OBJECTIVE—To develop a psychometric questionnaire to measure psychological barriers to insulin treatment in patients with type 2 diabetes.
RESEARCH DESIGN AND METHODS—Scale development was based on principal component analyses in two cross-sectional studies of insulin-naïve patients with type 2 diabetes. The structure of the questionnaire was developed in the first sample of 448 patients and subsequently cross-validated in an independent sample of 449 patients.
RESULTS—Analyses in the first sample yielded five components that accounted for 74.5% of the variance based on 14 items and led to the following subscales: fear of injection and self-testing, expectations regarding positive insulin-related outcomes, expected hardship from insulin treatment, stigmatization by insulin injections, and fear of hypoglycemia. In addition, an overall sum score of all values was calculated. The structure of the questionnaire was cross-validated in the second sample, with almost identical component loadings and an explained variance of 69.4%. An additional confirmatory factor analysis also indicated an acceptable to good model fit with root mean square error of approximation equal to 0.04 and comparative fit index equal to 0.97. Coefficients of reliability (Cronbach's α 0.62–0.85 and 0.78 for overall sum score) were acceptable, considering the very small number of items for each scale.
CONCLUSIONS—The Barriers to Insulin Treatment Questionnaire appears to be a reliable and valid measure of psychological insulin resistance in patients with type 2 diabetes. This short instrument is easy to administer and may be used by both clinicians and researchers to assess the psychological barriers to insulin treatment.
Despite the increasing body of knowledge regarding diabetes treatment (1), a majority of patients with type 2 diabetes are still in persistently poor glycemic control (2), a state that leads to higher risks of poor health outcomes (3). A variety of factors are responsible for poor glycemic control, including the inadequacy of therapeutic regimens (1) as well as various psychosocial aspects (4,5). In recent years, researchers also have focused on the reluctance of patients to take insulin and the resistance of health care providers to prescribe insulin (6). These negative attitudes toward insulin treatment contribute to unnecessarily long delays for initiating insulin treatment and, consequently, to extended periods of hyperglycemia (7,8). This so-called “psychological insulin resistance” (9) includes, among other factors, fear of injection and self-testing, hypoglycemia, and weight gain; a perceived loss of control over one's life; poor self-efficacy concerning insulin treatment; and perceived lack of positive outcomes related to insulin (9,10).
To overcome these psychological barriers to insulin treatment, first it is necessary to identify these barriers in specific patients in order to decide which interventions are appropriate. Thus, a well-validated diagnostic tool may be helpful to identify specific obstacles against the initiation of insulin treatment. There are some questionnaires that measure different aspects of satisfaction with treatment or diabetes-related burdens or stress (11–17), but presently, to our knowledge, no specific measurement of the psychological barriers to insulin treatment has been created, validated, and published.
This article describes the development and evaluation of the self-administered Barriers to Insulin Treatment (BIT) Questionnaire. The process is based on principal component analyses in two independent samples (n = 448 and 449, respectively). The aim of the BIT Questionnaire is to measure various aspects of psychological obstacles to insulin treatment in patients who have type 2 diabetes.
RESEARCH DESIGN AND METHODS
The development of the BIT Questionnaire was based on data from two independent German studies of insulin-naïve patients with type 2 diabetes. The results of the first study (sample A) formed the basis of the questionnaire development, and the dataset of the second study (sample B) was used for cross-validation of scale structure and consistency. Both studies were approved by the responsible ethical review boards, and written informed consent was obtained from all patients in the studies.
A total of 4,000 physicians were randomly selected from a countrywide database (pan-address, Planegg, Germany), which contained details of 31,155 general practitioners and 18,860 internists, to generate a representative cross-section of doctors in Germany. Physicians were approached between 2001 and 2003 by letter and were asked to recruit the next eligible patient during a routine appointment. Eligibility was based on the following inclusion criteria: clinical manifestation of type 2 diabetes, psychological and physical competence to participate in the study, and the patient's informed consent. Physicians provided data on the age, sex, height, and weight of the patient as well as the confirmation of type 2 diabetes diagnosis, duration of diabetes, and A1C levels. For the purpose of development of the BIT Questionnaire, only data from insulin-naïve outpatients (n = 448) were included.
This study was performed in 20 practices of general practitioners and internists all over Germany in 2005. A total of 449 type 2 diabetic outpatients were included who had an A1C suggesting more intensive therapeutic interventions were necessary including insulin treatment as defined by current treatment guidelines (18). Inclusion criteria were type 2 diabetes treated only with diet or oral antidiabetes treatments (OADs), age between 18 and 70 years, and insufficient oral antidiabetes treatment defined as A1C >8.5% (treated with diet alone), A1C >8.0% (one OAD), or A1C ≥7.5% (two or more OADs). The different cutoffs were chosen based on preexisting therapies and the current standards for the initiation of insulin treatment in Germany and other countries. Participants discussed treatment optimization options with a physician who recommended insulin treatment because of the failing OAD treatment. Subsequently, the patients were asked to make a theoretical choice about future diabetes treatment. The options were to continue with OADs or to move on to subcutaneous insulin. A set of medical data were measured that included the same variables as those in sample A (Table 1) (19).
Psychological barriers to insulin treatment assessed in sample A.
A pool of items, 35 regarding different attitudes toward insulin treatment, was created by an expert panel of health care professionals who were experienced in diabetes treatment based on patient interviews and current literature (9,10). The following attitudes toward insulin treatment were identified: positive feelings about the benefits of the treatment, fear of the consequences of diabetes, fear of injections, social barriers to the use of insulin, aversion to dependence on the drug, fear of insulin-related side effects, and negative feelings about one's competence to manage the insulin treatment. Each of the 35 items was presented as a statement, which the patient was asked to score using a 10-point Likert-type scale with the extreme scores labeled “completely disagree” (1) and “completely agree” (10).
Psychological barriers to insulin treatment assessed in sample B.
The psychological barriers to insulin treatment were assessed using the BIT Questionnaire developed in sample A (see results). The BIT Questionnaire includes 14 items, a total sum score, and the following five subscales: fear of injection and self-testing, expectations regarding positive insulin-related outcomes, expected hardship from insulin treatment, stigmatization by insulin injection, and fear of hypoglycemia.
To obtain a validated English version of the BIT Questionnaire, a linguistic validation (from German to English [U.S.]) was performed and certified by the MAPI Research Institute. This validation process included two forward translations; reconciliation; a backward translation, with review and discussion and retranslation as needed; clinician review and further changes as needed; cognitive debriefing with patients; and finalization and proofreading (20). An English and German version of the BIT Questionnaire can be found in an online appendix at http://dx.doi.org/10.2337/dc06-2042.
In a first step, a principal component analysis (with oblique rotation) that included all 35 items was conducted in an exploratory approach in sample A. The Kaiser-Guttman criterion (eigenvalue >1) was used to decide on the number of components to be retained (21,22). Subsequent principal component analyses were conducted to eliminate items or components until a result became available that had good test statistical values and was interpretable concerning its contents. The criteria for the item-respective component elimination were the following: items with ambiguous loadings on rotated component (in first-step items with a ≤0.40 factor loading and in subsequent-step items with a <0.60 factor loading were eliminated), items with multiple loadings on other components (>0.25), and items with an item selectivity ≤0.35. Components that contained fewer than two items with sufficient loadings were eliminated together with all their items. In some components, items that missed the criteria, but appeared necessary due to textual and test-related theoretical considerations, were retained after careful consideration.
Principal component analyses were conducted successively with the items remaining from the previous analysis until all objectives were met. The resulting components represent the subscales of the BIT Questionnaire, whose scale stability was verified in sample B. Finally, a second-order component analysis with oblique rotation was performed including the obtained components that represent the subscales of the BIT Questionnaire in order to analyze if the creation of a BIT Questionnaire sum score will be appropriate.
Using sample B data, a principal component analysis with the final BIT Questionnaire items of sample A was performed. This approach allowed for a demonstrative comparison of component loadings in both independent samples before the aggregated model was tested by a confirmatory factor analysis (CFA). Following current recommendations for a CFA, we report the comparative fit index (CFI) and the root mean square error of approximation (RMSEA) (23). Indicators of a well-fitting model would be evidenced by a CFI ≥0.93 and an RMSEA <0.05 (24,25).
To replicate the test of whether the creation of a BIT Questionnaire sum score will be appropriate, we performed a second-order CFA with the components representing the subscales of the BIT Questionnaire. Pearson's correlations of all subscales were calculated to evaluate interrelations among the subscales, and Cronbach's α was computed for the subscales and the total sum score to measure internal consistency reliability.
The predictive validity of the BIT Questionnaire scales was verified by comparision of subgroups of study B in the theoretical treatment choice scenario (described above). Using t tests, we compared the scores of patients who accepted to move on to subcutaneous insulin to those patients who have been offered insulin and decided to continue with the OAD treatment. Effect sizes for the differences between both groups were calculated using Cohen's d with pooled SD (26). Statistical analyses were performed with SPSS 12.0.1 (Chicago, IL), Amos 5.0 (Chicago, IL), and ClinTools (Brain Sciences Institute, Swinburne University, Swinburne, Australia).
Table 1 provides descriptive statistics for socioeconomic, medical, and psychological variables for samples A and B.
Scale development in study A.
As a result of the first principal component analysis with 35 items, seven components with an eigenvalue >1 could be extracted. We eliminated items with ambiguous component loadings and identified a component structure that explained 74.5% of the variance; this component structure included 14 items and five components for which no ambiguous component loadings had been observed (Table 2). The results of the second-order principal component analysis showed only one component with an eigenvalue >1 and an explained variance of 49.3%, indicating that it is appropriate to create a sum score of the total barriers to insulin treatment.
Replication of component structure in study B.
The principal component analysis in the sample used for cross-validation resulted in a clear replication of the questionnaire's structure. With an explained variance of 69.4%, five components with almost identical component loadings were identified (see Table 2). Results of the CFA also confirmed the structure of the BIT Questionnaire with an acceptable to good model fit with RMSEA = 0.04 (90% CI 0.03–0.05) and CFI = 0.97. The second-order CFA confirmed a good model fit with RMSEA = 0.04 (0.02–0.04) and CFI = 0.97, showing again that it is appropriate to create a total score.
The BIT Questionnaire scales were based on the items with the highest component loading in sample B and labeled according to the wording of the items as follows: fear of injection and self-testing, expectations regarding positive insulin-related outcomes, expected hardship from insulin treatment, stigmatization by insulin injections, and fear of hypoglycemia. Mean values were computed for each scale with a value range 1–10 for each scale (Table 3).
In addition, a sum score was created that summed up the values of the items of the BIT Questionnaire (items of the scale “Expectations regarding positive insulin-related outcomes” were inverted first). No ceiling effect could be observed, which was demonstrated by the percentages of patients with maximum values between 0% for the sum score and 3.8–18% for the different BIT Questionnaire scales.
Acceptable coefficients of reliability were observed for scales 1, 3, and 5, with Cronbach's α ranging from 0.78 to 0.85 and the total sum score with α = 0.78. Scales 2 and 4 showed only moderate values (α = 0.62 and 0.66, respectively). However, taking into consideration that each of these scales contains only three items, the results can be viewed as adequate.
Wording and statistical measures of the BIT Questionnaire items.
The analysis of the changes in reliability coefficients for the respective scales that would result if the single item were eliminated showed that only item three would result in a significant increase in reliability if it were eliminated (Table 3). Since the scale already demonstrated good reliability, we decided to keep the item in order to consider also the fear of blood glucose measures together with the fear of injection. This was done because all items measure aspects of “blood and injection phobia,” and we think that patients expect that insulin treatment requires more frequent testing of blood glucose. Item four was kept as well, since its elimination would yield only a marginal advantage (α = 0.68 instead of 0.66).
Intercorrelation of BIT Questionnaire scales.
The concept of the BIT Questionnaire assumes intercorrelated subscales, since each is intended to measure a different aspect of the same construct (barriers to insulin treatment). The intercorrelation of subscales was very low to moderate, with Pearson's r between −0.05 and 0.36.
The results of the t tests indicated a clear predictive validity of all BIT Questionnaire scales. Patients who opted for OADs consistently reported significantly higher barriers to insulin treatment than the patients who were willing to move onto subcutaneous insulin. The magnitude of these differences between both groups was medium for the subscales and strong for the BIT Questionnaire sum score according to Cohen's criteria (see Table 4) (27). No substantial increase in the predictive power of the BIT Questionnaire was obtained in a stepwise logistic regression analysis when including items that were excluded from the final BIT Questionnaire (data not shown).
Our report describes the development and evaluation of a questionnaire that measures barriers to the acceptance of insulin treatment in orally treated patients with type 2 diabetes. The BIT Questionnaire was developed first in a sample of 448 patients; the development was based on principal component analyses. Our results yielded an easily interpretable five-component solution based on only 14 items. These components were used to define the following subscales: fear of injection and self-testing, expectations regarding positive insulin-related outcomes, expected hardship from insulin treatment, stigmatization by insulin injections, and fear of hypoglycemia. In addition, an overall sum score of all values was calculated in order to summarize the 14 items of the BIT Questionnaire in a single score. The subscales of the BIT Questionnaire address a wide range of the most important psychological barriers to insulin treatment, as those barriers are described in the current literature (9,10).
In a next step, a cross-validation of the structure of the BIT Questionnaire was performed based on data from an independent sample of 449 insulin-naïve patients with type 2 diabetes. This cross-validation used the 14-item BIT Questionnaire developed in the first sample. The results of both the exploratory principal component analyses and the confirmatory factor analysis confirmed the structure of the BIT Questionnaire. The scales demonstrated adequate reliability and validity. The frequency distribution of the scores revealed a fairly normal distribution, with no ceiling effect; this distribution is especially important when assessing changes in retest measures (12).
We successfully performed a linguistic validation of the German BIT Questionnaire into the English version presented here in order to makes it possible to pool and/or compare the results obtained in Germany across English-speaking countries. This certifies that the language versions obtained are conceptually equivalent, culturally relevant, and acceptable to the target populations.
A possible limitation of our approach is that some of the topics that were excluded from scale construction may be highly relevant for some patients. On the other hand, we chose to keep in the BIT Questionnaire only those items that demonstrated a clear and unambiguous component loading. Another limitation at this stage of development of the BIT Questionnaire is that the test-retest reliability and sensitivity to change of this new instrument still need to be determined.
In summary, the 14-item BIT Questionnaire has gone through rigorous empirical development and offers reliable psychometric properties as well as an interpretable and relevant component structure. Our findings suggest that clinicians and researchers now can use this instrument in a valid and reliable way to assess and address psychological barriers to insulin treatment in insulin-naïve patients with type 2 diabetes.
This study was supported by Pfizer.
Parts of this study were presented in abstract form at the 66th Scientific Sessions of the American Diabetes Association, 2006; at the 41th Annual Session of the German Diabetes Association, 2006; and at the 42nd European Association for the Study of Diabetes Annual Meeting, 2006.
Published ahead of print at http://care.diabetesjournals.org on 15 June 2007. DOI: 10.2337/dc06-2042.
F.P. has received consulting fees from Pfizer Germany. T.F. and A.P. have received honoraria and grant/research support from Pfizer Germany.
Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/dc06-2042.
A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C Section 1734 solely to indicate this fact.