Psychometric validation of the Hypoglycemia Fear Survey-II (HFS-II) in Singapore.

OBJECTIVE

We conducted a cross-sectional study to adapt and validate the Hypoglycemia Fear Survey-II (HFS-II) for use in Singapore among persons with type 1 and 2 diabetes mellitus.

RESEARCH DESIGN AND METHODS

A total of 144 patients with type 1 or 2 diabetes on insulin therapy for at least a year completed the HFS-II between September and December 2013 in the Diabetes Center at Singapore General Hospital. We examined the validity (content, concurrent and discriminant validity, and construct validity) and reliability (internal consistency and test-retest reliability) of the instrument. Content validity was established using cognitive interviews. Construct validity was assessed using confirmatory factor analysis (CFA) followed by exploratory factor analysis (EFA) after the hypothesized two-factor structure was not confirmed by CFA. Measures of anxiety (Generalized Anxiety Disorder-7 (GAD-7)) and depression (Patient Health Questionnaire-9 (PHQ-9)) were used to establish concurrent validity; history of severe hypoglycemia and status of glycemic control were used to establish discriminant validity. Internal consistency was measured by Cronbach's α; test-retest reliability was measured by intracluster correlation coefficient (ICC).

RESULTS

Scores of the adapted HFS-II had moderate positive correlations with measures of anxiety and depression scores (r=0.41, p<0.01; r=0.37, p<0.01. Patients with a recent history of severe hypoglycemia had higher HFS-II scores than those without (mean difference=9, p<0.01). Patients with poor glycemic control had higher HFS-II scores than those with good control (p<0.05). The original two-factor structure was not confirmed in our sample. EFA results suggested a three-factor solution with the original Behavior subscale splitting into two dimensions. The adapted HFS-II displayed good internal consistency (Cronbach's α=0.93) and test-retest reliability (ICC=0.75).

CONCLUSIONS

The adapted HFS-II has good content, concurrent and discriminant validity, and reliability, but its constructvalidity was not proven with the Behavior subscale turning out to be non-unidimensional.