The GAD-7 measures one kind of anxiety well
Seven items, 0 to 21, two minutes, free. The GAD-7 is an excellent instrument for generalised anxiety and a mediocre one for several other anxiety presentations that clinicians routinely administer it to. Knowing which is which is the difference between a useful score and a misleading one.
The GAD-7 asks how often, over the last two weeks, the respondent has been bothered by seven problems — nervousness, uncontrollable worry, excessive worry, trouble relaxing, restlessness, irritability, and a sense that something awful might happen. Each is scored 0 to 3 on the same frequency scale as the PHQ-9, giving a total of 0 to 21.
Severity bands
| Total | Commonly described as | Typical implication |
|---|---|---|
| 0–4 | Minimal | Monitoring generally sufficient |
| 5–9 | Mild | Watchful waiting, re-administer |
| 10–14 | Moderate | Treatment plan warranted |
| 15–21 | Severe | Active treatment indicated |
A cutoff of 10 is the most commonly cited screening threshold for probable generalised anxiety disorder. As with any screener, the right threshold depends on setting and on the relative cost of false negatives versus false positives.
What it detects well, and what it does not
This is the part most worth knowing, and it is frequently skipped.
The GAD-7 was developed and validated primarily as a measure of generalised anxiety — the cognitive and physiological features of persistent, diffuse worry. Within that target it performs well and has reasonable screening properties for panic disorder, social anxiety disorder and PTSD, which is why it is often described as a general anxiety screener.
But "reasonable screening properties" is not the same as "measures the construct." Several presentations are poorly captured:
- Specific phobias. Someone with a severe, impairing phobia of flying may score near zero on a measure asking about persistent worry over two weeks, because outside anticipated exposure their anxiety is not persistent.
- Obsessive-compulsive disorder. The GAD-7 does not ask about intrusive thoughts or compulsions. OCD is no longer classified with the anxiety disorders and needs its own instrument.
- Panic disorder. Partly captured, but the defining features — discrete attacks and anticipatory fear of recurrence — are not what the items ask about. A client with weekly panic attacks and low baseline worry can score deceptively low.
- PTSD. Screened for, but poorly characterised. Use the PCL-5 where trauma is the presenting concern.
The practical consequence. A low GAD-7 in a client who is visibly anxious is not evidence they are not anxious. It is evidence they are not generally anxious in the way the instrument measures. Treating the score as a global anxiety index is the most common misuse, and it can lead to under-treating presentations the instrument was never built for.
Why it is paired with the PHQ-9
Almost universally, and for two reasons.
The clinical reason is comorbidity. Depression and anxiety co-occur at high rates, presentations frequently sit across both, and treating one while not measuring the other produces an incomplete picture — particularly when improvement in one domain masks deterioration in the other.
The practical reason is that the two share an identical response format and recall window, so administering both takes about three minutes and produces two scores on a comparable scale. There is no meaningful friction cost to running both, which is why the combined administration has become the default in outpatient behavioral health and primary care alike.
GAD-2
The first two items — feeling nervous or on edge, and being unable to stop or control worrying — form the GAD-2, scored 0 to 6. Like the PHQ-2 it exists for high-throughput screening where the full instrument is impractical. In outpatient behavioral health the full seven items are almost always worth the extra ninety seconds.
The unscored functional item
As with the PHQ-9, the GAD-7 carries a final question about how difficult the problems have made work, home life and relationships, and it does not contribute to the 0–21 total. Record it anyway. Functional impairment is frequently what makes the case for treatment intensity, and a moderate score with severe impairment is a materially different clinical picture from the same score with intact functioning.
Interpretation limits
- Somatic overlap. Restlessness, irritability and trouble relaxing are also features of hyperthyroidism, stimulant use, caffeine, akathisia and withdrawal states. Elevated scores in clients on medication or with medical comorbidity need interpreting.
- Cultural expression. The items are worded around Western idioms of worry. Anxiety expressed predominantly somatically may be under-detected.
- Two-week window. Captures current state only. Trajectory requires repeat administration.
- Validated primarily in adults. Adolescent use has separate literature.
Documentation
Record the total, the date and the administration method, and connect it to a decision. "GAD-7 15, unchanged from four sessions ago despite exposure work; discussed medication consultation and client agreed to referral" does what a bare score cannot — it shows the measurement changed something. See the golden thread.
Licensing
The GAD-7 is free to use, reproduce and translate without permission or fee. It was developed by Robert Spitzer, Kurt Kroenke, Janet Williams and Bernd Löwe, and — like the PHQ-9 — its open availability is a substantial part of why it is used so widely.
When the score and the presentation disagree
This happens often enough to plan for, and the resolution is diagnostic rather than psychometric.
A client who is visibly distressed with a GAD-7 of 4 is telling you something useful: whatever is driving the presentation is not persistent generalised worry. That points toward a specific phobia, a panic presentation with low inter-episode anxiety, OCD, trauma, or a depressive presentation with anxious features — each of which has a different treatment implication and a different instrument.
The reverse also occurs. A client reporting little subjective distress who scores 16 may be describing a baseline they have normalised over years. Both discrepancies are clinically informative, and both are worth a sentence in the note explaining how you resolved them, because a chart showing a low score alongside intensive treatment invites a medical-necessity question that the discrepancy itself answers.
Tracking through medication changes
Where a client is starting, stopping or adjusting a psychotropic, tighter GAD-7 administration is worth the extra minutes. Activation, akathisia and early-treatment anxiety increase are well-described in the first weeks of SSRI treatment, and a score that jumps at week two is information the prescriber needs quickly.
Coordinate on the timing rather than duplicating it — a therapist and a prescriber independently administering the same instrument a day apart produces two data points that disagree for uninteresting reasons. Agreeing who administers what, and sharing the result, is a small piece of split-treatment coordination that pays for itself.
Recording the pair
Where you administer the PHQ-9 and GAD-7 together, record both scores in the same place with the same date. Splitting them across separate fields or separate notes makes the comorbidity picture harder to read later and, in practice, means one of them stops being administered within a few months because nobody notices it is missing.
A note on the name
The "GAD" in GAD-7 refers to generalised anxiety disorder, and the instrument is frequently described in software and in payer documentation simply as "the anxiety screener." That shorthand is where much of the misuse originates. Calling it what it is — a generalised anxiety measure with useful but secondary screening properties for other presentations — keeps its limits in view for whoever reads the chart next.
Verified 29 July 2026. Instrument specifications are as published by their developers; interpretation thresholds vary by population and setting and are presented as commonly cited starting points rather than diagnostic rules. Screening instruments do not establish diagnoses. Primary references: VA National Center for PTSD; APA Services. This page is clinical reference, not clinical or legal advice.