The aim of this study was to compare identified regulation, self-efficacy and social support and their association with PA and HbA1c across divers socio-cultural environments. Comparison of these motivational constructs was possible since measurement invariance could be established across samples, suggesting a similar understanding across study populations. However, obtaining adequate measurement models required modification of the initial measures (e.g., items had to be dropped). Estimates of the motivational constructs were highest in Uganda and lowest in South Africa, with a substantial difference for barrier self-efficacy and social support. Structural models did not correspond across settings. Identified regulation was positively associated with vigorous PA in Uganda and with moderate PA in South Africa. In Sweden, none of the PA outcomes was associated with identified regulation. The strength of the association between social support and identified regulation was similar across settings. Depending on the setting, the association between social support and PA outcomes was weak to insignificant. The association between barrier self-efficacy and PA outcomes was not significant. Self-reported PA was highest in Uganda and lowest in Sweden. Vigorous PA was associated with lower HbA1c across countries, while this association was not significant for moderate PA.
The latent mean level of barrier self-efficacy was much lower in South Africa compared to both other settings, which may have occurred for several reasons. External barriers that are more prevalent in South Africa may hinder people from doing PA, including lack of security in the South African township setting [4]. South Africa’s historical context and pervasive social inequality may also have contributed to the perceived difference in levels of self-efficacy [4]. The lower socioeconomic status of study participants, including many internal migrants, may have affected people’s self-esteem [33], and consequently, their self-efficacy [34]. Potentially, this could also explain the lower self-efficacy level (compared to the Ugandan setting) in the Swedish setting, as the study sample was socioeconomically disadvantaged and included 60% migrants [35].
Social support was perceived to be much higher in the Ugandan setting compared to the other settings. This could be explained by the stronger social ties apparent in the rural Ugandan community versus the urban sample with many migrant workers in the South African and Swedish setting [4]. The proportion of participants that indicated to be married or co-habiting was also substantially higher in Uganda. In qualitative interviews conducted during the formative phase of the SMART2D project, participants from the Swedish setting reported to perform PA on their own, rather than with others, which corresponds to health and lifestyle being more individualized in Sweden [4]. We hope that future research can explore these factors in more depth.
The positive association between identified regulation and PA (vigorous PA in Uganda and moderate PA in South Africa) is in line with previous studies on SDT and PA [6]. However, this association was not found in Sweden, nor for other PA outcomes. We assume this lack of association can be explained by the PA-related questions covering all types of PA, including travel- and work-related PA. It is likely that most of the self-reported PA related to travel and work, especially in the African settings [36], which implies an important effect of triggers external to the activity. The difference between Uganda and South Africa could be explained by a different attitude towards the intensity types of PA, although this hypothesis was not tested in this study. In the rural Ugandan setting, participants may have been more used to performing vigorous activity as 69% of the participants were farmers [17], while in the urban township setting of South Africa, participants reported to perform less vigorous physical activity. As they seem less used to perform vigorous PA, they might perceive it as more demanding and hence, they might be more inclined to connect moderate rather than vigorous PA with autonomous forms of motivation. Other studies have highlighted the importance of the type of PA influencing the relationship between PA and the type of motivation [6]. For example, PA of a more repetitive nature was shown to be stronger associated with identified regulation than with intrinsic motivation [6]. Our study suggests that the association with more autonomous forms of motivation may depend on the perceived intensity of PA and that this association is context dependent.
From a statistical perspective, differences in dispersion of the outcome data may explain why certain associations did or did not occur. For instance, vigorous PA in Sweden (inter-quartile range = 1.5) and in South Africa (inter-quartile range = 2) had a lower dispersion compared to moderate PA in both countries (inter-quartile range = 6 and 4 respectively). In Sweden, 69% of the participants reported to perform 0 days of vigorous PA.
The construct of social support showed a similar positive association with identified regulation across the three country settings. Since this construct shows conceptual parallels with the concept of perceived relatedness [24], this finding supports the etic validity of the basic psychological needs theory, which posits that satisfaction of psychological needs fosters more autonomous forms of motivation and more sustainable behavioral outcomes (Deci and Ryan, 2000). However, this did not apply to the construct of barrier self-efficacy, which showed a positive relationship with identified regulation in Uganda, but not in South Africa and Sweden (although the latter could have been due to a lack of power). Moreover, barrier self-efficacy, which has been shown a consistent predictor of PA [16], did not show an association with PA outcomes. This may be explained by most of PA being related to work or travel, with an important influence by external triggers and not leaving much flexibility to participants. On the other hand, our data did provide support for an association, although small, between social support and self-reported PA. A potential explanation could be that participants included companionship at work or during travel in their conceptualization of social support.
Higher self-reported PA in the African sites of the study compared to a high-income Western country such as Sweden is in line with global trends and likely due to a higher level of travel- and work-related PA [36]. The high levels of PA in Uganda is also in line with a recent national survey [37]. In sub-Saharan Africa, urban regions have been associated with a more sedentary lifestyle which may explain lower self-reported PA in the South African versus Ugandan setting [16]. Besides these global trends, a substantially higher proportion of the Ugandan participants reported to be employed.
Self-reported frequency of vigorous PA showed a similar negative association with HbA1c across the three settings. Regression estimates for self-reported frequency of moderate PA were about half the size of vigorous PA estimates and non-significant. This is not in line with a recent meta-analysis of randomized trials which found changes in HbA1c driven by the duration of PA in a linear manner and independent of the type and intensity of the PA intervention [32]. If only duration and not intensity of PA would matter, a straightforward explanation of our findings would be that participants’ reporting of PA is dependent on the intensity of PA. In other words, participants’ perception to have performed 30 min. of moderate PA may be different from their perception to have performed 15 min. of vigorous PA. These findings from a real-life setting warrant further investigation as they would be crucial to consider in health promotion. Experimental trials using objective and self-reported measures can bring more insight.
Study limitations and recommendations
Comparison of motivational constructs was possible across settings, but measures had to be modified to obtain adequate measurement models. This incompatibility across very different contexts could be explained by translation to local languages altering certain nuances, a different contextual relevance of certain items and a different understanding by participants from different settings. As mentioned in the results section, two factor loadings of the barrier self-efficacy construct were very low for the Swedish site. However, sensitivity analysis after exclusion of the item with the lowest factor loading did not reveal any major differences.
We further need to acknowledge that sampling procedures and criteria were different in the three settings for pragmatic reasons. As we documented in the methods section, participants in the Ugandan setting were recruited through a random sampling procedure, while participants in the other sites were mainly recruited from health centers, at the cost of reducing external validity. Moreover, selection criteria also differed across sites: participants from the Ugandan and South African site had not been diagnosed with T2D for longer than 12 months, while this was 5 years for the Swedish site. Such differences may have influenced the mean levels and relationships between the motivational constructs in this study. Another limitation of our cross-comparison study was that we did not directly assess relevant cultural differences (e.g. via measures designed to assess individualism/collectivism) which could have been served as possible moderators of the effects we measured.”
The distinction between self-reported vigorous and moderate PA offered an interesting perspective, but also hindered the association of constructs with other factors that determined PA performance. A distinction between categories of PA (e.g., work-, leisure-, transport-related PA) and other categories of motivation from the SDT continuum may add further insight into the role of motivation. In addition, controlling for other factors (e.g. perceived safety, availability of sports infrastructure, etc.) may result in a more nuanced image of the role of motivation.
This study aimed to assess the cross-cultural validity of an SDT process model across different settings. However, the cross-sectional design of this study does not provide evidence for causal pathways or trends over time. While this study focused on people who were recently diagnosed, different dynamics may appear in people with long-standing diabetes. Studies collecting data at different time points and intervention trials can address these shortcomings.
Finally, we need to acknowledge that the use of self-reported measures exposes our findings to bias, including social desirability bias, recall bias (people who value PA more as beneficial for health, may also have reported higher values of PA) and interviewer bias. Shared method variance between measures may have led to overestimation of associations. Objective registration of PA through a pedometer or accelerometer could have made our findings more robust and challenged self-reporting.