Psychometric properties of the general self-efficacy scale among Thais with type 2 diabetes: a multicenter study

Introduction

Diabetes is widely acknowledged as one of the fastest growing chronic conditions worldwide with 1 in 11 adults (463 million) living with the condition, almost 80% of whom live in low or middle income countries (International Diabetes Federation, 2019). A large majority of people living with diabetes in the world (approximately 90%) have Type 2 diabetes(T2D) which is widely considered a lifestyle disease, and factors such as urbanization and increasingly sedentary lifestyles have been strongly linked to the increase of T2D (Hurst, Rakkapao & Hay, 2020). In Thailand, a country with a resource-limited health care system, the prevalence of T2D among adults has tripled over a 25 year period, rising from 2.3% in 1991 to 8.0% in 2015 (Euswas et al., 2021). This also represents a considerable burden on the Thai health care system and economy which is estimated to represent 21% per capita gross domestic product (Chatterjee et al., 2011).

One way to address this increasing burden of T2D is to augment the health care system by facilitating patient self-care. Diabetes is a disease where patients’ conscientious attention to their own disease can prolong, and even prevent, the onset of chronic diabetes complications such as diabetic retinopathy, nephropathy, neuropathy and peripheral artery disease. Indeed it has been suggested that strong diabetes self-management may be almost as efficacious as oral hypoglycemic agents in controlling blood glucose (McDowell et al., 2005), a clinical target widely acknowledged as being highly associated with preventing the onset of downstream chronic diabetes complications. It has been shown that among Thais living with T2D, each unit higher a patient scores on the diabetes self-care scale (when considered as a 10-point scale) doubles the odds of blood glucose control (Hurst, Rakkapao & Hay, 2020).

Self-care in those with diabetes has been shown to be related to diabetes management self-efficacy, the extent to which a patient believes they can manage their condition (McDowell et al., 2005). Diabetes management self-efficacy can be measured by the Diabetes Management Self-Efficacy Scale (DMSES), a well-established scale that has been validated in several populations including Thais living with T2D (Sangruangake, Jirapornkul & Hurst, 2017), as well as those in many other countries including the Dutch (Van der Bijl, Van Poelgeest-Eeltink & Shortridge-Baggett, 1999), Italians (Messina et al., 2018) and Australians (McDowell et al., 2005). The level of a patient’s diabetes management self-efficacy can have a profound effect on patient outcomes. Hurst, Rakkapao & Hay (2020) demonstrated that not only is diabetes management self-efficacy a major mediator of diabetes self-care in terms of controlling blood glucose, but it can even have an additional (direct) effect on blood glucose control. The reason why diabetes management self-efficacy may improve blood glucose control beyond its influence on diabetes self-care is unclear. We need to look more closely at mechanisms underpinning disease management self-efficacy.

The notion of self-efficacy was first proposed by Bandura (1977) and, in general, represents a person’s belief in their ability to perform particular tasks or practices. Importantly, Bandura believed that self-efficacy is a primary predictor of individuals’ behavior change, and people who show the highest levels of behavior change are those with high self-efficacy (Dehghan et al., 2017). Bandura speculated self-efficacy to be strictly task or domain specific (Bandura, 1986; Van der Bijl, Van Poelgeest-Eeltink & Shortridge-Baggett, 1999), whereas others believe that self-efficacy can be generalized to encompass a broader and stable sense of an individual’s personal competence to deal with a range of stressful circumstances (Schwarzer & Jerusalem, 1995; Sherer et al., 1982; Luszczynska, Scholz & Schwarzer, 2005). For instance, if an individual has high self-efficacy in one domain, it might be expected that they are more likely to have high self-efficacy in another. In this respect, self-efficacy can be conceptualized as having both a general (or global) and domain-specific component (Salsman et al., 2019). General self-efficacy may be considered a more stable personality trait and likely to be reflected by someone’s optimism or resilience, whereas more domain-specific aspects of self-efficacy (e.g., diabetes management self-efficacy) may drive more specific aspects of health behavior and therefore more likely to reflect a person’s particular experiences in this more narrowly defined domain (Salsman et al., 2019).

How domain-specific and general self-efficacy instruments converge and diverge may also provide insights into how to best craft intervention components in terms of domain-specific or general aspects of patient empowerment. For example, Luszczynska, Scholz & Schwarzer (2005) showed that GSE had varying levels of association with various domain-specific types of self-efficacy among different populations. For instance, swimming self-efficacy among swimmers was shown to be quite highly associated with the GSE, whereas smoking abstinence self-efficacy had only a weak association with the GSE. This might also flow through to sub-domains of domain-specific self-efficacy instruments. For instance, the extent to which general self-efficacy relates to the various sub-domains of diabetes management self-efficacy (diet-, exercise-, monitoring- and medical treatment self-efficacy) may provide some insights into how both general- and diabetes management self-efficacy are likely to act on patient self-care and patient outcomes, and furthermore, may shed some light on which aspects are more likely to be successfully modified through intervention. For example, in patients with poorer general self-efficacy, it may be easier to influence their adherence to a program of regular blood sugar monitoring than affecting major lifestyle changes such as dietary and exercise behavior.

The GSE is a widely used instrument and has been validated in many populations across the world. While it has been validated in some Asian populations including the Chinese (Zhang, Zhan & Liu, 2018) and Indonesians (Putra, Rahayu & Umar, 2019), no Thai language version of the GSE exists. The aim of this study is to demonstrate the validity of the General Self-Efficacy scale (GSE) in Thai adults with T2D. We then examine its psychometric properties including both its overlap (convergence) and discrimination (divergence) from the domain-specific Diabetes Management Self-Efficacy Scale (DMSES).

Methods

Results

Patient characteristic are provided in Table 1.

The average age of patients was 53.6 years old (sd = 21.6) with almost 58% of patients being women. Over half of the patients resided in rural areas (56.5%), and a majority of patients were married (80.8%), had primary or secondary school as their highest level of education (85.5%) and earned less than 15,000 Thai Baht (USD$480) per month (76.1%). Almost half of the patients had a family history of diabetes (45.5%) and 72.3% of them had some comorbidity with levels of hypertension and dyslipdemia being particularly high (49.1% and 35% respectively). The levels of blood glucose, blood pressure and low density cholesterol control was 39.6%, 31.1% and 41.1%, respectively.

Discrimination and convergence between general and diabetes management self-efficacy

Figure 1 shows the correlation of GSE with the total DMSES scale and its subscales and suggests there is substantial overlap among most of the DMSES subscales, although monitoring self-efficacy was only moderately correlated with the other DMSES subscales. In contrast, associations of GSE with the total DMSES scale and the DMSES subscales were considerably lower (correlations ranging from 0.181 to 0.258) providing some evidence of divergence. However, Pearson’s correlation coefficient does not account for measurement error in scales, so we used the disattenuated correlation coefficient (a measure of association between scale that accounts for their unreliability). Disattenuated correlation analysis between GSE and the DMSES provided strong evidence of discriminant validity (r_{disattenuation} = 0.283, 95% CI [0.214–0.352], p < 0.001).

Discussion

In this study we demonstrate the reliability and validity of the GSE in Thais living with diabetes, and explore its psychometric properties. We show that the unidimensional GSE is structurally valid in the Thai T2D population and that it has substantial overlap with the domain-specific DMSES, but it also has sufficient discrimination to measure aspects of global self-efficacy not captured by DMSES. This suggests that there is a considerable proportion of domain-specific diabetes management self-efficacy not explained by a person’s general self-efficacy, and consequently, diabetes management self-efficacy which has been shown to be a strong predictor of blood-glucose control among Thai T2D patients (Sangruangake, Jirapornkul & Hurst, 2017), may be improved by well-designed disease-focused behavioral interventions. Indeed, several studies have trialled interventions aimed at improving diabetes self-care through improving patient diabetes management self-efficacy in other populations, and have shown considerable success (Wong et al., 2015; Lin et al., 2020).

In terms of the correlation of GSE with the DMSES and its subscales, despite being statistically significant, we found only weak to moderate associations between the general and diabetes self-management specific measures of self-efficacy. In our sample we also observed that some domains of DMSES (e.g., Monitoring self-efficacy) correlated more highly with GSE, than others (e.g., Regimen self-efficacy). What this means in terms of the efficacy of current self-efficacy interventions in people living with diabetes, and what insights it might provide in the design of future interventions is something that needs further consideration. However, it does highlight that the relationship between general and different types of domain specific types of self-efficacy may be both subtle and complicated. This is something we see in other studies of the relationship between general self-efficacy with other domain-specific measures of self-efficacy. The magnitude of these associations is comparable to those found in Luszczynska, Scholz & Schwarzer (2005) who conducted a correlation analysis of GSE with other domain-specific measures of self-efficacy such as exercise-, nutrition- and smoking abstinence- self-efficacy. It is also worth noting the internal consistency reliability we found among Thais living with diabetes(0.87) was consistent with that found in other populations such as those from Germany (Schwarzer & Jerusalem, 1995).

In terms of how both GSE and DMSES vary with patient characteristics, we found region of residence, education and income to be the main drivers of both general and diabetes management self-efficacy. Interestingly, Northeastern Thais, a region with the highest prevalence of T2D (Sieng et al., 2015), had a considerably greater chance of higher general and diabetes-management self-efficacy. We also found that alcohol and tobacco use were also associated with both higher GSE and higher DMSES, although there may be a reverse causal association here. Indeed, it brings up the question of whether measures of self-efficacy may also reflect patient complacency. Some of those who feel confident in managing their disease may underestimate the difficulties and/or commitment in doing so.

We demonstrate that the GSE has excellent psychometric properties, and it is suitable for use in Thais living with T2D, and that this measure can now be used not only as either a health outcome or determinant, but we can now explore what aspects of self-efficacy may be amenable to change through intervention, and which might be more closely associated with a patient’s underlying personality traits. In short, we gain a clearer view of what aspects of self-efficacy may be amenable to change through intervention, and which are more likely to be immutable. It is also likely that GSE is likely to be valid in other Thai chronic disease populations especially in diseases that can have a lifestyle etiology such as hypertension and hypercholesterolemia. It is also likely that GSE would be likely to perform well in culturally similar T2D populations such as those of Vietnam, Laos, Cambodia, Malaysia and Myanmar. However, GSE measurement validity would still need to be formally established in these populations.

Although we did take care to obtain a geographically representative sample of the recently diagnosed Thai T2D population in this multi-center study, we only sampled patients from community hospitals which predominately service patients in the earlier stages of the disease, and even where chronic complications are present, morbidity is generally lower in patients attending these hospitals compared to those attending larger tertiary hospital outpatient clinics in Thailand (Sieng et al., 2015). It is also important to note, however, that community hospitals in Thailand service a large majority of T2D outpatients, especially those in the earlier stage of the disease (Sieng et al., 2015), where lifestyle interventions are more likely to be effective. Although our study was initially designed to include larger hospital outpatient clinics, the COVID-19 epidemic occurring at the time of sampling led to institution-specific administrative hurdles making access to their patients very difficult. Finally, given patients were surveyed only once, at their yearly check-up, we were unable to follow the COSMIN guidelines in their entirety. In particular, the cross-sectional nature of our study design meant we were unable to assess test-retest reliability of GSE, nor GSE’s responsiveness or predictive validity.

However, our study also had some major strengths. Our study design resulted in a nationally representative sample of people living with the earlier stages of diabetes in Thailand. As a result we feel our findings strongly confirm that the GSE is generalizable to all Thai’s living with the earlier stages of T2D. In this study we demonstrate that the Thai version of the GSE is psychometrically valid in the Thai T2D population. This patient-centered construct is likely to be useful in teasing out the complex interplay between patient disease knowledge, self-care attitude and practice. This understanding will provide strong guidance in designing well-grounded cognitive behavior interventions which in turn represent a cost-effective way of improving patients disease self-management, and subsequently, better patient outcomes in a resource-limited health care setting.

Conclusions

We demonstrated that the Thai version of the GSE is both a valid and reliable measure of general self-efficacy among Thais living with Type 2 Diabetes. We believe that it is also likely to prove valid among Thais with other metabolic syndrome related conditions such as hypertension and dyslipidemia. Furthermore, we believe that the GSE may be useful in being able to tease out the subtle interplay between disease knowledge, disease management self-efficacy, patient self-care, and how these relate to downstream patient outcomes. Stronger understanding of these relationships is likely to lead to the development of better focused and effective interventions that can help delay, and even prevent, the development of chronic diabetes complications.

Supplemental Information