The Clinician’s Illusion: A Powerful Source of Bias Three Decades On.

Share Button

30 years ago, two statistically minded psychologists published a brief but elegant paper describing a phenomenon they called “the clinician’s illusion”. Patricia and Jacob Cohen were a highly distinguished academic couple who are better known for co-authoring a dense book on statistical technique that every research psychologist surely has on their shelves.

The Cohens were concerned with the judgements of clinicians, which even at their best are necessarily distant from the ideal of the precision of a carefully conducted statistical study. They noticed that clinicians tended to overestimate the severity of serious mental health problems, and reasoned there may be some systematic bias going on.

As an example they noted that clinicians who talk about Schizophrenia seem to use a heuristic – the “rule of thirds” – which says approximately one third of people with the diagnosis will get better, one third get worse and one third will stay the same. The Cohen’s noted that existing follow up research actually suggested a picture rather more optimistic than this*.

The clinician’s illusion is thus the pessimistic bias clinicians show in their judgements of outcome, and it arises from the assumption that their personal experience accurately tracks reality. Specifically, clinicians are prone to base their reality-judgements on the characteristics of the population to which they are exposed rather than the population as a whole.

Imagine the average psychiatric inpatient unit. On any given day, it is likely to be populated by a higher proportion of the people who have been using the service over a long period of time than the people who have only brief contact. Extrapolating over the course of several months, a small handful of people with very long term problems are likely to become more mentally available to the clinician than the people who used the service for a short time and then moved on.

As Patricia and Jacob Cohen put it, there is a direct linear relation between length of stay and number of days on a ward, so people who spend more time on the ward are in the clinician’s consciousness for a greater number of days. Length of stay is thus proving to be an inherent confound in the clinician’s judgement about prognosis.

This illusion, like so many of the biases and heuristics that fool us every day, is deceptively simple. It might seem that once we know about it, we can make necessary adjustments and assume a more objective perspective. but our experiences are powerful and persuasive, and can override even the most firmly established statistical evidence.

Why bother re-visiting the Clinician’s Illusion? Over the past three decades, our understanding of mental health problems has improved somewhat, but the natural human tendency toward bias continues. We still hear the rule of thirds (and other simplistic heuristics) mentioned by clinicians in practice. The reality behind such figures may be obscured by clinician over-confidence and the sheer complexity of the literature-overwhelming for anyone trying to dig into it for the first time.

It is not just explicit biases (like the rule of thirds) which distort our thinking; a clinician’s attitude and expectations may also be impacted by their implicit judgements. Where this translates into unwarranted pessimism, the result may be that an unduly negative mood is fostered, impacting adversely on care.

This is no call for limitless optimism, or for an “anyone can do anything” approach to mental health. Presumably want realism from our clinicians as much as encouragement. Such realism requires a disciplined attention to research and the hard work of understanding it. As a patient or carer you should expect nothing less.


* It should be noted that the Cohens were drawing not only on US samples, which, prior to 1980’s DSM-III used a more liberal set of diagnostic criteria for Schizophrenia and might thereby be expected to be more optimistic. They base their judgement on three studies, one from the US, one from Germany and one from Switzerland. A more recent systematic review (from 2006) of 37 studies suggested a “good” outcome in 42% of cases of First Episode Psychosis and “poor” outcome in 27%, leaving 31% with an outcome somewhere in between.


Share Button

About the author

I am a PhD student and trainee clinical psychologist based in New York. My research is on the mechanisms of Auditory Hallucinations and the social epidemiology of psychosis. In addition to this I am also interested in the philosophical and conceptual problems around diagnosis and the history of Schizophrenia. Follow me on Twitter: @huwtube