Psychological and behavioral functioning was measured in a large sample of people with type 2 diabetes who were living in France by using a scale specific to diabetes. The original DHP-18 is a validated instrument for use with people with type 2 diabetes . Concerning the French version, in line with Meadows' preliminary results (based on the responses of people with type 1 & 2 diabetes) , analyses performed only on people with type 2 diabetes support its validity. Scores observed in our sample were close to scores observed in people treated by insulin, OHAs or diet in previous studies [17, 19]. As in our study, insulin therapy was associated with a marked decrease in psychological functioning.
Many factors were associated with the three DHP-18 dimensions in this sample: some were specific to the illness, some were related to socio-demographic factors, and some to health-related behavioral factors. The effect size, a distribution-based indicator, was calculated to determine whether a difference could be considered as important . Accordingly, major negative associations (at least a difference of 5 between groups in the DHP score, corresponding here to small to medium effect size) were observed for major microvascular complications (effect size of 0.18) -with little or no effect for macrovascular complications- and for severe hypoglycemia (effect size of 0.24), insulin treatment (effect size from 0.20 to 0.32 according to the DHP score), non-adherence to the treatment (effect size of 0.45), increasing weight (effect size of 0.18), at least one psychiatrist visit (effect size of 0.25), and surprisingly no alcohol consumption (effect size from 0.19 to 0.25 according to the DHP score). Finally, universal medical insurance coverage (effect size from 0.20 to 0.46 according to the DHP score), which, in France, permits free access to medical care for people with a low socioeconomic level, was negatively associated with psychological functioning, suggesting a higher toll of diabetes in people with low socioeconomic level. According to Cohen, effect sizes of 0.2, 0.5 and 0.8 are considered as small, medium and large, respectively . Therefore, in summary, the effect on PBF of non-adherence to treatment and universal medical insurance coverage can be considered as medium. The effect of insulin treatment, severe hypoglycemia, at least one psychiatrist visit and no alcohol consumption can be considered as small.
Our results are in line with those of others studies, that found a negative impact on HRQoL (especially on psychological functioning) of the presence of complications, comorbidities, depressive symptoms, insulin use, high BMI or a lower educational level [7–9, 19, 21]. Maddingan et al found that comorbidities (in particular depression and stroke) and markers of socioeconomic status were important factors related to HRQoL as measured by the health utilities index mark 3. To some extent, a lower educational level, longer diabetes duration, insulin use, higher BMI and non-practice of physical activity were also negatively associated with HRQoL in this study . Others underlined the negative role of micro- or/and macro-vascular complications on HRQoL . Surprisingly, compared to the impact of microvascular complications, the effect of macrovascular complications was limited (about a 3-decrease only in the BA dimension). Macrovascular disease was defined in this study as myocardial infarction or angina, or heart arteries surgery, and no distinction could be performed between myocardial infarction and angina. Moreover, no information about the time of occurrence was available. The presence of both complications was associated with a more important decrease than a single type of complication.
Diabetes complications contribute to excess morbidity and mortality and generate substantial costs. In order to provide timely treatment, it is essential that patients at risk for the development of diabetes complications are identified as early as possible. The normalization of factors such as blood pressure, blood cholesterol and plasma glucose can prevent or delay diabetes complications .
We found a major negative association between the DE score and having had at least one visit to the psychiatrist in 2001, which might reflect mental health problems. In particular, depression in diabetes can be approached in terms of depressed mood and anhedonia, cognitive symptoms and anxiety . Depression is a well-recognized determinant of HRQoL in diabetes [10–12]. Identification and thus optimal care of depressive symptoms is important as depression is associated with poor diabetes self-management, an increased risk for complications, a lowed use of health services, increased functional impairment and distress may well impact the course of the illness [11, 23, 24].
Surprisingly, compared to no consumption, daily alcohol consumption was associated with better psychological functioning (BA and PD dimensions). There is some evidence that moderate alcohol consumption is associated with a lower risk of mortality and coronary heart disease in people with type 2 diabetes ; other studies suggested a beneficial effect of moderate alcohol consumption on glycemic control but this has not yet been demonstrated. Moderate alcohol consumption may also be a marker of specific psychological profiles.
The most important deficits observed in our study suggest that prevention and management of modifiable factors (for example microvascular complications or adherence to treatment) could be essential to preserving or improving psychological and behavioral functioning among people with type 2 diabetes. A specific approach to type 2 diabetes management may also be required in groups with a low socioeconomic profile (i.e. people with a low educational level or for those who are on universal medical insurance coverage) or other non modifiable factors. Data issue from ENTRED showed that people with a lower socioeconomic status have more frequent macrovascular complications and a lower quality of diabetes care . Efforts to improve the prevention of complications, therapeutic education and diabetes management are required in this vulnerable population.
Diabetes management is complex and should be empowered early. To achieve this goal, a multidisciplinary approach is required. Medical interventions are needed in order to address a broader spectrum of outcomes such as patient-reported outcomes (e.g. HRQoL), personal models of illness and empowerment. The promotion of self-management using strategies that take into account the adaptation to the illness and its treatment (stress management for self-care of the disease, psycho-behavioral methods, psychosocial support...) that are not only limited to drug therapy, should be encouraged . Four groups of factors accounted for most of the variability in self-care behavior in patients with diabetes: patient characteristics, the patient family, the practitioner and the health system, and the community/work setting . Integrating such factors in the disease management can only be beneficial for patients.
There are some limitations in this study. The cross-sectional design did not allow us to examine any causal effect. Data were self-reported and people could under or over-report any of the collected conditions. However, when analyzing the same data collected through a medical questionnaire (sent to the responders' physician when its address was reported on the participant's questionnaire), Romon et al compared the prevalence of the macrovascular complications based on both the patients and physicians declarations . Estimates were similar overall, whether they were reported by patients or physicians. Moreover, Martin et al found that self-reports are reasonably accurate for certain chronic conditions and for routine screening exams . The algorithm used to distinguish between responders with type 1 and type 2 diabetes may have contributed to some misclassification of the different types of diabetes. This could affect the validity of our results if the factors associated with PBF differed between the two diseases.
External validity of our results is another important question. People treated with diet only could not be included because the sample selection was based on the database of the national health insurance system, which covered only reimbursed medical prescriptions. They may represent about 10% of people with type 2 diabetes  and have been reported to show similar HRQoL as people treated with OHAs . Our sample showed similar characteristics with other French samples in terms of age, sex, educational level, BMI, smoking [31–33]. The response rate was about 40%, a level generally observed in this type of mailing survey. The characteristics of the responders and the non-responders were compared using medical claims, available for all . In our sample, responders were younger and more often male, more frequently treated with insulin or with several OHAs than with a single OHAs, less frequently treated for a cardiovascular disease and received an overall better quality of diabetes care . Our results may therefore underestimate the true impact of diabetes on PBF.