Discriminant validity of the 12-item version of the general health questionnaire in a Swedish case-control study

Abstract

Background: The 12-item version of the General Health Questionnaire (GHQ-12) is widely used as a proxy for Affective Disorders in public health surveys, although the cut-off points for distress vary considerably between studies. The agreement between the GHQ-12 score and having a clinical disorder in the study population is usually unknown.

Aims: This study aimed to assess the criterion validity and to determine the sensitivity and specificity of the GHQ-12 in the Swedish population.

Methods: This study used 556 patient cases surveyed in specialized psychiatric care outpatient age- and sex-matched with 556 controls from the Stockholm Health Survey. Criterion validity for two scoring methods of GHQ-12 was tested using Receiver Operating Characteristics (ROC) analyses with Area Under the Curve (AUC) as a measure of agreement. Reference standard was (1) specialized psychiatric care and (2) current depression, anxiety or adjustment disorder.

Results: Both the Likert and Standard GHQ-12 scoring method discriminated excellently between individuals using specialized psychiatric services and healthy controls (Likert index AUC = 0.86, GHQ index AUC = 0.83), and between individuals with current disorder from healthy controls (Likert index AUC = 0.90, GHQ index AUC = 0.88). The best cut-off point for the GHQ index was ≥4 (sensitivity = 81.7 and specificity = 85.4), and for the Likert index ≥14 (sensitivity = 85.5 and specificity = 83.2).

Conclusions: The GHQ-12 has excellent discriminant validity and is well suited as a non-specific measure of affective disorders in public mental health surveys.

Keywords:

• Sensitivity and specificity
• screening
• GHQ
• area under the curve
• self-assessment
• affective disorder

Disclosure statement

Senior Statistician Peeter Fredlund at the Center for Epidemiology and Community Medicine (CES), Stockholm County Council, assisted in case-control matching. Forte took no part in the writing or interpretation of the findings. The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.