Birth and youth of ‘facts’: from identification of the bacteria to its population-level effects

Building our understanding of disease transmission is a gradual process. In this section, I will explore how the facts used to identify Hib, and the facts used to clarify its transmission in a population, were discovered. I call this emergence of facts birth and youth, and it represents the first phase in the life history of facts. I will provide a comprehensive picture of the early phases of Hib studies from a public health perspective. This perspective is concerned with the severe forms of disease caused by the pathogen in a population and is expressed as a big fact—a fact that comprises various aspects of a phenomenon or is built upon the ‘end-results’ of research. In a way, a life history of a fact shows how understanding a complex phenomenon develops gradually, through the unfolding of ‘facts’ that define the process. Moreover, I will show how the set of facts concerning population-level transmission of Hib diseases (F1–F3) emerge in epidemiological studies. The set of facts that are of greatest interest to this study are:

• F1: a fact that identifies Haemophilus influenzae and its strains (particularly strain b, i.e. Hib), and describes the correlation between the bacteria and the severe forms of disease it may cause.

• F2: a fact that clarifies immunity to Hib and shows its indirect, population-level effects.

• F3: a fact that explains the transmission dynamics of Hib within a population.

Understanding germs: identifying Hib

In order to understand how population-level transmission dynamics evolved over time and to conceptualize it as a life history of a fact, I will first introduce the discovery and identification of the pathogen, namely, Haemophilus influenzae type b bacteria, in the context of germ theory. The main focus of germ theory was on how transmission occurs between individuals. However, this study is primarily concerned with population-level transmission dynamics.

The formulation of germ theory created a new framework to identify factors behind infectious disease. At the core of this framework are Koch’s postulates,Footnote ⁴ which introduced the generalized principles with which to examine whether an organism can be accepted as a cause of a particular disease. This had important implications for epidemiological study. First, it paved the way for the development of vaccines and understanding immunity. Second, it made it possible to specify the concepts that describe transmission dynamics. Understanding transmission dynamics, in turn, led to studies that aimed to control and prevent disease transmission.

HibFootnote ⁵ was first identified as ‘Pfeiffer’s bacillus’ in 1892 by Robert Pfeiffer, who succeeded in isolating and identifying the bacteria, but was unable to establish the diseases it caused. In the 1930s, Margaret Pittman (1901–1995), working at the Rockefeller Institute for Medical Research, was able to define the distinct strains (a–f) that differ in the composition of their polysaccharide capsules (Pittman, Reference Pittman1931, Reference Pittman1933).

At this point, Hib had gained its identity: Once the strains of the Haemophilus influenzae were defined, the bacteria (especially the b strain) were soon identified as the causative agent of meningitis, a potentially fatal inflammation of the brain tissue. Furthermore, the bacteria were observed to circulate among young children, which increased the urge to provide protective measures against the disease. Understanding how the disease was transmitted between individuals was soon established. Upon successfully identifying the pathogen, and its mode of transmission, a mathematical theory on epidemics started to emerge. As Soper (1929: 34) emphasizes: ‘Perhaps no events of human experience interest us so continuously, from generation to generation, as those which are, or seem to be periodic.’ It is precisely this periodicity that led pioneers in the field of epidemiology to observe the patterns of transmission and formulate early models on them. For example, Kermack and McKendrick (1927) identified the basic categories of people (i.e. susceptible, infected and immune) observed in typical epidemics. Experimental epidemiology emerged as a platform for further investigation, expanding the focus of study towards understanding population dynamics of disease transmission (Amsterdamska, Reference Amsterdamska, Gaudillère and Löwy2001). For mathematical epidemiologists, the quest was to locate ‘global patterns of disease in time, space and population’ (Fine, Reference Fine1979). This search resulted in understanding the cyclicFootnote ⁶ nature of infection (Hamer, Reference Fine1906), refining the mathematical theory of epidemics that endeavoured to identify which factors govern the ‘spread of contagious epidemics’ (Kermack and McKendrick, Reference Kermack and McKendrick1927). These public health concerns, however, required further discoveries on the microbiological level to develop preventive measures, such as vaccines.

Pittman’s microbiological discoveries (Reference Pittman1931, Reference Pittman1933) paved the way for the development of vaccines. However, it was only in the 1970s that polysaccharide vaccines began to be used in population-wide vaccination campaigns, and, to improve their efficacy, conjugate vaccines came to market in the late 1980s. Population-wide vaccination programmes using conjugate vaccines against Hib began in the US in 1985, in Finland in 1986, and in the UK in 1992. Hence, Hib vaccine development proceeded in two phases. The earlier polysaccharide vaccines were capable of protecting populations against the disease, but they did not diminish the transfer of the micro-organism, whereas conjugate vaccines did both. This, of course, as we will observe when modelling Hib transmission and circulation in a population, affects the estimates of the basic reproductive rateFootnote ⁷ and optimal herd-immunity thresholdFootnote ⁸, which are important estimates for infectious disease epidemiology.

The birth of the fact—the identification of Hib and clarification of its severe disease forms—shows how knowledge produced in different fields, such as microbiology and mathematical epidemiologyFootnote ⁹, are used in other domains, such as broadening the perspective of transmission among individuals to a population level. In a way, the singular fact (F1) gradually grew into the ‘big fact’ of population-level transmission.

To prevent and to protect: public health concerns on transmission dynamics

Transmission puzzled historians of medicine, especially in the early days when the distinction between infection and heredity was not well understood (Gaudillère and Löwy, Reference Gaudillère and Löwy2001). In a simple classification of objects and patterns of transmission, horizontal is linked with infectious agents and vertical is seen as transmission of hereditary traits. But the story of infectious transmission is not that simple. Different infections are transmitted differently. Some are constantly present in a population, some occur in serious outbreaks but wane over time. The means of transmission may vary: some pathogens require physical contact and some are disseminated through the air. And it is sometimes not possible to specify a carrier, a susceptible, or an infected person, as some infectious agents do not always cause symptoms in their hosts. It seems that transmission is a stubborn piece of knowledge that requires thorough investigation in order to be incorporated into population-level transmission models.

In the case of Hib, it is transmitted from person to person via human excretions from the respiratory tract: by coughing, sneezing or kissing. However, Hib does not necessarily cause disease; the infected person may remain without symptoms,Footnote ¹⁰ capable of transmitting the bacteria but remaining healthy. This needs to be represented in dynamic Hib transmission models. Hib is therefore a good case study for explaining how the epidemiological features of a particular micro-organism may actually present challenges to mathematical representations of its spread. As described previously, the microbiological features of bacteria affect population-level transmission patterns. For example, an asymptomatic carrier is capable of spreading the pathogen, yet remains healthy. This is a challenge for effectively utilizing such knowledge.

In an effort to explain how disease transmission is related to population dynamics, I will discuss in detail the indirect population-level effects known as herd immunity. Herd immunity, according to Paul Fine (Reference Fine1993: 268), was based on the mass-action principleFootnote ¹¹ that was initially formulated on the basis of a chemical property that says: ‘the rate or velocity of a chemical reaction is a function of the product of the initial concentrations of reagents’. When applied to epidemiology, this means that changes in the number of susceptibles (infected and immunes) are expressed in successive steps. According to Fine (1993), herd immunity, as a concept, was developed in 1923 by Topley and Wilson, who studied infectious epidemics in laboratory mice and made herd immunity a topic of rigorous study (Fine, Reference Fine1993: 266).

The theory of herd immunity claims that when diseases are passed from person to person, it is more difficult to maintain the chain of infection when large numbers of the population are immune. This can be calculated as a herd immunity threshold, which is the point at which the vaccinated percentage of a population is such that it effectively stops the spread of the infection, because there are no longer sufficient numbers of susceptibles to contract the disease. To calculate this threshold, one needs to define the basic reproductive number, R₀, which is the average number of directly infected individuals. In infections that are transmitted from person to person, the potential for spread is called the reproductive rate, which depends on the risk of transmission per contact and on the frequency of contacts. The reproductive rate is determined by:

• the probability of transmission in a contact between an infected individual and a susceptible individual

• the frequency of contacts in the population

• how long an infected person is infectious

• the proportion of the population that is already immune.

As specified, all these characteristics can be expressed in mathematical equations to provide numerical estimates of the transmission dynamics in a population (Giesecke, Reference Giesecke2002). The next section explains in detail the challenge of mathematically modelling the general facts of disease transmission.

To summarize, through the birth and youth of disease transmission facts, the following details of Hib were defined: Hib bacteria is capable of causing life-threatening, invasive disease (such as meningitis, septicaemia or epiglottitis)—especially in young children. The severity of these conditions is a key factor motivating the quest for a detailed understanding of the dynamics of the infection in order to develop a means to control the spread of transmission. As briefly noted, infections caused by Hib have certain features that are challenging for the study of transmission. The natural locus of the bacteria is the human nasopharynx. This locus itself has two implications for transmission. First, it is difficult to define who is a carrier of the bacteria: there are no serological (blood) tests available, and taking a sample from the human nasopharynx is a much more complicated procedure than taking a simple blood sample. Second, transmission is easily spread through coughing, sneezing or close contact between a carrier and a susceptible. Moreover, an individual does not develop a life-long immunity to Hib. In fact, after infection there is only a short protective period before the individual is once more susceptible to re-infection. Hence, these general facts about Hib and its transmission in a population were established over many years of study on Hib diseases and vaccine development.

Thus, the birth and youth of facts shaped their identity. It seems that the facts of disease transmission largely led their lives in the mathematical expressions developed by early epidemiologists. The gradually built understanding of disease transmission encountered tension between the individual and the population, even in its infancy. Due to the complexity of transmission dynamics, the population-level effects of carriers or infected individuals can only be observed indirectly. These indirect observations are made by using transmission models that represent population-level dynamics. How these models mitigate the tension between the individual and the population is analysed in the following section.

Adulthood and (re)productive years: simulated ‘facts’ of transmission

This section explores how the set of general epidemiological facts (F1–F3), which form the big fact of population-level disease transmission, are expressed in probabilistic simulation models. In doing so, the aim of this section is to examine how facts about Hib transmission in a population are refined in models and to identify potential challenges encountered in this process. This section highlights the phase of adulthood in our life-history framework. Adulthood is often related to maturity, stability and the potential for reproduction. The potential for reproduction is elaborated at greater length when these facts are further investigated in a simulation model.

Iterating facts of transmission in probabilistic models

Facts about Hib transmission were established in four models before being translated into the simulation model, which integrated the different aspects of the individual-based disease transmission process in Hib. Let us consider these translations as a process of stabilizing the accumulated knowledge claims of transmission.

The set of facts (F1–F3) are represented in the following four models:

The main aim of the modelling exercise was not to develop further understanding of disease transmission as such, since it was well documented in the epidemiological literature. Instead, the aim was to assess the preventive and interventionist measures in the models. In other words, the aim was to understand the prevention of transmission and how it is reached by defining adequate immunity levels that result naturally (after the disease) or from vaccinations. This resulted in a ‘need to plan and evaluate different vaccination strategies’ (Auranen, Reference Auranen, Ranta, Takala and Arjas1999: 9). This need seems to be the reason why transmission is studied through investigations of immunity. Yet the main aim was accomplished via smaller steps (see Mattila, Reference Mattila2006a).

In developing the good-night kiss model, the aim was to study ‘the transmission in appropriate subpopulations as well as to assess the relative importance of subpopulation and population transmission’ (Auranen et al., Reference Auranen, Eichner, Käyhty, Takala and Arjas1996: 2235). In other words, the study examined transmission in terms of estimating transmission rates within a family and a community in relation to Hib infection.

Focusing on how population-level disease transmission evolves required that the model be built to estimate family and community transmission rates of Hib (M1), which reveals an important characteristic of maturation of facts. The relatively general facts about transmission (F1–F3) become more detailed when new knowledge about the carriage of Hib is added. The model affected the facts; it incorporated predictions on the prevalence and incidence of Hib carriage as a function of family size and age structure (Auranen, Reference Auranen, Eichner, Käyhty, Takala and Arjas1996: 2251). This means that what initially began as general features of transmission were later narrowed down and sharpened to show us that families with children of certain age-groups (those in day-care or at school) are more likely to harbour the infection, which also explains the apparently arbitrary occurrence of the infection in adult populations. The transmission happens within the family through casual physical contact, such as goodnight kisses (hence the model that bears this name).

The model to predict the duration of immunity to Hib (M2) established the decline in antibody concentration. It revealed the fact that the dynamics related to antibody concentration have important implications in predicting the consequences of different vaccination programmes (Auranen, Reference Auranen, Ranta, Takala and Arjas1999). The prevention of transmission by polysaccharide vaccines was observed by establishing the decline rate of antibodies in the model. Interestingly, the applicability of these observations extends to other bacterial infections with similar antibodies (i.e. pneumococcal and meningococcal infections). Hence, the facts of transmission became generalizable to other infections.

The dynamics of natural immunity model (M3) became an icon for later estimations of different levels of force of infection and evaluations of their impact on the duration of natural immunity. The principal idea was to explain the increase in invasive disease in unvaccinated cohorts. However, the estimations of the force of infection were useful because they helped parameterization of an agent-based model that simulated transmission. Once again, the exploration of natural immunity aimed at ‘optimizing vaccination strategies’, which was claimed to be the major aim of infectious disease epidemiology (Leino et al., Reference Leino, Auranen, Mäkelä and Takala2000: 583).

Furthermore, this general aim was translated into two specific steps (in M4): minimizing incidence of invasive Hib disease at a population level, and diminishing the number of colonizations (i.e. those environments where Hib protective antibody levels are below the estimate) by vaccination. This led to a model-based observation of transmission: there is a connection between the rate of decline of antibodies and the force of infection. This connection may have implications for population-level immunity (so called herd immunity). On the one hand, if the force of infection is low, vaccination may decrease the circulation of Hib bacteria and hence strengthen the herd immunity. On the other hand, if the force of infection is high, the vaccination may cause incidence of invasive disease in unvaccinated cohorts. Hence, transmission itself is altered, which has implications for vaccination planning (Leino et al., Reference Leino, Auranen, Mäkelä and Takala2000: 589). The model-based observation of transmission shows us how facts were iteratedFootnote ¹² from general epidemiological facts, which were established within theoretical and experimental epidemiology to model-specific facts, which highlight the specificities of transmission dynamics, carriage and immunity within a population.

Multiple interpretations of transmission

The practice of expressing the cluster of Hib transmission facts in an individual-based simulation model on transmission, immunity, and disease is a very detailed process in which the general facts are simplified through a process of integration. This process, in which facts about transmission constitute a special mode of mathematization in terms of dynamic modelling, is typically in a mode of question-oriented modelling practices (Mattila, Reference Mattila2006a, Reference Mattila2006b). This model is capable of capturing agent-based dynamics, addressing mixing patterns and including vaccination effects on the circulation and transmission of Hib. As a fine-grained, population-simulation model, it is thus capable of describing the full notion of Hib transmission, immunity and disease (Auranen et al., Reference Auranen, Eichner, Leino, Takala, Mäkelä and Takala2004). The ethnographic data on building, calibrating and testing the simulation model broadens the perspective on how facts about disease transmission are implemented within the simulation model.

The basic aim of the modelling activity was to come up with a transmission model. This task is neither obvious nor easy, even though it ultimately constitutes only one part of the more complex model. For those working on this problem, there is no guarantee that the process will succeed. As shown, population-level disease transmission is a manifold concept. For the purposes of modelling activities, it was disentangled, unpacked and, turned into smaller pieces with the help of auxiliary tools (e.g. representations of transmission patterns, estimates of natural immunity, etc.). One of these tools is called force of infection (defined as the rate at which susceptible individuals become infected, notated with ‘λ’). Auxiliary tools are essential, as they force the process of mathematization to take into account the contextual specificities, such as type of infection, population structure and age dependency. Beneath the transmission model there is a population model. But turning this into a dynamic, growing population implemented with an S–I–S mechanism (Susceptible–Infected–Susceptible) turned out to be problematic. When the modellers encountered a bug in the population simulation programme, they chose to ‘freeze’ the population (that is, to model it without growth dynamics). The S–I–S pattern was nonetheless introduced into the frozen population, but the S–I–S pattern was, at this phase, limited only to carriage (immunity was forced artificially to result from the disease). A month later, in order to specify the infection caused by Hib and infection caused by other bacteria, exposure to cross-reactive bacteria was added to the model. Exposure to cross-reactives is likely to boost an individual’s immunity. This interferes with the short period of immunity, caused by Hib (which will gradually turn back to susceptibility), as the S–I–S mechanism shows.

The interactional data show that, despite the long-term efforts to house facts about transmission within the models, it is still not quite clear to everyone what transmission actually is. The following extract, derived from my ethnographic observations, underlines this problem:

Thus, it seems that facts about transmission need to be actively accommodated in the model. Be that as it may, as facts they also reveal a surprising degree of sensitivity: just a couple of months later, in early 2003, the testing and calibration activities were resumed. The transmission model was now subjected to preliminary evaluation by comparing the model-based results with the existing (real-world) knowledge on Hib transmission as documented in the literature. And these comparisons again revealed new sides to the facts: they were sensitive to the population in which they occurred. So even though the discussion was limited to Hib transmission, factors such as different vaccination schedules, different age structures in day-care and school groups, and different levels of CR, varied depending on the country or region in which the study was conducted.

Therefore, it seems that the facts of transmission do not simply reproduce themselves. To do so, it would be necessary to rigorously mimic the trends of the Finnish population, the vaccination plan and the CR exposure on which the model was based. This was discussed in terms of how well the model explains the known specificities of transmission in different countries (in Gambia and in the UK).

As expressed in this discourse, determining transmission patterns in a specific population requires modelling practices that elaborate transmission dynamics in detail. However, if data are scarce or absent, or if behavioural and mixing patterns are not clearly documented, mathematization may be difficult, and transmission dynamics can only be explored at a more general level. This, in the end, may diminish the usefulness of the modelled evidence (or leave unintended uncertainty in the models). Some levels of maturity were reached in the models, even though they were tailored to represent only the Finnish population. The challenge of acknowledging a shared interpretation of the transmission facts reminds us that tensions within scientific work may not always originate from social dynamics or division of labour (Mattila, Reference Mattila2006c). Often discrepancies have their roots in epistemological ground, as this section showed.

Senior facts: the dual nature of disease transmission

The analogy with old age captures two aspects of the life of facts: either the facts reach a respected position or they are retired. In common terms, we might think of the elderly reverentially, as repositories of accumulated wisdom. In this section, the senior facts will be discussed by reflecting on the dual nature of disease transmission.

The dual nature of transmission addresses the phenomenon as a threat and a promise of prevention. It is a threat that must be well understood in order to carry out vaccine development, strategic planning and protection. Yet it is also a promise that public health can protect using preventive interventions. This duality may remain invisible to the public. Yet, for public health authorities, facts of transmission are senior facts, respected for their information content, and their capability of being a reference point in decision-making. Let us return to the case of Hib and the transmission facts (F1–F3) established by the four models. As discussed previously, the models are primarily concerned with the prevention of transmission and, therefore, the general facts of transmission of and immunity to Hib. These models emphasize the dual nature of disease transmission. Even though public health measures aim primarily to prevent transmission, transmission is also the force that induces immunity in the population. In order to optimize preventive measures (vaccinations), one needs to take into consideration possible decreases in natural immunity that result from the circulation of the bacteria.

Thus, these models elaborate, represent and reproduce facts about immunity—an immunity that protects against and prevents transmission. Each of these models tells a story of its own, but together they contribute to the development of realistic estimates to prevent transmission. In a way, they reveal the dual nature of transmission that must be fully understood. Transmission does not only pose a threat of severe disease, it also has a protective purpose, and interfering with this cycle (by vaccinations) may alter the delicate balance and cause new cases of disease in older age groups. Hence, this aspect of transmission clearly underlines the challenges of drawing firm conclusions about disease transmission and representing them in models. It also reminds us of the public health perspective, discussed earlier: it is not enough to explain the singular causes of disease, as it is necessary to expand the explanation to cover the causes and effects of disease transmission in a population.

Another way of looking at the obsolescence of facts is by reflecting on the unequal distribution of Hib vaccinations in national vaccination programmes. Thus, even after Western societies reduced the problem of disease transmission, it still continues to be a threat in many parts of the world. Simulated facts encourage us to keep the problem in mind, and urge us to develop tools for public health advisory work on reshaping vaccination policies. For example, the implementation of Hib vaccination in England and Wales can be considered a public health success story. The declining trend in Hib cases is indicative of the efficacy of the vaccine. Why call it a success story? To recap, Hib is the leading cause of bacterial meningitis, and other invasive diseases, such as pneumonia and epiglottitis. Worldwide, there are still approximately 3 million cases of serious disease each year, with approximately 386,000 Hib-related deaths in infants aged 4–18, and bacterial meningitis represents 52 percent of these infections. Societies continuously struggling with Hib diseases are in the developing world (WHO, 2006 445–452). Simulation models can be used as administrative tools to facilitate the alleviation of the disease burden in these countries.

The main use of simulation models is to help optimize vaccination strategies in the developing world. This was one of the aims of the Helsinki modelling project: to create a general tool for a variety of purposes in order to overcome the problem of unavailable surveillance or serological data. But the challenge remains to persuade developing countries to include Hib vaccinations in their strategies. In pursuit of these aims, simulation models may also function as flexible tools to carry the senior facts of transmission to other domains. In other words, as expressed in Leino (Reference Leino2003: 10):

This is clearly a strong reminder that a fact (about Hib transmission) is capable of adopting an expert position. It also explains the rationale behind using models to refine the facts, and to overcome possible shortcomings in available data. The modelled facts regarding decreases in transmission, the vaccination coverage needed for an optimal herd immunity level, and the falling number of severe cases should convince decision-makers of the necessity of vaccinations. It is through the seniority of these facts and their generalizability, despite the fine details of their modelled context, that they are imbued with sufficient persuasive power to fight against superannuation. Nonetheless, striking a balance remains an open-ended challenge. In summary: the characterization of the seniority of facts allows old facts to hold respected positions. With regard to the question of disease transmission, public health authorities seek advice and evidence for decision-making processes from senior facts, from knowledge claims that contain information applicable to different contexts.

Discussion: understanding disease transmission through the life history of facts

As a conceptual framework, what does life history highlight? Briefly, in each phase, knowledge claims—facts—are refined, sharpened, circulated and represented in a particular way, allowing them to carry on with their lives. Mathematical representations and models function as a key to these heterogeneous practices. From early observations of transmission in populations to fine-grained simulation models, the facts of transmission dynamics are expressed in mathematical terms. Life history, as an analytical framework, not only shapes the linear timeline of the developmental process, it also provides insights into how facts live their lives. Birth and youth emphasizes the gradual nature of the process, within which generally accepted epidemiological facts emerge and become accepted within research communities. Maturation towards seniority is expressed in terms of generalizability, which refers to the potential to broaden the epistemic scope of available knowledge. Life history, furthermore, invites us to explore the vita activa of facts in a more metaphorical sense: it shows the flux embedded in the continuum from knowledge production to dissemination. It seems that the idea of knowledge production, which is well documented and discussed, has led us to believe that once knowledge or facts are produced they are simply ‘out there’, ready to be used (e.g. Knorr Cetina, Reference Knorr Cetina1981; Latour and Woolgar, Reference Latour and Woolgar1986 [1979]). Yet, vita activa moves beyond these assumptions and reveals the continuous effort required to understand multifaceted disease transmission phenomena by expressing them in mathematical representations.

Life history, as a conceptual framework, is developed through the utilization of multiple sources of data. Thus, it facilitates the analysis of micro-level interactions between modelling practices and the more general public health studies of disease transmission. Hence, the parallel development of early theories on transmission dynamics, the difficulties encountered in first identifying the specific bacterial agent that causes Haemophilus influenzae, and growing awareness of population-level dynamics tell a different story. Through the life history of facts, some moments in the developmental process of epidemiological knowledge were traced and the interdependency between epidemiological knowledge and mathematical representations was established.