Here you'll find reviews of the latest research on "what works" in therapy.  We scour the literature as journals are published, take out all of the math, researcher-speak, and complicated statistics, and then translate the data into clear, practical, and empirically-supported principles for clinical practice.  Previous research reviews can be found in the Talkingcure.com archives.

Barry and Scott: You are all wet about this early change stuff!

Scott and Barry—you have been gallivanting around the globe asserting that therapists should monitor outcome with clients because early change is a robust predictor of ultimate change, but new research shows that early change does not necessarily predict later change. In fact, Lambert’s recent research suggests that the more and faster a client deteriorates in the first two weeks of therapy, the faster the client will improve in the further course of therapy. People are saying that the old adage that things have to get worse before they get better is really true after all.

Yes, Lambert's article "What is the predictive value of responses to psychotherapy for its future course?" published in May, 2006 in Psychotherapy Research does report data which suggest that early progress in therapy is not necessarily a good predictor of overall progress in therapy—for that relatively small portion of clients who stay in therapy and do not drop out despite poor results.This is a big qualifier that is often left out of descriptions of this study.  The study only speaks to a minority of clients and is not characteristic of most clients.

 Furthermore, like most research articles, you gotta read 'em to see if what is being touted as findings are actually in the study or are no more than wishful thinking supporting someone’s ideological beliefs (and we don’t mean Lambert, but rather those who seem to have a religious zeal to undermine the predictive value of early change). First, the most glaring aspect that undermines any generalization from this study to typical outpatient settings is that, of three separate data sets and analyses, the first two were inpatient samples! Therefore, the first two samples were a totally captive audience, a luxury that outpatient settings do not have—so the dropout issue, the biggest critique of this type of prediction, was sidestepped altogether (recall that drop out rates in the US exceed 47%). The prediction of a positive outcome from earlier deterioration or a lack of change is based on the notion that if you get people to stay in treatment, even if they are not doing well, they may eventually do well (this is often called the Random Walk phenomenon). This is likely true but we know that that almost half will drop out, which of course, an inpatient sample cannot reflect.    

 Another confound is the whole hospitalization experience. Was the initial deterioration due to the addition of medication and/or adjustment to being hospitalized, and therefore an artifact of such settings? Who knows?

 The third analysis, the only outpatient sample, really tells the tale and accurately depicts the issues at hand.  Of 1012 clients, only 260 fit the criteria set for analysis. Even though the average number of sessions in this setting was 3, they set the criteria for the study as those clients who completed at least 5 sessions and remained over the clinical cutoff (over the cutoff means a clinical population on the measure used in the study). So these were people who were not changing but were willing to hang out more than 4 sessions—a skewed group to be sure and one not representative of the majority of those who sought outpatient treatment at this setting or those that seek treatment elsewhere. The overwhelming majority of clients will not hang on like this 25% group, and even those will drop out at nearly a 50% rate thereafter. This small group will ultimately fit the pattern of the random walk where some will improve despite early deterioration or no change. However, the majority of clients will fit the robust finding that early change predicts ultimate change, especially when drop outs are considered.