The hidden gap between a positive PHQ-9 and treatment

A positive PHQ-9 is a data point, not a treatment. In a six-practice UCSF academic health system, 30% of adults who screened positive had no antidepressant order, no referral, and no documented follow-up eight weeks later (Garcia et al., JAMA Internal Medicine). In a 607,730-patient Veterans Health Administration cohort, only 32% of people newly identified with depression received timely follow-up within three months (Leung et al., JAMA Network Open). Most practices could not tell you their own version of that number, because almost nobody measures it.

Screening solved one problem, not the other

The US Preventive Services Task Force gave adult depression screening a Grade B in 2023, and the instrument question is settled — the PHQ-9 is validated, scored 0–27, and sitting in most EHRs already. Our companion piece on building the PHQ-9 follow-up workflow covers that operational build: how to score it, route positives, and structure a screening visit that survives a busy Tuesday.

This piece asks a different, quieter question: once the flow exists and the screen comes back positive, how often does anything measurable happen next? USPSTF’s own pathway-to-benefit language requires that people who screen positive be “evaluated further for diagnosis and, if appropriate, provided or referred for evidence-based care.” Nobody disputes that requirement. Almost nobody publishes their own rate against it.

That absence is the invisible cost. A clinic can screen every eligible adult and score every PHQ-9 correctly, and still be running a program that quietly loses a third of its positive results to nothing — because the instrument produces a number the EHR can count, and what happens after it does not.

What happens after a positive PHQ-9, measured

The clearest accounting comes from a UCSF Health study of 3,980 adults with a PHQ-9 score of 10 or higher and/or reported suicidal ideation, screened across six primary care practices between September 2017 and September 2021 (Garcia et al., 2024). The researchers tracked what was actually ordered or documented at the visit and over the following eight weeks — not what patients said they intended to do.

Read the funnel carefully — the honest reading is less comfortable than it looks. Seventy percent had something: an antidepressant order, a referral, or a documented follow-up contact within eight weeks. But only 38% had it at the visit itself; the rest of that 70% trickled in over the following weeks, in a health system with the staffing and EHR tooling to chase it. The remaining 30% — nearly 1,200 people in this cohort — had no antidepressant order, no referral, and no documented follow-up contact at eight weeks. Among the subset who had disclosed suicidal ideation on the questionnaire, only 44% got an order at the time of screening — most people who told their PHQ-9 they were having thoughts of self-harm still left that visit without an immediate plan on the chart.

The authors are explicit about what this measures and what it does not: an order is not a filled prescription, and a referral is not an attended visit. “We do not know how many patients declined or did not initiate treatment,” they write. The 30% gap is the floor of the problem, not the ceiling.

The same gap shows up in a system built to prevent it

If UCSF’s number came from a single academic system, it would be easy to dismiss as a staffing quirk. It doesn’t hold up as an outlier. The Veterans Health Administration runs one of the most systematized universal depression-screening programs in US primary care, with a shared EHR and an explicit guideline for what “timely follow-up” means. A study of 607,730 primary-care patients across 82 VA clinics between fiscal years 2016 and 2019 found the same shape of gap, at far larger scale (Leung et al., 2022; reported in STAT News).

Only 32% of the roughly 15,000 veterans newly identified with depression received three or more mental-health visits within three months — the guideline threshold the VA itself uses to define timely follow-up. The one-year measure is more forgiving and still not clean:

Even against that floor, better than one in five newly diagnosed veterans did not clear it within a full year. A system with integrated records, salaried clinicians, and no billing friction between primary care and mental health still lands close to the UCSF number on the tighter, more clinically meaningful three-month measure — the strongest evidence that this gap is structural, not a symptom of any one clinic’s understaffing.

Two quality measures, two definitions of follow-up

Part of why this gap stays invisible is that the quality measures built to catch it don’t agree on what “follow-up” means. NCQA’s HEDIS measure, Depression Screening and Follow-Up for Adolescents and Adults (DSF-E), scores follow-up care received within 30 days of a positive screen — a window that already sits well past the eight-week and three-month gaps the research above documents. The federal eCQM version, CMS2v12, sets a much lower bar operationally: a “follow-up plan” — a referral order, a medication order, or another documented intervention — logged the same day as the visit through two days after it. Neither measure asks whether the referral was attended or the prescription was filled.

That gap between what gets reported and what actually happened is not a conspiracy — it is a measurement problem. A clinic can score well on CMS2v12 by documenting a plan within 48 hours, and still be the same clinic where 30% of positive screens have nothing further happen by week eight, because “planned” and “happened” are different facts that different measures collect. Practice leadership sees the checkbox rate. Almost nobody sees the completion rate, because nobody is asked to report it.

What closing the gap actually looks like

Closing this gap does not require a new screening instrument or a new guideline — both already exist. It requires treating the eight weeks after a positive screen as an owned, measured process rather than a documentation event. In practice that means three things operating together: a registry of open positive screens that does not clear itself when the visit ends, outbound contact when a referral or medication start goes quiet instead of waiting for the patient to reappear, and a number — not a checkbox — that leadership reviews on a cadence shorter than a year.

None of that is exotic. It is the same operating discipline that access-and-outreach work already demands elsewhere in a practice, covered from a different angle in governing patient outreach — who gets contacted, when, and on whose authority. Specialty care runs into a structurally identical problem after a referral is placed rather than after a screen is scored; our analysis of referral leakage in specialty care sizes that adjacent gap. In both cases, the failure point is the same: an order gets written, ownership of what happens next does not.

An AI-driven voice and messaging layer fits this problem as the always-on contact mechanism — the thing that actually places the outbound call or message on day 3 and day 14 without a staff member deciding, each time, whether today is the day to chase it. It does not replace the clinical judgment in evaluating a positive screen. It replaces the silence that follows one.

What to measure this month

If you operate in the United States, the broader access and follow-up landscape for primary care is on Arbol in the United States.