The results presented above show that patients enrolled in the diabetes DMP receive a more evolved pattern of outpatient services both in terms of medical services provided by specialists and ambulatory health care centers as well as in terms of specific services as put forward by medical guidelines indicating an improved process quality of outpatient care.
As far as the overall number of outpatient services received (Table 2) is concerned, these findings are generally in line with other research: Villagra & Ahmed  present evidence for less frequent use of outpatient services with DMP patients accounting for significantly fewer office visits and Dall et al.  also observed that active participation in a diabetes DMP led to a decrease in the number of ambulatory visits. Contrary to the findings of Villagra & Ahmed  and Dall et al. , who also reported lower costs for overall outpatient care, the authors identified significantly higher costs for overall outpatient care of diabetes patients receiving treatment within the DMP (Table 5), which was also found by Buntin et al.  when analyzing insurers data on health care use and costs for more than 12,000 diabetics.
Analyzing the mere structure of the different outpatient services as presented in Table 2, it becomes clear that increased costs for outpatient care for the DMP group are mainly due to the higher share of enrolled patients receiving specific diagnostic services according to the treatment guidelines as opposed to the group of non-participating patients and is therefore in line with the intention of disease management. Higher frequencies of these services performed ultimately lead to higher frequencies of medical specialist and diagnostic services performed, which could also be observed by Sidorov et al. , who reported a higher number of outpatient office visits as well as higher figures of HbA1C testing and of lipid, eye and kidney screening for DMP patients. Similarly, other studies reported evidence for a significant increase in the frequency of specific examinations such as HbA1C testing, eye exams, foot exams and cholesterol testing due to participation in the DMP [5, 10].
The average number of inpatient stays as well as average length of stay (Table 3)reveals that patients enrolled in the DMP experienced fewer admissions (0.63 Vs. 0.80) and inpatient days (3.785 Vs. 6.03) per patient, reflecting a reduction in the latter of more than 40%. Again, these findings appear to be in line with the results presented by Sidorov et al. , Dall et al.  and Villagra & Ahmed et al. , who present evidence for a significant decline in inpatient days for DMP patients ranging from 20% to 40%. Considering the fact that the average hospitalization rates for regular diabetes patients observed in these US-based studies range between 1 and 1.5 days per patient per year [6, 7, 14, 15], it can be shown, that even within Austria’s hospital-driven health system, DMPs lead to similar relative reductions of hospitalization rates and hence appear to be able to counterbalance the adverse governance effects primarily put forward by the characteristics of the Austrian DRG-system .
As observed by Villagra & Ahmed  and Dall et al. , it is the overall reduction in inpatient days that represents the most important source of savings, which is also reflected by our results on overall cost effects of DMP enrollment as presented in Table 5. While participation in the DMP leads to substantial cost advantages for inpatient care due to steady figures on average days spent in hospital as opposed to substantial increases in hospital days for non-participating diabetics from 2006 to 2009, enrollment does not tend to have an overall impact on the amount and costs of prescribed medication. Villagra and Ahmed  and Sidorov et al.  observed similar effects suggesting that disease management is likely to increase the amount of medication due to higher adherence to pharmacological regimes on the one hand, while decreasing the amount of medication due to appropriate use and better disease control on the other.
Most important, in line with the results presented in recent studies on the effects of diabetes disease management in the UK  and in the US [6, 7], our findings indicate that even within the Austrian mandatory health insurance system, which has little competition and free patient choice in terms of outpatient care, type 2 diabetics currently enrolled in the DMP benefit from lower hospitalization rates as compared to non-participating patients. They also benefit from a different pattern of outpatient care promoting primary and specialist care associated with higher levels of medical assessments and tests according to guideline recommendations.
Yet one should acknowledge that participation in the DMP is still very low amongst type 2 diabetics covered by the Social Insurance Institution for Business and only reflects 2.5% of the target population in 2009. These low participation rates may be due to the specific characteristics of the insurer’s population, which mainly includes self-employed persons. Alternatively these figures may also reflect low overall participation rates throughout Austria, with the Social Insurance Institution of Styria leading the field with 8% of enrolled diabetics by the end of 2009 .
As put forward in the introduction, this study ultimately aims to evaluate the effects of diabetes disease management from an insurer’s perspective. The low DMP participation rates in Austria mean that health insurance providers seem cautious about promoting disease management. Even though our findings support the evidence for overall cost advantages for disease management, health insurance providers are very unlikely to benefit from disease management in financial terms, as more evolved outpatient services remunerated on a fee-for-service basis lead to higher costs whereas the insurance providers’ financial contributions to the centrally funded Austrian DRG-system remain constant regardless of changes in hospitalization rates. Hence, hospital operators tend to profit from disease management, as reduced hospitalization rates in terms of hospital stays cause higher remuneration for the remaining cases and consequently higher gross margins on the one hand. On the other, lower average lengths of stay may also contribute to the financial benefit of hospitals, as for the case of comparable main diagnosis groups within both study populations (secondary diagnoses do not have an impact on Austrian case groups), hospitals receive similar average lump-sum payments for enrolled and non-participating patients whereas costs associated with each particular inpatient stay vary due to the different number of days spent in hospital . One should therefore keep in mind that further promotion or even adoption of diabetes disease management into the basic benefit package would have to come with the implementation of adequate remuneration schemes in order to ensure that insurance providers benefit from cost advantages.
Our study has two basic types of limitations, the first relating to methodological concerns about the impact of the assessed outcome measures on overall health outcome. Several studies report evidence that diabetes DMPs are associated with better health related quality of life , reduced mortality  and in particular intermediate outcomes associated with better control of disease such as glycated hemoglobin levels, blood pressure and cholesterol levels [5, 8–10, 29, 30].
In our study we only had access to the insurance provider’s patient records including frequencies as well as the costs of the reimbursed health services. Information on intermediate outcomes such asHbA1C level, blood pressure, cholesterol level is not reported to health insurers (whether or not patients participated in the DMP) and can only be collected by either analyzing physicians’ own patient documentation  or direct inquiry of diabetes patients via questionnaires or interview [5, 8]. However, as McEwen et al. , Stark et al. , Snyder et al.  and Rothe et al.  observed improved healthcare processes in terms of frequency of specific assessments performed as well as improved intermediate outcomes, it can well be assumed that the better treatment conforms to the guideline, the better overall control of diabetes.
The second type of limitations reflects methodological concerns, which are due to the design of the study as an observational retrospective cohort-study. As put forward by several authors [11–13] selection effects are likely to occur in population based DMPs. The direction of these effects is not consistent however: whereas Linder et al.  and Schäfer et al.  report that patients with a better prognosis and lower risk status as well as higher self-activity and motivation are more likely to be included in the DMP, Buntin et al.  found evidence for higher utilization of health services and drugs before enrolment into the DMP which may be associated with a worse health status of DMP participants as well as higher motivation or a more stable life situation.
Due to the restricted access to other relevant socio-demographic characteristics such as income, education, profession or marital status and the lack of documented diagnosis in the patient records, the authors could neither control for socio-demographic differences other than age and sex between the group of participating and non-participating patients, nor identify a specific matching control group or matched pairs as suggested by Linder et al.  and Miksch et al.  respectively.
Schäfer et al.  used patient documentation of primary care physicians of randomly selected diabetes patients in order to determine differences not only in social-demographic characteristics, but also in cardiovascular risk profile and patient motivation. Again, these items are not covered by patient records held by Austrian mandatory health insurance providers and physicians are only obliged to perform extended documentation for patients enrolled in the diabetes DMP [18, 31].
Assessing the selection effect of our study, the specific target population of the insurance institution has to be taken into account: As pointed out in the presentation of the population studied, the Social Insurance Institution for Business only serves self-employed and retired self-employed persons as well as their close relatives (spouses, children) unless not covered by another insurance institution. As socio-economic factors such as household assets and education (along with race/ethnicity) have recently proven to be independent predictors of health decline amongst diabetics over 65  and due to the fact that variances within socio-economic characteristics within the group of self-employed can be deemed to be substantially smaller than for regional subsets of the whole population (apart from the fact that particular lifestyles associated with socio-economic factors favor the acquisition of diabetes in first place) selection effects due to economic factors are unlikely to account for the majority of the observed differences between the two study groups.
In terms of data on cost and utilization of the study population before the start of the DMP on the other hand, we did find substantial differences within the costs of the services that outpatient received, indicating a higher share of more expensive services provided to patients who later enrolled themselves in the DMP. However, as pointed out above when discussing the direction of possible selection effects, it remains unclear whether these differences at baseline are due to the higher patient involvement for the DMP group or a worse status of health within this group resulting in more expensive services. Considering this potential selection effect due to differences in individual as well as disease-related characteristics of diabetes patients, the comparison of the 2006 statistics with their corresponding figures for 2009 provides evidence of the efficiency of the Austrian diabetes DMP: Patients who decided to participate in the program experienced a reduction in overall days spent in hospital as well as lower increases in the number of outpatient services. This resulted in overall lower costs for care and medication as compared to their counterparts receiving regular diabetes treatment.