The PHQ-8, explained

01What the PHQ-8 measures#

The PHQ-8 measures how often eight depressive symptoms have been present over the preceding two weeks: loss of interest or pleasure, low mood, sleep disturbance, fatigue, appetite change, feelings of failure or self-blame, difficulty concentrating, and moving or speaking noticeably slowly or restlessly.

Each item maps onto a diagnostic criterion for a major depressive episode, a design inherited from the instrument it was cut down from. That one-to-one mapping is what allows a very short questionnaire to work as both a severity measure and a structured screen, and it is why it reads as clinical rather than conversational.

02How it is scored#

Each item is rated for how often the symptom has been present over the last two weeks: not at all (0), several days (1), more than half the days (2), or nearly every day (3). The eight items are summed for a total from 0 to 24.

A score of 10 or above is the conventional threshold for clinically significant depression, and the point at which further assessment is indicated. Kroenke and colleagues compared this simple cutoff against the more elaborate diagnostic algorithm and found they produced closely comparable results, which is why the cutoff is the version generally used.

03The item that was removed#

The ninth item of the longer version asks about thoughts of being better off dead, or of hurting yourself. Its presence turns a questionnaire into something with a duty of care attached.

In a clinical setting, a non-zero answer to that item triggers a defined protocol: direct risk assessment, and escalation where indicated. The answer is not scored and filed. It is acted on.

An instrument that can surface risk has to be deployed somewhere that can respond to it.

The PHQ-8 exists because a great deal of useful research happens in settings with no clinician attached and nothing to escalate to. Large population surveys are the obvious case. Asking a stranger by telephone whether they have thought about harming themselves, with nothing to offer afterwards, is not a neutral act. It takes on an obligation and then fails it.

Removing the item costs less than might be expected. In general population samples the ninth item is endorsed comparatively rarely, so the eight-item and nine-item versions produce closely similar scores and prevalence estimates.

04What it was validated on#

The PHQ-8 was evaluated on the 2006 Behavioral Risk Factor Surveillance System, a random-digit-dialled telephone survey of 198,678 people across the United States.

The sample matters as much as the result. Most depression instruments are validated on people who had already presented to a clinic, which is a population selected for being unwell and for having decided to seek help. The PHQ-8 was tested on the general public, which is a closer match to the population any consumer product is drawing from.

05Limits worth knowing#

• It is a screener, not a diagnosis. A score of 18 does not mean a person has major depression. It means someone qualified should look properly.
• Somatic items confound. Fatigue, appetite change and sleep disturbance are produced by plenty of things that are not depression: chronic illness, pregnancy, shift work, medication.
• Two weeks is a fixed window. Repeated fortnightly readings can miss shorter cycles entirely.
• Symptom framing is culture-bound. Distress is not described the same way everywhere, which is part of why wellbeing-framed instruments such as the WHO-5 travel more easily.
• It cannot see risk. Leaving out the ninth item is the point of the instrument and also a real limit. A PHQ-8 score of zero says nothing about whether someone is safe.
• Self-report under low mood is unreliable in a specific direction. Depression affects recall and self-appraisal, which are exactly the faculties the instrument depends on.

06Where it belongs, and where it does not#

The PHQ family is short, widely recognised and easy to obtain. That combination makes it tempting to drop into any product with a wellbeing feature, and a great many products have done exactly that.

Choosing the eight-item version is not a way around the responsibility. What it does is narrow the question to something a product can legitimately act on. A questionnaire that cannot surface an active risk disclosure can be answered outside a care setting without the people running it taking on a duty they have no way to discharge.

What it does not do is make a score meaningful on its own, or turn a product into a clinical service. A number tracked over time can show a pattern worth noticing. It cannot say what the pattern means, and it is no substitute for asking someone qualified.