The Science of Clinical Psychiatric Reasoning

Guest post: Heinrichs introduces his book “How Psychiatrists Make Decisions”

and

May 25, 2025

Recently I reviewed Adrian Kind’s 2025 book, “How Does the Psychiatrist Know?: On the Epistemology of Psychiatric Diagnostic Reasoning,” on this substack. I had mentioned that Kind’s book is the first systematic, in-depth, book-length philosophical investigation into how psychiatrists arrive at their diagnostic conclusions. However, we are not even in the second half of 2025, and we already have another systematic, in-depth, book-length philosophical investigation into the closely related topic of how psychiatrists make treatment decisions! Not only that, there is, reassuringly, a remarkable degree of convergence between the two independently-developed approaches. I am pleased to share my fellow colleague’s guest post introducing this new volume. Awais Aftab.

Douglas W. Heinrichs, MD, is a psychiatrist. He has spent 40 years in private practice, following nearly a decade of experience in clinical research and academic teaching at the University of Maryland. He is active in the area of philosophy of psychiatry and is a member of the executive committee of the Association for the Advancement of Philosophy and Psychiatry. He is the author of “How Psychiatrists Make Decisions: The Science of Clinical Reasoning” (2025) published by Oxford University Press.

“How Psychiatrists Make Decisions” is available in hardback in Europe starting May 27, 2025, and in the US starting August 27, 2025, through OUP and other online booksellers such as Amazon. It is already available in digital format at Oxford Academic and other online sellers.

This book grew out of reflections on my experience of over 40 years as a practicing psychiatrist, first as a full-time clinical researcher and teacher in an academic setting, and then in the private practice of general psychiatry, in which I treated a wide range of patients in both inpatient and outpatient settings. I accrued this experience during a period of rapid and fundamental changes in the way psychiatry understands itself and how it articulates standards of practice. This has left me with a deep-seated concern. The paradigmatic activity of the psychiatrist is the treatment of the individual patient who seeks relief from suffering or dysfunction. Central to this work is making rational and effective decisions about treatment interventions based on the psychiatrist’s understanding of each patient’s situation in all its unique complexity. Yet for all its importance, this decision-making process has not been well characterized. While it is informed by more generalized knowledge drawn from formal research or accumulated clinical wisdom, we often see the process by which we arrive at an understanding of a unique individual as the intuitive product of our experience, the “art” of medicine. Rather than being instructed in a specific methodology, this “intuitive” capacity is expected to be largely tacitly absorbed from supervisors and accumulated experience with patients during residency and beyond.

My undergraduate concentration was in philosophy. Among its most fundamental set of questions are those comprising epistemology: what are the methods by which we acquire valid knowledge about the world and how justifiable is our confidence in the resulting beliefs based upon those methods? This book is an attempt to answer those questions with respect to clinical reasoning and decision-making in psychiatry, specifically how we make specific treatment decisions for an individual patient. I argue that this is of considerably more than just theoretical interest. In my time in psychiatry, I have seen a growing determination in the field to see itself as scientific, whatever that is taken to mean, and of valuing the results of formal clinical and basic research over any unique understanding the clinician has of the individual patient seeking help. The rise to prominence of evidence-based medicine (EBM) and practice with its articulated methodology threatens to replace the traditional approaches to clinical reasoning as the gold standard as to how we should practice. To salvage any credibility for traditional modes of clinical reasoning, we need to articulate the underlying methodology and carefully examine the basis for our confidence in the alleged knowledge about our patients that comes from such reasoning as it guides our decision-making process in treatment. I argue that without such a methodology, EBM by itself fails to provide an adequate framework for translating research findings into good clinical care of the individual. I further argue that articulating a methodology of clinical reasoning can have benefits for enhancing the training of future psychiatrists and the enhancement of expertise in all mental health practitioners.

While the primary audiences for this book are psychiatric residents and psychiatric educators, I believe it also can be of considerable value to psychiatrists and other mental health professionals at all stages of their careers who are interested in enhancing their expertise. I also hope it has value for those philosophers with an interest in the conceptual underpinnings of psychiatry. Philosophy of psychiatry has, in fact, become a vigorously growing subspecialty in applied philosophy. Yet in my experience this vibrant discipline is largely unknown to most psychiatrists. It is easy to think about philosophy as an abstract and self-referential intellectual activity with little relevance to real-world concerns. But at its most basic, philosophy is an attempt to rigorously examine the largely tacit conceptual assumptions that underlie any complex human activity. A field like psychiatry is especially rich in these, given that we deal with some of the most complex issues about human nature. We cannot help having underlying conceptual assumptions about these issues that guide our concrete behavior as we do our work. If we choose not to wrestle with them explicitly—that is, to do philosophy—we will act on them implicitly. And often our implicit concepts are to varying degrees erroneous and incomplete, resulting in seriously misguided decision-making and behavior in our clinical work. From the other side, I find that many philosophers of psychiatry know psychiatry primarily through its official and academic outputs, such as the DSM, published clinical trials, textbooks and treatment guidelines. As a result, it is easy for them to lack a nuanced understanding of the problems the practicing psychiatrist faces in making sense of their work with individual patients. I believe this book can address some of these limitations and thereby enhance the fruitful collaboration between philosophy and psychiatry.

The 1970s saw a major shift in psychiatry’s emphasis from a psychodynamic framework to that of biological psychiatry, accompanied by a marked change in the way psychiatric teaching was done and the relative status of various activities in departments of psychiatry. Previously, most prominent department chairpersons and senior faculty were known as excellent psychiatric educators, with reputations built from their high quality of instruction of residents and their ability to develop an understanding of the individual case, articulated as a formulation. Following this transition, most are known for their research activities, with the clinical education of residents often falling increasingly to junior faculty. In concert with this has been a de-emphasis of a highly individualized understanding of each patient, supplanted by evidence-based medicine (EBM) and practice. The primary task of clinicians becomes how to best apply results from research studies, especially randomized controlled trials (RCTs) conducted on groups of patients judged to be similar to the individual the clinician is treating, usually based on a shared diagnosis. While lip service is paid to the overall complexity of treating the individual and the need for clinical judgment, these tasks are seldom discussed with much depth or with any articulation of the processes and methods involved. They are dismissed as the “art” of medicine rather than the science. Clearly the science is more valued, as contemporary medicine, including psychiatry, has become deeply concerned with being seen as scientific.

But what is science? What is the boundary that distinguishes the scientific method from other, presumably inferior, forms of thinking? It is a common assumption that there cannot be a science of the individual, since science is thought to involve applying generalizations drawn from large numbers of cases studied in a rigorous manner and subjected to statistical analyses. One of the arguments of this book is that limiting science to this sort of aggregated data is based on a very flawed notion of what constitutes valid scientific method. The underlying assumption is that science is a system for discovering laws of nature that are ideally universal, or at least applicable across some specified domain, by a method of inference involving large numbers, followed by applying deductive reasoning from these generalizations to the individual case—the hypothetico-deductive method. For many contemporary philosophers of science, this view of science is problematic and far too narrow. Since at least the 1970s a new model of science has evolved that argues that general laws play a much less important role in scientific thinking than the development of models that attempt to map on certain concrete pieces of the real world. I argue that the result of this approach is quite compatible with the methods of clinical reasoning in psychiatry that involve generating models for individual patients, and that what the clinician does, when articulated and done with rigor, is not outside of science but is in fact an instance of its paradigmatic activity.

Since EBM offers itself as an adequate clinical method for the practicing psychiatrist, I begin with a detailed examination of that method as laid out in representative texts. I conclude that while EBM is very useful in helping the busy clinician access relevant literature, it falls profoundly short of providing a clinical method for treating an individual patient.

While EBM is very useful in helping the busy clinician access relevant literature, it falls profoundly short of providing a clinical method for treating an individual patient.

I next argue that while clinicians have been employing methods of clinical reasoning about individuals for a long time, it has been a largely tacit process. It is largely learned by imitation of more senior clinicians during training with little critical assessment of how that process works and how to assess its validity and epistemic status. This adds to its current devaluation. To address this, I initially present a detailed phenomenological examination of four cases from my own practice, attempting to articulate the method of thinking that led to decisions about treatment. From this I argue that a general methodology can be articulated. I make the case that this is not a methodology unique to me, but is in fact the kind of reasoning in which all experienced psychiatrists engage, regardless of their theoretical backgrounds. Briefly, the psychiatrist generates a model of the individual patient built around the patient’s specific goals. The model is composed of a number of elements, labeled propensities, that may have their roots in any domain, be it biological, psychological, or social. The clinician then posits a series of interactive pathways of influence between these propensities, ultimately leading to the patient’s goals. The clinician hypothesizes potential points of meaningful intervention that may bring about those goals. I call this a patterns of propensity model (POP). This goes far beyond the traditional biopsychosocial model that gives little guidance on how to choose and weigh individual elements for each patient. In POP models, each element and pathway must be uniquely characterized for that patient, with due consideration for the kinds of evidence that would support it, and the articulation of specific predictions for any intervention proposed at any point in the pattern. Thus, any POP model generates hypotheses that must be tested by concrete predictions and adjusted if necessary.

Such a model has many interesting features that I discuss in detail: Models are built around the patient’s goals once they have been made explicit and negotiated together with the psychiatrist, thereby placing the patient’s values and preferences center stage. The domain of a POP model is an individual patient but the search for recurrent patterns in the patient’s life is central to model construction and assessment. The elements in a POP model are propensities; namely, conditional assertions of likely experiences and behaviors of the patient in response to certain categories of events. Such elements can occur at various explanatory levels in varying degrees of abstractness, integrating multiple levels of explanation. Proposed pathways of influence between these elements are articulated, thereby suggesting points of potential therapeutic intervention. To be useful, models must generate concrete predictions, including predictive responses to specific therapeutic interventions. Furthermore, feedback as to the predictive accuracy needs to be clear and available within a reasonable timeframe. Model construction is a pragmatic activity to be judged by its success in generating interventions that meet the patient’s goals. Multiple models derived from a range of theoretical backgrounds might meet this standard. There is not one correct model. The complex picture of the individual patient that emerges in a POP model goes far beyond the patient’s diagnosis. The importance of the diagnosis varies from case to case, but typically it provides a stand-in for certain hypothesized elements and interactions that are not definable in the individual case, but are probable given the clinical presentation that the patient shares with groups of similar patients. Of all the potential models that could be generated about a given patient, only a subset are viable candidates. The process of pruning away unacceptable models is discussed. Pragmatic considerations, acceptability to the patient considering their values and preferences, and ethical concerns are among the factors that can disqualify a potential model before it is even applied. POP models are bidirectional in that they should provide a potential explanation for the patient’s current situation as well as predict how new perturbations, especially therapeutic interventions, will impact the patient’s future condition.

A ‘patterns of propensity model’: The clinician posits a series of interactive pathways of influence between propensities at various explanatory levels and hypothesizes potential points of meaningful intervention.

Having characterized in considerable detail the workings of POP models, I next take a detour into a selective history of theories about human reasoning and deliberation in general, culminating in a consideration of recent research on how experts reason in situations of incomplete information, high ambiguity, high stakes, and limited time. The methodology emerging from this work fits precisely with POP modeling. This work also has developed pedagogical principles for helping novices become experts in a wide range of fields. I use these insights to suggest how POP modeling can significantly enhance the clinical training of psychiatrists and other mental health professionals to speed their transition into being experts. Included in this proposal is a preliminary taxonomy of the various sources of evidence relevant to assessing each step in a POP model. These ideas relevant to training can be the basis of a future pedagogical research project in psychiatric education.

My final task is to consider the epistemic status of knowledge of the individual patient that emerges from the use of POP models. There has been a long-standing battle between advocates of “scientific” psychiatry, who argue that a true science of the individual is not possible so the clinician’s essential task is to somehow apply more generalized knowledge to the individual patient, and advocates of narrative and hermeneutic methods, who argue that the validity of our knowledge of particular individuals requires other epistemic standards apart from those associated with scientific method. I maintain that both opposing viewpoints share the same antiquated view of what constitutes scientific method. They still largely operate under the influence of the hypothetico-deductive notion of science prominent in the first half of the 20^th century grounded in logical positivism. Since at least the 1960s the inadequacy of this model and the search for alternatives has led to a very different understanding of how science operates in real life. A prominent school of current philosophy of science argues that science proceeds by scientists developing concrete models to characterize specific pieces of reality, as they attempt to solve concrete problems. Such models do not fully reflect reality in the sense of being true or false. Rather they are constructed to sufficiently resemble a piece of reality in order to solve the specific problem at hand. If they do that, they are good models. These philosophers frequently compare such models to maps. No map is a completely accurate picture of the piece of geographic reality to which it applies. What features have to be accurate, and to what degree, depend on the purpose for which the map is to be used. A roadmap, for instance, typically characterizes towns by circles and leaves empty space between roads. Neither of these reflect reality accurately. However, if the goal of the map is to get one from one place to another using roads, it is a perfectly good map. Hence, generating a model is a creative act in which the scientists must decide what elements to represent, how the pathways between them should be construed, and what level of accuracy is needed. None of that can be done without having a clear goal in mind. For the psychiatrist, that goal is the shared goal developed with the individual patient.

These contemporary philosophers of science also emphasize the view that there is no single or optimal scientific method. Rather, each discipline must develop a methodology that suits its own subject matter. What is common is that all these methods must be rigorously applied to generate testable predictions that promote self-correction and progress in each discipline. To apply the methodology appropriate to one scientific discipline to a different one is in essence a kind of category error that will greatly impede progress in the discipline. In the case of psychiatry, I argue that it is very important to separate the discipline of psychiatry proper, whose paradigmatic activity is understanding and treating the distress of an individual patient, from other informing disciplines. Often the results of the latter are extremely relevant and useful to psychiatric practice, but these disciplines are distinct from it and have their own appropriate methodologies. This is fairly obvious in the case of disciplines such as neuroscience, cognitive science, biology, experimental psychology, anthropology, and sociology. But there are two other disciplines that I believe are often mistaken for psychiatry proper, with the result that these methods are often viewed as the appropriate methodologies for psychiatry proper—psychopathology and therapeutics. The focus of psychopathology is typically symptoms and symptom complexes that characterize groups of patients, often but not always identified by a shared diagnosis. Therapeutics likewise studies aggregate responses to treatment interventions in groups of similar patients, again often identified by a shared diagnosis. The fact that researchers in psychopathology and therapeutics are often people trained as psychiatrists does not change the fact that they are essentially engaged in a different sort of activity when doing this research than when treating individual patients. The error of EBM is to assume that the optimal methodologies in these disciplines simply are the optimal methodologies in doing psychiatry proper. As a result, other kinds of reasoning about the individual patient tend to be viewed as epistemically inferior stopgap activities, to be gradually replaced as more “scientific” knowledge becomes available.

I believe that the widespread and rigorous application of POP modeling as an explicit scientific methodology for psychiatry proper enhances the epistemic status of our knowledge of and predictions about individual patients. It provides tools to enhance our critical assessment of our own work, as well as providing a framework for the training and more rapid development of expertise of those entering the field. Properly understood, this constitutes a true application of scientific method to psychiatric practice. The outcome for patients will be enhanced, as well as our own pride in what we do as psychiatrists.

Psychiatry at the Margins is a reader-supported publication. Subscribe here.

See also: