The Efficacy and Effectiveness of Marital and Family Therapy: A Perspective from Meta-Analysis

Article excerpt

Judging from our recent meta-analysis (Shadish et al., 1993), it is clear that marital and family therapy works. The literature supporting that conclusion is at least as strong as it is for other forms of psychotherapy. However, that felicitous conclusion--along with many other interesting findings--may have been obscured by a barrage of technical statistics in that article and the several related reports (Shadish, 1992; Shadish & Heinsman, in press; Shadish, Heinsman, & Ragsdale, 1994; Shadish & Sweeney, 1991). To remedy this, our primary goal in the present article is a simple one, to summarize the main findings of our various meta-analytic inquiries in a more user-friendly way. We start with a general summary of the main substantive results, then discuss a host of important methodological issues in the MFT outcome research literature, and conclude with some needs for future research.

Some Background About Meta-Analysis

Meta-analysis is just one form of literature review (Cooper & Hedges, 1994). Its special feature is that it quantifies the characteristics of that literature and then analyzes the resulting data statistically. Because outcome literatures nearly always include diverse outcome measures, a key step is the computation of an effect size that converts each outcome measure to a common metric (i.e., a common mean of zero and standard deviation of unity). Like most psychotherapy meta-analysts, we used the standardized mean difference statistic as our effect size index:

(Equation omitted)

where (Equation omitted) is the mean of the treatment group on the outcome measure, (Equation omitted) is the mean of the control group on the outcome measure, and s estimates the standard deviation of the numerator, usually using the pooled standard deviation (Hedges & Olkin, 1985). The resulting statistic is interpreted like any other standard score. That is, an effect size of d = .5 means that the treatment group did half a standard deviation better than the comparison group on the outcome measure. For example, if a marital therapy had an effect size of d = .5 on the Marital Adjustment Scale, and the standard deviation of the MAS was 16, then the marital therapy group did about 8 points better than the comparison group on the MAS. In addition, a meta-analysis typically develops quantitative codes for study characteristics (e.g., published or not), treatment characteristics (e.g., treatment orientation, therapist experience), and outcome variable characteristics (e.g., self- versus other-report). In data analysis, the meta-analysis can then examine the relationship between these latter study characteristics and study effect size.

In the present case, the literature being reviewed consisted of 163 randomized experiments (n = 62 marital; n = 101 family) of the effects of marital and family therapy (MFT) with distressed clients published through 1988. Of these, 71 compared MFT to an untreated control group, and 105 compared MFT to either another MFT or some other kind of psychotherapy. These numbers sum to more than 163 because some studies contained both kinds of comparisons. In any event, the fact that 163 randomized trials on MFT had been done by 1988 is itself the first important finding. To judge from other psychotherapy reviews (Begin & Garfield, 1994), only a few other forms of psychotherapy have been studied this often with such a rigorous methodology. Marital and family therapists can be justifiably proud of the scientific scrutiny to which these therapies have been subjected. Those studies are far from perfect, of course; in fact, throughout this article we will note several imperfections we think most worth remedying. Nonetheless, compared to the outcome literature for other forms of psychotherapy, the outcome literature for MFT is quite impressive. Moreover, a number of important studies have appeared since our cutoff date of 1988, improving this literature even further and in some cases contributing to literatures that may have been underdeveloped by 1988, for example, in psychodynamic MFT (Snyder, Wills, & Grady-Fletcher, 1991) or in therapies with some humanistic components (Johnson & Greenberg, 1995). …