01
Insights
How should an ACO or health plan measure a transitional care management program?
Most measurement stops at whether a billable service occurred. The harder questions are what was found, what was resolved, and whether any of it changed what happened next.
The Short Answer
What does the standard measure actually tell you?
The common metric is the share of eligible discharges that resulted in a billed transitional care management service. It is a real number and it is worth tracking. It is also an activity measure, and it stops well short of the question a risk-bearing organization is actually asking.
A billed episode confirms that contact was made inside two business days, that a face-to-face visit occurred inside the required interval, and that the documentation supported the claim. It does not say what was wrong with the patient, whether anyone fixed it, or whether the fix held.
For an organization paid on utilization, that gap is the whole point. Two programs with identical completion rates can produce entirely different readmission curves, and completion rate cannot distinguish them.
The Visibility Gap
What claims data cannot tell you
Claims describe events. They are reliable about admissions, discharges, emergency department visits and readmissions. They are silent about everything that happened between those events.
A claims extract will not say who contacted the patient or how long after discharge. It will not say whether the medication list at home matched the discharge orders, what discrepancies were found, or which of them were resolved rather than merely noted. It will not say whether the primary care physician was told, whether a follow-up appointment was made, or whether the patient attended it.
Most consequentially, it will not say why a patient was never reached. A member who could not be contacted because the discharge record carried a disconnected number is a different operational problem from one who declined, and both are invisible in a claims file.
This is not a deficiency in claims data. It is what claims data is for. The point is that a program measured only through claims is measured only on its outputs.
Three Layers
What should be measured instead?
Useful measurement has three layers, and the third is the one most programs cannot produce.
The first is engagement. Time from discharge to first successful contact, contact attempts per patient, the share of patients reached at all, and the reasons the rest were not. Reasons matter because they are actionable in a way that a success rate is not.
The second is clinical. Medication reconciliations completed, discrepancies identified, discrepancies resolved, prescriber escalations raised, and patient-reported barriers recorded. The distinction between identified and resolved is the one that separates a documentation exercise from a clinical one.
The third is outcome. Emergency department use and readmission at seven, fourteen and thirty days, follow-up visit completion, and program completion by practice. Measured alone these are lagging indicators; measured alongside the first two layers they become interpretable.
Connecting the Layers
How do you connect an intervention to what happened next?
The useful question is not whether a program can prove causation in the way a trial would. It is whether the organization can follow a single patient from discharge through to outcome without the trail breaking.
That trail has a specific shape: the patient was discharged, contact was attempted and on which day, a medication discrepancy was identified, an intervention was made, the prescriber was notified, a follow-up visit was completed, and no readmission followed within thirty days. Each link is a fact someone recorded. The value is in their being recorded in the same place, about the same patient, in order.
Once that exists across a population, comparisons become possible that claims alone cannot support. Patients reached within twenty-four hours can be compared with those reached on day two. Episodes where a medication problem was resolved can be compared with those where one was identified and left open. Practices can be compared with each other on the same basis.
None of this requires a randomized design. It requires that the intervention layer be captured at all, which today it usually is not.
Vendor Diligence
What should you ask a vendor to produce?
Four requests separate a program that can be measured from one that can only be reported on.
Ask for reasons, not just rates. A vendor that can report why members were not reached is operating a program; one that reports only a contact percentage is operating a call list.
Ask for identified versus resolved as separate figures. If the two are reported as one number, they are not being distinguished in the work either.
Ask for comparability across practices. An organization with multiple practices needs the same definitions applied consistently, or the variation it sees will be measurement noise rather than performance.
Ask how intervention data reaches your own environment. Data that lives only in a vendor dashboard cannot be joined to claims, and joining it to claims is the entire point.
Common Questions
Related questions
Is TCM completion rate a bad metric?
No, it is a necessary one and a poor summary. It confirms the service happened and supports the claim. It cannot distinguish a program that resolves medication problems from one that completes calls, and those two produce different utilization.
Why not just measure readmissions?
Because readmission is a lagging, noisy, heavily risk-adjusted measure, and a program can move it without anyone being able to say which part of the program did so. Outcome measures answer whether something worked; intervention data answers what to do more of.
Does this require the vendor's software?
No, but it does require that intervention data be exportable in a form that can be joined to your claims. The practical test is whether you can answer a question the dashboard was not designed to answer.
How is this different from care management reporting we already get?
Most care management reporting describes volume: calls made, patients enrolled, tasks closed. The distinction here is between activity performed and problems resolved, and then between problems resolved and what happened afterward.
What is the minimum viable version?
Time from discharge to first successful contact, reasons for failed contact, and medication problems identified versus resolved, all reported per practice. Those four alone will usually reveal more variation than an organization expects.
Related reading
Talk it through with someone who runs these programs
Praventa operates care management programs with its own clinical team and licenses the same platform to practices running them in-house. Describe your situation and we will tell you which model fits.