Monday, 31 August 2015

Rethinking Therapist Drift

One of the strangest ways I have ever been evaluated in my career was while delivering a psychological intervention in a prison. My job was to stand in front of a room of young offenders and facilitate group discussions of their "thinking skills". The prison service has a clear idea of what this should look like, indeed so clear that every session I did was filmed. Auditors could then check that I was sticking to the protocol. 

This remarkable surveillance was an attempt to minimize what is commonly called "therapist drift", the process whereby a therapist ostensibly delivering an evidence-based therapy winds up doing something else instead. Psychotherapies are not easy things to administer, and in the face of diverse people and problems it's easy to see how one might end up straying from the guidelines outlined in therapy delivery manuals. But if you claim to be doing, say CBT when you aren't in fact doing anything of the sort, you might not get the same results.

For the most part, therapist drift is regarded as a bad thing. This makes sense (at least, it's internally consistent). Drift is a problem for people who are trying to research a therapy (because they don't end up testing what they intend to) and it's potentially a problem for people who are trying to deliver a therapy (because rather than delivering something that has been demonstrated to be effective, they do something which is not).

But there is a strand of thought in clinical psychology and psychotherapy that maintains suspicion about the notions of "adherence" and "drift", and of evidence-based therapy altogether. Critics of this stripe view evidence-based-approaches as overly rigid and formulaic, too focused on technique at the expense of relationship. There is some good quality criticism of manualized therapy (here's a good example), but also much exaggeration about manualized therapies making the process "robotic" (as though a set of instructive principles were incompatible with being human). 

Recently I looked into the topic (in a very non-systematic way) to see what research had found about the importance of adherence and drift. Most studies that have been done (and it is surprising how few there are that focus specifically on drift) seem to support the contention that "drifting" can lead to less impressive outcomes. However, one study had an intriguing result.

Examining CBT for panic disorder, researchers (Jonathan Huppert and colleagues) took measures of patient motivation (rated by the therapists) and adherence to the therapy manual (rated by listening to audiotapes of sessions). Perhaps counter-intuitively, the researchers found that among highly motivated patients, the therapist adherence did not have much impact on outcome (look at the graph and you can see the blue line only slopes upward a small amount; this difference was not found to be significant). However, among less motivated patients, adherence was associated with worse outcomes than drift.*

One possible explanation for this (one that Huppert and colleagues themselves suggest) is that patients with low motivation present an extra degree of complexity which cannot be adequately addressed by staying within the set protocol. An experienced therapist will depart from the standard protocol to address in some way the low motivation, before continuing with the planned process. Under this interpretation, the therapists who show the greater adherence with the "low-motivation" patients are paradoxically failing to do something with the low adherence therapists are succeeding at it. Although they are moving beyond the purview of the manual, it seems misguided to call this "drift". This squares with the extended discussion by Drew Westen and colleagues on the more tendentious implications of therapy manuals:

...manualization commits researchers to an assumption that is only appropriate for a limited range of treatments, namely that therapy is something done to a patient—a process in which the therapist applies interventions— rather than a transactional process in which patient and therapist collaborate. (p.639)

And yet still some idea of therapist "drift" seems important. Unless we believe there is no value to specific training for psychologists and psychotherapists, we want to have some reasonably defined sense of what we're up to; some sense of what it looks like to do the job properly. It is any deviation from this that can reasonably be considered drift. In other words: in the space between conforming, robot-like, to a predetermined protocol and doing whatever the hell you want, there lies a knowable range of skills which we ideally would want to adhere to. That set of skills is what constitutes being a good psychologist. This definition extends the realm of evidence based practice some way beyond the parameters of individual evidence based treatments

*Though the authors note: "it is important to keep in mind that even when therapists were rated as less adherent, adherence was still rated as “good” or better, suggesting that therapists did not cease doing CBT for panic, but likely incorporated other strategies into their armamentarium" p.202


  1. Thanks for the post – I think it’s important, and generous, for clinicians to write openly of their own experiences.
    In this case, the scenario in question couldn’t be more Foucauldian if it tried. You have prisoners being subjected to discipline, surveillance and micro-policing (“thinking skills”) by a clinician who is also being subjected to a lesser version of the same. This practice is clearly political/ideological rather than therapeutic in any sense of the term.
    In my view, matters are only clarified somewhat by the study on ‘drift’. Because, like most ‘scientific’ psychologists reject philosophy, these researchers speak of ‘motivation’ in meaningless, abstract terms, when we ought to recognise the Spinozist point that ‘’motivation’ (i.e. desire) is always a motivation for something. In this case, the ‘something’ for which motivation is low is adherence to the clinician’s imperatives.
    From these, I think there are at least two conclusions to be drawn. First, the manualised model of treatment is a bit stupid, even in terms of the medical model. Even the crudest medical interventions draw individual distinctions which are lacking in this one-size-fits-all, recipe book approach.
    Second, all of this is clearly an exercise in authoritarian discipline, even if couched in pseudo-medical or benevolent terms. The ‘treatment’ at stake is strictly speaking for the benefit of those who administer it, and not the recipients. The latter are regarded, literally, as measurable objects to be gazed and intervened upon. The only point at which this exercise even approaches something remotely ethical is in the ‘drift’, in which the imperatives are temporarily discarded, and the objects of intervention have a vague chance of being responded to as subjects. For the clinician to outsource his own personal responsibility to protocols is bad faith all round, and a stunted and technocratic (if popular) notion of what it is that psychologists actually do.

  2. Thanks David, always glad to have your comments here!
    I'll just say two things in response.

    1. I completely agree about the aptness of Foucault. This was a technological update of the panopticon. I never knew which, if any, of my sessions were viewed (only a percentage ever are), but it certainly had a disciplining effect! I left the prison service severely disappointed with that work. To the extent that prisons can be made more useful, I think it is through jobs training and education rather than interventions of this sort (psychotherapy proper could play a role, but not so focused on bending people to the will of the institution). As an extra tidbit of information, I can't resist sharing that the prison governor imposed a rule (motivated by a poorly thought through form of political correctness) that all the inmates be called "offenders" rather than "prisoners" because the latter carries stigma. It seemed never to occur to the powers that be that "offender" carries the stigma of having committed a crime, which some portion of them (those on remand pending trial, or those wrongly convicted) may not have done. "Prisoner" is a blunt statement of irrefutable fact.

    2. I think our views on the value of measurement in psychotherapy are all but incommensurable. No psychologist worth their salt believes that measures "get at" the final underlying thing; rather every measure is a kind of proxy; an attempt to roughly gauge the extent to which something might be present. Just as educational outcomes like grades, or financial indicators like salary could only ever stand in for some more important (and ultimately "unmeasurable") concept of how "successful" or "well-off" people are after, say, high school or college. Is this all an exercise in "authoritarian discipline"? Is it all " for the benefit of those who administer it, and not the recipients"? Does a treatment protocol preclude even "a vague chance of being responded to as subjects"? No, no and no in my view; none of those follow from any reasonable premise. That is all just so much entertaining rhetoric designed to shore up the omniscient benevolence of the outsider analyst figure, who is beyond the reproach or assessment of anyone who dares to question their process.

  3. I agree that there’s much that is irreconcilable in our views, though perhaps our aims are not so far apart.
    Arguably, clinical psychology is in the midst of an ethical and epistemological crisis. At least, there is some evidence for this. Whether this argument stands or falls, however, is independent of the merits of psychoanalysis. Foucault's critique, for example, remains as pertinent as ever, yet he was no sympathiser of analysis.

    The problem of unaccountable authority is rather inverse to your characterisation of omniscient, authoritarian analysts and epistemologically humble psychometricians/empiricists. The latter group grounds its procedures on claims of scientificity, all of which turn out to be themselves based on false assumptions. By contrast, the analyst calls himself into question continually, because his own analysis is the precondition of his practice. He at least makes an effort to understand what it is that impels him; the empiricist is not troubled by such questions. Moreover, the analyst does not deploy suggestion, or try to coach correct thinking (or ‘motivation’) and the like. In empirical psychology, all thinking and ethics can be (and are) outsourced to the pseudo-guarantee of a protocol, which licenses both ethical evasion, and any abusive practice whatsoever, as long as some committee or other has given a rubber-stamp. If you’re looking for evidence, you could do worse than check out the fans of the New Atheists for proof that scientism is perfectly compatible with authoritarianism.

  4. I think we might agree on this. Humans of any theoretical stripe are prone to believe that they are doing it the right way; prone that is to a certain over-belief in their own beneficence and superior judgement. I don't know how you feel about the fact that some portion of the evidence for this comes from experimental work in psychology, but no matter, as history and literature, philosophy and psychoanalysis all point to the same conclusion: we are endlessly self deluding. You and I seem also to agree some mechanism to work against that is required. I actually think you're right that analytic self-questioning is imperative for therapists; I endorse that whole heartedly and employ it (I hope) in my own work. I ALSO endorse (because I think even self-analysis has its limits) the use of guidelines to practice, and quantitative outcome measures, because I think these have amazing power to surprise us in our judgement of "how it's going". There is no final objective point from which one can "know" EXACTLY how it's going, but a kind of triangulation of observations seems reasonable. I find your invocation of the New Atheists to be a spot on example of scientism (the name of Jerry Coyne's website "Evolution is True" tells you a lot about the desire for a particular sort of certainty), but I also think you rather wishfully seek to extend it to *empirical psychology per se*, which is a mistaken and uncharitable leap. All of my most productive uncertainties come from people I consider heroes of empirical psychology.