Resistance to Insulin Therapy Among Patients and Providers

RESEARCH DESIGN AND METHODS

Data are from the multinational Diabetes Attitudes, Wishes, and Needs (DAWN) study (38,39). All data are self-reports gathered during 2001 by structured interviews conducted face to face or by telephone (depending on telephone availability). Interviews took 30–50 min to complete. Interviews were conducted in 11 regions (representing 13 countries): Australia, France, Germany, India, Japan, the Netherlands, Poland, Scandinavia (divided equally among Sweden, Denmark, and Norway), Spain, the U.K., and the U.S.

The study was conducted with three independent samples. Two samples consisted of providers who were treating at least five diabetic patients per month; only one provider was selected from a practice. The physician sample consisted of 2,705 respondents, with a quota of 250 per region, 200 in primary care and 50 specialists (endocrinologists and diabetologists with 2 years experience and treating >50 patients per month). Of this sample, 2,681 treated patients with type 2 diabetes and were used in the analyses (Table 1). The nurse sample consisted of 1,122 respondents, with a quota of 100 per region, 50 specialists (those caring for 50 or more patients with diabetes a month), and 50 generalists. Of this sample, 1,109 treated patients with type 2 diabetes and were used in the analyses (Table 1).

The third sample consisted of adults with self-reported type 1 or type 2 diabetes, with a quota of 500 per region (Table 2). Sample quotas were established to obtain equal numbers of people with self-reported type 1 and type 2 diabetes. This study examines only the 2,061 respondents who were not currently using insulin.

A variety of sampling frames was used in different countries to generate heterogeneous samples from the entire country (except India, where the sample was limited to five regions). Patients were recruited via random-digit dialing, clinic and practice rosters, and survey panels, supplemented by advertisements and snowball recruiting. Providers were recruited from professional directories and various listings.

Provider measures

The primary dependent variable for providers was insulin initiation delay (“I prefer to delay the initiation of insulin until it is absolutely essential”). In addition to age, sex, practice location, and years in practice, the following provider characteristics were assessed: professional roles (whether the provider was a specialist or primary care provider and the degree to which she/he was an opinion leader); percent of diabetic patients treated who had type 2 diabetes; perceived patient psychological distress; perceived patient adherence to diet, exercise, medication, self-monitoring of blood glucose (SMBG), and appointment scheduling recommendations; providers’ perceptions of patient attitudes toward insulin; provider beliefs about insulin efficacy and financial barriers to insulin therapy; and provider attitude toward delay of oral hypoglycemic medication. Provider characteristics and details concerning the measures appear in Table 1.

Patient measures

The primary dependent variables for patients were two attitudes toward insulin therapy: perceived efficacy of insulin (“Using insulin would help me to manage my diabetes better”) and self-blame for needing insulin (“Starting insulin would mean that I have not followed my treatment recommendations properly”). In addition to demographic and disease characteristics, other patient characteristics assessed include self-reported adherence to medication, SMBG appointment scheduling, diet, and exercise recommendations; perceived diabetes control; diabetes-related distress; and quality of relationship with provider. Patient characteristics and details concerning the measures appear in Table 2.

Statistical analysis

Statistical analysis used multiple regression models incorporating effects for country and respondent characteristics. This strategy identified country differences after adjusting for differences in the composition of country samples. The U.S. was the reference category for all analyses; country coefficients are interpreted as differences from the U.S. Analyses were conducted separately for each type of respondent (physicians, nurses, and patients). For both patient and provider analyses, country and individual variables were forced into the model with the following exceptions. For patient analyses, self-care adherence, perceived diabetes control, distress, and patient-provider relationship were entered only if significantly (P < 0.05) related to insulin efficacy or self-blame. For provider analyses, perception of patient attributes (adherence, distress, and insulin attitudes) and provider attitudes were entered only if significantly (P < 0.05) related to delay. For these exceptions, if a variable is in the final model, it was significant upon entry; any reduction in the relationship is due to mediation by a variable entered subsequently. Variables forced into the models that later became nonsignificant are indicated by a symbol in the tables. Unstandardized coefficients can be compared across models, and their significance levels are presented; standardized coefficients are presented so that the magnitude of relationships within each model can be compared. All analyses were performed with SPSS version 11.5 (SPSS, Chicago, IL).

RESULTS

Respondent characteristics for providers are shown in Table 1. Physicians were mostly men, while nurses were almost exclusively women. Physicians and nurses had similar views of patient adherence and patient attitudes toward insulin therapy, as well as similar attitudes toward insulin and oral hypoglycemic medication. Delay of insulin and oral therapy was high for both respondent groups, with insulin delay being higher.

Respondent characteristics for patients are shown in Table 2. Patients were on average middle aged, almost equally divided between men and women, and had diabetes for an average of ∼8 years. Patients reported higher levels of adherence for medication and appointments than for diet and exercise. Belief in the potential efficacy of insulin was low, and self-blame for having to take insulin was high.

Provider attitudes

Results for the analysis of provider orientation toward delay of insulin therapy are presented in Table 3. Total variance explained was 34% (physicians) and 54% (nurses). Delaying was higher among providers in the U.S. than in most other countries. U.S. physicians were significantly more disposed to delay insulin therapy than physicians in all other countries, except for India and Japan. U.S. nurses were significantly more disposed to delay than nurses in Australia, Scandanavia, and the U.K.; Only Indian nurses were significantly more disposed than U.S. nurses to delay insulin therapy.

Delay of insulin therapy was significantly less among specialists and opinion leaders and those who believed insulin was efficacious. Delay of oral hypoglycemic medication was the strongest correlate of insulin therapy delay and accounted for most of the explained variance (20% for physicians and 30% for nurses). Physicians (but not nurses) with higher insulin delay reported significantly higher levels of patient adherence to recommendations for medication. Nurses (but not physicians) with higher insulin delay reported a significantly lower percentage of their diabetic patients as having type 2 diabetes. For neither type of provider was insulin delay associated with perceived patient self-blame for insulin therapy.

Patient attitudes

Results for the analysis of patient attitudes toward insulin therapy are presented in Table 4. The total explained variance in perceived clinical efficacy was 14%. The total explained variance in self-blame was 9%.

U.S. patients reported lower belief in insulin efficacy than patients from all other countries and significantly lower belief than patients from Germany, India, Japan, Scandinavia, and Spain. Patients who reported high belief in insulin efficacy were significantly younger and had more complications. They also reported significantly poorer relationships with their providers, poorer perceived control, more SMBG adherence, and more diabetes distress.

U.S. patients reported more self-blame for insulin therapy than patients from all other countries and significantly more than patients from Australia, Germany, India, and the Netherlands. Patients who reported higher levels of self-blame for insulin therapy were significantly younger and had less duration of diabetes. They also reported significantly less exercise adherence and more diabetes distress.

CONCLUSIONS

Provider attitudes

Physicians and nurses in a given country tended to have similar attitudes toward delay of insulin therapy. For example, the belief that insulin therapy should be delayed until absolutely necessary was strongly held in the U.S. among both physicians and nurses; only in India was the tendency to delay higher (significantly so for both physicians and nurses before controlling for attitudes toward delaying oral medication). Although we cannot link specific patients to their own providers, this pattern closely resembles that observed for patients’ attitudes toward insulin; U.S. patients were among the lowest in perceived insulin efficacy and among the highest in insulin self-blame. This suggests that beliefs about insulin are related to the cultures and health care systems of the different countries, and understanding beliefs about insulin will require an understanding of how these factors operate.

Delay in prescribing oral blood glucose–lowering medication had the strongest relationship with delay in prescribing insulin. Providers tend to have a general orientation toward using glucose-lowering medications in managing diabetes. This orientation could reflect a lack of awareness about the need to keep glucose levels as close to normal as possible or it could reflect a more general orientation toward delayed prescribing of other medications for patients with diabetes or medications for other disorders. In any case, we know that propensity to use medications is an important determinant of prescribing behavior (30–32), which suggests that efforts to facilitate the timely prescription of insulin may need to address the more general reluctance to prescribe diabetes medication (or medication generally).

Clinical efficacy was the insulin-specific belief most strongly associated with provider inclination to delay insulin therapy. The scores on this measure were relatively low (just over half of physicians and nurses agree that insulin can have a positive impact on care), suggesting that efforts to improve awareness of efficacy are required. Although this belief was not strongly related to delay, other research (33,34) suggests that attitudes toward specific medications do influence prescription decisions. We believe that a more comprehensive measure of insulin attitudes would be more strongly predictive of insulin delay.

Specialists, opinion leaders, and those who treat a higher percentage of patients with type 2 diabetes were less inclined than nonspecialists to delay initiation of insulin, but belief in the benefits of insulin therapy was low among specialists and generalists alike. Specialists and opinion leaders are more likely to adopt new prescribing practices (40), and they can influence the practice behaviors of other providers (41,42), so the present study suggests that efforts to change provider beliefs might focus first on these groups.

Some have suggested that the prescribing behavior of providers is based on their perceptions of patients’ expectations (36,37), while others believe that providers are relatively autonomous in their decisions and may not be sensitive to the wants and needs of their patients. We found that providers’ tendency to delay insulin therapy was not strongly associated with their perception of patient beliefs about this therapy. While our findings suggest that providers’ general orientation to initiating insulin therapy did not take into account patients’ concerns, it is possible that a provider’s perception of the concerns of a given patient may affect prescribing for that particular patient (a hypothesis that this study could not address because provider perceptions were not patient specific). In contrast to patient beliefs, patient behavior was associated with prescription practices; providers were more willing to delay insulin initiation if they saw their own typical patients as more adherent to their medication or appointment regimens. Perhaps this reflects the hope that oral medications can be used to achieve adequate control if patients will take the medication or that closer teamwork between patient and provider will obviate the near-term need for insulin. Future research will be required to understand how provider perceptions of patient adherence are related to insulin delay.

In general, providers underestimated the number of patients who blamed themselves regarding the need for insulin. Other analyses of the DAWN data indicate that patients’ self-blame is associated with their greater worry about starting insulin (43). Providers can help patients who blame themselves for needing to initiate insulin therapy by making sure patients understand that diabetes is a progressive disease. Providers can also reduce self-blame as a barrier to insulin therapy if they avoid the strategy of encouraging more active self-care by pointing to insulin as a consequence of inaction. This strategy, which over half the providers in our study reported using, can make it more difficult to initiate insulin therapy subsequently.

Patient attitudes

Factors included in our study explained little variance in patients’ beliefs about insulin efficacy. However, there is a pattern to the relationships observed. Belief in insulin efficacy was stronger among patients who were in more negative situations: more distress, worse control, more complications, and poorer relationships with their diabetes care providers. As patients’ situations worsen, they may become more receptive to a treatment strategy they have previously avoided. Patients who perceived themselves as more adherent to SMBG also saw more benefit of insulin therapy, perhaps because more frequent monitoring made them more aware that their current regimens do not provide adequate glucose control. Alternatively, they might believe that more frequent monitoring will enable them to use insulin more effectively.

Although our study also explained little variance in patients’ self-blame for needing insulin, there is a pattern to these relationships as well. Those who report lower self-blame either have a biological rationale for why they need insulin (they are older and have had diabetes longer) or they may feel that they have done what they can to avoid the need for insulin (those more adherent to exercise have lower self-blame). Results (not shown) indicate that the adherence/self-blame relationship holds true if dietary adherence is substituted for exercise adherence in the model. In addition, dietary adherence has a marginally significant relationship with self-blame (P = 0.074) even with exercise adherence in the model. Thus, the greater the biological risk for needing insulin and the more done to take care of his/her diabetes, the less the patient feels that she/he is at fault.

Diabetes-related distress had the strongest association with insulin self-blame (as it did with perceived benefits of insulin therapy), with greater self-blame associated with more distress. Because the data are cross-sectional, we cannot say whether they are causally related, but it is likely that there is a reciprocal relationship in which distress increases self-blame and self-blame increases distress. Further research is needed to untangle these possibilities.

Limitations and implications

This study is cross-sectional, so its results show associations not causal relationships. In addition, providers and patients in the study were not linked, so we do not know whether provider perceptions of particular patients are related to their prescribing behavior for those patients. The sampling technique might not have generated representative samples of providers and patients; this could bias our findings in unknown ways. Finally, study outcomes are attitudes, and we do not know how strongly they are related to patient and provider behaviors, let alone clinical outcomes (such as A1C level or the development of complications). At this point, our explanations for the study’s findings remain largely speculative. Still, this study suggests a number of important implications for future research, educational interventions, and clinical practice.

Future research should consider influences on providers’ prescribing practices for oral hypoglycemic agents (given the strong association between these practices and insulin prescription practices). Our finding that many providers do not believe insulin can reduce the costs of diabetes care points to the need to better understand these attitudes. It would also be useful to know more about how providers interact with patients in the process of initiating insulin therapy. Finally, we need to understand what unmeasured factors account for most of the variance in patients’ attitudes toward insulin therapy and why diabetes-related emotional distress is strongly associated with a belief in the benefits of insulin.

Our findings suggest that educational interventions to make initiating therapy easier should enhance awareness of insulin efficacy and the role of insulin therapy in type 2 diabetes given the nature of progressive β-cell failure. These findings also suggest that providers seeking to facilitate the initiation of insulin therapy should avoid using insulin as a threat in an effort to encourage more active self-care and identify and address the specific beliefs that support an individual patient’s resistance to insulin therapy.

Appendix

The International DAWN Advisory Panel members are Ib. Brorly, Denmark; Ruth Colagiuri, Australia; P.H.L.M. Geelhoed-Duijvestijn, The Netherlands; Hitoshi Ishii, Japan; Line Kleinebreil, France; R.L., Germany; T.L., Denmark; D.M., U.K.; A. Ramachandran, India; R.R.R., U.S.; F.J.S., The Netherlands; and Giacomo Vespasiani, Italy.