Measuring the work
How to measure health coaching outcomes that count
The short answer
Measuring health coaching outcomes takes three layers: engagement, whether the between-visit work happened; patient-reported change, whether the experience shifted; and clinical change, whether the physiology followed. Capture engagement weekly and patient-reported measures at baseline and weeks four, eight, and twelve. Set your change threshold before the first score arrives.
The essentials
Key takeaways
The core of this article, in a few lines.
- Between-visit support needs three separate measurement layers: engagement, patient-reported change, and clinical change. Collapsing them into one number hides real progress.
- Capture engagement weekly and patient-reported measures at baseline and weeks four, eight, and twelve. Anything collected earlier is process data, not outcome data.
- Behavioral sleep work often makes daytime symptoms worse for one to three weeks before sleep consolidates, so a week-three score reads a working program as a failing one.
- The literature has not settled how much change counts. Reported Insomnia Severity Index thresholds run from roughly 4 to 8 points, so a practice has to state its own number in advance.
- A defensible measurement plan is three instruments and a schedule, written down and dated before the first patient enrolls.
Three layers, one schedule, and a number you set first
Knowing how to measure health coaching outcomes inside a practice comes down to three layers: engagement, whether the between-visit work happened; patient-reported change, whether the patient's experience shifted; and clinical change, whether the physiology followed. Capture engagement weekly, patient-reported measures at baseline and weeks four, eight, and twelve, and clinical markers on your normal schedule. Then decide what size of change counts as change before the first score arrives.
Most practices skip that last step. They add a coach, run it a quarter, then evaluate it the way they would evaluate a supplement: by asking the patient whether they feel better. That is an impression, and impressions do not survive partner meetings.
Behavioral support is in fact the most measurable service in a practice. Behavior is observed weekly, not inferred quarterly. The data exists. What is missing is a structure to hold it, the same gap behind why patients do not follow treatment plans.
Why your visit notes cannot answer 'is the coaching working?'
Picture the review: twelve weeks of the new service line, a stack of visit notes, and a partner asking whether it is actually working. You have anecdotes. Three patients are sleeping better. One dropped out. What you do not have is a column anyone else can check. Visit notes were never built to measure health coaching outcomes. They record what happened in the room, which is a different job.
Worse, the few numbers you have point the wrong way. Insomnia Severity Index scores from the week-three visit are higher than at intake, and nobody can say whether the program is failing or doing exactly what an effective protocol does at week three.
That ambiguity is not a program failure. It is a measurement design failure, correctable upstream.
Three layers of measurement, and why practices collapse them into one
The most common error when practices try to measure health coaching outcomes is treating the service as one thing with one number. It is three, and each answers a different question.
Layer one is engagement. Did the work happen? Sessions attended, diary days completed, messages returned, weeks active. This is the fastest signal you get. If engagement is low, nothing downstream means anything, and the problem is delivery design rather than clinical effect. This is where improving patient adherence between visits becomes numbers instead of impressions.
Layer two is patient-reported change. Did the experience shift? Use a validated instrument. The Insomnia Severity Index is short, sensitive to treatment effects, and has published interpretation data behind it (Morin et al., 2011). Add diary-derived sleep efficiency, time asleep divided by time in bed, in the Consensus Sleep Diary format so your data stays comparable to the literature (Carney et al., 2012).
Layer three is clinical change. Did the physiology follow? Blood pressure, HbA1c, weight, mood scales, whatever you already track. It moves last and slowest, which is why it cannot be your only measure, and it links the behavioral work to what improves when sleep improves.
Why coaching outcomes look worse before they look better
Sleep restriction therapy, the active ingredient in most behavioral insomnia work, deliberately reduces time in bed to consolidate sleep. In adults with insomnia disorder, Kyle et al. (2014) found objective total sleep time dropped by an average of 91 minutes on night one, 78 minutes on night eight, and 69 minutes on night 22. Daytime sleepiness rose and measured vigilance was impaired across those first three weeks. Both returned to baseline by three months.
Read that curve at week three and you will classify a working intervention as a failing one. A follow-up landing at week two or three samples the trough of a trajectory, not the result of one. Learned skills produce trajectories, and one that starts flat or dips is not a failure. Treat anything collected before week four as engagement data.
How much change counts as a real outcome?
This is the number nobody publishes, and the field has not settled it. Wu et al. (2024) reviewed 81 insomnia randomized trials and found only 31, or 38.3%, reported any minimal important change or minimal clinically important difference at all. Among those that did, Insomnia Severity Index thresholds scattered: the most common minimal important change was a 6-point drop, followed by 8-point and 7-point values, while the most common clinically important difference was 4 points. Pittsburgh Sleep Quality Index thresholds clustered near 3.
Morin et al. (2011) separately found a change of roughly 8 points corresponded to moderate improvement judged by an independent assessor. So no single correct threshold exists, and any source handing you one without a range overstates the evidence. What your practice needs is its own stated number, written down before collection starts.
A defensible default: a 6-point drop is meaningful improvement, a 4-point drop is a real but partial response, anything under 4 points is noise. Put that in the protocol and the week-twelve review becomes a comparison against a standard, not a negotiation.
The uncomfortable step most practices skip
Name your threshold before you collect anything. It removes your ability to be pleased by whatever the data turns out to show. Learning how to measure health coaching outcomes honestly means accepting in advance that the program might not clear your bar.
It also means deciding what you do at each result. Clears the bar, expand it. If engagement is high but scores are flat, the protocol needs review. Low engagement points at the delivery model instead. That branching is where cadence matters as much as instruments, the finding behind whether contact frequency improves patient outcomes.
What not to do when you measure health coaching outcomes
Five errors that make good between-visit work look like nothing happened.
- 01
Do not measure only at the next scheduled visit
Visit intervals are set by clinic logistics, not by the biology of behavior change, and they land in the dip.
- 02
Do not average all patients into one number
A group mean hides two responders and three who never engaged. Report engagement bands first, then outcomes within each.
- 03
Do not treat wearable sleep data as your outcome measure
Consumer devices help patients self-monitor, but estimated sleep stages are not validated endpoints.
- 04
Do not add instruments you will not sustain
Two well-collected measures beat six abandoned ones, and measurement that stops in month two looks like evidence of failure.
- 05
Do not conflate satisfaction with outcome
Patients can like a coach whose scores are not moving.
A measurement plan you can start Monday
You do not need a research department. Three instruments and a schedule will do it.
- 01
Write the protocol page
One page: the three layers, the instruments, the collection points, your thresholds. Date it. This is what you show at the partner meeting.
- 02
Set the baseline at enrollment
Insomnia Severity Index, seven days of diary, the clinical markers you already follow. No baseline, no outcome.
- 03
Track engagement weekly
Sessions completed, diary days returned, weeks active. A partner should send this unasked, and a remote model makes it easier to capture, as a telehealth behavioral support model shows.
- 04
Re-measure at weeks four, eight, and twelve
Same instruments, same wording, same delivery method. Consistency of administration matters more than sophistication.
- 05
Review against the written threshold
Compare each patient to the number you set in advance, then read the panel by engagement band.
Progress over precision
None of this is unusually rigorous. Fortney et al. (2017) found only 17.9% of psychiatrists and 11.1% of psychologists routinely administer symptom rating scales, and without them providers detect deterioration in about one in five patients who worsen. Choosing to measure health coaching outcomes at all puts you ahead of most of the field.
At Clementine, the embedded coaching service line we run inside practices returns this dataset to the chart by design: weekly engagement, Insomnia Severity Index at set review points, diary-derived sleep efficiency. The clinician keeps the diagnosis and the plan. Delivery and reporting sit with the behavioral layer.
Two questions sit outside this article by design: whether the service line pays for itself, which is the economics briefing for this work, and who should deliver it, which is the staffing comparison.
A plan does not have to be perfect to be defensible. Knowing how to measure health coaching outcomes well enough to defend them takes three things: write the plan down, run it on schedule, read it against a threshold you named in advance. Do that for a quarter and the week-twelve conversation stops being a debate.
A measurement plan only matters if the work it measures actually happens between appointments. For why sound plans come apart there, start here.
Frequently asked questions
How often should I collect patient-reported outcomes between visits?
Is the Insomnia Severity Index or the Pittsburgh Sleep Quality Index better for tracking sleep in practice?
Can I use wearable sleep data as an outcome measure?
How long should I wait before expecting to see a change in patient outcomes?
How do I tell a real improvement from normal week-to-week variation?
Sources & clinical context
- 1.Carney, C. E., Buysse, D. J., Ancoli-Israel, S., Edinger, J. D., Krystal, A. D., Lichstein, K. L., & Morin, C. M. (2012). The Consensus Sleep Diary: Standardizing prospective sleep self-monitoring. Sleep, 35(2), 287-302.
- 2.Fortney, J. C., et al. (2017). A tipping point for measurement-based care. Psychiatric Services.
- 3.Kyle, S. D., Miller, C. B., Rogers, Z., Siriwardena, A. N., MacMahon, K. M., & Espie, C. A. (2014). Sleep restriction therapy for insomnia is associated with reduced objective total sleep time, increased daytime somnolence, and objectively impaired vigilance: Implications for the clinical management of insomnia disorder. Sleep, 37(2), 229-237.
- 4.Morin, C. M., Belleville, G., Belanger, L., & Ivers, H. (2011). The Insomnia Severity Index: Psychometric indicators to detect insomnia cases and evaluate treatment response. Sleep, 34(5), 601-608.
- 5.Wu, J., et al. (2024). The gap between statistical and clinical significance: Time to pay attention to clinical relevance in patient-reported outcome measures of insomnia. BMC Medical Research Methodology, 24, 177.
