Beyond Diversion Rates: The Metrics That Actually Describe In-Home Acute Care Performance

Programs delivering acute care at home are most often evaluated on a single number, usually the share of encounters that avoided an emergency department visit. That metric is easy to produce and easy to misread.

A more complete evaluation requires several measures read together, each addressing a different way a favorable headline figure can be produced without corresponding performance.

Why Is a Single Metric Insufficient?

Any single operational metric can be optimized independently of the outcome it is meant to represent. A diversion rate can be raised by narrowing acceptance criteria rather than by improving clinical capability.

Reading several metrics together makes selective optimization visible. Improving one at the expense of another shows up in the combination.

What Does Acuity Distribution Reveal?

Acuity distribution describes how sick the accepted patients actually were. A program reporting excellent disposition on a low-acuity case mix has demonstrated something modest.

The population detail in bridging the access gap for complex patients reports acuity alongside chronic condition burden precisely because disposition figures are uninterpretable without that context, and a favorable rate in a healthy population means very little. Acuity is the first variable that makes a diversion rate meaningful.

Programs handling higher acuity with comparable disposition are demonstrating more. The comparison only works when both figures are available.

Why Does Population Complexity Matter?

Population complexity indicates whether a program is serving the patients most likely to generate avoidable utilization. Chronic condition counts and complexity flags describe that directly.

A program serving relatively healthy members may perform well operationally while affecting very little of the cost base plans care about. Complexity connects performance to financial relevance.

What Does Response Time Add?

Response time determines whether the program was available at the moment the alternative decision was being made. Slow response produces patients who left for an emergency department before the visit occurred.

Response figures broken out by acuity tier are more informative than blended averages. A strong overall average can conceal poor performance on the highest acuity cases.

Why Is Follow-Up Utilization the Hardest Measure?

Follow-up utilization asks what happened over the subsequent weeks rather than on the day of the visit. It requires claims data and an appropriate comparison group.

This is the measure that distinguishes deferred utilization from avoided utilization. A patient who presented to an emergency department three days later was not diverted.

What Should a Complete Evaluation Include?

A defensible assessment generally covers several dimensions together:

• Disposition rate with acuity distribution attached
• Population complexity measures including chronic condition counts
• Response times reported by acuity tier
• Emergency department and admission utilization across a follow-up window
• Referral source mix and unique referring provider counts
• Patient and clinician experience measures

No single item establishes program value. The combination is considerably harder to construct selectively than any one figure.

How Do Experience Measures Fit In?

Patient experience measures indicate whether the model will retain the population it serves. A program with strong clinical outcomes and poor experience will see declining voluntary utilization.

The Massachusetts program reported a Net Promoter Score above eighty and a 4.9 out of 5 rating across more than 1,600 reviews. In a directly accessible model, patient willingness to call again is an operational input.

Why Does Clinician Experience Belong on the List?

Field clinician retention determines whether accumulated operational familiarity persists. Programs with high turnover repeatedly rebuild capability they had already developed.

Reported clinician satisfaction, such as the 8.8 out of 10 post-shift figure from the same program, is a leading indicator of workforce stability. It predicts future performance more than it describes current output.

What Makes These Metrics Hard to Game Together?

Narrowing acceptance criteria to raise disposition rates lowers acuity distribution and reduces volume. Loosening criteria to raise volume tends to lower disposition.

Reporting that presents these dimensions together, as the 2025 care-in-place outcomes data does by pairing encounter volume, acuity tiers, population complexity, and utilization reductions, makes it considerably harder to present one figure favorably at the expense of another. The tension between metrics is what gives the combination its value.

That built-in tension is the reason to require the full set. Any individual number can be produced by design.

What Should Plans Request?

Plans should request the full metric set at contracting rather than accepting a headline figure. Defining measurement terms upfront avoids disputes about interpretation later.

Programs confident in their performance generally supply this readily. Reluctance to provide context is itself informative.

How Long Should a Follow-Up Window Be?

Thirty days is the most common follow-up window for utilization analysis, aligning with readmission measures used elsewhere in healthcare. Shorter windows risk crediting deferral as avoidance.

Some analyses use both a seven-day and a thirty-day window to distinguish immediate deferral from durable change. Reporting both is more informative than choosing one.

What Comparison Group Makes Sense?

Comparison against a matched group of similar members who did not use the program is stronger than comparison against the same patients’ prior history alone. Prior-period comparisons are vulnerable to regression toward the mean.

Matching on chronic condition burden, prior utilization, and age produces a more defensible comparison. Plans with claims data are generally positioned to construct this.

Diversion rates describe one dimension of a multidimensional program. Read alone, they support conclusions the underlying data may not justify.

For anyone evaluating these programs, the practical step is to require acuity, complexity, response time, and follow-up utilization alongside any disposition figure. The combination is what makes the headline number interpretable.