The initial hospitalisation and mean length of stay for five ambulatory care sensitive chronic medical conditions was substantially higher on average in the Danish healthcare system, compared with the Kaiser Permanente Integrated Delivery System. Subsequent readmission rates also were higher in the DHS for angina compared to the KP benchmark, but lower for diabetes compared to KP. The findings on mortality were mixed. There was a higher mortality in the DHS for patients with heart failure and lower mortality for chronic obstructive pulmonary disease, but no statistically significant differences for the other three conditions. In short, patients in the Danish healthcare system appear more likely to require preventable hospitalisations associated with chronic medical conditions, and have longer hospitalisations on average, compared with patients in the KP system.
There are several strengths of the study, including the quite large populations observed and the multi-year timeframe; our adjustment of analyses for differences in population characteristics and the timing of hospitalisations; and our inclusion of a condition that is not sensitive to ambulatory care as a control. Other strengths include the focus on two health systems with excellent capture of hospital data within electronic databases.
Comparing clinical outcomes across systems, however, is challenging, even for relatively straightforward hospital events and survival. We used hospitalization rates for selected ACSC as an indicator for the quality of primary care to patients with chronic conditions. However, there are a number of potential alternative explanations for differences in hospitalisation rates between the two systems. These include unmeasured differences in health and culture of the populations of the two healthcare systems, the level and quality of data capture, variations in access to primary care due to formal and informal barriers, and practice patterns within each system. In several cases, we would expect a bias towards finding no differences in our outcomes; in other cases, the effects are difficult to predict. The main threat to the findings is potential different use of hospitals for equal diagnoses in the two systems.
Unmeasured differences in the clinical and socio-demographic characteristics between the two populations almost certainly contribute to some of the observed findings. The DHS sample consists of the entire Danish population aged 65 and older whereas the KP population is a non-random sample of the US population that is enrolled within a single health plan. Our ongoing work, however, suggests that the KP population closely resembles the US population aged 65+ with respect to predicted risk. The US population in general also may have larger range in income levels compared to the Danish population where the progressive taxation policy means that few are very rich and few are very poor. Unmeasured differences in the available public and private social services also could contribute to some of the observed findings. Public transportation, public food services and other public funded services are available in Denmark whereas such services are private in the US. More public services available could favor lower readmissions in the DHS compared to KP. However, it is difficult to predict how differences in clinical and socio-demographic population characteristics overall affect the outcomes because of counterbalancing forces.
There are a number of social contextual issues that could influence the results and for which there is an extensive literature. Factors associated with genes, cultural norms around diet and exercise, and individual behavior such as smoking all could contribute to population level differences in disease prevalence, disease severity, and numbers and types of co-morbid diseases. The direction of such factors, however, is difficult to predict. For example, prior studies showed that Danes tend to both exercise and smoke more that the KP population and the KP population has higher self-reported prevalence of chronic conditions than the Danish population .
Differences in data capture among the systems may also to some degree contribute to some of the differences between the healthcare systems. Diagnoses may be recorded differently between the systems, which would affect our findings. The intensity of diagnosis might also differ, though the direction of the net effect is unclear with more hospital use in the DHS resulting in greater opportunity for hospital related diagnoses, but potentially higher outpatient diagnostic intensity in the United States .
Another potential explanation for the differences in hospitalization rates between the two systems is different use of hospitals. Danish hospitals have a high number of beds available compared to hospitals in KP, and supply is a powerful determinant of utilisation, which also could lower admission thresholds [25, 30]. Correspondingly, the two systems could have different thresholds for admission to- and discharge from the hospital. Differences in admission thresholds alone, however, should result in lower deaths in the system with more initial hospitalisations. Higher hospitalisation rates combined with lower death rates in a healthcare system might suggest potential inefficiency either because of unnecessary hospitalisations or hospitalisations preventable with earlier intervention in the outpatient setting. Higher hospitalisation rates combined with higher death rates in a healthcare system, however, might suggest more serious quality problems.
Subsequent readmission rates were higher in the Danish healthcare system for angina compared to the KP benchmark, whereas the readmission rates for diabetes were lower in the DHS compared to KP. There is amount of literature stating that readmission rates may not serve as a valid indicator for quality of care. E.g. a report by Williams and Fitton state that readmission, perhaps on several occasions, may be generally preferred to permanent admission, both by the patient and by the system  and other studies have identified associations between readmissions and underlying physical conditions [32, 33]. We were not able to adjust the analyses for underlying chronic conditions as the data did not include enough secondary diagnoses to do proper case-mix adjustment. Consequently, differences in readmission rates between the systems may not be a good indicator of quality differences among the systems.
Factoring in all of these issues, we believe that our results are consistent with other studies suggesting problems in the quality of chronic disease care within the DHS. This prior work includes findings of a more comprehensive and systematic approach to disease detection and disease prevention in KP compared to the DHS. KP provides more medical (secondary and tertiary) prevention to its members and more self-management support is provided in KP compared to the DHS. Additionally, disease treatment and complication prevention within the healthcare systems will affect hospitalisation rates. The KP system has structured chronic care management programmes that integrate multiple elements, such as clinical guidelines, disease registries, proactive outreach, reminders, multidisciplinary care teams, and performance feedback to providers . Also, KP's integrated IT system, the medical centers in KP housing GPs and specialists as well as aligned financial and non-financial incentives throughout the system in KP make the interactions between providers easier, leading to better coordination and more follow-up which we believe result in lower initial hospitalisation rates and lower readmission rates.
Programmes to improve chronic care management have only recently been introduced in the DHS and are still in the implementation phase and were not widespread when the data from this study was obtained. Additionally, the DHS is a more fragmented system with general practitioners, hospitals, and preventive and rehabilitation services being paid from different public sectors, without aligned incentives or a proactive approach to prevention. Thus, prior studies conducted in the DHS have indicated that lack of acute services in the municipalities responsible for home nursing care and nursing homes to some extent caused undesirable hospitalisations . Further, prior studies conducted in the DHS suggest a substantial amount of mistrust and lack of cooperation between physicians in the different settings in the DHS [36–38]. In addition to these clashing cultures, there also is a pervasive lack of information integration across settings and clinicians within the DHS. Accordingly, previous studies show that the coordination of care between GPs, hospitals, and municipalities has been insufficient [36, 39]. Comparing hospitalisations for ACSCs within healthcare systems can serve as a surveillance mechanism to identify problems, but are not very precise in terms of identifying how to target the cause of that problem. However, together, these findings combined with previous studies on care coordination differences between DHS and KP , and on quality improvements within the KP system [34, 40], suggest substantial opportunities to improve the quality and efficiency of care in Denmark for patients with chronic medical conditions, compared to the KP benchmark. In addition to providing a benchmark for potential quality improvement, the findings also suggest room for efficiency gains. Over the six year period from 2002 to 2007, the hospitalisation rates did decrease within DHS, but on average remained several fold greater in magnitude than the rates in KP, thus suggesting an upper bound for improvement. In other words, in 2007 alone, among the 32,001 persons hospitalised in Denmark for a preventable hospitalisation, 19,300 (60%) of them would not have been hospitalised, had the DHS rates been comparable to those in KP; there were 26,662 excess hospitalisations (i.e., 61% of the 43,521 observed hospitalisations in DHS in 2007 would not have occurred if the DHS rate was equal to that of KP's), and 5,599 excess readmissions (i.e., 61% of the 9,139 observed readmissions in DHS in 2007). Reallocating these resources from the hospital to preventing disease exacerbations in the outpatient setting could yield welfare gains for patients and their families, without requiring substantial new investments.
While additional research using individual level data on patient characteristics would improve the estimates of rates within each system, these longitudinal estimates within each system provide useful benchmarking information that can guide future reform efforts in the DHS as well as track the effects of any new reforms. Based on our results obvious areas for future reform efforts in the DHS may be improving the integration of services, improving structured care to persons with chronic conditions. However, it is critical to assess whether approaches from one healthcare system can be directly transferred to another system and whether major or minor changes should take place to obtain the desired effects . Prior studies of implementation of technologies have shown that a technology, policy or function can be transformed in a new context and that the new context will influence how this approach is implemented and how it works . Thus, caution must be exercised before transferring ideas or approaches used in KP to the Danish healthcare setting. Fireman et al. investigated savings resulting from the use of chronic care management programmes in KP. Actual cost savings were elusive, but programs could have sizable potential savings . The study only focused on healthcare costs and savings. There is insufficient evidence that this approach will achieve the same improvements in the DHS; however it can be hypothesized that investing in efforts to improve the quality of chronic care by strengthening outpatient care settings in the DHS will lead to fewer preventable hospitalisations. Implementation of chronic care management programs in the DHS cannot be expected to create immediate savings in the healthcare budget, but the potential for improved quality of care and long term savings at the society level seems to be substantial.
External benchmarking can be a valuable tool for healthcare system reforms striving to improve performance as it can shed light on areas with potential for improvements and provide inspiration for how to reform organisation and delivery systems. As an example the study published by Feachem et al. in 2002 comparing cost and performance in Kaiser Permanente and the NHS was followed up by additional studies and played an important role in the decision about implementing chronic disease management approaches in the NHS .