Evaluation of an integrated care contract for the care of patients with chronic heart failure

A matched pair analysis based on GKV routine data

Heart failure is one of the most medically and health-economically important diseases in Germany (Neumann et al. 2009; Gensichen et al. 2004; Zugck et al. 2010; Willich et al. 2013; Buhr et al. 2007). The prevalence of heart failure is strongly age-dependent, so it can be assumed that the costs for diagnosis and treatment of heart failure will increase in the coming years in view of the aging German population. A large part of the treatment costs for heart failure arise in the inpatient area (Zugck et al. 2010). Against this background, effective integrated care programs have the potential to improve the care of heart failure patients and at the same time prevent decompensations, which are particularly often treated in the inpatient setting. Effective new care programs make it possible to optimize treatment processes, especially in the outpatient area, and to actively involve patients in the treatment concept (Buhr et al. 2007).

>> Based on the HeartNetCare-HF™ (Angermann et al. 2012; Gelbrich et al. 2014; Güder et al. 2015), the company health insurance companies (BKK) of the Central Regional Association that participate in the selective contract have been offering insured people with chronic heart failure an additional integrated option since 2011 supply program. The medical networks GO-LU, Ludwigshafen/Rhine, and PRAVO, Speyer, the social insurance for agriculture, forestry and horticulture in Hesse, Rhineland-Palatinate and Saarland as well as the IKK Südwest concluded an integrated care contract in accordance with Section 140a SGB V, hereinafter referred to as the integrated care model .
Patient care should be improved and the hospitalization rate reduced through continuous care by the family doctor and cardiologist and the establishment of coordinated treatment paths, taking into account the current status of valid national and international guidelines. A specialized heart failure nurse should monitor medical parameters that indicate decompensation of heart failure and prevent unnecessary hospitalizations through early interventions.
In a controlled longitudinal study based on routine data, the Institute for Applied Health Research Berlin (InGef) examined how the patient parameters in the intervention group (IV) compared to a control group consisting of patients in standard care (RV) in a follow-up period of 12 months have developed. The results presented here aim to investigate any connections between additional care provided by the integrated care model and patient-side parameters. The focus was particularly on questions regarding the use of care services by considering treatment costs, hospital admissions and allocations from the morbidity-oriented risk structure adjustment (Morbi-RSA).
methods
The study examines the effectiveness of the integrated care model compared to standard care. The study was conducted as an observational study with a control group design. The inclusion criteria for the two groups examined are shown in Table 1.
Anonymized routine data from 1,035 participants in the integrated care model of the BKK Landesverband Mitte and from around four million insured people in the InGef research database, from which the potential comparison group was recruited, were evaluated. Data from the years 2009-2014 were available for both groups.
Various patient-side parameters were examined over a follow-up period of 12 months, including total treatment costs, inpatient treatment costs, number of hospitalizations and level of referral by the Morbi-RSA.
The patients from the IV were assigned a group of comparison partners from the RV using propensity score matching (PSM). These are similar to the participants in terms of various observable sociodemographic and clinical characteristics, so that the two groups only differ in terms of participation in the nursing model.
The propensity score (PS), which reflects the conditional probability that an insured person was enrolled in the integrated care model, was formed in a pre-observation period of 12 months before the index quarter. The probability of participation was modeled depending on various sociodemographic and clinical characteristics of the patients (Rosenbaum and Rubin 1983). Age, gender and the highest NYHA stage in the previous period were taken into account as independent variables. A complete list of all covariates considered can be found in Figure 1.

The formation of the paired matches from IV and RV took place in two steps:
1. Determine the PS for each patient using logistic regression. For this purpose, the probability of participation in the integrated care model was modeled using observable covariates (=log(p/(1-p)).
2. Selection of subjects from IV and RV with similar PS. The PSM was carried out without replacement and using a caliper of ≤0.2 standard deviations, i.e. the absolute difference in the PS of the matched patients in IV and RV could no longer be 0.2 standard deviations of the PS of the entire study population.
To check the balance between the paired groups, the standardized mean difference (SMD) was created. An SMD >10% was considered to represent inadequate balance between groups (Normand et al. 2001).
Since the PSM was unable to achieve balance for some variables, a hierarchical linear model (HLM) was then calculated, which can be used to analyze hierarchically structured data (Farin 2005; Nelder and Wedderburn 1972; Sascha Walter and Oliver Rack 2006). In our use case, it was assumed that the pairs of the participant and non-participant groups each form a group. This procedure makes it possible to adjust for additional variables that were not taken into account during the PSM or could not be balanced. The group membership (IV vs. RV) as well as the incidence status of the heart failure diagnosis and the respective utilization parameter examined in the previous period were included in the HLM. The following target variables were modeled: total service costs and the costs from three main service areas (outpatient, inpatient, medication), Morbi-RSA allocations and number of inpatient treatment cases.
Results
For the above-mentioned evaluation period, the routine data of 1,035 insured persons from the BKKen of the Central Regional Association and the HRI research database were available.
A total of 8,438 patients met all inclusion criteria for the analysis, of which 891 were in the IV and 7,547 in the RV (see Table 2).
After the PSM, 791 pairs could be analyzed. While there were clear differences in the socio-demographic and clinical characteristics between the IV and RV before the PSM was carried out, the SMD for the majority of the covariates after the PSM was below 10% (see Fig. 1). IV and RV after PSM continued to differ only with regard to the incidence of heart failure diagnosis in the pre-observation period. It appears that particularly patients with an incident heart failure diagnosis were included in the intervention group. With regard to various utilization parameters measured in the previous period, such as the total number of hospitalizations, no major differences in SMD were observed after PSM (see Figures 1 and 2).
After PSM and additional adjustment for the incidence status of the heart failure diagnosis using HLM, there were significant differences in the total treatment costs, inpatient and outpatient treatment costs (see Table 3). The inpatient treatment costs were significantly lower in the IV than in the RV. However, this difference was not due to inpatient treatment costs due to heart failure. There were no significant differences in this regard. A detailed analysis of the main discharge diagnoses of fully inpatient treatment cases revealed clear differences in the indications between the two groups. The observed inpatient cost differences in favor of IV are mainly due to indications outside the cardiovascular spectrum. The largest differences in mean costs between RV and IV were for ICD Chapters A*/B* (Certain infectious and parasitic diseases), D* (Diseases of the blood and hematopoietic organs and certain disorders involving the immune system) and J* (diseases of the respiratory system) were observed (see Fig. 3).
The outpatient treatment costs were significantly lower in RV than in IV. The mean number of hospitalizations and hospitalizations due to a heart failure diagnosis were significantly lower in IV than in RV. No significant differences were observed regarding the mean drug costs and Morbi-RSA allocations during the follow-up period.
The significant differences in the total treatment costs can largely be attributed to the differences in the inpatient treatment costs (see Table 3 and Fig. 3).
discussion
The analyzes show that differences can be demonstrated between participants and non-participants in the integrated care model. Significant differences in favor of IV were demonstrated for total service costs, inpatient treatment costs, hospitalizations and hospitalizations due to heart failure. The outpatient treatment costs, however, were significantly lower in the RV. Parts of the significantly lower outpatient treatment costs in the RV can be attributed to the additional costs for the integrated care model in the IV. Almost half of the observed cost difference can be attributed to this additional compensation.
The significantly lower total treatment costs in the IV are largely due to the lower inpatient treatment costs (see Table 3). Since the inpatient treatment costs due to heart failure were similarly high in both groups, the inpatient cost advantages cannot be attributed to the improved care of heart failure through the integrated care model in the narrower sense. Rather, it has been shown that parts of the differences in expenditure can be attributed to other indications, such as inpatient treatment for certain infectious and parasitic diseases (ICD 10 code A*/B*), diseases of the blood and blood-forming organs (ICD 10 code D *) as well as certain disorders involving the immune system and diseases of the respiratory system (ICD 10 Code J*).
The lower inpatient treatment costs could also be due to a selection-related higher comorbidity of patients in the RV. Selection effects that contribute to enrollment in the integrated care model and that affect not only the severity of heart failure but also various comorbidities and for which appropriate adjustment was not made using the measured covariates cannot be excluded. It is therefore possible that heart failure patients with severe morbidity are less likely to participate in the integrated care model and that this selection effect explains the observed cost differences.
It would also be conceivable that the care of the patient by the heart failure nurse would have a positive effect on the patient's other treatment needs. Angerman et al. (2012), evaluating the HeartNetCare-HF™, observed that 80% of patients sought advice from their heart failure nurse practitioner more often for non-cardiovascular health problems than for cardiovascular disease. A patient survey of participants in the integrated care model by the participating medical networks showed that a similar spill-over effect may have occurred within the evaluated integrated care model. Overall, 95% of the patients surveyed felt that they received better care through the integrated care model and 93% of those surveyed found it helpful that the heart failure nurse regularly inquired about their health. This illustrates the importance of comprehensive care models that take into account the generally increased care needs of multimorbid heart failure patients through continuous personal approach and involvement of patients in their personal health management - even outside the narrow care indication of heart failure. <<