This is a forest plot. It is looking at normalized or standardized mean differences of improvement between computerized CBT and controls. The controls are not specified. And if you look at it, you get a significant improvement if you do computerized CBT.
This, to a busy clinician or clinical team, sounds very attractive. Computers are cheap, work 24 hours a day, and go nowhere. They can be accessed via the internet. If we can deliver effective therapy quickly and cheaply, then this may allow for more people to be treated and (even, perhaps) lead to talking therapy being accessed before the use of medications. It all sounds fairly good.
But there are some problems. At present CBT or cognitive behaviour therapy is generally given by psychologists or people who have had a fair amount of training subsequent to getting a clinical degree (in nursing, social work or medicine). These people do not come cheap. And to tell them that their painfully acquired skills may be matched by a silicon chip is… a challenge to an insult. Many therapists say a computer cannot do as well as they can. and this paper does not help: the two comparisons were “treatment as usual” — where anything could be done, or a wait list control.
One of the basic rules in setting up clinical trials is to try to make everything you can even. If you are doing individual therapy in the trial group, give the same amount of therapy time in the control group. A more correct design would involve two sham or placebo inputs — a sham computerized therapy, and a sham individual therapy. Now, if we accept (and there is fairly good evidence to do this) that CBT is generally better than supportive therapy, you could randomize to: supportive therapy and Computerized CBT OR CBT and say, computerized activity scheduling and problem solving therapy (which have been used in, say adjustment disorder). And, although you cannot blind the therapist or the client, you can make sure that the person doing the ratign scales does not know what therapy the person is getting.
So let’s look for RCTs comparing one with the other…. but I cannof find any that have been published. Doing a trial like the one we have outlined os too expensive, too difficult, and too politically fraught for most people.
However, it is needed. We need to ensure our most skilled and trained clinicians and therapists work with the most disabled. We need to have some form of effective treatment in primary care that is less people intensive. And we need to use less medications. The computer should help break down the guild, so that effective help can be given to many, but the intervention has to work. It needs to dismantle the ideologies about individual therapy being the gold stnadard, and that cathedral of “best practice” — that has limited or no trial data that supports it.