Scientific Papers

The current state of complex systems research on socioeconomic inequalities in health and health behavior—a systematic scoping review | International Journal of Behavioral Nutrition and Physical Activity

Description of Image

After removing duplicates, 4059 abstracts were screened. Of these, 383 full texts were screened and 36 were included according to the eligibility criteria. An additional 6 studies were identified by screening citations via hand searching, resulting in a total of 42 studies published between 1987 and 2023 included in the review. The PRISMA flow chart details the identification, screening, and inclusion decisions made (Fig. 2). In the full text screening phase, the most common reason for exclusion was study type, meaning that studies did not report applying a complex systems approach to develop a conceptual or simulation model.

Fig. 2
figure 2

Key study and model characteristics

In the 42 included studies, 4 contained both a conceptual and simulation model [29, 33, 36, 37], 18 only contained a simulation model [34, 38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54], and 20 only contained a conceptual systems model [32, 35, 55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72]. While all studies included at least one measure of SEP and stated, broadly, that a complex systems approach was applied, about a third of the included studies mentioned socioeconomic inequalities (N = 12) or a complex systems approach (N = 15) in their study aim.

Table 1 shows an overview of key study and model characteristics. Types of simulation models included agent-based models (N = 15) [29, 34, 36,37,38,39,40, 46,47,48,49,50,51,52,53], system dynamics models (N = 6) [33, 41, 42, 44, 45, 54], and a dynamic microsimulation (N = 1) [43]. The types of conceptual systems models were more varied, and CLD (N = 7) [35, 59, 67,68,69, 71, 72] was the most common.

Table 1 Key study and model characteristics of studies included in the review

A wide variety of health and health behavior outcomes were considered in the included studies, though some were more commonly modelled than others. Out of 42 total studies, 37 [29, 32, 34,35,36,37,38,39,40, 42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64, 67,68,69,70, 72] modelled one or more health behaviors. Diet or eating behavior (N = 17) [29, 35, 38, 39, 46, 47, 49, 51, 52, 56, 58,59,60,61,62, 64, 70] and physical activity or sedentary behavior (N = 12) [36, 37, 40, 43, 47, 48, 56, 57, 64, 67, 70, 72] were the most commonly modelled health behaviors. Other behaviors, including smoking (N = 4) [43, 57, 68, 70], alcohol consumption (N = 4) [42, 43, 53, 57], and sleep behavior (N = 2) [67, 70] were less commonly modelled. 30 studies [29, 32,33,34,35, 37, 41, 43,44,45,46,47, 50, 53,54,55,56,57, 60,61,62,63,64,65,66,67,68,69,70, 72] modelled one or more health outcomes, with general outcomes like health or well-being (N = 19) [29, 32, 34, 37, 43, 44, 50, 54,55,56,57, 60,61,62,63, 66, 67, 69, 70], mental health outcomes (including stress) (N = 12) [32, 34, 35, 37, 45, 53, 57, 64, 67,68,69,70], and chronic disease (N = 10) [34, 41, 44, 53, 54, 60, 65, 67, 70, 72] the most common. Obesity, cardiovascular disease, and oral health were present in the literature, but these were the least frequently modelled health outcomes [29, 43, 46, 61].

In addition to the categories of determinants included in the CSDH framework, individual-level determinants of behavior change (e.g., psychosocial factors, knowledge, skills, attitudes, and preferences) and the economic environment (external economic factors influencing consumers and businesses) were used to organize model elements. Overall, 41 out of 42 models (all except [33]) included at least one measure of intermediate determinants of health. These intermediate determinants included the health care system (N = 16) [29, 32, 34, 41, 43,44,45, 50, 54, 55, 57, 60, 61, 65, 68, 69], material circumstances (N = 37) [29, 32, 34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54, 56,57,58,59,60,61,62, 64,65,66, 68,69,70, 72], behaviors (N = 38; also includes behavior not directly related to health) [29, 32, 34,35,36,37,38,39,40, 42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64, 66,67,68,69,70, 72], individual-level determinants of behavior change (N = 27) [29, 34,35,36,37,38,39,40, 48,49,50,51,52, 56,57,58,59,60,61, 63, 64, 67,68,69,70,71,72], biological factors (N = 21) [32, 34, 37, 40, 48,49,50, 53, 57, 60, 61, 63,64,65, 67, 68, 70,71,72], and social cohesion (N = 26) [29, 32, 35,36,37, 40, 41, 44,45,46,47,48,49, 57, 60, 61, 63,64,65,66,67,68,69,70,71,72]. The most commonly included measures of material circumstances were related to the general physical environment (N = 20) [32, 35,36,37, 42, 48, 50, 54, 56, 58,59,60,61, 64,65,66, 68,69,70, 72], finance-related circumstances (N = 17) [32, 35,36,37,38,39,40, 42, 49, 51, 52, 59,60,61, 64, 68, 69], and the food environment (N = 16) [29, 32, 35, 38, 39, 42, 46, 47, 49, 51, 52, 59, 61, 62, 65, 66]. In addition to SEP, other measures of social stratification considered in the models were sex and gender (N = 13) [32, 37, 40, 43, 44, 48, 49, 51, 53, 57, 61, 64, 70] and ethnicity (N = 4) [32, 44, 52, 61]. Overall, 16 [32, 35, 42, 55,56,57,58,59,60,61,62,63, 66, 68, 69, 72] out of 42 models (38.1%) included at least one measure of structural determinants of health. These included governance (N = 2) [32, 61], macroeconomic (N = 5) [56, 57, 59, 61, 66], social (N = 2) [32, 69], and public (N = 6) [32, 42, 57,58,59, 68] policies, the economic environment (N = 5) [35, 59, 61, 68, 69], and culture and societal norms and values (N = 14) [32, 35, 42, 55, 56, 59,60,61,62,63, 66, 68, 69, 72].

We assessed the quality of reporting on the evidence complex systems models were based on and the extent to which key concepts of a complex systems approach were applied. About half (N = 23) [29, 34, 35, 37, 39,40,41,42, 44, 49, 52, 54, 56, 57, 59, 61, 63, 64, 67,68,69, 71, 72] of the included studies clearly described how the modelled relationships were based on literature, empirical study, or iterative model building processes. Table 1 contains descriptions of the evidence each model was based on.

The extent to which the models in the included studies applied key concepts of a complex systems approach is shown in Fig. 3. All but one model [58] explicitly contained heterogeneous elements. Other key concepts of a complex systems approach were explicitly applied by between 38.1% (emergence) and 66.7% (relationships between elements) of studies. There were no discernable patterns in the extent to which key concepts of a complex systems approach were applied in terms of health behaviors and health outcomes included in the models or in terms of study publication dates. Studies that applied all key concepts of a complex systems approach were more likely to report polarity of the model relationships than studies that did not apply at least one key concept (80.0% vs. 53.8%).

Fig. 3
figure 3

Quality assessment of the application of key concepts a systems approach

A visualization of the current state of research in a summary conceptual systems map

The direction of relationships was reported in 36 studies [29, 33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49, 51,52,53,54,55, 58,59,60,61,62, 64,65,66,67,68,69, 71, 72], polarity was reported in 26 studies [29, 33, 35,36,37,38,39,40,41,42, 44, 46,47,48,49, 51, 52, 54, 59, 60, 64, 67,68,69, 71, 72], and magnitude was reported in 10 studies [29, 33, 38,39,40,41,42, 44, 49, 51]. Simulation models contained more detail about the modelled relationships than the conceptual systems models, though conceptual systems models were not expected to report the magnitude of relationships (direction: 95% vs. 75%, polarity: 77% vs. 45%, magnitude: 45% vs. 0%). Relationships between model elements were extracted from most studies (N = 35), though relationships were too vague to extract from 7 studies containing conceptual systems models (17%) [32, 36, 56,57,58, 63, 70]. For example, the conceptual framework presented by Chastin et al. [56] contains a list of determinants of sedentary behavior belonging to different levels of influence, but it was not possible to extract specific relationships between these determinants or levels of influence.

The summary conceptual systems map of complex systems research on socioeconomic inequalities in health and health behavior, shown in Fig. 4, contains 66 elements and 399 relationships between these elements. The elements in the summary conceptual systems map are the element groups derived from the categorization process. The map includes relationships for which polarity was specified and consistent in the literature (positive or negative) but does not show relationships for which polarity was unspecified or inconsistent in the literature. Focusing on this subset of the relationships represented in the literature narrows the scope of the summary conceptual systems map in a way that favors studies that applied the key concepts of a complex systems approach, since the models in these studies were more likely to report relationship polarity. Figure 4 visualizes the complexity and interrelatedness of elements at different levels of influence, and the numerous displayed elements belonging to the material circumstances category illustrate the prominence of material circumstances in the literature. For more detailed insights, we encourage readers to view an interactive version of the summary conceptual systems map on the Kumu website: https://kumu.io/amudd/mudd-et-al-2024-summary-conceptual-systems-map-public [73]. An interactive version of the map that includes the additional 400 relationships for which polarity could not be deduced or was conflicting in our analysis can also be found on the Kumu website. The interactive version includes functionalities such as zooming in on the full map, zooming in on specific model elements (and their relationships with other elements), and filtering based on type of element or relationship. Supplementary File 4 lists the references for all relationships in the summary conceptual systems map and for relationships with unspecified or conflicting polarity, which are only visible in the interactive version of the summary conceptual systems map.

Fig. 4
figure 4

Summary conceptual systems map of complex systems research on socioeconomic inequalities in health and health behavior

Solid arrows represent relationships with positive polarity, and dashed arrows represent relationships with negative polarity. We encourage readers to view an interactive version of the figure at https://kumu.io/amudd/mudd-et-al-2024-summary-conceptual-systems-map-public

Social capital, income, financial strain, the built environment, and health-positive attitudes, beliefs, and preferences were direct drivers of the largest number of other elements in the summary conceptual systems map, meaning that these elements had the most outgoing arrows towards other elements. Health-positive attitudes, beliefs, and preferences, health-positive tendencies and habits, general health, and healthy diet had the most incoming arrows from other elements, meaning that they were directly driven by the largest number of other elements in the summary conceptual systems map.

We identified 15 direct shared drivers of socioeconomic inequalities in health and health behavior, many of which were material circumstances (N = 4) or individual-level determinants of behavior change (N = 4). General health, financial strain, the cost of resources, healthy diet, and health-positive attitudes, beliefs, and preferences were direct shared drivers of socioeconomic inequalities in the greatest number of outcomes. For example, financial strain was a direct shared driver of employment-based inequalities in general health, mental health, stress, and healthy diet, which is depicted in Fig. 5. In the literature, better employment led to less financial strain, and more financial strain led to worse general and mental health, more stress, and a less healthy diet.

Fig. 5
figure 5

Financial strain as a shared driver of socioeconomic inequalities in multiple health and health behavior outcomes

Solid arrows represent relationships with positive polarity, and dashed arrows represent relationships with negative polarity

Description of Image

Source link