Anyone who knows me appreciates that I was destined to be no more than a limited 'futurist' as compared to the best futurists. I lay the blame on the Catholic Church, becoming an altar boy, studying Latin for more years than I wish to count (including a year of self-study), and gaining too much respect for early origins and roots. Think about me as a conservative radical, conserving the roots. I might be too 'conservative' to be a good futurist.
At the same time, I also am known as an iconoclast, because I do not hesitate to challenge bogus concepts when they creep into our field of epidemiology, among which I can specify 'risk factor' and 'reverse causality' as special within-field jargon terms that have almost no communication value when we, as epidemiologists, try to describe our evidence to audiences beyond the local 'tribe'. (Worse yet, there are non-tribe members who actually think they know what we mean when we use these jargon-terms, as has occurred when my non-baseball friends have thought I'm talking about mental health when I use the term 'screwball' when describing a baseball pitcher.)
With that background in mind, you can imagine my response when I was giving an end-of-semester talk to my research trainees and students about some challenges that epidemiologists are going to face during the next few decades. I thought back to two issues raised when I was early in my studies of epidemiology back in the late 1970s: (1) an idea that David Mechanic and Bill Eaton mentioned about epidemiology having run out of ideas and theories, and (2) an idea that Bruce Dohrenwend offered about epidemiology being little more than a toolkit or collection of methodological approaches.
For my students, I framed my response to a hypothetical question about the challenges they might face in relation to these two ideas, and I offered this statement: "The most serious challenge you will face in your career is running out of substantive ideas and theories that deserve testing."
For the past day, I have been thinking about what I told them, and I realized that there might be an empirical indication of the degree to which epidemiology, as a discipline, has started to run out of ideas and theories, and has entered a terminal (?) stage of being a methods-oriented field of study as might be distinguished from a science, akin to the way we'd differentiate 'operations research' as a field of study without thinking about it as a 'science' in the generally accepted sense of the term.
My realization of an empirical approach required a reflection on (a) an early expression of Wade Hampton Frost's idea that epidemiology has to do with infective agents and infectious diseases, and he was uncertain about whether it made sense to think about epidemiology as applied to other conditions that did not originate with an exposure to an infective agent, (b) Mort Levin's assertion in a Milbank Memorial Fund session that Ernie Gruenberg had organized, which amounted to this statement: "There can be an epidemiology, and there can be a set of psychiatric disorders, but I'm not sure there needs to be an epidemiology of psychiatric disorders," and (c) a general principle I first learned from my PhD advisor Len Schuman at University of Minnesota, which also later was stressed by my local colleague and friend Nigel Paneth, along the lines of the advantages gained by restricting the prefix adjective for epidemiology to the domain of individual disease entities or disease classification names, as in 'cholera epidemiology' and 'infectious disease epidemiology,' as opposed to alternatives based on prefix adjectives that shifted attention from the epidemiological response under study in the direction of the causal explanations under study (e.g., 'genetic epidemiology').
When I first appreciated this point as it was being made by Len Schuman in my early years after starting graduate studies in epidemiology, I had not read Wade Hampton Frost nor had I met Mort Levin. I was out in Minnesota, ignorant of Mechanic's medical sociology program at University of Wisconsin, and completely ignorant about Bruce and Barbara Dohrenwend's social psychological research program at Columbia. Nonetheless, it was not long until I was reading the Lilienfeld-Gifford 'Chronic Diseases and Public Health' book assigned by Schuman, and I recognized a clear distinction being made between 'Infectious Disease Epidemiology' and 'Chronic Disease Epidemiology' (hereinafter abbreviated as ID and CD).
[At this point, I stopped for refreshment and I pick up the thread down below.]
Now that I have drawn attention to this separation of 'ID epidemiology' from 'chronic disease' epidemiology, I can offer a thought experiment that some enterprising young epidemiologist might try to pursue. (The most valuable thought experiments are those that can be investigated empirically once the experiment has been specified.)
The empirical work involves the study of time trends in appearance of adjectives used to modify the term 'epidemiology' as can be seen in traces such as Web of Science or Google Search or JSTOR or other bibliographic search tools that have emerged as a consequence of internet advances.
An initial interesting time trend will depict the relative frequency of 'infectious disease epidemiology' and 'chronic disease epidemiology' as these terms appeared in the published literature, year by year, over time. Here, the expectation would be an increasing frequency of both terms, but in addition, even with de-trending, the ID/CD ratio might be expected to decline, at least until the mid-1980s, given (1) increasing prominence of work in CD epidemiology until (2) emergence of HIV/AIDS research as priorities for NIH before 1990. (I have not studied these trends so these expectations are based solely on my 'seat of the pants' reckoning, and I might well be wrong.)
A savvy approach will put into play de-trending research approaches that compensate for generally increasing frequencies of 'epidemiology' as a search term, based on knowledge that NIH and other funding sources increased their investments in epidemiology across these years. A more savvy approach will take other interesting issues into account, such as might be illuminated by studying citation counts as opposed to what can be learned via 'topic' or 'title' restricted searches over these years.
But I leave these issues up to others in order to focus on the possibility that epidemiology's de-celerating rate of innovative ideas might be reflected by studying (1) the relative frequency of 'disease entity' adjectives positioned as prefix terms before 'epidemiology' appears as a noun, and (2) the corresponding relative frequency of 'non-disease-entity' adjectives positioned as prefix terms. Here, by 'disease entity' adjectives, I'll use as examples 'cancer epidemiology' and 'dementia epidemiology.' For 'non-disease-entity' examples, I'll use examples 'social epidemiology' and 'genetic epidemiology' or 'nutrtional epidemiology' or 'life-course epidemiology.'
It is fair to ask why the relative frequency of these prefix adjectives might convey something important about the content of epidemiological research and the inherent generativity of epidemiology as a source of new ideas and theories. I have little more than a simple answer of the following form, expressed in relation to 'Types' of epidemiology:
(Type 1) When the prefix adjective is focused on the response variable, such as a cancer or a dementia, then the domain of potential explanations is expanded almost infinitely to cover whatever might be accounting for variation in the incidence rates and prevalence proportions for that response. The epidemiologist is faced with the challenge of drawing from multiple disciplines in an effort to understand what might account for the population-level variations in the epidemiological parameters. Let's think of the specification of the response variable as one dimension of epidemiology.
(Type 2) When the prefix adjective is focused on a domain of explanation (e.g., 'social' or 'nutritional' or 'genetic'), there is a greater constraint on the range of sources of variation. The epidemiologist is faced with a more circumscribed challenge, which is to ask whether and under what circumstances that domain of explanation might be more or less important in attempts to account for population-level variations in the epidemiological parameters. Let's think of the specification of the explanatory domain as a second dimension of epidemiology, when 'second' denotes a time sequence rather than my own pre-judgement (prejudice) about the ordinal ranking of importance or public health significance.
(Type 3) When the prefix adjective is focused on a domain of process (e.g., life-course'), there is a third dimension that can be considered quite separately from the first two dimensions, and the third dimension has a correlation with time. This dimension might reflect age (considered subgroup by subgroup, as in a cohort-sequential study), or aging (considered as a progression from one birthday to the next), or as developmental (as in the relatively fuzzy sets that we use to distinguish developmental stages along the lines of (a) pre-term, full-term, etc., with respect to measured gestational age of a newborn, or (b) 'pre-pubertal,' 'pubertal,' etc., along the lines of the Tanner stages, or (c) pre-senile versus senile, as in the differentiation of age of onset of dementia syndromes based on when the syndrome was detected, earlier for those with trisomy-21, at mid-life for those with certain familial histories of early-onset, or later for those formerly known as 'sporadic' due to the absence of the family history of mid-life onset of the dementia syndrome. This is a third dimension, separable from the first two dimensions.
It is my speculation that an increasing frequency of prefix adjectives of type (2) and (3), relative to type (1), over time, can serve as an indication of the degree to which epidemiology, per se, has stopped generating its new ideas and theories, and instead is turning in the direction of other disciplines to fill a gap of imagination or generativity that otherwise would not require turning toward these other disciplines.
I have no idea whether this speculation has merit. I offer it as a suggestion for epidemiologists and historians of science who might wish to seek quantitive evidence about the degree to which a discipline has 'run out of its own ideas and theories' versus the alternative of borrowing from other disciplines.
I look forward to comments and to any empirical evidence on these matters.
This comment has been removed by the author.
ReplyDeleteHas epidemiology stopped generating new ideas and theories?
ReplyDeleteI would not think so.
Here's my slightly more positive spin.
I've always thought of epidemiology as a collection of toolkit to be applied to something substantively meaningful. Is there theoretical epidemiology, and what does that mean?
Is pre-1980 era the golden age for epidemiology? It surely gave epidemiology its fame. That said, I think no matter what we do, we need to be adaptive to stay current and relevant. Maybe that is why 'non-disease-entity' prefixes appeared and thrived for epidemiology. I am not arguing that the disease-driven approach is not a good approach, but in reality, one thing is related to a lot of others. Some people choose to be expert in outcomes; some people choose to be expert in determinants; and some others are open to both. In my mind, the latter is the most practical simply because everything is related to everything else. For example, if we see alcohol drinking as an outcome, we will never discover the various diseases due to drinking. But should they be encompassed in 'alcohol epidemiology'? I have no doubt. And did we make progress on that since 1980, we sure did!
Is it wrong to take a determinant-driven approach? I'd rather keep an open mind. Maybe it is even more relevant for mechanistic research because of its focus on the explanatory variable. One variable may lead to a variety of different outcomes via a common initial reaction. To me, it is certainly possible. A lack of sunshine will kill many different types of plants. Should we focus on one plant, not the sun exposure?
Circle back to the question whether epidemiology has run out of steam. I think there are always plenty of things to explore if we keep our eyes open and look around. It is a wonderful and complicated world. The beauty of epidemiology lies in the population- and health-oriented approaches which open up a whole range, almost an infinite number, of environmental heterogeneities. Do we have to proclaim ourselves distant relatives to cholera epidemiologists to survival. I think not. I'd rather think early epidemiologists would be happy to see how we're trying to learn about how agents (e.g., drugs) interact with the genome and the envirome. Shape shifter is not necessarily a bad thing.
Epidemiology has been generating copious interesting ideas based on empirical observations. One simple example is why male-female ratios in drinking vary so widely around the world? Gender roles or social norms may never come up if we were to focus on bench sciences or clinical treatment.
The slow pace on cracking chronic diseases is, at least partially, due to the fact there are usually multiple competing pathways to a common phenotype, not as simple as the infective agent-disease relationship. Without epidemiological approaches, I do not know we will be able to grasp the complete picture for a given chronic disease.
In sum, I do not think it is time to bite the dust. Rather, It is time to thrive with borrowed caliber from other sources.
This comment has been removed by the author.
ReplyDeleteJim's reply, part 1:
ReplyDeleteVery thoughtful ideas, Hui.
But I don't think a scientific discipline can have it both ways.
At its origins, epidemiology sought to understand sources of variation in the disease outcome, and entertained hypotheses from all sectors.
(I'll insert 'chronic disease epidemiology' as the exemplar that is noteworthy precisely because it ignores the fact that infective agents can be causes of chronic diseases (e.g., tuberculosis as prime example.)
This tradition was sustained through the mid-1900s, after which epidemiology began to see proliferation of what I call 'the pledge of allegiance' to specific domains of causal influence. (Here, I'll insert 'social epidemiology' and 'genetic epidemiology' as exemplars.)
Your example of 'alcohol epidemiology' is in the second tradition, in that it's an outgrowth of temperance movements, and what is to be done when it is discovered that a small amount of drinking alcohol actually might have a beneficial effect on some (but not all) disease outcomes.
This example is a beautiful illustration of the wisdom of orienting epidemiology around the disease outcomes and not around the suspected causal influences. For one disease outcome, alcohol might be beneficial. For another, it might be harmful. The essential discoveries place the influence of alcohol drinking within the context of the larger conceptual model (theory) of what explains a disease outcome, and alcohol might have one valence in relation to its influence on outcome Y1 but a different valence in relation to its influence on outcome Y2.
But I've already acknowledged my position as a 'conservative radical' and in this context I don't think anyone has a monopoly on the truth.
The only question is how should epidemiology organize itself in the 21st century, and which is the most public health promoting foundation-stone?
In my own view, epidemiology's most solid foundation-stones are created when epidemiologists declare their efforts to understand as much as possible about the broad array of potentially modifiable determinants of a disease outcome, and set about to eradicate that outcome, without pledging allegiance to any specific source of variation.
I appreciate that there are epidemiologists who wish to stand on a different foundation-stone such as 'social influences' or 'genetic influences' and focus on research that comes as close to proof as we can get to the definitive evidence that 'social influences' or 'genetic influences' are important. This is the 'pledge of allegiance' I mentioned. If so, then there seems to be an inherent bias in favor of that domain of causal influence, of the type that needs to be confessed in a conflict of interest statement.
Here is the thought experiment to consider:
Let's suppose we have the same study replication conducted by a social epidemiologist (pledge of allegiance to social influences) with respect to an outcome such as liver fibrosis stage 4, and also in a counterfactual world by the general epidemiologist (no pledges of allegiance to any domain of causal influence).
The same study is conducted in two essentially exact replications, one with social epidemiologist as PI and the other with general epidemiologist as PI.
The key association under study shows p =0.051, and the journal's statistical editors require a frequentist approach.
Which of these two epidemiologists is going to try to get away with the following statement:
"The estimate, with p = 0.051, trends toward statistical significance."
And which will say "The association is null at p>0.05."
Leaving aside the fallacy of the journal's homage to frequentist statistics, I think I'm offering an example that accomplishes what I'm trying to accomplish, which is to say that pledging allegiance to a specific domain of causal inference is not necessarily a step forward for epidemiology. It might be a step backward.
Jim's reply, part 2:
ReplyDeleteIn my own instance, I was drawn to epidemiology because it offered a way to investigate and to make a difference with respect to public health actions to reduce occurrences of disease-states, and I was not required to pledge allegiance to any domain of causal influence. I could let the chips fall as they may.
I do not have the perspective of a 'social epidemiologist' or of a 'genetic epidemiologist' or of an 'alcohol epidemiologist' but what worries me that those who enter the field with these perspectives have a finger placed on the scale in a way that is not necessarily conducive to the best science nor to the best inferences upon which public health action should be based.
At minimum, I see a conflict of interest. And as with any conflict of interest, the best course of action is not to deny its existence. The best course of action is to declare it, and to let the world know that it is present. (This is why the Donald Trump policy of not showing his tax returns and not releasing visitor logs is offensive. The problem is not that he might have dodged taxes in a legal way. Or that he has a stream of visitors who seek to influence his decisions one way or the other. The problem is that he is not willing to make a public disclosure of the conflict of interest and to let us take it into account. If you catch my drift, then those who are aligned with a specific domain of causal influence owe it to the world to disclose the conflict of interest when they publish their results.)
But I respect the possibility that I am way off base here. I will await education and enlightenment if I'm worrying about something that epidemiologists really should not be worrying about.
Again, thanks for your comment. I really enjoyed reading it even though I'm not sure we agree about the inherent conflict of interest that was created when epidemiology lost its bearings and orientation to specific disease outcomes, and instead headed off in the direction of pledges of allegiance to specific domains of causal inference (e.g., 'social' or 'genetic' or 'nutritional' epidemiology) or suspected influential processes (e.g., 'developmental' epidemiology).
Conflict of interest... That is an interesting point.
ReplyDeleteDoes being a 'social epidemiologist' or 'genetic epidemiologist' blur one's vision and place one domain of causes at a higher order than anything else? They should not. They should be aware and acknowledge the existence of other domains even when they have no passion for them. In the end, the ultimate goal is to improve health or contribute to knowledge, not to win a meaningless turf battle. The only downfall I see is that they might deliberately choose to be an ignorant about other sets of causes. Would a conflict of interest statement cure that? Maybe. But is it fair to them? Don’t we all have conflict of interest in that sense though? Sometimes, people want to prove their hypotheses so eagerly, they overlook things, somethings crucial things. In my mind, that lies in the training of epidemiologist. How to put out our best effort, and how to take the responsibility of our own work.
I ain't no conservative, not at all. Maybe I am just a radical with sometimes too open a mind...
Possibly the best adjective prefix for epidemiology is no prefix at all.
ReplyDeleteFrom a pedagogical standpoint, the curriculum for a prefixed epidemiology would have to yield mastery of the adjective area as well as mastery of epidemiology.
The "social epidemiologist" would master and profess in sociology or one of the social sciences in addition to mastering and professing in epidemiology. Would be as familiar with Marx, Weber, and Durkheim as with Frost, Snow, Farr, and Lilienfeld.
Ditto for the "genetic epidemiologist" with respect to genetics, Johansen, Wright, Watson, Crick, and Mullis.
Ditto for the "viral epidemiologist" with respect to virology and its experts.
But I'm reminded of hedgehogs and foxes, each with a different and equally respectable worldview.
We can agree to disagree, but I'm thinking you agree with me that conflicts of interests and one's favoritism toward a specific kind of subject matter evidence should be declared.
I do agree on that. I usually say I am an epidemiologist, and I study such and such...
ReplyDeleteDitto for COI!