Prescribable depression apps work, but the evidence is one app deep

Design

systematic review and meta-analysis of 19 randomised controlled trials

Population

4,078 adults with depression or generalised anxiety disorder, pooled mean age 38.7 years (SD 12.1), in German prescription-eligible programmes

Primary outcome

change in symptom severity versus control conditions

Effect

depression SMD −0.49 (95% CI −0.65 to −0.32) post-intervention and −0.35 (−0.46 to −0.29) at follow-up; generalised anxiety disorder evidence limited to two studies

Germany has prescribed digital mental health applications on its statutory health system since 2020, which makes it the one place where the question 'do prescribable apps actually work' can be answered from trials rather than marketing. A systematic review pooled 19 randomised trials covering 4,078 participants with a mean age of 38.7 years.

For depression the effect is real and moderate: a standardised mean difference of −0.49 (95% confidence interval −0.65 to −0.32) immediately after the intervention, across five applications and 17 comparisons. At follow-up the effect was smaller but held, −0.35 (−0.46 to −0.29), though that rests on a single application across four comparisons. For generalised anxiety disorder there were only two studies, which is not enough to say anything.

The important caveat is concentration. The authors say plainly that the evidence base is dominated by a small number of applications, particularly Deprexis, and should not be read as representative of every product on the German list. That is directly relevant wherever apps are being recommended without a national approval process — including India, where nothing equivalent to the German scheme exists and the products a patient will find are almost entirely untested. A moderate effect for one well-studied programme is not a licence to recommend whatever the patient downloads.

• Ask which specific application, not whether the patient is 'using an app' — the evidence does not transfer between products.
• Treat a digital programme as an adjunct with a review date, the same as any other intervention you start.
• Check engagement at follow-up: the trials measured what participants completed, not what they installed.
• Do not extrapolate the depression effect to generalised anxiety disorder — two studies is not an evidence base.
• Where no national approval scheme exists, look for the trial behind the specific product before recommending it.

Why it matters

'Digital therapy works' is the wrong conclusion to draw from this; 'one or two programmes work' is the right one.

Don't overread it

This measures a handful of German-approved programmes, chiefly one of them — it says nothing about the thousands of mental health apps a patient can download unprescribed.

The statistics, in plain English

A standardised mean difference of −0.49 is a moderate effect by convention, and the confidence interval (−0.65 to −0.32) sits entirely below zero, so the direction is not in doubt. The follow-up estimate of −0.35 looks more precise (−0.46 to −0.29) but is the weaker of the two: it comes from one application across four comparisons, and a narrow interval around a single product tells you about that product, not about digital therapy. The generalised anxiety disorder result is not reportable at all with two studies.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.