Stopping therapy early is common, and being asked what you want may help

Many people who start therapy stop before it is finished. Across hundreds of studies, about 1 in 5 people stop early. If you are in therapy, you may wonder whether stopping means something went wrong, or whether you are allowed to say what you want first. If you are a therapist, you may wonder why a client stopped coming. Both people want the same thing: therapy that fits, and an ending that leaves the client with a plan.

I want to be clear from the start that this is not a post about blame. Stopping early is not a failure by the client or by the therapist. Some people stop because they feel better. Others stop for practical reasons. What the research does suggest is that some early endings are linked to things a client and therapist can talk about openly, above all the client's own preferences.

The first post in this series looked at how therapists may not notice when a client is getting worse, and clients who get worse may simply stop coming. This post looks at stopping early itself.

What the research found

The largest summary of this question is a meta-analysis by Swift and Greenberg (2012). A meta-analysis pools the results of many studies. They combined 669 studies, with 83,834 adult clients, and found that, on a weighted average, 19.7% of clients stopped early (95% confidence interval 18.7% to 20.7%). In their words, that is "about 1 in every 5 clients". Researchers call this dropout, and studies define it in different ways.

A narrower meta-analysis, by Cooper and Conklin (2015), looked only at randomized trials of individual therapy for depression. Across 54 trials and 5,852 participants, the rate of stopping early was 19.9% at the study level and 17.5% in the therapy conditions. That is close to the broader figure.

Here, a careful reading matters. About 1 in 5 is an average across hundreds of studies that defined stopping early in very different ways. It is not a prediction for any one clinic, therapist or person.

Why the number moves

If about 1 in 5 is an average, the next question is what sits underneath it. The main answer is the definition. Studies count stopping early in many different ways, and Swift and Greenberg (2012) found that the definition used changed the rate. When a therapist judged whether a client had stopped early, the average rate was 37.6%. When studies counted anyone who did not reach a set number of sessions, it was 18.3%, and when they counted anyone who did not complete a treatment protocol, it was 18.4%. The authors note that the higher figure may partly reflect the kind of studies that relied on therapist judgment, which more often looked at everyday practice. The type of study also changed the rate: 17% in efficacy studies, which are tightly controlled trials, against 26% in effectiveness studies, which look at everyday practice.

An older review had put the overall rate at 47%. Swift and Greenberg (2012) found no sign that rates had fallen over time. They suggest the gap may come from the different studies each review included, the larger share of tightly controlled trials in their own review, and different statistical methods.

This fits with a further point, which I offer as an inference. Trials select their participants and support them to take part, so stopping early is probably more common in everyday practice than in trials. One review of routine-care CBT for depression found that about 1 in 4 clients stopped early (Hans & Hiller, 2013), though studies define it differently, so this is not a direct comparison. Cooper and Conklin's figures come from trials only, and so they may understate what happens in ordinary care.

Shifting numbers may be one reason the topic gets little attention. Therapists may also never learn why a client left, so the reasons stay out of view. Public conversation tends to focus on getting into therapy, not on staying or leaving well. As a rough signal only: in one week on X (English posts, retweets excluded, exact phrases), phrases about stopping, such as "quit therapy" or "ghosted my therapist", appeared in 62 posts, against 3,029 posts urging people to go to therapy or find a therapist. Exact phrases miss paraphrases, and Instagram, TikTok, YouTube and Reddit were not measured, so coverage there is unmeasured, not zero.

Who stops more, without blame

Averages also hide differences between groups, so it helps to look at who stops more often. Swift and Greenberg (2012) found that stopping early was more common for younger clients, for clients with a personality disorder or eating disorder diagnosis, and for clients seen by trainee clinicians. In the depression trials, Cooper and Conklin (2015) found higher rates in conditions with more patients who also had a personality disorder.

These are patterns across groups. They do not predict what any one person will do, and they do not mean that clients in these groups care less about their therapy. A higher rate may say something about how well services fit a group. In the full paper, Swift and Greenberg (2012) offer some possible reasons, which they present as speculation, not findings. For clients with a personality or eating disorder diagnosis, they point to the slow progress often seen in treatment. For trainee clinicians, whose clients stopped early more often (26.6%, against 17.2% for experienced clinicians), they suggest that therapists become more responsive and more focused on the relationship as they gain experience. They also note that trainees more often see younger clients in university clinics, where rates were higher too, so these patterns overlap. They offer no specific reason for the higher rate among younger clients. A broader review by Barrett et al. (2008) adds two points: clients and therapists often expect therapy to last different lengths of time, and therapists are less likely to know why a client left when the reason is a negative one. Swift and Greenberg (2012) are clear that this is not about blaming clients: treatment, setting and therapist factors also predicted who stopped early.

It is just as useful to note what did not change the rate. In Swift and Greenberg's (2012) analysis, the rate did not differ by therapy orientation, or by individual versus group format. In Cooper and Conklin's (2015) depression trials, orientation did not explain the differences either. This does not mean that the type of therapy never matters, only that it did not explain the rates in these studies.

What is linked to staying: preferences

If who stops is partly a question of fit, the next finding speaks to fit directly. Swift et al. (2018) pooled 53 studies with more than 16,000 clients. Accommodating clients' preferences was linked to fewer people stopping early (OR = 1.79) and better outcomes (d = 0.28), compared with giving a treatment the client did not prefer. Preferences can be about things such as the type of therapy, the format or how often sessions happen.

The two numbers need a plain explanation. OR stands for odds ratio. It compares the odds of stopping early in one group with the odds in another, and here a value above 1 favours the clients whose preferences were accommodated. An odds ratio is not a percentage, so 1.79 cannot be read as a percentage drop in people stopping early. The second number, d = 0.28, is a standardized difference in outcomes between the two groups. By common rules of thumb, it is a small effect.

The link held across groups: the effect did not differ by the type of preference, or by client age, gender or education.

It is important to note what this study does not show. It shows a link, not proof that accommodating preferences keeps any given client in therapy. The effect varied with study design, timing, type of outcome measure and diagnosis. Many studies compared a preferred treatment with a non-preferred one, rather than testing a conversation about preferences in ordinary practice.

What researchers suggest

Preferences are one piece of a larger picture. In the discussion of their meta-analysis, Swift and Greenberg (2012) list strategies that other researchers had proposed for reducing early endings. These include talking early about what therapy involves and how long it may take, addressing motivation, repairing strains in the therapeutic relationship, using feedback on progress, addressing client preferences, and offering time-limited therapy. Barrett et al. (2008) review several of these in more detail. They note that weaker alliances are linked to more early endings, and that therapists who respond to conflict in a flexible, non-defensive and direct way can improve the relationship.

These are suggestions drawn from the literature. They are not each proven to reduce early endings in ordinary practice, and they have not been tested as a package.

If you are in therapy

For clients, the practical message is that you have more say than you may think. Stopping early is common, and it is not a personal failure. You are allowed to say what you prefer: the type of therapy, how often you meet, or whether sessions are online or in person. Research links having your preferences accommodated to fewer people stopping early and slightly better results.

If therapy does not feel right, it can help to say so before you stop. A change of approach may help. Saying "this isn't working" or "I want to stop" is useful information for your therapist, not a failure on your part.

If you raise a concern, a good therapist should be able to take it in. Research suggests that therapists who respond to concerns in a flexible, non-defensive and direct way can strengthen the relationship (Barrett et al., 2008).

In my view, part of a clinician's work is being able to take in feedback. Even criticism can be useful information for shaping better interventions and treatment plans. If a clinician struggles to hear your feedback, they may not be the right fit for you, as it can be a sign that they are finding it hard to manage their own internal reactions. The flip side is that feedback delivered poorly, or used to avoid real clinical issues, can cause similar problems.

If you decide to stop, a short final session or message can help you leave with a plan, and may make it easier to come back later.

Some clinics ask you to fill in short questionnaires. Your answers can show you both whether things are moving.

Questions you can take to your therapist

• "How will we talk about my preferences for type of therapy, format, or how often we meet?"
• "If I want to stop, how can we plan a short final conversation so I leave with a clear next step?"

These questions are not a test. They make preferences and endings something you plan together.

If you are a therapist

For clinicians, the same findings point to a few things you may wish to consider. This is general information, not advice about any particular client or setting.

• Ask early about preferences: the type of therapy, the format, how often to meet, what the client hopes for and what they want to avoid.
• Explain early what change usually looks like, including that it is not always steady, so that a hard week is less likely to end therapy.
• Talk about endings at the start. Agree that if the client wants to stop, a short final conversation is welcome.
• Expect strains in the relationship, and name them early. Reviews describe ruptures and repairs as a common part of therapy, clients often keep negative reactions to themselves, and responding in a non-defensive, direct way can improve the relationship (Barrett et al., 2008).
• Check progress from more than one angle. A fuller picture can come from a quick scaling question (for example, 0 to 10), a standard questionnaire such as the PHQ-9 or GAD-7, the client's own account in their own words, and, only with the client's explicit consent, observations from people in the client's life outside therapy. The PHQ-9 and GAD-7 are screening and tracking tools; they do not diagnose. Noticing a stall early gives you a chance to adjust before the client leaves. Swift and Greenberg (2012) list using feedback on progress among proposed strategies, but no source here shows that this combination reduces early endings.
• Keep an eye on your own pattern of early endings over time, especially with younger clients. This is an inference from the findings above, not a tested step.

My own view is that clinicians should get used to ruptures, and do so early. Beating around the bush and avoiding difficult topics so as not to cause a rupture, even when asking or pushing is clinically relevant, is poor clinical work. The beauty of ruptures is that they can be repaired, and repair in a clinical relationship is extremely valuable because it models a healthy relationship for clients.

Consent and ethics

Clients can say what they prefer, and they can decide to end therapy. Saying so is useful information, not a failure. Outside reports from family or friends need the client's explicit consent and careful handling of privacy. This post is general information only, not individual clinical advice.

Limitations

Each source has limits worth stating plainly.

• Swift and Greenberg (2012): stopping early was defined in many ways, and the definition changed the rate. The analysis mixed tightly controlled trials with studies of everyday practice, and it covers adult clients only. The reasons the authors give for group differences are their own suggestions and were not tested.
• Cooper and Conklin (2015): some possible predictors, such as socioeconomic status, could not be tested because trials rarely report them. It includes randomized trials only, so it probably understates stopping early in routine care (an inference).
• Hans and Hiller (2013): routine-care CBT for adult depression only, from nonrandomized studies, with stopping early defined differently across studies, so it can't be compared directly with the trial figures.
• Swift et al. (2018): the effect varied by study design, timing, type of outcome measure and diagnosis. Many studies compared preferred with non-preferred treatments rather than testing conversations about preferences in ordinary practice (an inference).
• Barrett et al. (2008): a narrative review, not a meta-analysis, and many of the studies it covers are older. Its suggestions come from the literature and are not each tested as ways to reduce early endings.
• Swift and Greenberg (2012) and Barrett et al. (2008) were read in full. For Cooper and Conklin (2015), Hans and Hiller (2013) and Swift et al. (2018), the figures come from the published abstracts.
• About 1 in 5 is an average across very different definitions of stopping early, not a rate for any one clinic or therapist.
• The X counts cover one week, English posts and exact phrases only. They are a rough signal of coverage, not a census, and other platforms were not measured.
• Two passages marked as my own view are clinical opinion, not research findings.

Leaving well is part of therapy too

Taken together, the research suggests that stopping early is common, that its rate depends a great deal on how it is measured, and that being asked what you want is linked to staying. None of this makes an early ending a failure. It makes stopping something a client and therapist can talk about openly, from the first session to the last.

If you are in crisis in Canada, call or text 9-8-8 any time.

General information, not clinical advice.

References