demonstrated that patients with better knowledge and positive attitudes tended to have higher adherence levels. Knowledge is known as an in uencing factor that triggers motivation for adherence behavior among all kinds of patients (52,53). Stroke survivors who are more knowledgeable can more positively accept the importance of physical activity as they are more aware of the bene ts of physical activity and exercise adherence. However, path analysis results showed that knowledge only exerts an indirect positive effect on adherence through its direct effects on self-e cacy. This could be have been caused by individuals who has disability, the positive effect of better knowledge on healthy behavior may be eliminated by the fear or restriction of participation (13). Studies highlighted that patients often feel unprepared do deal with long-term disabilities, self-e cacy has been proposed as important determinant of adaptation process in individuals after stroke (24,42,54). Researches also demonstrated that patients who proactively and willingly participated in exercise had high adherence, even without any supervision (55,56).
In this study, social support as a protective factor in the adherence pathway demonstrated a direct, positive, and signi cant association with adherence and indirectly associated with adherence through selfe cacy. This nding is supported by the results reported in Zhao's study (57). Our ndings thus suggested that social support was more important than other external factors like environmental factors for adherence among community-dwelling stroke survivors. This matches the results of Damush et al., who reported that social support was a facilitator of exercise after stroke (19). Effective social support could come from physicians, family members, friends, other patients or volunteers, etc. However, from the viewpoints of health care professionals, many barriers exist that interfere with their ability to provide help or support to stroke survivors (58), among them time constraints was the main barrier (59). Especially in China, with limited health care service in community and lack of effective social support system, it's a big challenge using social factors to improve stroke survivors' physical activity and exercise adherence at home.
Finally, but most importantly, self-e cacy presented as a key and predictive factor that can both directly affect physical activity and exercise adherence and mediate other variables. In our study, the results showed that self-e cacy is not only an in uencing factor but is also an important determinant of physical activity and exercise adherence for stroke survivors. Our results were consistent with those of two previous studies by Ivey et al. (60) and Frederick et al. (61), who suggested that patients with higher self-e cacy tend to have better adherence. However, other researchers have reported con icting results, such as in a study of sports prescription lead by Jones et al. (48). They found that patients with high self-e cacy do not necessarily comply better than others because they may have higher expectations of recovery; when these patients were not satis ed with their results, they retreated, and their levels of persistence, enthusiasm and adherence changed. Therefore, we propose that the synergies and mechanism of knowledge, attitude, social support and self-e cacy in PAE adherence behavior should be clari ed and applied in future practice.
This project has several limitations. First, a relatively small group of Chinese stroke survivors participated in the study, so the results cannot be directly extrapolated to all stroke survivors. Second, we conducted the investigation at one time point, but it will be superioris to conduct a longitudinal study of rst-ever stroke survivors to identify the trajectory of physical activity adherence. In addition, other barriers and facilitators should be considered in study, or a qualitative study is needed to explore how variables like self-e cacy, et al. in uences adherence among stroke survivors. Such a study would provide more detailed guidance for further intervention development if the sample size was larger or the study design was more comprehensive.

Conclusions
In summary, we explored factors that might associate with the adherence to physical activities and exercise among community-dwelling stroke survivors in China. The results showed that overall adherence was moderate and that adherence to seeking advice was the lowest-scoring aspect. The most signi cant discovery was that stroke survivors' adherence could be in uenced by personal, social and environmental factors, including self-e cacy, attitude, social support, self-care level, time since initial diagnosis, community rehabilitation experience, and knowledge. Among these, attitude and self-e cacy are the most important. In particular, self-e cacy is a major determinant for community-dwelling stroke survivors because of its direct effect on adherence and its mediating effect on other variables. Speci c interventions should be performed to encourage stroke patients' internal motivation and thus improve their adherence to prescribed physical activity and exercise, including frequent health interviews in the community, proper monitoring, reminders about scheduled exercise and other tactics. Declarations Ethics approval and consent to participate The protocol was evaluated and approved by the Research and Health Ethics Committee of Zhengzhou University.

Consent for publication
Not applicable.
Availability of data and materials  The constitute ratio of three levels of (a) Total, (b) PAEP, (c) PAEM, and (d) PAEA.