Why clinical supervision doesn't improve client outcomes, and five ways to make it better

Since Freud, supervision has been an essential part of training new therapists. The junior learns from the master and gradually becomes an expert in their own right. It is so pervasive that it’s built into university training and almost every ongoing professional development program.

But does it work? Recent research suggests that supervision, as we currently do it, has close to zero impact on client outcomes.

The evidence

The first hint came from a large study by Tony Rousmaniere and colleagues, published in 2014: more than 6,500 client outcomes, 175 trainees and 25 supervisors. Using hierarchical linear modelling, they asked how much of the variation in outcomes depended on which supervisor the therapist had. The answer: less than 1%.

Science relies on replication. In 2020, Whipple and colleagues repeated the exercise at a training centre in the Rocky Mountains, with 3,030 clients, 80 therapists and 39 supervisors. Their result: again, less than 1%.

Why it doesn’t work

Imagine coaching a Premier League football team like this. Once a week you meet the players for a chat about the weekend’s game, which you never watched. You discuss tactics, then send them out for the next match without anyone ever lacing up a pair of boots to work on what needs improving.

That’s supervision. It runs on an Oracle of Delphi model: the supplicant visits the supervisor and hopes to receive wisdom. Every other field where people actually get better runs on a coaching model.

Five ways to make it better

1. Supervise the right cases. Therapists are poor judges of which cases need supervision. Talking about cases that are already going well can’t add value. Talking about cases that are impossibly stuck probably won’t either. Use the data to find the cases that should be improving but aren’t.

2. Supervise what actually matters. Supervision tends to focus on formulation, diagnosis and treatment planning in whatever model supervisor and supervisee happen to favour, CBT, ACT or otherwise. The research suggests the model matters about as much as placebo. We wouldn’t spend supervision on the placebo effect, because it would be pointless. Focus instead on the parts of the relationship that matter, above all goal consensus. When did you and your supervisor last discuss how you make sure you and your client agree on the goals of therapy?

3. Stop ignoring the elephant in the room. Around 30% of clients drop out before session three. Supervision should actively monitor dropout rates and put strategies in place to improve them.

4. Observe directly. We can’t rely on supervisees to tell us whether they’re doing a good job. Without live supervision or video review, there’s little point supervising something you’ve only been told about.

5. Skills and drills. Once you identify a weakness, build a protocol to train improvement: role-play, verbal scripting, even talking to a mirror. There’s no point telling someone to go away and do something differently without telling them exactly what to practise between now and the next time they see a client.

A different model

These problems are why, at STAP, we have been leading the move to a different model of supervision: one built on deliberate practice and client outcome data, where the supervisor works more like a coach than an oracle. If you supervise and want to try it, it’s what I teach in STAP’s masterclass on coaching with outcome data.