This bit of the conversation is about the transition from Era 2 to Era 3, and how machine learning helps us to think about that transition. But before I can get to that I need to set out two more bits of context. The first, set out in this input, is to describe an evidence based programme called Functional Family Therapy, which I will use as a case example to describe where we may be going.
Functional Family Therapy is designed for young people with conduct disorders. It trains a qualified psychologist or social work in a set of additional methods. The therapist meets with the young person and his family members. The therapist helps the family members to re-frame their problems, and find solutions that work for all.
There have been over 30 trials of Functional Family Therapy meaning it is, by every existing definition, an evidence based programme. The Washington State Institute of Public Policy, generally regarded as the safest set of hands when it comes to cost-benefit analysis, reckon that investment in Functional Family Therapy will generate $7,098 of economic benefits for every participant enrolled. To use the U.S. shorthand, there isn’t much not to like about the programme.
It has an additional feature. Around 6,000 practitioners trained in Functional Family Therapy around the world keep their records on software called Care4. They use it to record:
- administrative data such as when appointments are due, if they are missed et cetera
- measures, not unlike the Strengths and Difficulties Questionnaire, of the young person’s mental health and the family functioning
- fidelity measures checking on whether the practitioners are doing the programme as planned, with facility of supervisor feedback, and
- ad hoc notes on how the practitioner, young person and family members feel at each session, what the practitioner saw et cetera.
The data from the 30 or so trials, largely before and after measures of intervention and treatment as usual group are good for era 2. The Care4 data offer some opportunities for era3.