Predictors of dropout in the German disease management program for type 2 diabetes

In Germany, disease management programs for type 2 diabetes (DMP DM 2) were introduced on a nationwide scale in 2003. All German statutory health insurance (SHI) funds, which cover about 85% of the German population [1, 2], are required, by law, to offer these programs, though participation is voluntary for doctors and patients. Over the years, DMP participation has been increasing continuously. To date, there are about 3.4 million patients enrolled in the DMP DM 2 in Germany, which are approximately 40% to 60% of all diabetic patients in Germany [3, 4]. Key components of German DMPs are evidence-based treatment guidelines, patient training, the regular documentation of disease development and treatment goals, as well as provider feedback. Usually, general practitioners (GPs) act as coordinating physicians: they enrol patients into the program, organize patient training, and negotiate individual treatment goals with patients [5, 6].

The evaluation of German DMPs is a legal requirement: it is carried out according to a pre-determined design based on data from routine documentation forms that doctors have to submit at regular intervals for every enrolled patient [7, 8]. This official evaluation program has been criticized for its methods, such as the lack of a control group, which makes it difficult to assess effectiveness of the programs [9, 10].

Recently, several studies showed improvements in care, such as adherence to treatment guidelines and the completion of regular routine exams, as well as patient and clinical outcomes such as quality of life and mortality [5, 11‐14] (but also see [15]).

However, in many of the studies potential biases are insufficiently addressed. One aspect that has been ignored in the official evaluation program, as well as in some of the recent studies, is the rate of attrition [5, 8, 14]. Understanding the mechanisms of patients' dropout from DMPs is important to evaluate potential bias due to attrition in studies that look at the effectiveness of DMPs [16, 17]. Furthermore, improved knowledge of how many and which kind of patients are likely to drop out of the DMP is important to decide whether and how the design of these programs could be improved to avoid losing patients who could potentially benefit from disease management.

Therefore, in this study we wanted to specifically focus on dropout from DMPs. We aimed to investigate, firstly, how many people who initially enrolled in a DMP left the program again at a later time. Secondly, to better understand the underlying reasons, we looked at which variables assessed at the time of enrolment were predictive of later dropout.

Overall, 11,933 patients enrolled in the DMP Diabetes mellitus type 2 between 1.7.2004 and 31.12.2005. 304 patients died and 443 patients left the insurance fund before the end of 2007 - they were excluded from the analysis. Of the remaining 11,186 patients a further 197 cases were excluded due to missing data (see figure 1). From the final group of 10,989 patients, 5.5% (604) dropped out of the DMP within the observation period.

In this study we identified a number of factors measured at the time of patient enrolment into the DMP DM 2 that were associated with later dropout from the program. We found that patients from North Wurttemberg tended to dropout from the DMP more frequently compared to patients from other regions. This result might reflect the different levels of acceptance of DMPs in different regions in Germany [20]. In the region North Rhine for example, similarly structured programs for the management of diabetes had existed long before the current DMPs were introduced [21]. Here, the acceptance of disease management programs was high and implementation quick. Most patients who had participated in the structured diabetes programs were then enrolled in the DMP DM 2. On the contrary, in North Wurttemberg, resistance against the introduction of DMPs from doctors was high, which might have also affected continuity in patients' DMP participation. Dropout was further associated with insurance status, number of secondary diseases, presence of a disabling secondary disease, treatment goals requiring behavioral/lifestyle changes (such as the recommendation to stop smoking or seek nutritional counselling) and the completion of regular exams required by the DMP guidelines. Based on an additional analysis stratified according to insurance status (see table 5), we could observe different trends for the group of patients who were still working and those who were retired. Among the employed, those who were younger, healthier and recommended by their doctor to change their lifestyle were especially at risk of later dropping out of the DMP. Among the retired however, predictors of dropout were older age, the presence of a disabling secondary disease, an abnormal foot status or routine exams not being carried out. These could possibly reflect two very different reasons for dropout. First, for younger people who are still fairly healthy and pre-occupied with their working life, DMPs might not be perceived as beneficial, but rather the requests to make changes to their lifestyle might be perceived as an inconvenience. Second, retired people of older age might also be at higher risk of dropout, since other health-related issues might obtain higher priority; patients might be close to the end of their life, spend more time in hospital or move to long-term care. The observation that not carrying out the required routine exam(s) was also associated with dropout in this group could possibly also be a reflection of the doctor's decision that due to other circumstances in the life of the patient, these exams might not be necessary anymore. Further research would be needed to explore these trends of dropout in more detail.

Overall, we found the percentage of patients dropping out of the DMP DM 2 to be low (5.5%) in our study. However, we only followed patients over a short time period and our ability to measure dropout accurately might have been limited. Dropout rates in international studies on disease management vary widely. In a recent meta-analysis including 41 randomized controlled trials on disease management interventions with varying durations from 1.5 to 48 months, dropout rates ranged between 1.1 and 39.8% (mean = 14.0%) [22]. The results from this meta-analysis refer to dropouts of clinical trials, which might not reflect real-world conditions and might therefore not be comparable to dropout from disease management programs outside of study conditions. However, similarly varied results can be found in studies looking at attrition rates from diabetes education services, either at physician-led diabetic clinics or diabetes education centres usually lead by diabetes educators. In studies carried out in Britain, the United States, Canada, Ireland, and Japan, dropout rates ranged from 4 to 57% per year [23, 24]. Due to the diversity of programs as well as variations in assessing attrition, comparison between studies is difficult. Enrolment in the German DMPs is voluntary. Furthermore, inclusion criteria for the German DMPs are the willingness and ability of the patient to actively participate in the programme and an expected benefit for the patient in terms of an improvement in quality of life and life expectation. The GP as coordinating physician decides about a patient's suitability for the programme. Therefore DMP participants are likely to be a selected group in that they might be more motivated and able to reach their treatment goals, take responsibility to self-manage their chronic disease, and are adherent to treatment recommendations. Patients with low motivation or ability to follow recommendations to improve the management of their disease would be enrolled in the DMP less frequently because of a lack of potential benefit. This could explain why the overall rate of drop-out was low in our study.

Regarding predictors of dropout, our results are consistent with the results from international studies. As in our primary analysis, most studies did not find a relationship between attrition and gender, age or diabetes duration [24]. If we stratified for retirement status, however, we found an association between attrition and age. As in our results, retirement status itself was significantly related to dropout in other studies [24, 25]. Also, in qualitative studies, in which patients were asked why they discontinued program participation, conflicts with work was a reason frequently given [23]. Other reasons given were low perceived concern for the disease, not considering attendance important, not feeling sick, but also having too many other health conditions [23, 25]. In our data we found a similar trend, as patients suffering from a disabling secondary disease as well as patients with fewer secondary diseases were more likely to dropout of the program. The association of dropout with recommendations to stop smoking or seek nutritional counselling has not been explicitly investigated in other studies; however, results on smoking status and dietary treatment in relation to the likelihood of dropout were contradictory.

Other variables, which have been shown to be related to patient dropout and were not addressed in our study, were socio-economic status, patients' perceptions, attitudes and beliefs as well as variables related to the program provider, such as physician-specific variables, practice structure and organization [23‐25].

Overall dropout from the German disease management program DM 2 seems to be low, but variables assessed at the time of enrolment are predictive of future dropout. We could identify two trends of dropout: firstly, retired patients of old age, who possibly drop out of the DMP due to other health care priorities towards the end of life and secondly, employed people of younger age who have not yet developed many secondary diseases, but were recommended to change their lifestyle. Both patterns of dropout are important for the evaluation of the effectiveness of DMPs, since they could bias the results if not taken into account appropriately. The second trend should be of particular interest for the design of DMPs, since it applies to a group of patients who could benefit the most from preventive measures offered as part of the DMP. Further research is needed to investigate in more detail to what extent the DMP DM 2 can successfully include patients at earlier stages of disease and if this may help them to take preventive measures to avoid disease progression.