Depression screening has become one of the cheapest things a health system can do. The PHQ-9 is nine questions, self-reported, and takes under three minutes. Treatment is the expensive part. Across high-income systems, the gap between how well depression is detected and how quickly it is treated has widened rather than closed, and the metrics in use mostly track the detection side.

Screening Tools Outpace Treatment Capacity

The nine-item Patient Health Questionnaire was introduced in 2001 to screen adults in primary care. It asks about low mood, sleep, appetite, concentration and related symptoms over the previous two weeks, and each item is scored from zero to three. The total runs from zero to 27, with cutoffs commonly set at five, ten, fifteen and twenty for mild through severe.

Its design assumption was primary care. A clinician with a short consultation needs a fast, standardised way to decide whether a patient's presentation warrants further attention. The PHQ-9 does that job well enough that it has been folded into routine workflows, annual wellness visits and quality-reporting schemes in several countries.

What the instrument was not designed for is the follow-up loop. A score tells you someone is likely depressed today. It says nothing about whether the clinic has anyone to send them to, how long that takes, or whether the referral survives the first month. Screening mandates have expanded faster than the capacity to act on what they find.

The pattern repeats across programmes. In England, the Improving Access to Psychological Therapies service was built explicitly around referral-to-treatment measurement, and it publishes waiting figures that most mental health systems do not. Even there, the numbers describe people who entered the service, not everyone whose positive screen was supposed to send them. The distance between a referral written and a referral received is where most of the loss sits, and almost nobody counts it.

The PHQ-9 Measures Symptoms, Not Biology

There is no blood test for depression. No biomarker, no imaging finding, no lab value that confirms the diagnosis. The PHQ-9 measures reported symptoms, which means it measures distress as the patient describes it, filtered through their own interpretation of questions like whether they have been "feeling down, depressed, or hopeless."

That is not a flaw so much as a boundary. Cutoff scores vary by setting and population. A threshold that performs acceptably in a general adult primary care clinic may behave differently in a postnatal ward, a cancer service or an older-adult clinic, where physical symptoms overlap with the somatic items on the scale.

Every cutoff trades sensitivity against specificity. Set the bar low and you catch more genuine cases while flagging many people who are distressed but not depressed. Set it high and you miss people. The score reflects severity of reported symptoms, not their cause, and two patients with identical totals can need very different things.

Item nine, which asks about thoughts of self-harm, deserves separate attention because it changes the stakes of the whole exercise. A positive response on that single item is meant to trigger immediate clinical assessment regardless of the total score. Whether that happens depends on whether the clinician reads the item-level responses or only the summary number, and on whether the clinic has a same-day route to urgent assessment. A total of fourteen with a zero on item nine and a total of fourteen with a two on item nine describe two different clinical situations, and a screening dashboard that records only totals cannot tell them apart.

Referral Is Not a Treatment Pathway

Primary care detects. Specialty care treats. The handover between them is where most of the loss happens. In the UK, Improving Access to Psychological Therapies services publish referral-to-treatment figures, and waits of several weeks to a few months are common depending on the area and the therapy offered.

In the US, the constraint is often the insurance network rather than a queue. A positive screen may generate a list of in-network providers, some of whom are not accepting new patients, some of whom have waits of their own, and some of whom are listed inaccurately. Narrow networks make the list shorter, not the need smaller.

Warm handoffs, where the referring clinician speaks directly to the receiving service, are rare in routine practice. More often the patient leaves with a phone number. Whether they call, whether the number works, and whether they get an appointment are all unmeasured by the screening system that generated the referral.

Attrition compounds at each step. A proportion of patients never make the call. A proportion call and reach voicemail. A proportion book and then cannot attend because the appointment falls during work hours or requires transport they do not have. Each of these is a plausible, ordinary reason for a referral to die, and none of them appears in the quality metric that rewarded the original screen. Systems that have tried to close this gap usually do it with a dedicated coordinator role whose job is to chase referrals, which works and costs money and is therefore the first thing cut when a budget tightens.

What Happens After a Positive Screen

Stepped-care models are the standard rationing mechanism. Mild presentations get guided self-help or digital tools, moderate ones get brief talking therapy, and severe ones get specialist input. The logic is sensible: match intensity to need and spend scarce therapist time where it does most good.

In practice, steps can become queues. A patient stepped to a digital intervention may wait for a place on a course, and the evidence base for many app-based tools is thinner than for face-to-face therapy. Medication-first defaults appear in systems where prescribing is faster than securing a therapy slot, which shifts the treatment decision toward what is available.

Follow-up intervals are often unspecified. A positive score may generate a scheduled review in four weeks, or none at all. Where review is built in, it tends to be another PHQ-9, which measures whether symptoms have shifted but not whether the patient actually received the treatment the score referred them to.

Stepped care also assumes the steps exist in the right proportions. A service with plenty of low-intensity capacity and very little specialist capacity will step people up into a bottleneck, and the waiting time for the severe tier becomes the real constraint on the whole system. Clinicians who know the specialist tier is full start managing severe presentations at the moderate step, which is rational locally and corrosive systemically. The screening score did its job. The stepped-care model did what its capacity allowed.

The Measurement Gap in Mental Health

Screening metrics track detection. They count how many people were asked the questions and how many scored above threshold. Payment and quality incentives in several systems reward those counts, which makes the screening step visible and the treatment step invisible.

Denominators shift when capacity is fixed. If a service can only treat a set number of people per month, improving detection does not increase that number. It lengthens the list. A related piece on Kenya's mental health fund describes the same arithmetic from a financing angle, where clinics are left off reimbursement lists and the treatment gap widens regardless of how well cases are found.

Few systems publish treatment wait times as a routine statistic. Recovery is measured, where it is measured at all, through symptom scales that were built for screening. The result is a data environment that can report how many people were identified as depressed and cannot easily report how many were treated.

There is a version of this that would be fixable with better indicators. A service that reported the interval from positive screen to first treatment contact, and the proportion of referrals that resulted in at least one attended session, would be measuring the part that matters. Those two numbers are harder to collect than a screening count, which is precisely why they are rare. They require linking records across primary and specialty care, and most systems have no shared identifier to link them with.

Practical Steps for Clinicians and Patients

The questions worth asking sit after the score, not before it. A clinician who has just recorded a positive PHQ-9 can ask the receiving service directly what its current wait is, rather than handing over a number and hoping.

• Ask what happens next and who is responsible, naming a specific service and a contact rather than a general referral route.
• Request a named follow-up contact and a date, so the review does not depend on the patient re-presenting.
• Check waiting-list length before referring, and tell the patient what it is, so they can plan around it.
• Document the date treatment actually started, not just the date the screen was completed.
• Push for outcome measures beyond the PHQ-9, including functioning and whether the planned treatment was delivered.

None of this closes the capacity gap. It does make the gap visible to the people standing in it. A screening programme that cannot say what its referrals lead to is measuring the easy half of the problem.

This article is informational and does not constitute personalised medical advice. Anyone concerned about their mental health should consult a qualified clinician.