Soporific wonkery

As part of my continuing series discussing what I actually do in the real world, I will now confess that about half of what we call my "percent effort" in the eerie netherworld of sponsored research is devoted to a system I developed, with major input from colleagues, to code and analyze the interactions between physicians and patients. Other such systems already exist, the most popular being the Roter Interactional Analysis System. Why do we need a new one?

Well, obviously, any method of categorizing and measuring any components of reality, be they subatomic particles, species of beetles, planets, stars, personalities, national economies, or hate-radio hosts, can answer some questions and not others. The Roter system is concerned principally with the affective and interpersonal properties of interaction, as are most other systems. But my colleague was conducting a trial of an intervention to improve physicians' awareness of the medication-taking behaviors of their patients with HIV, and hence, it was hoped, their discussions of medication adherence with their patients and ultimately the control of their patients' HIV disease. He needed a method that could also provide detailed, specific information about the content of the interaction, that is, what the people were talking about, not just how they were talking about whatever it was.

I also had some ideas about how to make interaction analysis more rigorous and interpretable. Systems such as the RIAA were really developed pragmatically, and don't have a clear underlying theory. They allow only one code to be assigned to each unit of analysis -- and the units themselves aren't defined as clearly as I would like -- and end up mixing together concepts that pertain to what is being discussed (topic), with concepts that pertain, in a loosely defined way, to the process of the interaction.

So, we went back to the beginning and started with a theoretical basis in Speech Act Theory -- a set of ideas developed, in their modern form, by the philosopher John Searle. We can observe that human conversation consists of the exchange of specific kinds of social resources, no matter what we are talking about. For example, I can give you information, express my feelings, ask you to do something, command you to do something, make a promise, ask a question (which is a specific kind of request), praise you, insult you, etc. So we have a unit of analysis -- a completed speech act -- and a way of labeling each of these units. Then, we can go back and code them according to the topic. This is a question by the doctor about symptoms. This is a request from the patient for an analgesic. This is a promise by the patient to quit smoking. And so forth. It's a bit more complicated than that, but you get the idea. Some speech acts are directly about the interaction process -- introducing or closing topics, for example. Some are specifically about interpersonal affect such as empathy and reassurance, but these are actually quite rare.

The taxonomy of speech acts is always the same -- across cultures and languages, I am convinced -- but the list of topics that you use, and how you organize them, depends on what you are interested in. However, for medical visits, it was possible to develop a short list of broad categories, which investigators can then unpack into greater detail as they choose, depending on the current research question. We have also developed software to make it easier to implement the system.

The results will be quantitative data -- numbers of speech acts of various kinds by each participant, cross-tabulated with topics. While I am hopeful that we can learn a good deal about how the content of medical visits is related to certain patient outcomes, I'm still very much aware that all such attempts at quantification squeeze the juice out of life, that ultimately each interaction is a story, a complex, self-contained event that cannot be understood without taking it in as a whole. In order to make sense of human interaction, we need to be able to work between levels, to view interactions through the lenses of both holism and reductionism. Ultimately, interpersonal effectiveness in the healing professions will probably elude complete, rigorous description. But at least we hope to come up with some helpful hints.