Background Despite remarkable progress in the scientific understanding of the biological characteristics of the pathogens, pathogenesis and immunology, human sexual behaviour and population transmission dynamics, there are still considerable knowledge gaps regarding the heterogeneity and determinants of epidemics of sexually transmitted infections (STI) and HIV. To understand more fully the causes of STI and HIV epidemics it is necessary to reconcile individual and population approaches and bring together sociological and biomedical streams of research.
Methods This study examined the implications of approaching the study of STI and HIV epidemics from the perspective that individual and population-level characteristics and interactions result in emergent properties and structural patterns that cannot be easily predicted. In addition to offering examples from the research literature, female sex work is analysed as an example of a complex adaptive system and the implications for STI and HIV epidemics are examined in that context.
Results Previous research in this field has provided compelling examples of how the complex interplay of individuals and resulting structural patterns including sexual networks can influence the patterns of emergence and the trajectory of STI and HIV epidemics.
Conclusions Approaching the study of STI and HIV epidemics as emergent phenomena arising from complex interactive systems with diverse structural patterns offers a promising avenue for developing a more coherent understanding of these epidemics. It would also promote consilience between sociological, population and biomedical disciplines that could open new vistas for the science of public health.
- Sexually transmitted infections
- emergent properties
- complex adaptive systems
- sexual networks
Statistics from Altmetric.com
Fundamentally, science that is meant to underpin the prevention of sexually transmitted infections (STI) and HIV must ultimately provide insight into what explains the heterogeneity in the epidemics of these pathogens in diverse human populations. The past quarter century has seen remarkable progress in the scientific understanding of the biological characteristics of the pathogens, pathogenesis and immunology, human sexual behaviour and the population transmission dynamics of sexually transmitted pathogens.1 Moreover, there has been substantial research into the social, economic and cultural contexts in relation to STI and HIV.1–4 However, despite this progress substantial questions remain. In this regard, a recent review by Padian and colleagues5 is illuminating. They summarised and discussed the results of 37 randomised controlled trials (science's ‘gold standard’ for evaluating efficacy) of HIV preventive interventions. Only five of these trials demonstrated a protective benefit, three focused on the same intervention (male circumcision), and four of them assessed prevention at the individual level and not the population level. As Padian and colleagues5 point out, part of the problem has been the focus on the discovery of a ‘magic bullet’ solution for HIV prevention, whereas there is increasing interest in shifting to combination prevention packages that could capitalise on synergies among interventions.6 7 This more comprehensive approach presumes that the causes of epidemics within specific contexts are known and that the correct intervention mix can therefore be selected. However, as Merson and colleagues8 point out, there remain considerable knowledge gaps in this realm: ‘Even more urgently, there is a need for reliable evidence-based research to better guide the selection of available behavioural and structural interventions in specific areas or populations’. As Pisani and colleagues9 have observed, it is somewhat troubling that much of the empirical and conceptual inference in relation to HIV epidemiology moves from effect to presumed cause; from epidemic amplitude to presumed epidemic drivers. Therefore, much research has focused on understanding the population distribution of HIV in different settings, rather than on understanding the differences in sexual structures and transmission dynamics that can guide intervention priorities accordingly. Despite the considerable scientific advances in various research domains and their integration, there remains a lack of a comprehensive and coherent scientific framework by which to study, describe and explain the sources of epidemic heterogeneity, and more importantly to predict the trajectory and amplitude of epidemics in different populations. To develop a more comprehensive and robust scientific framework for understanding the determinants of STI and HIV epidemics, it will be important to reconcile more fully the sociological and the biomedical sciences and bring together population and individual research paradigms. In this paper we briefly review the progress and remaining challenges inherent in developing such a framework, discuss how structural patterns and emergent properties must be incorporated into such a framework, and provide suggestions as to how the development of such a framework might be approached.
In many ways, the recent history of STI and HIV research echoes themes from the broader history of epidemiology and public health science. As Terris10 has recounted, central among these themes has been the contrast between sociological and biomedical views on the causation of epidemics. In the 19th century, this was reflected in the competing views of ‘contagionists’, who sought to improve public health by containing the spread of disease, and ‘anti-contagionists’, who contended that disease was related to unhealthy social and environmental conditions (miasma). Following the emergence of the germ theory of disease and well into the 20th century, the emphasis in public health science and application was on the microbiological aetiology of disease and the control of infectious diseases. Eventually, this was countered in large measure by the ‘social medicine’ movement in the Great Britain, which placed emphasis on non-communicable diseases (‘diseases of prevalence’) and the role of social conditions and occupations in the occurrence of these diseases. Terris10 indicates that this heralded the ‘second epidemiological revolution’, with the emergence of epidemiological approaches and statistical methods that still predominate in much of public health research, which focus on deriving associations between risk factors and disease based on probabilistic inference from data on groups of individuals or populations.
In STI/HIV research there has been a similar tension and interplay between sociological and biomedical perspectives and individual and population approaches.11 The dramatic advances in understanding the biology of STI and HIV have been parallelled by the burgeoning of research to obtain a better understanding of population transmission dynamics.12 13 Epidemiological research into the behavioural risk factors for STI and HIV has been accompanied by the growth of research into the role of social and sexual networks in their spread.14 15 Perspectives on STI and HIV epidemics transitioned from a temporally static concept to an appreciation for the ‘dynamic topology’ of epidemics as they evolve over time.16 However, despite widespread recognition of the importance of both sociological and biomedical research to understand more fully the determinants of STI and HIV epidemics, there remain considerable obstacles to achieving consilience between these scientific streams.
Multilevel approaches to STI and HIV research
Recently, there has been a movement in STI/HIV research to understand epidemics in relation to the combined and interacting effects of individual and population factors through the use of multilevel conceptual and analytical approaches.1 14 17–20 Conceptually, the multilevel approach offers promise for bringing together the sociological and biomedical scientific domains by incorporating determinants at multiple levels, including population level (macro), community or social group level (meso), and individual level factors. As Diez Roux and Aiello19 highlight, this approach moves away from viewing disease risk as being directly related to individual attributes and behaviours, and from viewing populations as simply collections of individuals. Instead, it elaborates and articulates how individuals and groups are embedded within larger social contexts, and attempts to assess simultaneously how individual and various levels of contextual factors affect the disease risk of individuals and disease rates across populations. Despite its obvious appeal, few studies have been published that employed multilevel modelling in STI and HIV research, and these have generally not been directed to understanding the characteristics and dynamics of epidemics.19 There are substantial theoretical and empirical challenges in applying multilevel modelling methods to the study of infectious diseases, including STI and HIV.19 21 Chief among these are the difficulties in modelling the dependencies between individuals within a population such that an individual's risk of acquiring or transmitting an infectious pathogen is not only dependent on that individual's characteristics and behaviours, but also on patterns of disease and exposure among other individuals and networks in the wider population. Studying these relationships in detail requires the use of mathematical models that can simulate transmission dynamics in populations. In addition, substantial theoretical and empirical research has shown the importance of characterising the properties of social and sexual networks and how they relate to transmission dynamics.22–24 Whereas multilevel models can incorporate some of these network characteristics as population level variables, they are still constrained by an inability to reflect adequately the dynamic changes that occur within a population over time when features of the population or its environmental context changes.19 In this respect, it has been observed that the essential linearity of hierarchical regression models render them insufficient for understanding the complex and often non-linear relationships that occur within social networks and in the reciprocal relationships between individuals and networks and the social environment.21
Emergent properties and structural patterns
More than a decade ago Albrecht and colleagues25 posited that ‘human health is the outcome of complex processes that operate within and across physical, psychological, social and ecological systems’ and that ‘our way of thinking about health problems will need to reflect this complex interrelatedness’. More recently, Glass and McAtee26 proposed a conceptual framework to integrate biological and social processes more fully into the study of behaviour and health, recognising the complexity of the interplay between these processes. Diez Roux27 has summarised some of the complexities of social–biological interrelations. These include the interdependence of heterogeneous individuals, the presence of non-linear relationships and feedback loops in the interactions between individuals and in the dynamic interplay between individuals and their environment. In addition, these complex interactions often exhibit emergent properties, which Diez Roux27 defines as ‘properties of the system that arise from the functioning of its interdependent components but are not simple aggregates of component-level properties’. The concept of emergent properties is not new in the scientific literature, spanning the basic, applied and social sciences since the mid-19th century. McLaughlin28 has traced the history of scientific thinking and discourse in relation to ‘emergence’ and summarises the important early thinking, citing the work of Stuart Mills, Alexander Bain and George Henry Lewes, among others. He relates that Lewes coined the term ‘emergent’, referring to what Mills described as a ‘heteropathic effect’, which McLaughlin defines as ‘an effect that is not the sum of what would have been the effects of each of its causes had they acted separately’.
In addition to emergent properties, complex systems often exhibit group-level or population-level structural properties that can be measured and used as predictors of individual-level and population-level disease outcomes. For example, the structure of social contacts in a group or the configuration of a sexual network are group-level structural characteristics that are influenced by the social context and the psychosocial characteristics of individuals, and might also directly influence population transmission dynamics. These structural properties could be relatively stable over time, or change in response to changes in other components of the system. Figure 1 provides a schematic diagram illustrating the interplay between population-level and individual-level variables in relation to STI/HIV epidemics. In this framework dynamic interactions between individual and societal factors co-create structural patterns such as mixing patterns and sexual networks among individuals. These structural patterns, along with individual characteristics and behaviours, affect the transmission dynamics, and thereby influence the epidemic trajectory and amplitude in a population. The main differences between this framework and others that incorporate population-level and individual-level variables17 are the explicit identification of structural properties (eg, sexual networks) that are important determinants of transmission dynamics and are influenced by the interplay of individual and population-level factors, and the emphasis on feedback loops between various components of the model.
Emergent properties often lead to apparent paradoxes and inconsistencies when considering relationships between aggregate measures of individual characteristics or behaviours and population-level phenomena. For example, the observed relationship between socioeconomic deprivation and the risk of HIV at the individual and population level is not straightforward, with a positive association in some contexts, and an inverse association in others.29 Despite the fact that by 2002 the proportion of young African-American women in the USA reporting multiple partners within the past year had reduced to a level equivalent to young non-Hispanic white women, their incidence rates of gonorrhoea and syphilis were approximately 20 and 30-fold higher, respectively.30 Adimora and Schoenbach14 have pointed to the complex demographic, social and network properties in African-American communities in the USA to explain this profound public health fact. Similarly, Low and colleagues31 have highlighted substantial differences in gonorrhoea and chlamydia rates between black ethnic groups in London, pointing beyond individual behaviours to factors such as sexual mixing patterns and networks to explain higher rates among those of black Caribbean origin.31 In Pakistan, the prevalence of HIV among injection drug users in different cities varies by 10-fold without any consistent association with average injecting behaviours, including injection frequency, duration of drug use and needle sharing.32 Potential explanations include the existence of larger injection drug user clusters in cities with high prevalence,32 33 and more ‘open’ injecting networks in cities with lower prevalence.34
Female sex work as a complex system
To illustrate further the challenges and opportunities presented by the integration of social and biological science in STI/HIV research, we examine female sex work as a complex adaptive system. Although female sex work is a dominant driver for many HIV epidemics globally, there is substantial heterogeneity in the amplitude of HIV epidemics related to female sex workers (FSW).35–37 Moreover, there is substantial diversity in the social organisation of sex work across contexts and over time.38–42 Differences in the social organisation and structural patterns have important implications for the transmission dynamics of STI and HIV in sex work networks, and in the wider population.36 43 In figure 2 we illustrate how female sex work can be conceptually summarised as a complex system. The macro-level societal context, consisting of the particular socio-cultural and economic milieu exerts influence on the size, socioeconomic, and demographic characteristics of the FSW and male client populations. Detailed social mapping studies in India and Pakistan have demonstrated the substantial variability in the relative size of the FSW population, even across small geographical areas.37 44 For example, a study in northern Karnataka found that the number of FSW per 1000 men varied from 9.6 to 30.5 in six regions of a largely rural district, with the regions with the highest FSW population being irrigated and more affluent, possibly due to increased client demand in those regions.37 Similarly, the relative size of the client population varies by context and is largely shaped by macro-level societal factors. The societal context also influences the organisation of sex work, including features such as the location and venues of sex work (eg, brothels, ‘red light’ areas, indoor establishments, etc.), how FSW and clients connect (eg, by pimps or brokers, through open solicitation, by cell-phone, etc.), and who controls the work conditions of FSW (eg, individual autonomy, tight controls by madams and pimps, etc.). For example, in some cities in Pakistan where sex work is highly censured, the organisation of sex work is such that it is unobtrusive and in many respects hidden from the authorities and the general population. This often entails the involvement of pimps or ‘brokers’ to connect clients to FSW who often work in private homes, or the establishment of ‘kothikhanas’, which function as small mobile brothels wherein FSW circulate between apartments or small houses that are established by madams. Madams move the ‘kothikhanas’ frequently, often keeping a relatively stable group of clients and changing the FSW periodically. In contrast, in some major Asian cities where sex work is more openly accepted and there are large migrant male populations, sex work has been organised into large static brothels and/or red light areas to facilitate large client turnover.
The characteristics of FSW and clients and the organisation of sex work result in variable properties that are relevant for STI and HIV transmission dynamics. These can be represented as either aggregate properties of the sex worker and client populations or as structural patterns of interaction between sex workers, clients and the general population. Generally, research focuses on the aggregate properties such as average client volume, condom use rates, age distribution of sex workers and duration in sex work. While these parameters are useful and accessible through standard survey methods, there is growing consensus and theoretical research that indicate that differences in more complex structural patterns are key drivers of heterogeneity in STI and HIV epidemics. For example, Boily and colleagues36 have shown that in HIV epidemics that are largely driven by female sex work, if the total number FSW–client contacts is held constant then the HIV prevalence among low-risk women is inversely associated with the relative size of the FSW population, probably because the smaller number of FSW creates denser transmission networks. Simulation models of client–FSW partnering patterns by Ghani and Aral45 demonstrated that the sharing of clients by FSW produced a substantially higher prevalence of both gonorrhoea and herpes simplex virus type 2 among FSW and clients, but a lower prevalence in the general population, compared with scenarios in which clients repeatedly visited the same FSW. Interestingly, the models also predicted that FSW–client partnership structures wherein clients were ‘clustered’ or unequally distributed across FSW would result in a somewhat attenuated prevalence of gonorrhoea and herpes simplex virus type 2 in FSW and client populations, but little change or a slight increase in the prevalence of these pathogens in the general population. It has since been shown empirically that FSW–client partnering patterns can vary substantially such that in some contexts there is a very even distribution of clients across FSW, whereas in other settings a large proportion of all client encounters occur with a small proportion of the FSW population.41 Another example of a structural pattern that is not represented by a simple derivation from individual-level characteristics is the extent to which new and young sex workers assemble in ‘cohorts’, as opposed to admixing with sex workers who are older and have been working for longer time periods. In large villages and towns of northern Karnataka, India, three sex worker zones have emerged; one for the youngest FSW, another for those of medium age and duration in sex work, and a third zone for the oldest FSW. The young FSW have a higher client volume and use condoms less often.46 By clustering together socially, the young group of FSW could establish social norms that are not conducive to condom use. Moreover, from the perspective of HIV transmission dynamics, a dense network of new FSW and clients with low condom use could result in rapid transmission facilitated by a high number of acute HIV infections within the network. Indeed, bio-behavioural surveys among FSW in Karnataka have shown that a high percentage of FSW living with HIV were probably infected within the first 2–3 years of sex work.47
Although we have proposed how female sex work can be conceptualised and examined as a complex adaptive system and provided some illustrative examples, we hasten to note that much more work is required to realise fully the potential benefits of this conceptual approach. As yet, there is a paucity of information on the basic characteristics of sex work in different contexts. Moreover, although some research has been conducted to categorise sex work by operational typology,48 more theoretically based and empirically supported approaches are required to describe the social organisation of sex work to permit comparative analyses. In particular, there has been very little research into the patterns of FSW–client partnering and sexual networks. Although there have been a number of mathematical modelling studies examining the role of FSW in the generation and maintenance of HIV epidemics in the broader population, there is a scarcity of research examining how different sexual network patterns influence the transmission dynamics of STI and HIV within sex work networks and the epidemic trajectory and amplitude in the wider population. Table 1 illustrates how an approach that focuses on structural patterns would add new parameters to examine heterogeneity more fully in transmission dynamics between contexts. For example, instead of representing the demography of FSW populations in terms of the average age, an approach that focused on structural patterns would measure and study parameters such as the synchrony of age/duration cohorts within a population to assess the extent to which the FSW population is clustered by age group. In addition to developing parameters such as the overall level of knowledge or attitudes about risk reduction based on an aggregation of individual-level data, measuring features such as social cohesion or the prevailing norms around risk reduction (eg, ‘condom-friendly environment’) might add important information because even those individuals who might be less inclined to engage in risk reduction might be influenced by their broader environment. Instead of only measuring the relative size of the FSW population, it might be important also to measure the level of clustering or dispersion of FSW, because this might relate to the density of sexual networks. Beyond measuring average client volume, modelling studies suggest that the way in which FSW–client partnerships are structured could have substantial impacts on the transmission dynamics.
These are not merely academic issues. The lack of capacity to study and explain heterogeneity in FSW-driven epidemics frustrates attempts to predict which epidemics are likely to grow most rapidly. Once interventions are initiated, there is a lack of conceptual clarity about the extent to which interventions should be targeted to FSW–client subsets. In those cases in which epidemics appear resistant to prevention programmes the current state of science often precludes the provision of a convincing explanation.
Challenges and opportunities
Although there is growing interest in moving the field of STI/HIV research towards greater integration of sociological and biomedical approaches and studying STI and HIV epidemics through the lens of complex systems, considerable challenges remain. Although in our example of FSW we have focused substantially on emergent properties in relation to sexual structures and networks, we recognise that there are many other ways in which emergent properties influence health. These might include, for example, the ways in which vulnerable populations and their risks overlap. We have not touched on the importance of social and economic hierarchy and inequality and how that relates to human health and behaviour.49 As this field of research expands, diverse aspects of emergent properties will need to be explored.
A further conceptual challenge relates to the extent to which the complex system phenomena that govern the trajectory of STI and HIV epidemics are approached as exhibiting patterns of ‘strong’ or ‘weak’ emergence. As Bedau50 explains, the concept of ‘strong’ emergence refers to a system property that is irreducible to the properties of the individual components or structural patterns of the system, and supervenes on the system through a ‘direct (‘downward’) determinative influence on the pattern of behavior’ involving the system's individual parts. In contrast, the concept of ‘weak’ emergence refers to a system with diverse structural properties or ‘macrostates’, which are ‘constituted wholly out of its microstates’.50 In our view, both conceptual approaches have potential value in the study of STI and HIV epidemics, with a systematic empirical study of different epidemics and contexts helping to understand relationships from the ‘top down’, whereas modelling approaches focus on a better understanding of how individuals and networks contribute to diverse ‘macrostates’. In this regard, we emphasise the need to integrate research systematically at various levels, including context-specific individual/behavioural research and network studies in addition to broader research programmes spanning multiple contexts to develop more coherent theoretical and empirical bases for understanding epidemic drivers and diversity.
Another main challenge is the development of a comprehensive and coherent analytical approach. Galea and colleagues21 have argued that we should move away from a traditional social epidemiological mindset and eschew multilevel regression models on the grounds that such analytical approaches cannot take into account the dynamic and reciprocal interrelationships between ‘exposures’ and ‘outcomes’ and the ways in which relationships might change over time. Instead, the authors promote the adoption of complex systems dynamic computational models to study complex systems in public health.21 This approach would entail not only a paradigm shift but also the collection of much more comprehensive and diverse data to parameterise these models. This approach might therefore be difficult to apply to a wide range of epidemic types and contexts, at least in the near term. Still, this area offers a rich field for endeavour and discovery and should be embraced and joined by public health scientists.
Diez Roux and Aiello19 have proposed an approach that we believe has considerable merit. The authors recommend bringing together the conceptual methods and tools of social epidemiology with mathematical modelling of disease transmission dynamics. They propose that simulation models of transmission dynamics could provide insights into which group-level or structural parameters are relevant for understanding, and that these parameters can be incorporated into multilevel analyses to understand better what individual and population-level factors are associated with these structural parameters. So, using FSW as an example, transmission dynamic modelling studies could help to define the most important aggregate properties and structural patterns that drive the transmission dynamics (see the lower section of figure 2), and empirical studies of the social organisation and structural characteristics of sex work in diverse settings could provide insight into the determinants of these structural patterns. Such a research process would require close collaboration between social scientists, epidemiologists, biomedical scientists and mathematical modellers to design and implement research programmes. If this is done successfully, it would offer new opportunities for understanding the complex causes of epidemics. It would also promote consilience between sociological, population and biomedical disciplines, bringing together knowledge from these fields in ways that will lead to new coherent theoretical paradigms that could open new vistas for the science of public health.
STI and HIV epidemics are shaped by a complex interplay of individual behaviours and the structural patterns of social and sexual networks.
The organisation of female sex work exemplifies the importance of considering how these structural patterns are formed, and their importance in determining transmission dynamics.
New research approaches that are drawn from scientific concepts of emergence and complex adaptive systems and bring together social and biomedical sciences are needed to obtain a better understanding of these epidemics and how to control them.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.