Gnothi seauton, “Know Thyself” - inscribed at the Temple of Apollo at Delphi Medice, cura te ipsum, “Physician Heal Thyself” - Luke 4:23 (KJV)

Despite significant improvements in somatic, pharmacologic, and psychotherapeutic treatments during the last century, clinical outcomes in psychiatry over the last few decades have stagnated [1–3]. While many practitioners passively blame this on our inability to counter “treatment resistance,” and impatiently wait for new treatments that can, data shows us that the largest measure of the problem appears to be the quality of clinical practice we are employing [4,5].

Over half of current psychiatric treatments today have disappointing results: 20–60% end in failure or suboptimal outcomes [6]. For the last decade I have focused on describing the most effective methods of clinical practice [7–9], as well as the factors that allow persistent clinical error [10–12].  While a significant body of knowledge about these matters is available to clinicians, I also find that only a small number of us at all levels sufficiently appreciate the problem and feel an urgency to correct it.

When I presented a talk on “addressing treatment resistance” at a large national meeting, overflow rooms had to accommodate the very appreciative crowd. When I lectured with much of the same content one year later described as “Reducing Cognitive Errors in Psychiatric Practice,” only a dozen people showed up. It often takes me several sentences of introduction before clinicians realize I am talking about a problem with providers, not patients. When metacognitive solutions are mentioned, too often their eyes glaze over with the assumption that they already know what I’m going to say.

The Dunning-Kruger effect describes how the less we know, the more we think we know [13–15]. Our confidence peaks early and unrealistically in our learning curve and, unfortunately for many of our patients, limits learning and accurate self-assessment [16]. We, therefore, too often fail to sufficiently appreciate our inadequacies and to apply the necessary steps to correct or compensate for them. The correlation between reality and self-estimation of performance (ideally 1.00) has been measured at 0.47 for athletes, but only 0.17 for those performing medical related tasks [17].

This overconfidence may result from the prior knowledge that practitioners do have. In the logic of Bayesian inference, strong “priors” (existing knowledge that precedes the evaluation of new data), help us cautiously evaluate unexpected findings in the context of much previous data to the contrary. Though this method is quite helpful with clinical reasoning, it appears that lower performers persist in believing only their prior knowledge, as they lack the cognitive skills to properly evaluate it [18]. Maintaining strong priors without the proper use of logic impedes learning.

Fourth year psychiatry residents at several institutions often respond to my lectures with, “Wow, I’ve never heard that before.” This would indicate that metacognition (thinking about how we think) is not being taught sufficiently well enough during the 12–13 years of higher education a psychiatrist receives.

We can clearly describe with our existing technology common causes for our failures to accurately diagnose and prescribe more effective treatment plans, along with excellent solutions that we could all be currently employing. What is apparently missing is the self-knowledge that this is possible and necessary. Merely demonstrating a person’s errors through feedback is insufficient to improve their skills or provoke change [15]. A psychiatrist must learn metacognitive skills and procedures (including reflection, review, planning, use of counterfactuals, patience, and data management) in order to more accurately assess and be able to improve their performance [12]. These must be taught at all levels, from undergraduate to graduate to continuing education.

How can we reach the majority of clinicians who don’t realize we are the most common limiting factor in patient outcomes? In every field there are some people who are more effective than others, but we must not be content to allow our current broad range of practice quality in Psychiatry. Providers are the anchor leg of the clinical team that begins with basic research, is carried forward with clinical research, and that ends with clinical application. If we fail to correctly apply the knowledge handed to us, even the best and most appropriate advances never reach the patient [19].

Overcoming the high level of treatment failure in psychiatry is possible now, with our current tools.  Tertiary care centers report that in three quarters of the cases referred to them as “treatment resistant,” valid treatment methods have yet to be applied [20]. To overcome this, all practitioners must accept that our brains developed to survive and promote the species, not to solve complex medical problems. Allowing ourselves to operate mostly on default mechanisms, i.e., auto-pilot, will never improve outcomes.  We must remember how our minds work, not only those of our patients.

We must be able to tolerate uncertainty until higher level cognition can reduce it [12]. However, being aware we do not quickly know what to do can scare us into believing that we are powerless to help our patients. Then, we unconsciously return to immediate, easier, and less complex explanations. The ability to tolerate uncertainty is not only an excellent definition of mental health and maturity, but also essential behavior for practitioners; it is not the same as indecision [21].  When we admit uncertainty, we are open to broad differential diagnoses, and to myriad and complex solutions to the multi-faceted problems we seek to solve in healthcare.

When our brains default to rapid and repetitive diagnoses, quickly assigning the least complex solutions, this often results in a sense of satisfaction for the average clinician [22,23]. But then we are wrong half the time. When we are trained to improve our metacognitive processes, however, our self-assessments will become more accurate, and we will be able to appreciate and anticipate more of our limitations; this will prepare us to make fewer mistakes and more quickly correct the ones we do [13].

We must begin with the expectation of our fallibility and error, anticipating and seeking to uncover it, rather than suppressing the possibility. Intelligence is inversely associated with the speed of solving complex problems. The more gifted take longer to assess and diagnose [24]. Similarly, when faced with a treatment failure, they reexamine in minute detail their initial assessment and conclusions, rather than simply moving on to another, similar treatment option, or, worse, to a treatment for “treatment resistance.”

One of the primary reasons that the concept of “treatment resistance” has become so popular, and that the appearance of the term in psychiatric literature has increased by 75% over the last twenty years [6], is that it provides a simple, prepackaged explanation for clinical impasses. And, yet, treatment failures result from a wide variety of causes – no one is likely to be the same as another. If we cannot examine and challenge our assumptions, altering our models and problem solving to fit the new data available to us, how will we ever see improvement in clinical outcomes? Probably never. Labeling transcranial magnetic stimulation (TMS) or ketamine merely as indicated treatments for “treatment resistance” misunderstands not only the therapeutic landscape, but also how cognitive errors are created and resolved. The common question “What do you use to treat treatment resistant [insert psychiatric diagnosis]?” ignores the obvious oxymoron and masks the true question: Why did this treatment fail this time?

Practicing psychodynamically-oriented psychotherapy requires us as therapists to be constantly monitoring our own counter-transference, thoughtfully considering every response we make to a patient as our ongoing treatment plans evolve within our discourse. Similarly, our interactions with patients when providing pharmacological and somatic therapies also involve two humans trying to communicate effectively. To achieve that, and reduce error, we must be as assiduous in self-examination for these “medical” therapies as we are with the psychodynamic. We must constantly be under self-observation and evaluation, questioning every assessment and decision we make. Disappointing results will still arise; then we must repeat this process, arriving at new models of the problem that point to different and, hopefully, more effective solutions (that will then also be tested and examined). Otherwise, we continue as before with self-delusion and over-confidence. In all of medicine, only psychiatrists are formally trained in psychotherapy - we really have no excuse for avoiding self-assessment.

When the next treatment outcome is disappointing, or even suboptimal, begin by looking first at yourself, your process, and your work. Obtain multi-source feedback and reflect on all of this newly gathered information, before presuming to have found the definitive solution. Only though creative and competitive hypothesis iteration are we likely to improve our rate of problem solving [25,26]. 

We always owe the patients trusting and depending on us our best effort. To give them any less is a breach of their trust and our ethical duty. Physicians, first know thyself [27].