How to measure home exercise adherence: diaries, EARS and counts

Ask a patient how the home exercises went and you will get an answer. Whether it matches what happened at home is a question the research has not solved. Here are the main ways to measure home exercise adherence, what each one actually records, and how to combine them. For why patients skip exercises in the first place, see why patients skip home exercises.

Why home exercise adherence is hard to measure

The developers of the Exercise Adherence Rating Scale open their abstract bluntly: "There is no gold standard for measuring adherence to prescribed home exercise." Nobody watches at home, so every measure relies on the patient's memory or on a device.

Self-report depends on recall, and many patients want to come across as good patients. Newman-Beinart and colleagues name "inaccurate recall and self-presentation bias", along with a lack of standardization, as limits of the commonly used self-report diaries. Devices avoid recall, but they only record what they can see.

A systematic review in BMJ Open by Bollen and colleagues found 58 studies using 61 different self-report measures of adherence to unsupervised home exercise: 29 questionnaires, 29 logs, two visual analog scales and one tally counter. Only two measures scored positively on any psychometric property, and most had no reported validity or reliability testing at all. A later review in chronic low back pain by Uzawa and Davis found that none of the eight included studies had validated self-reported frequency, duration or intensity of home exercise.

None of this research is specific to hand therapy. Read it as general rehabilitation evidence.

Exercise diaries: what a hidden sensor showed

Diaries are the default: cheap, and they double as a reminder. But how close do they come to the truth?

Nicolson and colleagues tested this in 54 adults with chronic knee pain on a 12-week home strengthening program. Participants kept a paper diary and, every two weeks, rated their own adherence on an 11-point scale. An accelerometer was concealed in the ankle cuff weight they used for the exercises.

The diaries overestimated adherence: a median of 220 exercises in the diaries against 176 recorded by the accelerometer. Diary and sensor were moderately correlated (r = 0.52), but agreement varied widely between participants. The rating scale did worse, with poor to fair correlations with the sensor (mean r between 0.23 and 0.39) and less than acceptable test-retest reliability. The authors concluded that diaries showed "questionable validity".

So a diary is useful, but on average it runs high, by an amount that differs per patient.

The Exercise Adherence Rating Scale (EARS)

EARS was developed at King's College London in people with chronic low back pain. Starting from a 16-item version completed by 224 patients, the authors arrived at six items that formed a single scale, with acceptable internal consistency and high test-retest reliability.

The six short statements are rated on a five-point agree-disagree scale, for example "I do my exercises as often as recommended" and "I forget to do my exercises". The total runs from 0 to 24; higher means more adherent.

The limits are worth knowing:

• Ceiling effect. In a Swedish adaptation in patients after shoulder surgery, Söderberg and colleagues found a ceiling effect on all six items, which limits how well the scale shows change over time.
• Still self-report. The same authors note the risk that patients overestimate and give the answers they think are acceptable. EARS structures self-report; it does not remove its bias.
• Developed in back pain. It was not built for hand conditions.
• Not a gold standard. Its own developers say none exists.

Objective counts: sensors, app logs and camera reps

Objective measures replace memory with a record, but each records something narrower than "did the patient do the program".

• Wearable sensors, like the hidden accelerometer above, record movement of the object they sit in. In that study the sensor was built into the exercise weight for the purpose.
• App logs record that an app was opened, or that a patient tapped "done". That is a timestamp, not a movement.
• Camera-counted reps are hand movements detected in front of the camera during a session and matched against the prescribed exercise. They show that a session happened and how many reps were counted in it.

None of them records exercise done away from the device, how well each repetition was performed, or whether the program suits the patient.

What a webcam can and can't tell you about hand movement

Consumer hand tracking such as Google's MediaPipe finds the hand in each video frame and estimates the positions of points on it. That makes it possible to tell an open hand from a fist, or to see a thumb touch a fingertip, and so to count reps.

It is not a clinical measurement. Sprague and colleagues compared MediaPipe and the Leap Motion Controller 2 with a marker-based lab motion capture system while 11 people did six hand tasks at three speeds. Both camera systems captured finger joint motion in real time, but less accurately than lab motion capture. MediaPipe had higher error and lower correlation than Leap Motion, and its error grew at faster speeds. The authors conclude that the level of error "may not be acceptable for certain applications".

The practical reading:

• A webcam count can tell you whether a session took place and how many reps were counted.
• It cannot measure finger joints for clinical purposes, see exercise done away from the camera, or judge quality.

A practical mix for your clinic

Uzawa and Davis point to the WHO recommendation to use more than one adherence measure at the same time. Each method has a different blind spot. A workable mix:

1. Ask. Use EARS, or its questions informally, at check-in. It captures the patient's own view.
2. Log. A simple diary gives the patient a record and a prompt. Expect it to run high.
3. Count where you can. A sensor or camera count shows the sessions and reps the device registered. It misses anything done away from the device, and it can miscount.
4. Talk about the gaps. When the diary says daily and the count says twice a week, that difference is the conversation. Maybe the patient exercises away from the laptop; maybe the diary was filled in afterwards. Neither is a reason to blame anyone.

Measuring is a means to that conversation. None of these studies shows that measuring adherence more precisely improves results on its own. For what the evidence says about helping patients actually do the exercises, see reminders, apps and gamification.