Prognostic value of physicians' assessment of compliance regarding all-cause mortality in patients with type 2 diabetes: primary care follow-up study

Type 2 diabetes mellitus (DM) is the most common endocrine disease in developed countries and prevalence is increasing worldwide [1]. In Germany, about 5% of the population are affected and the prevalence is increasing dramatically with age [2]. Patients with type 2 diabetes have an approximately threefold risk for all cardiovascular diseases [3, 4] and their risk of death from all causes is increased by 75% compared to patients without diabetes. The morbidity and mortality resulting from complications of type 2 diabetes such as micro- and macrovascular disease cause considerable costs for the individual and the society [2, 5] and necessitate special care and attention [6].

It is well known that the likelihood of diabetic complications, and subsequently risk to die, in patients with diabetes is associated with glycemic control [7‐11] and adherence to treatment and successful self-management. Whether the physicians' assessment of patient compliance is a valuable prognostic information to identify patients who are at increased risk for subsequent complication or death, or merely reflects measurement of various treatment parameters such as HbA_1C or other laboratory markers is unclear yet. However, as most patients are treated in primary care setting, this information would be critical and would eventually allow to identify subjects in need for improved diabetic therapy and help to target preventive measures.

The objective of this prospective cohort study including all patients with type 2 diabetes who were under medical treatment in the primary care practices of family physicians and internists in a geographically defined region was to investigate various aspects and the prognostic value of the physicians' assessment of the patient compliance with respect to all-cause mortality during a one year follow-up period.

In total, 1014 patients with known DM aged 40 and over were included in this study. Response among all eligible patients with DM was 88%. 882 (87%) of the patients presented themselves in the 11 participating general practices and 132 (13%) were seen in nursing homes or visited at home by the physicians.

As table 1 shows, 55% of the patients with DM were women. Mean age of patients with DM was 69.0 years (SD 10.4, range 40 to 91 years), two thirds of them (66%) were between 60 and 80 years old, and 65% were married. The time since diagnosis was less than 5 years among 34% of patients, 5 to 10 years among 32% of patients, and more than 10 years among 30% of patients. The physicians' assessment of patient compliance was in general more favourable for the patients presenting at the surgery. Overall, 18% of patients were judged as "very good", 45% as "rather good", 31% as "rather bad", and 6% as "very bad" regarding their compliance.

During the one year follow-up period 48 (4.7%) of the patients died. Table 2 shows the all-cause mortality according to various factors. Mortality was much higher in patients visited in nursing homes or at home compared to patients presenting at the surgery (18.2% vs. 2.7%. In bivariate analysis gender was not associated with all cause mortality. However, mortality strongly increased with age and was highest in widowed subjects. Time since diagnosis of DM, and presence of macrovascular or microvascular complications at baseline were clearly associated with a higher proportion of subjects who died during follow-up.

The physicians' assessment of patient compliance during baseline was also associated with an increased mortality (see table 3). Patients who's compliance was judged at baseline as "very bad" showed a 1-year mortality of 13.1% compared to a 1-year mortality of 4.2% ("rather bad"), 3.8% ("rather good"), and 4.0% ("very good") in the respective categories (p = 0.01). By contrast, a high HbA_1C-value (≥ 8%) and a cholesterol value ≥ 257 mg/dl (≥ 8^th percentile) at baseline showed no statistically significant association with all-cause mortality during the one year follow-up (see table 3).

We then used unconditional logistic regression and a stepwise backward selection procedure to find the most parsimonious model of factors associated with all-cause-mortality (see table 4). Patients seen at nursing homes or visited at home had a higher odds to die within the follow-up period compared to patients presenting at the surgery (OR 3.92, 95% CI 1.85–8.30). In addition, female patients showed significantly lower odds to die compared to males (OR 0.34; 95% CI 0.17–0.67). In addition, there was a clear increased risk to die with increasing age. Furthermore, the presence of macrovascular complications was a statistically significant determinant of subsequent risk to die (OR 3.17, 95% CI 1.57–6.39). The physicians' assessment of the patient compliance was also an important factor associated with all-cause mortality. Patients whose compliance was assessed as "very bad" had an OR of 2.67 (95% 1.02–6.97) compared to patients assessed as "very good" or "rather good". The fit of the model did not improve and the observed association did not change if HbA_1C-values, total cholesterol, prescription of ACE-medication or lipid-lowering drugs, or the blood pressure values from baseline were added in the model (data not shown). None of the latter four variables met the criteria for inclusion in the model.

Finally, we carried out a small survey among the fifteen physicians to find out which aspects of compliance were most important for them when making this assessment on a individual basis. For most of them self-acceptance of disease, treatment adherence, patient's interest in physician's explanations, attendance at appointments, a good self-management, and a good physician-patient relationship were key elements. Patient's co-morbidity, a high HbA_1C value, a good social support, and quality of health related life were other, sometimes less important elements in their decision making.

In this study physicians' assessment of patient compliance was found to have an independent prognostic value for subsequent risk of death, which persisted after control for other factors. Compliance may reflect many aspects of successful self-management of diabetes and the identification of patients in need of improved compliance may help to target preventive measures.

It is well accepted that an optimal glucose control, adherence to a treatment regimen, and successful self-management minimizes the likelihood of diabetic complications, and subsequently, the risk of death [7‐11]. However, many different components seem to be related to successful self-management of this chronic disease; provider characteristics, the patient-physician relationship, and patients' personal characteristics and beliefs, as well as environmental, social, and cultural factors act as important contributors to this complex issue [12]. Although physicians' assessment of patient compliance appears to be strongly related with glycemic control [13], its prognostic value was found to be independent of glycemic control in this study and prognostically more relevant than gylcemic control.

The key element in management of chronic diseases is self-management. Unfortunately, many patients are not willing or able to take full responsibility for their care [14]. Successful treatment should be regarded as a team approach, incorporating patients, care givers and family members and the social context of the individual [15]. Factors such as a good team spirit in medical practice can improve management of patients with chronic diseases [16]. However, attempts to increase the frequency of appointments or the use of other measures in patients with poor compliance may be counterproductive. Many of these patients will react with even more resistance. A promising approach may be to strengthen the "sense of coherence" by taking aspects of comprehensibility, manageability and meaningfulness from the patients' social and personal perspective into account [17]. As earlier work suggested individual differences in perceived control and coping strategies may strongly interact with management of chronic diseases [18]. However, clearly more studies especially in primary care setting are needed to further determine which factors pose obstacles to an integrative conception of a successful treatment of chronic diseases and what compliance means from the physician's as well as the patient's view. As shown by the cross-national Diabetes Attitudes, Wishes and Needs (DAWN) study, psychosocial problems in patients with diabetes are common according to reports of patients and providers, and are perceived as important barriers to effective self-care. The regimen adherence of patients was less than optimal in all ten participating countries (Germany included) and the providers believed that mainly psychological problems were the cause. However, providers also reported that they often do not have the critical resources for addressing these problems [19].

If looking at the results of this study the following strengths and weaknesses should be considered: Relatively few patients were assessed as having a "very bad" compliance and only for this group the prognostic relevance was evident. Unfortunately, we had no comparison with another related sample or norm groups. Furthermore, we did not measure a specific component of compliance such as treatment adherence (e.g. pills counts) or the compliance with a special diet. However, we measured physicians' assessment of the patient compliance as an integrative measure (including also aspects of self-management behaviour) and from our small qualitative survey among the physicians we got important clues which aspects of compliance were most important for them when making this assessment on an individual basis. Notably, outcome parameters such as HbA_1C, the social network, and the health related quality of life were less important elements in their decision making with respect to assessment of patient compliance.

This study was conducted in the framework of a quality assurance program of the participating practices including all patients within a defined geographic region, irrespective of their insurance affiliation. As in Germany almost all patients have access to health care, as all primary care physicians within a coherent region participated, and as most patients with DM usually consult their primary care giver once in three months, we are confident that the study population comprised almost all patients with type 2 DM in primary medical care in the study region. However, patients not seeking medical care and patients hospitalized were certainly underrepresented in this study.

Having the aforementioned caveats in mind, our study suggests that the primary care physician's assessment of patient compliance is a valuable prognostic marker for the risk of subsequent death among patients with type 2 diabetes. Identification of patients in need of efforts to improve compliance may help to target preventive measures.