Outcomes & measurement-based care

Measurement-based care: how the Outcomes page works

See where the numbers on the Outcomes page come from, what each tile means, and how to read the trend, stratification, and CMS coverage views.

Updated
Updated
Reading time
6 min read
Tags
outcomesmbcphq-9gad-7mips
A look at Outcomes
Measurement-based care

Every screener, one dashboard

PHQ-9, GAD-7, AUDIT - the scores your patients already submit, rolled into one live view of your panel

Measurement-based care (MBC) means asking patients to fill out validated screeners — PHQ-9, GAD-7, AUDIT, WHODAS — on a regular cadence and using the scores to steer treatment. The problem MBC solves is well known: clinicians are excellent at noticing acute change in front of them, and worse at noticing slow drift across a 200-patient panel. Without numbers you cannot see who is plateauing on a med, which modality is working, or whether you would pass a CMS quality measure. The Outcomes page turns the responses your patients already submit into a single live dashboard you can answer those questions from in under a minute.

Where the data comes from

Every screener your clinic has enabled (Establishment → Surveys) gets sent to assigned patients on its cadence. When a patient completes a screener — in the portal, by SMS link, or in the room with a clinician — the response and its computed score land in the patient chart and roll up into the establishment-wide Outcomes page automatically. There is no separate "publish" step. If a screener is sent and a patient answers, you see it on Outcomes the next time you load the page.

Picking a window

Window:30d90d180d365d
312 patients enrolled in MBC
Window selector - pick a 30, 90, 180, or 365-day lookback

The window selector at the top of the page controls every number below it. 30d is your week-over-week pulse — useful when you have just changed protocols. 90d (the default) is the clinical sweet spot: long enough that random noise washes out, short enough that a med change two months ago still shows up. 180d and 365d are for board reports and payer attestations. The "patients enrolled in MBC" pill on the right is a count of distinct patients with at least one active screener assignment, regardless of whether they have responded yet.

Reading a screener tile

PHQ-9 (depression)

9-item Patient Health Questionnaire. Lower is better.

0 - 27

9.4/ 27

Mean score, last 90 days

-2.7vs prior

Enrolled

184

Response rate

82%

Reliable improvement

38%

A single screener tile - headline mean, trend sparkline, response stats

Each enabled screener gets its own tile. The big number is the mean score across every response in the window — for PHQ-9 lower is better, for WHODAS lower is better, etc. The arrow next to it compares the current window mean to the prior window of the same length, so a 90d view compares the last 90 days against the 90 days before that. The sparkline shows weekly buckets of the mean across the window so you can spot trend shape, not just direction. Enrolled is how many patients have an active assignment for that screener; response rate is the percent of those enrolled who completed at least once in the window; reliable improvement is the percent of patients (with both a current AND prior response) whose score dropped by at least the screener-specific reliable-change threshold (≈20% of score range, floor of 3 points).

Stratification — what is actually moving the needle

PHQ-9 mean change · stratified

Compare modalities to spot which interventions are moving outcomes most

Virtual visit

n = 86

-2.8

In-person

n = 112

-2.1

No active treatment

n = 14

0.0
Stratification - same screener, split by visit modality so you can spot what is moving outcomes

A mean by itself does not tell you *why* outcomes are moving. The PHQ-9 stratification chart splits the same panel by visit modality (telehealth vs in-person) using each patient's predominant visit type in the window. A bar to the left of center means that segment is improving (negative delta = lower depression score). A bar to the right means the segment is getting worse. Tiny n? Treat the bar as directional, not statistical. We surface modality because that is the dimension every practice can act on this quarter; richer stratifications (med + therapy vs med-only, by provider, by diagnosis) are on the roadmap.

CMS / MIPS quality coverage

CMS2v12 coverage

312 of 348 eligible patients (89.7%) have a documented PHQ-9 in the last 12 months - meeting the MIPS quality measure benchmark of 75%.

Meeting
CMS2v12 coverage - documented PHQ-9 in the last 12 months versus eligible patients

CMS2v12 ("Depression Screening: PHQ-9") is one of the highest-leverage MIPS quality measures for behavioral-health-adjacent practices. The coverage card answers a single question: of the patients who actually came in over the last 12 months, what fraction has a documented PHQ-9 in that same window? The MIPS benchmark is 75%. We render a Meeting badge in green when you are at or above the benchmark and a neutral badge when you are below it, so you can see at a glance whether you would pass the measure if your reporting period closed today.

Who can see this page

Outcomes is gated by the managesurveys role. By default that means providers, clinical staff, and practice managers (anyone who can already assign a screener). Front-desk and billing-only roles do not see the page in the sidebar. If you are an admin and want to grant access to a non-default role, edit the role under Establishment → Roles and check managesurveys. Outcomes is also a paid feature — if your plan does not include it, the page renders as a blurred preview behind a See plans overlay so you can decide whether to upgrade without leaving the app.

Troubleshooting common views

Empty state ("No screeners enrolled yet") — your establishment has no enabled practice_surveys. Go to Establishment → Surveys and enable PHQ-9 (or whichever screener fits your specialty) to start collecting data. Tile shows enrolled but 0% response rate — assignments exist but no patient has completed the screener in the current window. Send a reminder from the patient chart, or shorten the window to confirm whether responses just happened to fall outside it. Stratification empty — you do not yet have paired responses (one in the prior window, one in the current window) within each modality. Wait one more cadence cycle.

Frequently asked questions

Is this view real-time?
It is recomputed on every page load, so a response submitted ten seconds ago is included the next time you refresh.
How is "reliable improvement" calculated?
Per patient: take the latest score in the current window and subtract it from the earliest score in the prior window of equal length. Count the patient as reliably improved if that drop is at least max(score_max × 0.2, 3) points. The percentage shown is improved patients / patients with both windows of data.
Why are scores rendered as `mean / 10` in the headline?
Purely visual normalization so tiles for different scales (PHQ-9 0–27, GAD-7 0–21, AUDIT 0–40) all read at a glance. The underlying mean and the trend sparkline use the screener's native scale.
Will this export?
Export PDF at the top of the page renders the same view as a one-page provider-friendly report. CSV export and HL7 v2 outbound are on the roadmap; if you need raw data today, the GraphQL outcomes(window_days:) query returns the same shape.
Why telehealth-vs-in-person and not med-vs-therapy?
We only stratify on dimensions we can derive from the chart with full confidence today. Visit modality is on every CalendarEvent. Treatment plans (med + therapy, med only, therapy only) require a structured care-plan model that is on the roadmap; we will not stratify on a dimension you have not explicitly recorded, because that produces misleading lines.