Ch 3 - Causes & determinants in PH Practice.docx

Chapter 3: Causes & Determinants in Public Health Practice This chapter may be conceptually more demanding than some of the others and it is certainly longer. At times, the concepts may seem rather abstract but, writing as someone whose career has primarily in public health practice, I am convinced...

Chapter 3: Causes & Determinants in Public Health Practice This chapter may be conceptually more demanding than some of the others and it is certainly longer. At times, the concepts may seem rather abstract but, writing as someone whose career has primarily in public health practice, I am convinced of the utility of these various ways of understanding *cause* for public health practice. **Outline** - Why does it matter? - Modelling causality - Causes of individual cases, causes of population incidence - Single causes & multiple interacting causes - Systems concepts - Determinants of Health - Conclusion Why does it matter? =================== If what we want is effective action, of course we can take a stab in the dark and we might just get lucky; our blind effort could be followed by the improvements we were hoping for. But---as a general rule---the likelihood that our work will accomplish its intended purpose is enhanced if we have a sound notion about what's going on out there in that system we're trying to influence. So, we want to get a handle on what's actually happening. Another way to express this is: we want to understand what's causing things to be as they are... and what might cause things to bend in the direction of some desired change. We need to be able to come up with a well-informed *theory of the problem*, on the basis of which we can identify candidate *critical leverage points*, where strategically applied force may move the system (see Archimedes' lever, below). And, in determining what that force (or intervention) should consist of, we formulate a *theory of the solution*. (Some folks refer just to "theory of change" to cover what I'm calling theories of the problem and of the solution, but I think it's helpful to separate them out.) Given our understanding of the issue and its relevant context (i.e., the system), how do we expect our intervention to work, causally, to bring about the desired result? How do we expect the pillow to get to where it needs to be, when it needs to be there (in the Rube Goldberg cartoon above)? Our theories both of the problem and of the solution consist of a set of factors that we propose come together and bring about some result; that is to say, a *set of causes*. To increase the likelihood of a successful effort, we need to approach our work critically, thinking through what we think is happening in the system in which we are intervening (our theory of the problem) and how we think our intervention will influence that system (our theory of the solution). Other near synonyms you'll encounter include: *program theory, logic model, program impact pathway,* and---as we have noted*---theory of change*. We seek to understand causes to allow us to intervene effectively, to bring about improvement. In public health, another of the terms of art we've used to talk about causes, particularly since the 1980s, is "determinants". This is particularly in discussions on *population* health, i.e., patterns of health and disease in communities or populations. The word is most often used in public health discourse for socio-political factors affecting the health of populations or segments of populations although, in principle, there's no good reason not to use the word "determinants" also to cover physical or biological factors. By using terms like *theory*, *hypothesis*, *model*, or *construct*, I'm implying that we don't have to have this fully worked out; we don't need perfect knowledge. We can try something out, see if it works, and then---as necessary---we can adjust our theory of the problem and reformulate strategy. One final introductory task would be to link our discussion with *epistemology*. You'll recall that epistemology concerns the question: what does it mean to "know" something? When we're trying to figure out what to do to bring about some desired change---whether in public health or any other field---we operate on the assumption that we have a reasonable handle on what causes what. We commit to a theory or an explanation of how things work and, on that basis, we then formulate plans and put them into motion. Given that many of the problems we're dealing with in public health are inherently complex, rather than merely complicated, we should not be surprised to see things work out differently than how we predicted, requiring us to go back and revise our causal models. Discerning patterns & making predictions ---------------------------------------- [David Hume](https://socialsciences.mcmaster.ca/~econ/ugcm/3ll3/hume/enquiry.pdf) proposed, almost 300 years ago that, "we may define a cause to be an object followed by another, and where all the objects, similar to the first, are followed by objects similar to the second. Or, in other words, where, if the first object had not been, the second never had existed." The wording here is a bit opaque and archaic but the key principles Hume proposes are: 1. *Regularity*: X, routinely, is followed by Y, e.g., if the cue ball hits a stationary ball, the second (formerly stationary) ball will suddenly begin to move, and 2. The *counter-factual*: if X doesn't happen, neither does Y. As human beings, we are hard-wired to look for or to impose patterns; we need narrative; we need to connect the dots, to make sense of things. So---all the time---whether we think about it that way or not, we are constructing causal models. In public health, we survey the field or system where we're working, we seek to discern meaningful patterns; and, as we search for regularities, we try to work out rules for predicting what will happen next. All of these mental functions involved in trying to understand what's happening concern speculating about and coming to at least tentative conclusions about *cause*. With our concern about health in populations, we observe and try to make sense of variations we see *by person, place, and time*. Why is it that this person is sick and that person is not? What variations do we see, comparing one place vs. another, and what do those differences mean? How are patterns changing over time and what does that tell us? What do these changes or variations tell us about what may be going on causally? As we've noted, we don't necessarily need to have perfect knowledge about what's going on. We need just good enough knowledge, i.e., causal explanations that have some predictive utility or power. Back in 1854, the English physician John Snow was confronted with a bad outbreak of cholera in London, in a part of town just south of where the London School of Hygiene and Tropical Medicine now is. This was before Lister and Pasteur formulated germ theory. But Snow studied what was going on, trying to make sense of it in terms of *person, place,* and *time* (see panel below). He may not have described what he was doing as epidemiology but, in fact he was engaged in the study of the distribution of disease (and potential drivers of disease) in a population. He looked at who was getting sick, who wasn't, and how they differed. In doing so, he was able to make the connection (i.e., recognize an association) between place of residence of those getting sick and where they were getting their water. He didn't know anything about the *Vibrio cholera* bacterium but he was able to come up with a good enough causal explanation: occurrence of cholera had something to do with the water supply. Armed with a plausible causal account, he was able to persuade the local council to quit drawing water out of the Thames downstream from where sewage was flowing in and to disable the Broad Street pump (from which many of those who'd fallen ill had been drawing their water)... after which the epidemic abated. **On person, place and time...** Epidemiology can be understood as the study of the variations in occurrence of disease / injury within and between populations, in terms of *person, place* and *time*. Although we might dismiss such study as merely descriptive, the *patterns* discernable from such analysis can give us important hints on what may be going on causally or mechanistically that could account for the observed differences in disease occurrence across these dimensions of personal, spatial and temporal. **Person** -- who is getting sick? - Age - Sex / gender -- biological and non-biological factors - Race / ethnicity -- mainly non-biological factors, with some important exceptions, e.g., sickle-cell disease, melanoma, hemoglobinopathies - Socio-economic status - Nutritional status **Place** -- where are instances of disease / injury occurring? Is there spatial clustering? If so, what does that suggest about what may be causing it? **Time** -- how does the number of cases differ from what would have otherwise been expected over this period of time? What else is going on temporally that may align with emergence of this problem? In public health practice, we are constantly engaged in tackling real-world problems and trying to get to better outcomes. How effective our efforts turn out to be depends, in large part, on understanding what's going on, and that means getting at least an approximately valid sense of what's causing what. We care about causality in public health... because we care about improving things out in the complicated---or should I say "complex"---real world. **To summarize**: public health action can be understood as hypothesis-testing: we make a tentative diagnosis, put in place a treatment or intervention that seems to be a reasonable fit with our diagnosis, in our particular setting---predicting / placing a bet on effectiveness---and then see whether or not it produces the desired change. To do so, we need adequately well-supported ideas about: 1. what's driving or causing the problem we're trying to do something about (what's happening in this system)---I describe this as our ***theory of the problem***; and about 2. how, causally, we expect our intervention / treatment to bring about its desired effect (how we expect our intervention to change the behavior of the system)---what I call our ***theory of the solution***. Paul Batalden---building on insights from quality improvement pioneer and systems thinker W Edward Deming---has noted that "every system is perfectly designed to get the results it gets"... which leads us to the question: what intervention or set of inputs will alter the function of the system to give rise to the desired change in output or improvement in performance of the system? Modelling causality =================== *"All models are wrong; some are useful"* (George EP Box, British statistician) Our work requires that we posit causal relationships. As we've noted in the case of the cholera example above, we may have a very incomplete, imperfect understanding which can---nevertheless---inform effective action. Algebraic modeling ------------------ Mathematical modeling can sometimes help us move towards better understanding and more effective action. In such modeling we have to make choices or trade-offs, balancing comprehensiveness / predictive power vs., explanatory clarity. One application of such modeling in public health is for understanding the spread of infectious diseases. [Anderson and May (1979)](https://www.ncbi.nlm.nih.gov/pubmed/460412) offer a simple formula that can be used to predict the behavior of an infective agent in a population. R~0~, (R-naught) in this formula, is the "basic reproduction number", the mean number of new cases arising from each infection. If R~0~ exceeds 1, the infection will continue spreading; its incidence in a population will increase. If it's less than 1, and remains so, over time it will fail to propagate and will fizzle out. In the formula below, R~0~ is driven by just three variables: R~0~ = βcD, where β is the likelihood of transmission per contact with someone susceptible c is the number of such contacts per unit time (e.g. per day) D is the average time spent infectious (say, in days) But, in the real world, we are almost always dealing with more complex situations involving a larger number of variables. Mathematical modeling can still be helpful but, with more comprehensive mathematical models that better reflect this complexity, there's a price to pay with regard to explanatory clarity for non-specialist audiences. Early in the COVID pandemic, some of you may have followed the work of groups such as those at Imperial College, London, and the University of Washington / IHME that were using more elaborate compartment versions of this, e.g., the 4-compartment Susceptible--Exposed--Infected--Recovered ([SEIR model](https://www.idmod.org/docs/emod/hiv/model-seir.html)), to project possible future trajectories for the COVID-19 epidemic, based on different sets of assumptions. It was based on projections from this type of calculation that it was determined early in the pandemic that, in order to avoid overwhelming the hospitals, measures needed to be taken to reduce likelihood of transmission per contact, β, (e.g., by wearing masks) and reduce the number of such contacts per unit time, c, (e.g., by working remotely), ultimately with the expectation that this would "flatten the curve" of new cases and new hospital admissions. Another example: how to reduce incidence of syphilis in a given population? From this formula, three different types of strategy can potentially reduce R~0~ to below 1 (i.e., causing incidence to decline): - reducing the likelihood of transmission per contact with a potentially susceptible person---by increasing condom use in your target population - reducing the number of (new) contacts per unit time---by addressing social determinants contributing to pattern of multiple sexual partners - reducing the average time infectious---by doing effective outreach encouraging seeking healthcare for symptoms suspicious of a sexually transmitted infection and getting appropriate treatment on a timely basis. Boxes and arrows ---------------- In our program work in public health, it's not often that we model causality algebraically. But often we do so *diagrammatically*, to capture the program logic of our interventions or efforts. This, typically, is how we represent our causal models, "results frameworks", "theories of change", or "program theory", what I have referred to as our *theory of the solution*. Each arrow in such a framework or model implies a direct causal relationship. As a management tool, such frameworks can be a helpful way to organize our work teams and reporting. But, when used uncritically---grossly oversimplifying what may happening causally---such frameworks encourage lazy thinking and let us off the hook from thinking critically and questioning our assumptions. In this example, what is the basis for believing that simply by training health workers we will necessarily see an improvement in service quality? There's no avoiding overly simplistic results frameworks like this; they're widely used in program work in public health and in other areas of work. But you need to view them with some skepticism. I think of the appropriate stance as *sitting lightly on our boxes and arrows*. I have found that a helpful way to think about evidence and causality that respects the tentativeness that's always inherent to our "knowledge" of the systems we're engaged with is in terms of *"double-loop learning"*. This idea comes from Chris Argyris and Donald Schön. In our ongoing daily work, we are always monitoring the immediate result of what we're doing, making adjustments as necessary to move things along. This is what Argyris and Schön refer to as single-loop learning. But it is also necessary to find ways of periodically stepping back, seriously reviewing what's happening and then---as necessary---revising or reformulating our original assessment of the situation, i.e., our theory of the problem and then, as necessary, making changes in strategy (double-loop learning). This can also be understood as reflective practice. For more on these concepts, I'd strongly recommend viewing this 3-minute [video](https://www.youtube.com/watch?v=kBeFOXhWxyo). The military has institutionalized this way of thinking in its adoption of [after-action reviews](https://www.mindtools.com/pages/article/newPPM_73.htm). Causes of individual cases, causes of population incidence ========================================================== In public health, we are interested in both what causes *individual cases* and what accounts for patterns across *populations,* including changes over time. Sometimes, insights arising from between-population differences (from "[ecological studies](https://pubmed.ncbi.nlm.nih.gov/7639884/#:~:text=An%20ecologic%20study%20focuses%20on,environmental%20measures%2C%20or%20global%20measures.)") can help inform our understanding of what's happening causally at the level of individuals. You might think that what causes a lot of cases is the same as what causes one case, just a lot more so. But that's not necessarily so. Geoffrey Rose is another of those British physicians who've been a parent of epidemiology and public health. In a [classic paper](https://apps.who.int/iris/bitstream/handle/10665/70950/bu1409.pdf?sequence=1&isAllowed=y), first published in the International Journal of Epidemiology in 1985, he's given an explanation of the different forces that account for: - variability *within* a population with regard to health and disease (i.e., who, within a population, gets sick and who doesn't) vs. - variability *between* populations (what accounts for differences between populations in the burden of specific conditions). In this paper, he uses the example of hypertension among London civil servants and Kenyan nomads, making the point that the questions: - Why does this individual have hypertension? (causes of cases), and; - Why do some populations have a lot of hypertension while others have little? (causes of rates)\... are not different ways of asking the same thing. In this example, it may be that *genetic variability* plays an important role in who---*within* a given population---is affected by hypertension whereas *nutrition* (notably sodium intake) may be the most important driver of the observed differences *between* these two populations. An analogy from ecology may be helpful here: a botanist may have a very sophisticated grasp of the processes at work within a single tree. But if her knowledge is limited to these processes, she may be completely in the dark with regard to other causal processes going on that involve aggregates of many trees, together with all the other organisms they're linked with. Single causes and multiple interacting causes ============================================= Sometimes "cause" is pretty straightforward and we are able to identify a single *exposure* or precipitating cause and a single *outcome* that appears clearly linked to the cause. I touch the pot on the stove---not having realized it was hot---and then pull my red and painful fingers quickly away. But in many instances, whether in public health or in daily life, we are dealing with a more complex picture. In this section, we will build up progressively from simpler, more straightforward situations that may involve a single cause towards more complex causal patterns. Pragmatic judgement ------------------- Austin Bradford-Hill was another of those mid-20^th^ century British pioneers in public health and epidemiology. One of his contributions (along with Sir Richard Doll) was a series of studies establishing the relationship between smoking and lung cancer. For the issue of interest in this chapter (causes), an important related contribution was a set of pragmatic principles to consider when trying to establish whether or not there's a direct causal relationship between a particular exposure and an outcome. They are: +-----------------------------------+-----------------------------------+ | - strength (effect size, i.e. | - dose-response gradient, | | how strong is the statistical | | | association), | - biologic plausibility, | | | | | - consistency | - coherence, | | (reproducibility), | | | | - experimental evidence, and | | - specificity, | | | | - analogy. | | - temporality, | | +-----------------------------------+-----------------------------------+ Read more [here](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1898525/pdf/procrsmed00196-0010.pdf). Bradford-Hill's pragmatic rules of thumb bear some similarity to those of Robert Koch---a German pioneer in bacteriology, active in the late 19^th^ century---who proposed the following "postulates," as criteria for making a judgement on whether a particular organism is the cause of a particular infectious disease: - The bacteria must be present in every case of the disease. - The bacteria must be isolated from the host with the disease and grown in pure culture. - The specific disease must be reproduced when a pure culture of the bacteria is inoculated into a healthy susceptible host. - The bacteria must be recoverable from the experimentally infected host. Risk factors ------------ Several times in this chapter, you've encountered the terms: *exposure* and *outcome*. One example I gave was having my hand come into contact with a hot pot: that was the exposure; the burnt fingers were the outcome. In the simplest case, all we have is: E O. I should note that "exposure," as used here, is a very generic term; it can be equated with the term "independent variable", used in the social sciences (with "outcome" equating with "dependent variable"). In public health and medicine, the term "exposure" can be used both for a possible cause (or "etiology") of a disease as well as for a treatment or intervention. Referring back to Bradford-Hill's criteria, seeing a regularity (A is generally followed by B) is helpful in trying to determine whether there may be a simple, direct E O causal relationship. But in many instances, a lot of what's going on causally may still fall in a black box, obscured from view. It's one thing noting that there appears to be a strong statistical association between smoking and lung cancer (criterion 1: strength). But what's the *causal mechanism*? How could smoking bring about this result? In the mid-20^th^ century, as efforts were being made to better understand what was driving incidence or risk of major non-communicable diseases, a series of important (mainly retrospective, case-control) studies were done, notably those of [Doll and Bradford-Hill](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1814562/pdf/brmedj02630-0017.pdf), documenting statistical evidence of a link with certain exposures. But this was incomplete evidence. Consider another of Bradford-Hill's criteria: consistency (or reproducibility). If smoking causes lung cancer, why is it that most smokers never develop it? And why is it that there are people who develop lung cancer without ever having smoked? So, it's evident that we have a more complex causal picture here than a hot pot and burnt fingers. The notion of "[risk factor](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5686931/?report=printable)" (coined by WB Kannel, principal investigator of the seminal Framingham cohort study), refers to an exposure or characteristic that is associated with the outcome of interest, where there appears to be some kind of causal relationship.[^1^](#fn1){#fnref1.footnote-ref} Use of this term can be seen as an admission that we haven't got things causal entirely worked out; we don't necessarily know the causal mechanism(s). We see some kind of a regularity and we have strong suspicions that the exposure plays an etiologic role. But we haven't yet got it all figured out. As recently as the early 1980s, physicians and epidemiologists would talk about risk factors for peptic ulcer disease and stomach cancer---referring for example to smoking and alcohol use, regularly using aspirin or other non-steroidal anti-inflammatory drugs, consuming spicy food, and experiencing emotional stress---because epidemiologic studies had shown statistical associations between these exposures and ulcers and stomach cancer. But from the 1980s, when I was in medical school, evidence began to emerge that the most important driver of these disease conditions was actually chronic infection of the lining of the stomach and duodenum by *Helicobacter pylori* bacteria. In North America, it used to be that much of the population had low-grade chronic helicobacter infections and this is still the case, today, in some parts of the world, lacking a safe water supply. So, in a population where almost everyone harbors this infection---mostly simmering on with minimal or no symptoms---variability on other factors like smoking and alcohol use may track with prevalence of ulcer *symptoms*. Now, with a more complete understanding of the etiology of peptic ulcer disease and stomach cancer, we still acknowledge that these other practices or conditions can contribute to symptomatic gastritis, but we can now explain causation of these conditions drawing on more than just "risk factors", but also drawing on knowledge of the *pathological mechanisms* involved. Isolating single causal factors ------------------------------- Observing a statistical association between some exposure and an outcome can certainly warrant speculation that there may be some kind of direct causal relationship. But, by itself---as we've seen---it is not sufficient evidence. There is a rich and developing literature in epidemiology, philosophy of science, and other fields on the relationship between association and causation, which is worth your digging into. A very important concept in epidemiology that relates to circumstances in which association most certainly does not equal (direct) causation is *confounding*[^2^](#fn2){#fnref2.footnote-ref}. It may be that an observed statistical association is explained by some third variable that is independently associated with both the exposure and outcome of interest, and it is the third variable which actually has the direct causal relationship with the outcome. The third variable may be a cause of the exposure or may be associated with exposure but without a direct causal relationship, such that the exposure is a mere marker or proxy for the third variable. An example: - exposure---place of residence, Alberta vs., Northwest Territories - outcome---mortality rate - third variable---age structure Ignoring the third variable, Alberta appears to be an unhealthier place than NWT. But this observed association is accounted for by age-structure which, in this relationship between variables, is a confounder. Note that with confounding, it's not all or none; often we have a situation where our exposure of interest does have some direct causal influence on our outcome of interest but the observed statistical association may overstate that causal relationship, due to confounding by a 3^rd^ variable. In a different comparison between populations where one has a much higher mortality rate than the other, after statistically adjusting for age structure, the difference may be reduced but still present. Another possibility is that the 3^rd^ variable may lie *along the causal path* *between exposure and outcome.* Such variables are called *mediators*. An example: - exposure---obesity - outcome---cardiovascular disease - third variable---hypertension In this case, at least part of the way that obesity contributes to risk of cardiovascular disease is through its effect on the likelihood of hypertension. An important question to consider in such situations is whether *all* the observed association between exposure and outcome is mediated through the third variable or some of it results from a more direct E O causal effect (or through other mediators, e.g., in this case, diabetes). When we are trying to establish whether or not there's a direct causal relationship between two variables, we want to be able to rule out confounding (and other biases) as possible explanations for an observed association; we want to isolate specific causes from other factors that could be acting in the picture. The logic of an experimental study (a randomized controlled trial), is that we hold all else equal and vary just one factor, checking to see if this results in a different outcome. To the extent that we've been able to hold everything else constant, we can attribute any observed difference in outcome to that factor. That gives us confidence to assert a direct causal relationship. It doesn't mean that's the only factor that can influence the outcome. But such a study design does allow us to isolate a single factor and look at its independent causal influence. There is no question that well-designed and conducted randomized controlled trials can give us firmer evidence, than other study designs, for the causal force of a specific exposure on an outcome. But in practical public health programs, there are many instances where plausible evidence (short of the more definitive evidence from a randomized controlled trial) is good enough, as discussed in [this useful paper](https://watermark.silverchair.com/280010.pdf?token=AQECAHi208BE49Ooan9kkhW_Ercy7Dm3ZL_9Cf3qfKAc485ysgAAAlQwggJQBgkqhkiG9w0BBwagggJBMIICPQIBADCCAjYGCSqGSIb3DQEHATAeBglghkgBZQMEAS4wEQQMhsFxScggDep3aCf1AgEQgIICB9apWXgyknFW7q3oqkr0GfjD8-NHrASFWJnTPuxWIgGmGqE0vd1hnkU4maVKGTl4L_H8AOKHAQb4Om1NIdH5zm-0TsE8GouGwmfAvpxj0lhVqoenuYwe5wBT0hyYmBzs3l4NFnW6Lm7x_YXnrYBEyXjAOhvDHgZyPwQeCsfyCr3Kof2LSX-rnPnDui7GTbyL9Sh7RwhY842Aj6NbiG8_Z9i2v4Re3oWcQe09XBCu2rgO1hfuu9jQtQWN2MNiZJulwnzIoPS9LexWAEH-RJcdgD_0iBPnZD5QBSBTBFDOR34UIOeB2VyKvhFyDRVn_YWkmNyA7F5RBTE2HtY-5FRatxDH-61tPEcJAoApA0ZrpfKMtBTPPd9RFMh1r5sEyPFHGA-NVVzc_KhBOHpJqPTKm_YbPXpR3CBWENL_fqPwjMWkUAvert7AXAqcNdIcmCDyKA6eBIu9qKv2exY1HiqrWbrWAJNRH0FeNnfbHnxwYXz_z5_NQ28P_yGFgOXa8MdiZRgPiVX0mNenxLAtkTlGDp5NFT-XimPG1YgIiP6DOFFj7YZkeBx8rX62MLZq_lvZyKdu3QIq9QNr1RlCyIYiV3yLenvLlFZsFm8JcqPCPc9OyLa8gBALqto0pX6hrJ_s9KpDmwLh49MzclOq4226G6503bQnMniYg8J-GVdk7G7mXbswcAO-nQ). Now, it may be that the causal relationship we observe between an exposure and an outcome holds for people with some particular characteristic but not for those without. Patients with a particular genetic marker may benefit from a certain medicine used to treat a condition, but those not having the marker do not see a similar benefit. We have two more or less interchangeable terms to refer to these other factors that may influence a causal relationship: *effect modification* (the third variable modifies the effect of the exposure on the outcome) or *interaction* (there is an interaction between the exposure and the third variable, which affects the outcome). When we're trying to tease out the possible effects of a variable that may modify a relationship between an exposure and an outcome, one way to do so is by *stratifying* our analysis, so we look independently at the effect---disaggregating by the modifying variable. An example: - exposure---taking an iron supplement - outcome---hemoglobin level - third variable---baseline iron status In this case, taking an iron supplement (the exposure) will result in higher hemoglobin (the outcome) for those suffering iron-deficiency anemia at baseline but will not have this effect for those whose baseline iron status (the effect modifier) was normal. Nature & nurture (including fetal life) --------------------------------------- One of the ways we talk about cause in daily life---when we're thinking about why people are as they are, or why certain people suffer from a health condition and others don't---is to frame the question as "nature vs., nurture", meaning: was that person just born that way (all in the genes) or did that characteristic develop as a result of something that person was exposed to in life? A fancier way of saying that would be: to what extent is *phenotype* driven or determined by *genotype*? Of course, for most characteristics (including health / disease conditions), it's not one or the other: they're the result of an interaction between the two. There are various types of research that have helped disentangle these relationships. For decades, there have been studies done comparing twins (including identical twins separated at birth or comparing identical and fraternal twins), to try to assess the relative contribution of genetics vs., environment. There have also been studies of migrants, demonstrating---for example---that for conditions like cancer and heart disease, first-generation migrants have patterns of disease intermediate between country of origin and the country into which they've migrated whereas their children, second-generation immigrants, have patterns of health / disease that more closely resemble those of their non-migrant peers. To further complicate the nature / nurture picture, exposures experienced by the fetus during pregnancy may have important lifelong consequences. Those of you with health sciences backgrounds may have heard or read about the TORCH infections (Toxoplasmosis, Other (syphilis, varicella-zoster, parvovirus B19), Rubella, Cytomegalovirus, and Herpes) which, if contracted during pregnancy, can result in congenital anomalies. Nutritional exposures during pregnancy can be similarly important; over the 20^th^ century, one of the most important public health success stories, globally, has been the near-elimination of cretinism---serious cognitive handicap associated with inadequate iodine during fetal life---achieved through universal salt iodization. It is not only deficiencies in specific micronutrients that can have long-term consequences but also more global undernutrition in pregnancy; in low and middle-income country settings where childhood stunting remains common, approximately ¼ of stunting is already locked in by birth, due to fetal growth restriction. Such fetal growth restriction has been found to result in "metabolic programming" (per the "Barker hypothesis"), with higher risk through adult life of obesity, insulin resistance, hypertension, dyslipidemia, and associated atherosclerotic disease. For more on this, read [this article](https://ogscience.org/Synapse/Data/PDFData/3021OGS/ogs-60-506.pdf). In addition to effects later in the lives of these individuals, girls whose growth is stunted will be of shorter stature as adults. This, in turn, increases the risk, in their offspring, of fetal growth restriction, resulting in generation-to-generation transmission independent of genetics. Another important category of exposures during fetal life that may have important long term consequences for growth and brain development is toxic, chemical exposures, notably: nicotine and other compounds associated with maternal smoking, alcohol, heavy metals (mercury, lead); and a range of air pollutants, including particulate matter, NO~2~, and polycyclic aromatic hydrocarbons. "Nurture," in this context, can also be framed as *environment*. And, whether we're considering the factors influencing how we develop in life or---more specifically---disease and health conditions, it is important to recognize a range of dimensions of the environment that includes not just the physical / biological environment, but also the social, cultural, institutional, political and economic environment (as we will be exploring in subsequent chapters). Nutrition (we are what we eat) ------------------------------ One important influence on health and disease falling into the "nurture" or environment category that we've already touched on (which helps to account for the findings of migrant studies) is *what we eat*. In the child survival literature from low- and low-middle-income countries, there is a well-established nexus of nutrition and infectious disease (see [Scrimshaw](https://apps.who.int/iris/handle/10665/41782) 1968): malnutrition predisposes to infectious illness, which---in turn---undermines nutrition status, rendering the child even more vulnerable to severe infections. This is an example of a vicious circle feedback loop (see further discussion later in this chapter on causal loop diagrams). A significant proportion of the major non-communicable diseases affecting those living in high-income countries (and an increasing proportion of those in LMICs) can be attributed in part to *over*-consumption (not just of calories but also of specific components of the diet, notably sugar, saturated fat, salt, and red and processed meats). In contrast, even now, the health of populations of LMICs is very much affected by *under*-consumption. Of course, the relationship between nutrition and health is complex. There are relatively few instances of single nutrition-related causal factors and associated disease outcomes, the exceptions being certain conditions associated with deficiencies of specific micronutrients, e.g., scurvy and vitamin C. What we eat is, itself, a function of a complex set of factors, including culture, how we organize our time, and the economics (and politics) of the food industry. Component causes ---------------- As we've noted earlier, though there are simple cases where there's just one cause involved, more often we are dealing with *sets of causes*. The notions of *necessary* and *sufficient* conditions are helpful here. A *necessary* condition is one that is absolutely required to produce the outcome of interest. But, on its own, such a condition may not be *sufficient*. Ken Rothman (a late 20^th^ century epidemiology pioneer) has provided a useful discussion on causality, found both in his textbook, *Modern Epidemiology*, and in this [paper](https://blogs.kent.ac.uk/jonw/files/2015/03/Rothman-Greenland-05-Causation-and-Causal-Inference-in-Epidemiology.pdf). One of his key points, he illustrates as [causal pies](http://www.angelfire.com/bug/berberian12/article/rothman.htm), where each such pie represents a set of conditions sufficient to produce an outcome. The slices represent specific causal components that---together---produce a full sufficient cause. In the figure, looking at the specific components, "A" is common to all three pies, but the other causal components are found only in one or two of the full sufficient-cause pies. Rothman's pies could represent the various different sets of conditions that could result in active, symptomatic tuberculosis illness. In every case, the person has to have been infected at some time in the past (component "A"). But a range of different sets of conditions (all including A as one component), coming together, can provide sufficient conditions for the development of active disease, e.g., malnutrition, alcoholism, use of immunosuppressive drugs, HIV-associated immune suppression. Population attributable fraction -------------------------------- For outcomes with multiple potential causes, an important concept is **attributable fraction**---the proportion by which the risk for the *outcome* of concern would be reduced if *exposure* to the factor of interest were eliminated. Attributable fraction for a population (AF~p~) is the proportion by which incidence of the outcome *in the entire population* would be reduced if the exposure were eliminated. It can be estimated using the following formula: where P~c~ is the exposure prevalence among cases and RR is the relative risk for the outcome, comparing exposed vs., unexposed. Let's work through an example. Uncontrolled hypertension is an important known risk factor for stroke, with risk rising with increasingly high levels of blood pressure. Although it's an important risk factor for stroke, there are also cases of stroke among those with normal blood pressure. For the purpose of this example, let's say that among stroke cases in our population of interest: - 20% have no history of uncontrolled hypertension (they have normal blood pressure), - 60% have a history of mild uncontrolled hypertension, and - 20% have a history of moderate to severe uncontrolled hypertension. Let's also say that the relative risk for stroke among those with mild uncontrolled hypertension is 3.5 (i.e., they are at 3.5 times higher risk for stroke than those with normal blood pressure), and let's say relative risk among those with moderate-to-severe uncontrolled hypertension is 6. Plugging these numbers into our formula above, we calculate the population attributable fraction for mild hypertension---i.e., the % of stroke attributable to mild uncontrolled hypertension in our population---as 43%. And, for moderate-to-severe uncontrolled hypertension---17%. So, even though the relative risk of stroke for mild uncontrolled hypertension is lower than it is for moderate-to-severe (3.5 vs. 6), because mild uncontrolled hypertension is much more common in our population, it accounts for a significantly larger number of stroke cases than does moderate-to-severe uncontrolled hypertension. Context, mechanism & outcome ---------------------------- Much of what we're working on in public health, whether it concerns health / disease conditions at the individual level or health / disease patterns in populations, is complex and often---to a considerable degree---context-specific. Even when we test an intervention or strategy using a strong research methodology (a randomized controlled trial or close approximation), if it's a complex multi-component intervention, our RCT design does not allow us to open the black box to unpack how the intervention produced its effect; that is to say, it doesn't tell us what the *mechanism* was. Furthermore, the fact that an intervention produces an effect in one population or setting is no guarantee that it will produce the same effect somewhere else, where the mix of causal factors at work may differ. Pawson and Tilley have pointed to the importance of looking at how programs play out in the real world, characterizing---as best we can---what was the *mechanism* and how did that mechanism interact with the features of a particular *context* to produce the observed *outcome*. This [2005 paper](https://www.researchgate.net/publication/7691447_Realist_review_-_A_new_method_of_systematic_review_designed_for_complex_policy_interventions) is well worth reading. Also, I would strongly recommend watching this [short video](https://www.youtube.com/watch?v=YL39DAVNLBE), and the three videos following it. Behaviour / environment interaction ----------------------------------- Clearly, an important influence on health outcomes is *what we do*, in other words---our behaviours. If I am sedentary and eat poorly, I am likely to move along a different health trajectory in life than I would if I were physically active and consumed a healthy diet. Noting the relationship between specific personal practices and health outcomes, we may be tempted to believe that the solution is just to get suitable messages out and, if our target audiences fail to respond appropriately, to point the finger of blame. But behaviour itself is complex and subject to many interacting influences. [Ross and Nisbett](https://www.amazon.ca/Person-Situation-Perspectives-Social-Psychology/dp/1905177445/ref=sr_1_1?crid=8E7G1ZUAZ2IZ&keywords=the+person+and+the+situation&qid=1564002090&s=gateway&sprefix=the+person+and+%2Caps%2C197&sr=8-1) marshal persuasive evidence that: 1) the circumstances we're in exercise potent influence on our behaviour, and 2) we systematically underestimate how potent that influence is (what they refer to as *fundamental attribution error*). See also my list of drivers of behaviour, entitled "[we decide or act...](https://drive.google.com/file/d/1taZZpZtBhGPDRd3Mik4vj3Bbkevahmlf/view?usp=sharing)", which we will be looking at later in this book. Larry Green and Marshall Kreuter have developed a comprehensive approach to problem characterization, program planning, and evaluation ([PRECEDE-PROCEED](https://www.amazon.ca/Promotion-Planning-Educational-Environmental-Approach/dp/B00SQBTEQG/ref=sr_1_1?keywords=health+promotion+planning+an+educational+and+environmental+approach&qid=1564002645&s=gateway&sr=8-1)) focused on both the *independent* contributions of behaviour and environment to health, and their *interactions*, which acknowledges the insight from Ross and Nisbett that our behaviours happen within a context. PRECEDE-PROCEED has been widely used in public health. Find more information [here](https://ctb.ku.edu/en/table-contents/overview/other-models-promoting-community-health-and-development/preceder-proceder/main). PRECEDE-PROCEED  The epidemiologic triad ----------------------- As we saw in Chapter 1, Hippocrates­­, as long ago as 400 BC---in his work *Air, Water and Places*---noted the important relationships between place, season, and health outcomes. Ideas from Hippocrates about *miasma* (bad air, marsh exhalations), as an underlying cause for illness, were accepted well into the 19^th^ century, until they were largely displaced by germ theory. But current use of the *epidemiologic triad* points to the continuing recognition of complex interactions between the environment, potentially noxious agents in the environment, and the host organism (us, in this case), and their aggregate effect on health status. These relationships are commonly represented as a triangle consisting of *host, agent* and *environment*, all interacting. To this, I've added an arrow for *time*, reflecting changes over time in both the states of host, agent, and environment and their interactions. Although this framework is most often used to look at causal factors involved in infectious diseases, it is certainly applicable more broadly. Indeed, the agent in this formulation can be understood to be physical, chemical, or biological. This framework can be extended to cover the situation in which the agent can also be found in another host species, e.g., malaria, rabies, ebola, avian influenza, corona-viruses. When an infection can be transmitted from that other host to a human, we refer to it as a zoonotic disease. If there is a third species as an intermediary (as in malaria, where the third species is a mosquito), we refer to it as a *vector-borne* disease.[^3^](#fn3){#fnref3.footnote-ref} A field of public health has emerged called [One Health](https://www.who.int/features/qa/one-health/en/), (discussed in more detail in Chapter 7, on physical environment) which works at the nexus of food, human health, animal health, and the environment. The epidemiologic triad can also be applied to problems with an important psychosocial component. It is part of the [conceptual toolkit that public health brings to problematic substance use and addiction](https://www.ncbi.nlm.nih.gov/books/NBK232973/). Attributing dangerous use or addiction to the specific substance or to the user---alone---misses the fundamentally important role of context or environment; in this case, it is particularly the *social* environment that matters. For example: many Vietnam war vets addicted to heroin, returning home to rural America had relatively little difficulty---in this very different social context (where the product was very difficult to source and their peers weren't using)---staying off heroin, although we could certainly imagine that the outcome could be very different today, with much more widespread access to and use of opioids in rural North America than was the case in the 1970s. An important extension of the epidemiologic triad is [**Haddon's matrix**](https://web-a-ebscohost-com.login.ezproxy.library.ualberta.ca/ehost/detail/detail?vid=2&sid=1a2828d6-2f4a-40f3-9165-e6fa937efb7e%40sessionmgr4008&bdata=JnNpdGU9ZWhvc3QtbGl2ZSZzY29wZT1zaXRl#AN=7735113&db=a9h), developed specifically for assessing the causal factors contributing to injury risk, and determining appropriate intervention strategies. Haddon has explicitly incorporated a time dimension (before, during, and after the "event"), and has replaced "host" with "human" and "agent" with "equipment / vehicle". Recall the Rube Goldberg cartoon at the beginning of this chapter, which concerned mitigating the potential consequences of a tumble on an icy sidewalk. Using Haddon's matrix, we can recognize the solution Goldberg proposed as falling within the "event" time-point. In the table above, a solution involving getting a pillow quickly in place to reduce risk of injury would correspond to controlling the pattern of release of energy to reduce harm (like air-bags in cars). Now, whether or not the contraption sketched out by Mr. Goldberg would actually work is another matter. Both the triad and the matrix draw attention to the interaction between a host organism (human, in this case) and the environment. This is an important feature of an *ecological perspective*. It can be helpful to look at various phenomena seen in ecosystems to better understand how comparable processes play out, as regards human health. In studying the relationships between different kinds of organisms in an ecosystem, biologists have characterized a variety of types of symbiosis (living together) by the effects of the relationship on its partners. - When both benefit, this is categorized as *mutualism*; - when one benefits and the other is not affected, positively or negatively, as *commensalism*; and - when one benefits and the other is harmed, as *parasitism*. As we will discuss in more detail in chapter 4, the human organism is in such continuing relations with a variety of other organisms; this includes bacteria, viruses and fungi (and other life forms) living on or within our bodies. These organisms---in the aggregate---constitute a complex ecosystem or biome with organisms having relationships not just with us but also between each other. And, recalling the time arrow in the epidemiologic triad we looked at earlier, relationships and roles can change over time. For example, when the status of the immune system of the host (that's us) changes, a commensal relationship---which neither helps nor harms us---can become parasitic (i.e., one that is harmful to us). That's what happens if our immune competence is compromised, for example by HIV or by taking certain immunosuppressive drugs, and we develop opportunistic infections, like pneumocystis pneumonia. Nested interacting systems -------------------------- We'll continue our move from simple towards increasingly complex causal influence by turning now to causal interactions unfolding across multiple levels, to produce an outcome. From the field of community psychology and child development, Uri Bronfenbrenner offers us an ecological systems model, as first elaborated in his book *[The Ecology of Human Development](https://www.hup.harvard.edu/catalog.php?isbn=9780674224575)* (1979). The framework comprises: micro, meso, exo, macro, and chrono-systems, later extended to include the biological. Building on Bronfenbrenner's model, McIlroy and colleagues ((https://www.researchgate.net/publication/20088489_An_Ecology_Perspective_on_Health_Promotion_Programs)) propose a model consisting of interacting levels of potential influence on health-related behaviour, including: *intrapersonal* (e.g., knowledge, attitudes, behavior, self-concept, skills), *interpersonal and primary groups, institutional* (including formal and informal rules), *community* (including relationships among organizations, institutions, and informal networks), and *public policy*. A similar multi-tiered framework is used by Smedley et al. in [Promoting Health: Intervention Strategies from Social and Behavioral Research](https://www.nap.edu/catalog/9939/promoting-health-intervention-strategies-from-social-and-behavioral-research). As we see in this figure---similar to the Bronfenbrenner and McIlroy models---the framework consists of interacting, nested systems (to which we could add the physical environment and time dimension). All of these models have the value that they can bring a little more coherence into a very big set of influences acting at many different levels of scale, from the micro- to the macro-.  I find the Smedley model (with the illustration of nested, colored cups) useful, in what it includes. And there\'s some logic to the scaling. The model by McDermott (figure below, from Ch.4 in [Nolan, 2023](https://ajph.aphapublications.org/doi/book/10.2105/9780875533377)) has the additional strength that it explicitly represents the time dimension (\"over the lifespan\"). However, both of these models tend to force-fit important domains into just one level of the micro-macro spectrum, e.g. biological factors at the most micro level. And certain domains end up de-emphasized. For example, biology gets short shrift here, and McDermott packs a lot into "living and working conditions". If we do too much lumping, we can lose sight of important categories of influence, making such conceptual frameworks somewhat less useful as thinking tools. Another potential problem with such models is that some folks use them with the assumption that what gets categorized as \"macro\" can only have effects on health, mediated through domains represented here as more \"meso\", e.g., societal level influences on health ONLY mediated through behavior, for example. We have the same problem with notions of upstream/downstream and proximal/distal, as [Nancy Krieger](https://www.ajph.org/doi/10.2105/AJPH.2007.111278?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%20%200pubmed) has pointed out; use of such metaphors can lead us to uncritically assume that macro-level factors influence human health *only mediated through* more micro-level (proximal or downstream) factors. Systems as webs --------------- A nested hierarchy of categories of influence is one metaphor that can be used to characterize an interacting system of systems. Another fruitful metaphor is a *web*. Such web models are used in ecology, illustrating what eats what, and how energy and other resource flow within a system. On the next page, we see a complex representation of a causal web influencing the development of obesity.  UK Government (2007) [Tackling Obesities: Future Choices](https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/287937/07-1184x-tackling-obesities-future-choices-report.pdf) -- Project Report, 2^nd^ Ed Systems concepts ================ Whether we are concerned about health services targeting individuals or initiatives intended to produce health improvements at population scale, we are dealing with *systems*. Recall the quote cited earlier, from Batalden, that "every system is perfectly designed to get the results it gets." The performance, results or output of the system are what we're interested in influencing. Our actions or interventions are intended to produce an effect, i.e., to give rise to a different result than would have obtained had we done nothing. We make a prediction and then place our bet on a particular action or set of actions; then we see what happens to the system. As we've noted at the beginning of this chapter, we can take a stab in the dark but---generally speaking---we stand a better chance of getting the results we're aiming for if we can begin to get a handle on what's going on in that system. As a basis for action, we need to discern patterns, and form tentative hypotheses on causation. We make sense of, or render intelligible, patterns and processes out in the real world, using constructs, theories, models or schemata, all of which---at least implicitly---posit causal relationships between the constituent elements of the systems we are concerned with. A system can be seen as a set of elements in some dynamic, functional relationship with each other; so, *interaction* is a fundamental characteristic of systems. As Williams ((https://search.library.ualberta.ca/catalog/7505277)) points out, in trying to understand what's going in systems terms, we also need to define *boundaries*, specifying what piece of the infinite world out there is the focus of our gaze (What's in? What's out?). Furthermore, depending on who we are and our particular interests or investment in the issue, we will bring a specific *perspective*; when looking at a tree in the forest, an ecologist will be looking at somewhat different constituent elements and processes than will a forester. One systems tool mentioned earlier in this chapter is causal loop diagrams. As described in Williams ((https://search.library.ualberta.ca/catalog/7505277)), such diagrams are often developed through a participatory process that taps implicit knowledge of stakeholders familiar with some particular process, in which their understanding of the relationships between elements in the system is elicited. A particular feature of causal loop diagrams is *feedback* processes, which either reinforce / amplify a phenomenon (positive feedback), e.g., malnutrition \> weakened immune defenses \> infectious illness \> reduced nutritional intake \>\>\>, or dampen, moderate or regulate (negative or balancing feedback), e.g., the response of a thermostat, switching heat on or off when a certain temperature threshold is met. Below is an example of a causal loop diagram, from [Ozawa 2016](https://bmchealthservres.biomedcentral.com/track/pdf/10.1186/s12913-016-1867-7.pdf), outlining factors contributing to building trust in immunization. Causal loop diagrams are generally used qualitatively. They are intended to offer some explanatory purchase on a problem of interest, not to generate mathematical predictions. Other somewhat tools (like the stock-flow models used in systems dynamics, structural equation modeling, and other forms of path analysis) have been developed for such mathematical modeling. Another useful causal model with some similarities to causal loop diagrams, though graphically simpler, is Kurt Lewin's Force-Field analysis, which considers---for any particular phenomenon (though most often applied to social phenomena or intentional change efforts)---what are the forces potentially helping (driving) or hindering (restraining) moving something in the desired direction, and what is their relative weight or importance. Such analysis can be useful in determining where, strategically, to direct one's efforts. Image result for force-field analysis There is a growing literature relevant for a range of disciplines, including public health, on *complex systems*. One useful concept from this literature is the distinction between *simple*, *complicated* and *complex* problems. In practical, real-world problem-solving, failing to recognize which type of problem we're dealing with can result in inappropriate, unproductive strategies. [Glouberman and Zimmerman](https://www.alnap.org/system/files/content/resource/files/main/complicatedandcomplexsystems-zimmermanreport-medicare-reform.pdf) (2002) offer the following useful illustration... Unfortunately, a common occurrence is that we misread the situations we are dealing with as merely *complicated*, assuming that with good detailed plans---as good technocrats---we should be able to accomplish our goals when, in fact, what we are dealing with is a *complex* problem in which outcomes can vary enormously depending on very small baseline differences; there may be reinforcing or extinguishing causal loops, threshold effects (i.e., tipping points), or non-linear relationships between inputs and outputs (or exposures and outcomes); and where we may see emergent phenomena arising at higher levels in the system (recall our earlier discussion of the conceptual frameworks proposed by Bronfenbrenner and McIlroy). When dealing with problems in complex systems, plans are still likely to be necessary, but we need to track the actual performance of the system and be ready to modify our plans in response to unpredicted developments. We need to be flexible. Good to keep in mind that *no battle-plan survives contact with the enemy*. One key reference that can serve as a good starting point for systems theories, methods and tools, and their application to public health is Peters' 2014 [commentary](https://health-policy-systems.biomedcentral.com/track/pdf/10.1186/1478-4505-12-51) in the journal, Health Research, Policy & Systems. Other recommended readings on systems thinking and public health: - National Collaborating Centre for Health Public Policy (2013). [[Wicked problems and public policy]](http://www.ncchpp.ca/docs/WickedProblems_FactSheet_NCCHPP.pdf) - Conklin, J. (2005). [[Wicked problems and social complexity]](http://cognexus.org/wpf/wickedproblems.pdf) - Figure A. [[Final form of the CSDH conceptual]](https://www.who.int/social_determinants/corner/SDHDP2.pdf) framework in World Health Organization Geneva (2010). Social Determinants of Health Discussion Paper 2. - [[Five examples of systems mapping]](https://drive.google.com/file/d/14oDXa7e597GJ_BqziNazP6Vsbhp5PRA6/view?usp=sharing) Determinants of Health ====================== There have been important Canadian contributions to thought on what influences health states or causes disease. Released by the federal ministry of health in 1974, when Marc Lalonde was minister (therefore referred to as the Lalonde report), [A New Perspective on the Health of Canadians](http://www.phac-aspc.gc.ca/ph-sp/pdf/perspect-eng.pdf) introduced the concept of *health fields* (i.e., major categories of determinants), which consisted of: +-----------------------------------+-----------------------------------+ | - Lifestyles / health-related | - Human biology / genetics | | practices | | | | - Healthcare organization | | - Environment / physical and | | | psychosocial | | +-----------------------------------+-----------------------------------+ As the authors of the report explain, although the bulk of expenditure for health is on the last element---healthcare---much of what causes disease falls into the other three. One consequence of this prominent report was to direct the attention of policy makers to the contribution of these other "fields." Some of the language in the report, however, was unhelpful. Notably, the lifestyle element was described as consisting of the aggregate of *"decisions by individuals* which affect their health and *over which they* more or less *have control...* *personal decisions and habits that are bad*, from a health point of view, create *self-imposed risks"* (italics added). This assertion that behaviour can be largely understood in terms of individual agency, without reference to the forces impinging on the individual from their environment runs counter to insights we have discussed earlier (for example from Ross & Nisbett, and Green & Kreuter). This unfortunate individualizing perspective on behaviour can lead to victim-blaming and---by failing to recognize important behaviour-environment interactions---it directs attention away from potentially effective intervention strategies. The [government of Canada](https://www.canada.ca/en/public-health/services/health-promotion/population-health/what-determines-health.html) lists the 12 following areas as "determinants" or influences on health: 1. 2. Income and social status 3. Employment and working conditions 4. Education and literacy 5. Childhood experiences 6. Physical environments 7. Social supports and coping skills 8. Healthy behaviours 9. Access to health services 10. Biology and genetic endowment 11. Gender 12. Culture 13. Race / Racism As we have seen, different groups organize the determinants in various ways. Over the next several chapters we will be looking at the determinants grouped under the broad headings of: biology, behaviour, social environment, physical environment, and healthcare. See below:  [Gruszin 2011](https://www.health.nsw.gov.au/hsnsw/Publications/classifications-project.pdf) uses essentially the same outline for grouping the determinants (with healthcare not included, in this case), as we see below: Conclusion ========== I've thrown a lot of concepts at you in this chapter and much of what I've presented has been very much in summary form. What I've tried to do is to get you at least a bit acquainted with some of the key issues and concepts related to *causation*, as relevant for public health work. What I have included here is what I have found useful in my own public health practice. But, as broad as this has been, what I've presented here doesn't go very deep. So I would encourage you---as time permits---to dig deeper, taking advantage of the embedded links. Below I have also listed some key texts you could start with. **Additional suggested readings:** Canadian Medical Association. [Health care in Canada. What makes us sick?](https://nccdh.ca/resources/entry/health-care-in-canada) Town Hall Report. 2013 Cartwright N, Hardie J. [Evidence-Based Policy](https://global.oup.com/academic/product/evidence-based-policy-9780199841622?cc=ca&lang=en&)*.* Oxford Univ Press, NY NY, 2012. Read part 1, section IB 7.1 (8 pages: 49-58). Evans RG, Barer ML, Marmor TR. 1994. [Why are Some People Healthy and Others Not: The Determinants of Health of Populations](https://www.degruyter.com/document/doi/10.1515/9783112421628/html?lang=en). Aldine de Gruyter. NY NY. Keyes KM, Galea S. [Setting the Agenda for a New Discipline: Population Health Science](https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/26959265/). Am J Public Health. 2016 Apr;106(4):633-4. doi: 10.2105/AJPH.2016.303101. Glouberman S, Zimmerman B. [Complicated and Complex Systems: What Would Successful Reform of Medicare Look Like?](https://www.researchgate.net/publication/265240426_Complicated_and_Complex_Systems_What_Would_Successful_Reform_of_Medicare_Look_Like) Commission on the Future of Healthcare in Canada, discussion paper \#8. Ottawa; 2002. Greenhalgh T. 2021. Miasmas, mental models and preventive public health: some philosophical reflections on science in the COVID-19 pandemic. Interface Focus 11: 20210017. Havranek EP, Mujahid MS, Barr DA, et al. [Social Determinants of Risk and Outcomes for Cardiovascular Disease](https://www.ahajournals.org/doi/pdf/10.1161/CIR.0000000000000228) Circulation. 2015 Sep 1;132(9):873-98. doi: 10.1161/CIR.0000000000000228. James JE. Chapter 1 - The Origins of Health. The Health of Populations. Academic Press. [http://dx.doi.org/10.1016/B978-0-12-802812-4.00001-1 2016](http://dx.doi.org/10.1016/B978-0-12-802812-4.00001-1%202016). On The McKeown Thesis Keyes KM, Galea S. 2016. Population Health Science. Oxford Univ Press. Oxford, UK. Kindig D, Stoddart G. [What is population health?](https://ajph.aphapublications.org/doi/10.2105/ajph.93.3.380?url_ver=Z39.88-2003&rfr_id=ori%3Arid%3Acrossref.org&rfr_dat=cr_pub++0pubmed) Am J Public Health. 2003 Mar;93(3):380-3. doi: 10.2105/ajph.93.3.380. Krieger N. Epidemiology and the web of causation: has anyone seen the spider? Soc Sci Med. 1994;39: 887--903. Krieger N. [Proximal, distal, and the politics of causation: what\'s level got to do with it?](https://www.ajph.org/doi/10.2105/AJPH.2007.111278?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%20%200pubmed) Am J Public Health. 2008 Feb;98(2):221-30. doi: 10.2105/AJPH.2007.111278. McKeown T. 1979. [The Role of Medicine: Dream, Mirage, or Nemesis?](https://press.princeton.edu/books/hardcover/9780691643663/the-role-of-medicine) Princeton Univ. Press, Princeton, NJ. Pawson R, Greenhalgh T, Harvey G, Walshe K. [Realist review---a new method of systematic review designed for complex policy interventions](https://www.researchgate.net/publication/7691447_Realist_review_-_A_new_method_of_systematic_review_designed_for_complex_policy_interventions). J Health Serv Res Policy. 2005 Jul;10 Suppl 1:21-34. Pawson R, Tilley N (1997). [Realistic Evaluation](https://www.amazon.ca/Realistic-Evaluation-Ray-Pawson/dp/0761950095/ref=sr_1_1?keywords=Realistic+Evaluation&qid=1565045534&s=books&sr=1-1). SAGE Publications. London. Pearl J, MacKenzie D. 2018. [The Book of Why: the New Science of Cause and Effect](https://dl.acm.org/doi/10.5555/3238230). Basic Books. NY. Rose G, Khaw K-T, Marmot M. 2008. [Rose's Strategy of Preventive Medicine](https://academic.oup.com/book/2993). Oxford Univ. Press. Oxford, UK. Rothman KJ, Greenland S. [Causation and causal inference in epidemiology.](https://blogs.kent.ac.uk/jonw/files/2015/03/Rothman-Greenland-05-Causation-and-Causal-Inference-in-Epidemiology.pdf) Am J Public Health. 2005;95 Suppl 1:S144-50. Young TK. Chapter 4: Modeling Determinants of Population Health, in Population Health: Concepts and Methods (2nd edn). Oxford Academic, 2004. Zimmerman FJ. [Population Health Science: Fulfilling the Mission of Public Health](https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/33320388/). Milbank Q. 2021 Mar;99(1):9-23. doi: 10.1111/1468-0009.12493. **Suggested viewing:** An introduction to realist evaluation, by Ray Pawson (\~30 min total). *[Part 1](https://www.youtube.com/watch?v=YL39DAVNLBE&t=186s) (7 min), [Part 2](https://www.youtube.com/watch?v=Zg87MXwbtKg) (9 min), [Part 3](https://www.youtube.com/watch?v=YZfb7vXOzNo) (8 min), [Part 4](https://www.youtube.com/watch?v=qRuqG0BZO-c) (7 ½ min)* On double-loop learning watch: (16½ min) Here's a good [short video](https://www.youtube.com/watch?v=eXdzKBWDraM) with Peter Senge on systems thinking. And see this webinar, [Equiate webinar](https://www.youtube.com/watch?v=qUHgUIf2nZU), on Systems Thinking and Public Health Another useful resource on causality and public health: **Other resources:** - Government of Canada. Social and economic influences on health. The "determinants of health": - National Collaborating Centre for Determinants of Health. (2022). Understanding Indigenous health inequalities through a social determinants model. - Material on the National Collaborating Centre on the Determinants of Health website: - Rutter, H, Savona, N, Glonti, K, Bibby, J, Cummins, S, Finegood, DT, et al. The Need for a Complex Systems Model of Evidence for Public Health. The Lancet. 2017; 390(10112):2602-4. - Zukowski, N, Davidson, S, Yates, MJ. [Systems Approaches to Population Health in Canada: How Have They Been Applied, and What Are the Insights and Future Implications for Practice?](https://www-ncbi-nlm-nih-gov.login.ezproxy.library.ualberta.ca/pmc/articles/PMC6964537/pdf/41997_2019_Article_230.pdf) Canadian Journal of Public Health. 2019; 110(6):741-51. - Riley, BL, Robinson, KL, Gamble, J, Finegood, DT, Sheppard, D, Penney, TL, et al. [Knowledge to Action for Solving Complex Problems: Insights from a Review of Nine International Cases](https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/25970804/). Health Promotion and Chronic Disease Prevention in Canada. 2015; 35(3):47-53. Note CEPH competency, to be assessed: "Apply a systems thinking tool to visually represent a public health issue in a format other than standard narrative" Note, also, related CEPH foundational learning objective: understand the interaction among social, community, environmental, ecological, & global factors in the onset of & solution to public health problems using systems thinking (CEPH LO) Will add relevant competencies from PHAC, once they're released (probably late in 2024) ::: {.section.footnotes} ------------------------------------------------------------------------ 1. ::: {#fn1} In Porta's Dictionary of Epidemiology, "risk factor" is defined as: a factor that is causally related to a change in the risk of a relevant health process, outcome, or condition. The causal nature of the relationship is established on the basis of scientific evidence... and causal inference. The causal relationship is inherently probabilistic.... If the relationship is non-causal, the factor is just a risk marker. Risk factors... may be a socioeconomic characteristic, personal behavior or lifestyle, environmental exposure, inherited characteristic, or another trait. Risk factors for human health often have individual and social components; even when individual and social risk factors can be separated, they often interact. A determinant that can be modified by intervention is a modifiable risk factor. The term risk factor became popular after its frequent use in papers from the Framingham study.[↩](#fnref1){.footnote-back} ::: 2. ::: {#fn2} Note that for any of these epidemiologic terms, it may helpful to consult the definitions in Porta's [Dictionary of Epidemiology](https://www-oxfordreference-com.login.ezproxy.library.ualberta.ca/display/10.1093/acref/9780199976720.001.0001/acref-9780199976720). For example, this is what Porta has to say about *confounding*: "Loosely, \[confounding is\] the distortion of a measure of the effect of an exposure on an outcome due to the association of the exposure with other factors that influence the occurrence of the outcome. Confounding occurs when all or part of the apparent association between the exposure and the outcome is in fact accounted for by other variables that affect the outcome and are not themselves affected by exposure." Or, more concisely, confounding is "an association between the potential outcomes under alternative exposure or treatment possibilities and the actual exposures or treatments received by the persons under study."[↩](#fnref2){.footnote-back} ::: 3. ::: {#fn3} Note that vector-borne diseases also include those only involving humans and the vector, with no third species involved.[↩](#fnref3){.footnote-back} ::: :::