The General Health Questionnaire: how many items are really necessary in population surveys?

This paper seeks to investigate whether only a few questions selected from the General Health Questionnaire (GHQ) may be used to measure the degree of mental distress in population surveys. Data from 2112 men and women, 18 to 70 years old from two cross-sectional studies conducted in northern Norway and the island Spitzbergen in the Arctic, were used. Correlation analysis of Likert scores from a 20-item version of GHQ (GHQ-20) with Likert scores based on four and six items selected by multiple regression analysis or by competent physicians was performed. The correlation coefficients between the scores from the subsets of four items and the full GHQ-20 questionnaire were high (greater than 0.80) in all examined subgroups of the populations. Increasing the number of questions from four to six only marginally increased the correlation coefficients. Thus, a simple linear sum of Likert scores based on a few GHQ items can be used to measure the degree of mental distress in populations surveys.