Critical pathways and factors of non-health outcomes affecting self-care behaviors in patients with type 2 diabetes and their differences in patient preference in medical decision-making

The effects of patient sustained self-care behaviors (SCB) on glycemic control are even greater than the effects of medical treatment, indicating the value of identifying the factors that influence SCB. To date, these factors have not been placed in one model to clarify the critical path affecting SCB. The aims of this study were to explore the relationships of these factors and the differences by decision-making preference. A cross-sectional study was conducted among outpatients with type 2 diabetes at a regional teaching hospital. Purposive sampling was adopted to recruit 316 eligible patients via self-administered questionnaires. Partial least squares structural equation modeling was used for analysis. patient empowerment Indirect pathways were from HL to SCB via or PE. The pathway from HL to SE was significantly stronger in those preferring shared decision-making than in those who preferred physician decision-making.


Introduction
Another factor well known to be associated with SCB in patients with diabetes is selfefficacy (SE), which is "the belief in one's capacity to organize and execute the courses of action required to manage a prospective situation" [19]. SE is the main factor that directly affects health behaviors [20]. Patients with high SE have better compliance with SCB [21], both general SE and disease-specific SE can affect SCB [22]. To improve the SCB of patients with diabetes, high SE is necessary.
Diabetes self-management is a complex lifelong journey whose prerequisites for success are the patient's active and responsible participation in the process. Patients must be internally motivated rather than externally motivated [23]. Patient empowerment (PE) is a process designed to facilitate self-directed behavior change [24]. Therefore, PE is another widely researched determinant of healthy behaviors [18]. The WHO defines empowerment as "a process through which people gain greater control over decisions and actions affecting their health" [25], which demonstrates that PE "doesn't mean 'giving' people power. Rather it's about 'enabling' them to recognize and use their power" [25]. The empowerment approach has been proven to improve SE and SCB [26]. Patients with different health literacy levels may respond differently to PE. The Health Empowerment Model [27] proposes that PE is deeply interwoven with HL to affect the health outcomes of patients. Findings of the interactions between PE and HL in terms of health outcomes in patients are not consistent [28].
In summary, patients with diabetes require a high level of responsibility and promise to implement a new lifestyle of SCB [29]. HL, SE, and PE all directly or indirectly affect SCB, but they have not yet been integrated into one model to clarify their pathways of influence on SCB or their relationship to patient preference in terms of medical decisionmaking. This study used partial least squares structural equation modeling (PLS-SEM) to analyze these relationships. The aims of the current study were to: 1) investigate the pathways between HL, SE, and PE on SCB in patients with type 2 diabetes; and 2) compare the differences in these pathways by preference in making medical decisions. These findings may identify key pathways to provide a more effective reference for professional interventions, and improve the SCB of patients with type 2 diabetes, thereby helping them achieve effective control of their blood glucose and avoid the complications associated with uncontrolled disease.

Study participants
The study was conducted with the participation of outpatients who were diagnosed with type 2 diabetes for more than 1 year (primary diagnosis included up to three diagnostic codes in the International Classification of Diseases, Ninth Revision, Clinical Modification: 250) upon visiting the Department of Metabolism of a regional teaching hospital in Hsinchu City, Taiwan. Purposive sampling was adopted to recruit eligible patients with consent during all clinic sessions from June through September 2017. Researchers described the purpose of this study briefly prior before acquiring written informed consent and distributing the questionnaires. A total of 372 questionnaires were distributed, and 316 completed self-administered questionnaires were collected, for a rate of valid questionnaires of 85%. The questionnaires include the scales of HL, SE, PE, and SCB, and the patient's preference in making medical decisions. The SE scale used in this study was developed based on the Diabetes Self-Efficacy Scale [31] and Perceived Diabetes Self-Management Scale [32], which included 19 items on a 5point Likert scale (from 1 = strongly disagree to 5 = strongly agree). It had a Cronbach's α = 0.83 for reliability.
The PE scale used in this study referred to the Chinese Diabetes Empowerment Process Scale [33], which included 15 items with 5-point Likert scale (from 1 = strongly disagree to 5 = strongly agree). It had a Cronbach's α = 0.95 for reliability.
The SCB scale used in this study was developed with reference to the Diabetes Self-Management Questionnaire [34], containing 14 items on 5-point Likert scale (from 1 = strongly disagree to 5 = strongly agree). The Cronbach's α = 0.83 for reliability.

The patient's preference decision scale was based on the Patient's Role Preference in
Decision-Making questionnaire [35]. Patients selected their preference for making medical decisions, selecting one of these five items: 1 = like to make treatment decisions on their own; 2 = like to make treatment decisions on their own after listening to physician's opinion; 3 = like to make treatment decisions together with the physician; 4 = like physician to make treatment decisions after talking to the patient; and 5 = like physician to make treatment decisions alone. Upon analysis, patients were categorized into three groups: patient decision-making (answers 1 and 2), shared decision making (SDM) (answer 3), and physician decision-making (answers 4 and 5).

Data analysis
The PLS-SEM incorporates into canonical correlation concepts the important statistical analysis techniques of regression analysis, principal component analysis, and path analysis. PLS-SEM can be applied to Mediation and Moderation analysis, and has gained universal attention in the field of health care [36]. It can not only help the researcher handle measurement problems of variable reduction but also addresses the structural problems of predicting and interpreting relationships in the research hypothesis. In interpreting the relationship of the latent variable to the dependent variable in regression analysis, this type of analysis is not affected by multi-collinearity, being distribution-free, requires only a small sample size, and can be applied to either formative or reflective measurements [37]. PLS-SEM was used to investigate the pathways between HL, SE, and PE on SCB in patients with type 2 diabetes and multiple group analysis was used to compare the differences in these pathways by decision-making preference. Descriptive statistics, including mean, standard deviations, and frequency, were used to analyze the distribution of patient characteristics, HL, SE, PE, SCB, and decision preferences.

Study participant characteristics
Among the 316 participants, 39.6% were male and 60.4% were female. Those aged ≥

Reliability and validity
In this study results were analyzed by measurement mode. The factor loadings of each dimension were between 0.592 and 0.939 (Fig. 1). The factor loading of FHL was 0.339.
According to Hair et al. [38], the factor loading of each dimension is recommended to be ≥ 0.5, so the FHL dimension was excluded from measurement mode analysis. Table 2 shows that the Cronbach's α and Composite Reliability (CR) values of each dimension were greater than 0.7, indicating good construct reliability and high internal consistency. The Average Variance Extracted (AVE) was greater than 0.5, showing convergent validity.
Fornell and Larcker's test shows that the correlation coefficient was lower than the value of the diagonal element (√AVB), indicating that the measurement mode had the required discriminant validity [39].

Pathways differences by preference in medical decision-making
To compare the pathways differences in terms of preference decisions (SDM and physician decision), Hensler et al. [40] pointed out that Measurement Invariance Assessment is necessary, which includes three steps: configural invariance, compositional invariance, and the equality of composite mean values and variances. Using the Measurement Invariance Assessment in SmartPLS usually automatically establishes configural invariance (Step 1). Table 4 shows the correlation coefficient of Step 2. In Step 3, the P value of both the mean value and the variance were greater than 0.05, which is not significant. These results show that the measurement invariance was established for the scales when measuring different groups. Table 4 Results of the measurement invariance assessment

Construct
Step 2 Step Next, multi-group analysis was used to compare the pathways between the SDM and the physician decision-making groups. In the five paths shown in Fig. 1, the path of HL directly positively affecting SE was significantly stronger for the SDM group than for the physician decision group (Table 3, path coefficient 0.249, P < 0.01).

Discussion
This study is the first to place HL, SE, PE, and SCB in patients with type 2 diabetes in the same model to explore their pathways used PLS-SEM, and their differences in terms of the patient preference for decision-making.
These findings clarify the relationships of HL, SE, and PE to SCB, which are: HL directly positively influences SE and SCB, and SE directly positively influences SCB. HL can also indirectly influence SCB through SE. These relationships were consistent with previous researches [7,16,17]. Patients with higher HL can better promote their own healthrelated behaviors, and they may feel more confident in their ability to complete SCB [41,42]. In particular, IHL and CHL have a greater impact on SE than FHL [7]. The more patients can enhance their SE, the more they may feel empowered to handle their situation [43], so HL plays an important role in the impact of SCB, and SE is also an important predictor of SCB [22].
This study also further clarified the effect of HL and PE on SCB, which is: HL directly positively influences PE, and PE directly positively influences SCB. HL can also indirectly influence SCB through PE. Studies have shown that HL and PE are deeply interwoven [21], and each independently affects SCB [44], but limited HL is a threat to PE and selfmanagement [45]. Wang et al. [28] proposed that PE may promote SCB in patients with high IHL and CHL, but may have no effect on SCB in patients with low communicative and critical health literacy (CCHL). Obviously, no matter patients are empowered externally or internally, these empowerment may sustained only when patients have adequate HL.
Increasing HL is an antecedent of PE [46][47][48]. Strengthening PE without adequate HL may lead patients to harm their health condition by making uninformed decisions, and HL plays a bigger role than PE in determining health status [18].
In this study, the mean PE was 4.03 and the mean SE was 3.82, indicating that patients tended to have more PE than SE. The combined care plans for patients with type 2 diabetes under the National Health Insurance System in Taiwan encourages patients to be empowered by health providers, so PE and SE are generally high. However, the mean of the HL scale was 3.42, indicating that patients' HL was obviously insufficient, and the mean of the SCB scale (3.66) was not high as well. Although these patients had the beliefs and actions to perform health behaviors and wanted to control their own health behaviors, they still felt a strong sense of powerlessness. Therefore, the self-management behaviors of the patients relied too heavily on the health care system to take active responsibility for SCB. The mindset of these patients must be changed, and their self-improvement in HL is the cornerstone by which to promote SCB.
In terms of the differences in preference in decision-making (SDM vs physician decision), HL directly positively affected SE in the five paths of Fig. 1 significantly more for the SDM group than for the physician decision group. Because HL can improve the ability of the patient to perform SCB, the patient then is better able to participate in SDM [49], further influencing clinical decisions. Patients will be more confident to take on self-management when they have more health-related knowledge, feel they can seek out resources and applications, and have positive interactions with health care professionals. As a result, the self-efficacy of patients is also improved. The study by Brabers et al. [49] showed that HL was associated with patient involvement in SDM, especially CHL. Patients participating in SDM have an increased commitment to health behaviors [50] and greater awareness and confidence to start their treatment [51]. Also, because patients with HL are more likely to play an active role in clinical decision-making, patients with HL are much more likely to show behavioral change [52]. Because HL involves all aspects of health care, such as prevention, screening, diagnosis, and treatment, it is considered the basis of the health care delivery system [53].
Limitations of this study include the fact that participant inclusion was based on patient consent, which may have introduced selection bias into the study sample. This study sample of a single regional hospital may not be generalizable. Also, the sample included only outpatients with type 2 diabetes; those with other types of diabetes or more advanced disease may have different outcomes. Further study is needed to explore the specific factors that influence HL, in order to develop an effective promotion strategy and the associated application tools. Overall, improving the HL of patients with diabetes can enhance their self-care ability.

Conclusions
This study provides insight into HL as a critical factor in improving the SCB of patients with type 2 diabetes, and clarified the pathways between HL, SE, and PE on SCB. HL was found to have a direct positive effect on SE, PE, and SCB. HL also had an indirectly positive influence in SCB through SE and PE. Separately, SE and PE each directly positively affected SCB. In the SDM group, HL directly positively affected SE significantly more strongly than in the physician decision group. Therefore, the development of an effective promotion strategy and application tools to improve the overall HL of patients with type 2 diabetes is the cornerstone of enhancing SCB to improve health and reduce complications in this patient population.