Hypoglycemia is one of the main limiting factors for achieving adequate glycemic control in patients with type 1 diabetes mellitus (DM1).1 In diabetic patients, the unpleasant symptoms of hypoglycemic episodes, as well as their negative consequences, are often associated with the development of the fear of hypoglycemia (FH), which has a negative impact on the quality of life of both diabetic subjects and family members.2,3 This fear is of clinical relevance in diabetes management, metabolic control, psychological well-being and quality of life.4

Individuals who develop FH take counterproductive actions to prevent potential hypoglycemic episodes, such as reducing the insulin dose or consuming large amounts of carbohydrates, which leads to worsened blood glucose control and an increased risk of long-term complications.2 On the other hand, people with DM1 and FH experience severe hypoglycemia (SH) and asymptomatic hypoglycemia (AH) more often, thereby further increasing FH and complicating disease control.5

Fear of hypoglycemia can be assessed using the Hypoglycemia Fear Survey (HFS), published in 1987 by Cox et al., which quantifies FH-related worries or concerns and the behavior in people with DM1.6 The HFS has been widely used in clinical trials, and has been translated into more than 50 languages, demonstrating its reliability and validity.5 Following the first version, the same authors have recently developed two more versions.5,7 The FH questionnaire, and specifically the original version (HFS), has been translated and adapted in many countries, though not in Spain. Being able to use this questionnaire in Spain is important, since it remains the gold standard diagnostic tool in this field.

The present study was designed to translate into Spanish and adapt and validate the HFS in its initial version, and to analyze the variables associated with FH in a population with DM1.

This was a cultural adaptation and evaluation observational study of the psychometric properties (feasibility, reliability, validity and stability) of an FH questionnaire–the HFS–in a population with DM1 representative of the corresponding healthcare area, together with an assessment of the clinical variables associated with the outcomes of the HFS.

A total of 167 subjects with DM1 agreed to participate in the study, though four patients failed to complete the questionnaires. A total of 163 subjects were thus finally included. Table 1 shows the sociodemographic and clinical characteristics of the study sample. The mean age (SD) was 36 (10) years; 24.6% used continuous subcutaneous insulin infusion (CSII); 54% had suffered some SH episode in their lifetime; and 35% had AH.

Following the exploratory factor analysis and confirmation of the sample suitability of the 27 items of the translated original questionnaire (HFS), the number of items was reduced to 24 in the final questionnaire (EsHFS). The items not included from the original HFS were items 1, 2 and 5 corresponding to the behavioral subscale. The final 24-item version of the EsHFS increased the reliability of the translated original questionnaire, Cronbach's α 0.927, and for all items we confirmed that the item-total correlation was >0.3 (Appendix A, Table e-1). The Kaiser–Meyer–Olkin sample suitability measure for the selected items was 0.904 (p<0.001).

Factorial analysis (Appendix A, Table e-2) identified three components or subscales: subscale 1 or the worry scale (items 1–17); subscale 2 or behavior associated with the avoidance scale (items 18–21); and subscale 3 or behavior causing hyperglycemia (items 22–24). Of these three components, subscale 1 accounted for 42% of the global questionnaire variance, subscale 2 accounted for 6.5%, and subscale 3 accounted for 6.1% of the global variance.

The reliability of each subscale measured by means of Cronbach's α was 0.939, 0.6 and 0.728, respectively, for subscales 1–3 (Appendix A, Table e-1).

The feasibility of the EsHFS as assessed by the percentage of responses was 98.8%. Only two subjects did not answer item 10, referring to vehicle driving. The mean response time (SD) was 8min (1.3), and the questionnaire was completed in under 10min by 162 subjects (99.8%).

Table 2 shows the validation of the questionnaire based on the correlation between the results of the EsHFS and the EsDQOL. The subscales with the highest correlations corresponded to subscale 1 (worry) of the EsHFS, and with the social worry and worry about DM subscales of the EsDQOL.

The stability of the questionnaire assessed by test–retest correlation was: global EsHFS, r=0.92 (95%CI 0.84–0.96), p<0.001; subscale 1, r=0.91 (95%CI 0.82–0.96), p<0.001; subscale 2, r=0.46 (95%CI 0.13–0.70), p=0.009; and subscale 3, r=0.91 (95%CI 0.82–0.96), p<0.001.

The EsHFS questionnaire in its final form is shown in Table 3. A score of 1–5 is assigned to each item, making 24 points the minimum score for the global EsHFS and 120 points the maximum. Table 4 shows the distribution of the questionnaire and each of the subscales in our sample.

The association of the results of HFS with the variables analyzed was explored using univariate and multivariate models. In the univariate analysis (Table 5), higher scores were seen in subscale 1 in the group of subjects that performed more SMBG, the presence of polyneuropathy, SH and AH. In turn, higher scores were seen in subscale 2 in the group of subjects that performed more SMBG, in those who had more NSH, women, those who used CSII, and those with polyneuropathy.

A subanalysis was performed to check for differences in EsHFS between subjects with very poor glycemic control (4th quartile, HbA1c≥8.5%) versus better control (1–3rd quartile, HbA1c<8.5%). Higher scores on the hyperglycemia-causing behavior subscale were found in subjects showing poorer control, with 7.2 (3.1) versus 6 (2.6) (mean [SD]); p=0.022.

The multivariate analysis (Table 6) showed FH measured as worry to be associated with AH, a lower educational level, older age, and a greater frequency of SMBG. Avoidance behavior was more common in females, in subjects with NSH, and in patients with more frequent self-monitoring. Hyperglycemia-causing behavior was associated with not living with a partner, and increased in subjects with a history of SH and in older patients.

The present study translated, adapted and validated the HFS, which is regarded as the gold standard for quantifying FH in diabetic populations. The version obtained (EsHFS) has adequate psychometric measures, such as feasibility, reliability and stability over time, and has been validated in a population with DM1 representative of the Spanish population. The association with certain sociodemographic and clinical variables, supported by studies using this questionnaire translated into other languages,2,13 warrants its use in the population with DM1 in our setting.

During the translation and backtranslation process, as well as in the final adaptation of the instrument, we attempted to be faithful to the original version in terms of content and intention of items, seeking always to use language comprehensible to any person regardless of educational level. The elimination of three items from the original questionnaire allowed for an improvement of Cronbach's α, reflecting excellent internal consistency, with a value of >0.9.14

Factorial analysis identified three subscales, one coinciding with the original questionnaire (the worry subscale), and two referring to FH-associated behavior such as avoidance or actions producing hyperglycemia. The original study identified a worry or concern subscale and a general behavior subscale. However, within behavior, the actions clearly differ and do not always coincide in subjects with FH. Recently, the analysis of a sample of 1460 adults with DM1 from 5 countries using the most recent version of the HFS,5 also known as HFS-II, translated into their respective languages, obtained the same subscales as in our study.13 However, the two behavior subscales showed less reliability, though they can be regarded as acceptable considering that the number of items in each subscale was less than 10.15 This aspect–the lower reliability of the behavior subscales–was already observed in the original questionnaire,6 and has been repeated in subsequent adaptations of the instrument.5,16 In parallel, both subscales explain a lower percentage of the variance of the global questionnaire (6.5% and 6.1%, respectively) than the worry subscale (42%). This aspect is what could have caused some investigators to prefer to only use the worry subscale in their studies.17

Feasibility has not been reported in the studies that have translated and validated this questionnaire. In the present study, it took subjects an average of 8min to complete the questionnaire, and it was administered in the clinical practice setting: these features support its use in routine practice. The stability of the questionnaire over time, with test–retest correlation coefficients higher than those reported in the original study,6 warrants its adequate psychometric properties.

The final questionnaire (EsHFS) was validated by analyzing its correlation with the EsDQOL, though this process was not necessary, since we were dealing with the translation and adaptation of an instrument that had already been validated. Most studies use questionnaires that measure the emotional impact associated with diabetes, such as the Problem Areas in Diabetes scale18 and/or questionnaires such as the Stait-Trait and Anxiety Inventory (STAI).19 However, such tools may be distorted by other factors not associated with diabetes. Quality of life and FH are interdependent variables,20,21 with adequate correlation being found in our study, particularly with the more specific spheres of diabetes such as impact, social worry, and worry about diabetes.

Analysis of the association between the included variables and the final version of the EsHFS allowed us to test the questionnaire for the first time.

Among the sociodemographic factors related to the development of FH, only female gender4,5,22,23 and age13 were studied and correlated to the HFS outcomes. Other factors such as educational level and marital status or type of family cohabitation were not evaluated. In our study, older people and women independently showed greater FH in some of the subscales, namely age in the worry subscale and the hyperglycemia-related behavior subscale, and female gender in the avoidance behavior subscale. These findings may be explained by taking two aspects into account: (i) over the years, as people gradually accumulate life experiences they tend to become more conservative; and (ii) women have greater FH in the same way as is seen in other depressive or anxiety disorders.24

Among the clinical factors, the frequency of SMBG was higher among the subjects with greater FH, and an independent association was found between these two parameters. This association has not been analyzed in the literature, and we interpret it as a consequence of having greater FH. With regard to the degree of control, we only detected an association between HbA1c and FH (the hyperglycemia-causing behavior subscale) in the subanalysis comparing those subjects with poorer control, 4th quartile versus 1st–3rd quartile, though not in the multivariate study. This aspect was reported in the original study and in an international study in 5 countries.6,13 The sample size of this study perhaps was not large enough to identify this association with greater statistical robustness.

The most commonly implicated clinical factor in FH is probably hypoglycemia, in terms of both frequency and severity, particularly SH.2 The presence of SH is the factor most commonly described in the literature as being associated with FH measured by means of the HFS4,17,22 and by another recently developed questionnaire (FH15).25,26 Asymptomatic hypoglycemia, a factor often preceding SH,27 was also associated with FH particularly in the worry subscale, though the p-value only reached the limit of significance in the multivariate analysis. These data suggest higher scores to be associated with greater worry, more avoidance behavior, and more hyperglycemia.

The present study has limitations:

• -

  Although data collection was prospective, there were parameters such as the frequency of hypoglycemia as reported by the patient during the interview which could have been underestimated.

• -

  Subject classification referring to AH is based on a question. At the time of this study, we did not have the Clarke questionnaire recently translated into Spanish.28

• -

  The classification of marital status to establish the type of cohabitation (with or without a partner) did not correspond to the clinical variable proposed in the study design (i.e., the determination of whether stable living with a partner or family was associated with greater or lesser FH).

• -

  The short version of the HFS-II was published at a later date7 and therefore could not be used.

The study has a number of strong points:

• -

  It is the first study to translate and validate the HFS questionnaire, which is the reference for quantifying FH, and also the most widely used instrument in most studies that address this subject.

• -

  We included any patient with DM1, without excluding individuals with SH as is often done in studies.

• -

  The operating procedure, as well as the methodology used, are widely supported by similar studies.

• -

  Study participation in terms of the completion of the questionnaire was high.

In conclusion, the Spanish version of the HFS questionnaire developed in this study (EsHFS) has good psychometric properties, and represents a useful tool for assessing FH in Spanish-speaking patients with DM1 in our setting.

The authors declare that they have no conflicts of interest.