Anxiety Test (GAD-7)

Robert Spitzer, Kurt Kroenke, Janet Williams and Bernd Löwe published the GAD-7 in 2006 in the Archives of Internal Medicine. Spitzer, Williams and Kroenke had already built the PHQ questionnaires for primary care, and the GAD-7 followed as a short measure of anxiety. An unrestricted educational grant from Pfizer paid for its development.

The team began with 13 items: nine that matched the DSM-IV symptom criteria for generalized anxiety disorder (GAD) and four taken from other anxiety scales. They kept the seven items that correlated most strongly with the total. Each item asks how often a problem bothered you over the last 2 weeks, from 0 (not at all) to 3 (nearly every day).

The 2006 study ran in 15 US primary care clinics between November 2004 and June 2005. A total of 2740 adults completed the questionnaire, and 965 of them also had a telephone interview with a mental health professional within a week. The average age was 47.4 years, 65% were women and 80% were white non-Hispanic. Internal consistency was high (Cronbach's alpha .92), and test-retest reliability was 0.83.

Against the interview diagnosis of GAD, a score of 10 or higher had 89% sensitivity and 82% specificity. Yet the positive predictive value was 29%. Only 73 of the 965 patients had GAD by interview, and 23% of all patients scored 10 or higher, so most high scorers didn't meet the interview's criteria, which required six months of symptoms.

A 2007 paper from the same clinics found that the GAD-7 and its two-item short form both worked as screens for panic disorder, social anxiety disorder and posttraumatic stress disorder as well as GAD, with areas under the curve from 0.80 to 0.91. A 2016 meta-analysis of 12 samples (5223 participants) found acceptable accuracy for GAD at cut-offs from 7 to 10, with the best balance at 8 (sensitivity 0.83, specificity 0.84). In a nationally representative German survey of 5030 people, about 5% scored 10 or higher.

The cut points of 5, 10 and 15 come from the authors, who wrote that they "might be interpreted" as mild, moderate and severe. The paper tested cut-offs from 8 to 15 against clinician interviews, but not the cut-off of 5. At 15, sensitivity was 48% and specificity 95%, so many people with GAD score below 15. The "Minimal" label for 0 to 4 is a common convention, not a term from the 2006 paper.

The GAD-7 was built to screen for generalized anxiety disorder. Its authors call the result a probable diagnosis that needs further evaluation, and they noted that it covers one anxiety disorder while social phobia and posttraumatic stress disorder also need attention. Their cut-offs came from adults in US primary care clinics, 80% of them white non-Hispanic. The study can't tell you how the scale performs for everyone, or for people who take it online.

Your answers cover only two weeks, so a score can't separate a rough patch from long-standing anxiety. It also depends on how you rate yourself that day. Anxiety and depression frequently occur together. In the original study they showed up as separate dimensions, yet many patients had both.

Treat your score as a snapshot of the last two weeks. The authors suggested 5, 10 and 15 as mild, moderate and severe levels, and a score of 10 or higher is the usual point for further evaluation. A doctor or therapist can ask how long the symptoms have lasted and what else could explain them. A score under 10 doesn't mean your worry is fine. If it keeps you from sleeping, working or seeing people, it's worth asking for help at any score.

If you retake it, use the same two-week window and look for a pattern across several tries. For something to do about worry, see the worry time technique or the beginner's guide to CBT for anxiety.

More on this topic: Worry time technique, CBT for anxiety: a beginner's guide, Stress test (PSS-10), Depression test (PHQ-9).

Your answers are scored in your browser. If you sign in and choose to save a result, we store the score and its label, not your individual answers. Without an account, nothing is saved.

• Spitzer, Kroenke, Williams and Löwe, Archives of Internal Medicine, 2006 (JAMA Network): the full article: item pool, sample, reliability, cut points 5, 10 and 15, stated limits, Pfizer grant. Table 1 (sensitivity, specificity, positive predictive value by cut-off) was read in a PDF copy of the same article.
• Same paper on PubMed (abstract): abstract only: 2740 adults, 965 interviews, sensitivity 89% and specificity 82%.
• Kroenke et al., Annals of Internal Medicine, 2007 (PubMed): abstract only: area under the curve 0.80 to 0.91 for the GAD-7 and its two-item short form across four anxiety disorders. The sensitivity and specificity figures for each disorder were not checked.
• Plummer et al., General Hospital Psychiatry, 2016 (PubMed): abstract only: meta-analysis of 12 samples, best balance at a cut-off of 8.
• Löwe et al., Medical Care, 2008 (PubMed): abstract only: general-population survey in Germany, about 5% scored 10 or higher.
• PHQ Screeners instruction manual: the developers' scoring bands, the yellow flag at 10 and red flag at 15, and the statement that these measures are in the public domain with no permission required.
• GAD-7 form on phqscreeners.com: official item wording and the line that no permission is required to reproduce, translate, display or distribute.
• PHQ Screeners terms of use: Pfizer's disclaimer and the note that the screeners are free for download and use as stated on the site.