Patient retention in physical therapy: the evidence

Every practice-software blog carries the same statistic in some outfit or other: 70% of physical therapy patients never complete their plan of care. Sometimes it appears as “only 7 to 25% attend every recommended visit,” sometimes as “20% quit inside three visits.” Try to find the study. The trail runs through marketing pages citing each other and never reaches a journal.

That matters, because patients really do stop coming, and the peer-reviewed literature on why is genuinely useful. It is just quieter than the folklore. Here is what it supports.

Why do patients quit physical therapy?

For decades nobody had simply asked the patients who vanished. A 2025 study in Musculoskeletal Science and Practice finally did. Its authors interviewed 26 patients at a large regional health system who attended a PT evaluation, went 30 days without a follow-up visit, and had nothing else on the books. By the authors’ account, no previous study had examined that population from the patient’s side.

Five reasons came out of the interviews. Access problems led at 26.9%: getting to the clinic, fitting visits into a working life. Improving at an acceptable rate came next at 23.1%, tied with not seeing the value or deciding the exercises could be done alone. Another 15.4% had a different medical path take over, and 11.5% pointed at the relationship with the provider.

Twenty-six interviews is a small base, so treat those percentages as themes rather than benchmarks. But look at the shape. Roughly a quarter of these patients left because they felt better. Half left over things a clinic can influence: whether visits are bookable around a job, and whether anyone explained what weeks three through eight were for. The my-therapist-and-I-didn’t-click story that retention advice obsesses over was the smallest slice.

Why is there no reliable dropout rate?

Because the research cannot agree on what it is measuring. A British Journal of Sports Medicine review pulled 86 studies of therapeutic exercise for musculoskeletal pain and found that only 40% defined adherence at all. Just 12% used the same definition as any other study. The most common bar for “adherent” was completing 80 to 99% of the prescribed dose. By that math, a patient doing three-quarters of their exercises is a non-adherence statistic in one study and invisible in the next.

So published estimates swing, and the honest ones say so. A systematic review in Disability and Rehabilitation notes that non-adherence to home-based physical therapy “can reach 70%.” That is the closest thing to the folklore number the literature contains. Notice what it is: a ceiling, about home exercise programs specifically, not a completion rate for episodes of care. When someone quotes you a single dropout percentage with no definition attached, you are reading marketing.

Who is most likely to drop out?

The predictor research is far more consistent than the prevalence research. A systematic review in Manual Therapy covering 20 studies in outpatient musculoskeletal settings found strong evidence that poor adherence travels with low physical activity before treatment, low self-efficacy, depression, anxiety, helplessness, and weak social support. Two more factors made the strong-evidence list: a longer roster of perceived barriers, and pain that increases during exercise.

None of this is new. Physical Therapy published a Dutch study in 1993, built on 1,681 patient questionnaires from private-practice PT. Its top three correlates of noncompliance: the barriers patients ran into, the lack of positive feedback, and helplessness. Thirty years of research keeps circling the same two families of cause. One lives in the patient’s head: this hurts, and it isn’t working. The other stands between the patient and your door.

The second family includes money. In a Physical Therapy analysis of low back pain episodes across 80 clinics of a single US provider, higher out-of-pocket payment per visit predicted fewer visits per episode of care. The study cannot say why, but any owner can. A copay explained at the evaluation is a plan; the one that ambushes the patient at visit four is an exit. That puts the patient payment experience squarely on the retention list.

A dropout is not a no-show

A no-show is one broken appointment. A dropout is the patient who stops coming entirely, usually without telling anyone; the utilization literature calls it self-discharge. Keep the numbers separate, because the fixes differ: benchmarks, reminder-sequence evidence and fee policy live in our no-show rates guide.

The two problems share one bridge. Nobody has published a curve showing when PT patients drop out of an episode. What is published points at the schedule: in the PLOS One study of 444,995 PT patients, 73% missed at least one visit per episode, and the strongest predictor of missing the next visit was how many the patient had already cancelled. Dropout announces itself. The missed visit is the smoke; the quiet exit is the fire.

What actually improves adherence?

Less than the listicles imply, and the reviews are refreshingly blunt about it. The Cochrane review on exercise adherence in chronic musculoskeletal pain covered 42 trials with 8,243 participants and concluded that supervised or individualised exercise therapy and self-management techniques “may enhance” adherence. May. It also found the trials measured adherence inconsistently, and that where adherence did improve, the link to better clinical outcomes was conflicting.

The sharpest positive finding is small and specific. A BJSM meta-analysis of adherence interventions in older adults with low back pain or hip and knee osteoarthritis identified 3,899 studies and found nine eligible trials. The one effect that survived pooling was booster sessions, scheduled return visits with a physical therapist after the main program: a small-to-medium effect, on moderate-quality evidence. Individual trials backed motivational strategies and behavioural graded exercise.

Read the pattern in that. The interventions with evidence behind them are mostly more structured contact with the therapist. The trials never test copays or scheduling friction, because you cannot randomize a patient’s insurance plan. For those, you have the observational evidence above, and it points the same direction.

The levers a practice controls

Map the research onto a Tuesday at the front desk and four levers fall out.

Access first, because the never-returned patients ranked it first. Book the follow-ups before the patient leaves the evaluation. When one slips, let it be rebooked at 9pm from a couch instead of waiting on a phone line that closes at 5. The case for that window is in our self-scheduling data piece.

Cost clarity second. The out-of-pocket finding means the money conversation belongs before visit one, not on the first statement.

Feedback third. Sluijs flagged “lack of positive feedback” in 1993, and the not-seeing-value quitters in the 2025 interviews are the same patient a generation later. Retest the outcome measure and say the number out loud; a patient who hears their progress has a reason to protect visit nine. Booster sessions, the best-evidenced intervention on the list, are structured proof that contact with the therapist is the product.

And fourth, watch the smoke. The second cancellation is the moment somebody should pick up the phone, because the data says the third is already forming.

Orion’s share of this work is the unglamorous half. Reminder sequences, a recovery message when a visit falls through, and waitlist refill for the emptied slot all run as automations off the schedule. Eligibility runs at booking, so coverage surprises surface before the evaluation instead of at visit four. None of that substitutes for a therapist saying “your shoulder gained 20 degrees this month.” It makes sure logistics never get the first shot at ending the episode.

The folklore number says 70% of your patients will vanish, and implies you should feel bad about it. The literature says something you can use: patients leave when they cannot get in, cannot afford it, or cannot see it working. And they signal it one cancellation at a time. You cannot fix a made-up percentage. Tuesday is fixable.