Abstract
Background:
There is no common definition for the Dementia Care Network (DCN). They are heterogeneous and there is no general, longitudinal evidence for the effects of DCN.
Objective:
We describe changes in utilization of health services by people served by dementia care networks in Germany and factors associated with those changes over time.
Methods:
Primary data was assessed in 560 people with dementia (PwD) and their caregivers supported by DCN in Germany; sociodemographic and clinical variables, utilization of services; DCN were characterized according to governance. The design: observational study with face-to-face interviews at two time points over a period of one year. Data was assessed via semi-structured interviews at the participants’ homes.
Results:
Utilization of health services in this study is consistently higher than reported for the general population and does not significantly change over time. The strongest predictor of utilization of any service after one year was the use of this service at baseline (OR from 3.23 to 44.16). Higher activities of daily functioning increased the chances to utilize specialist physicians (OR = 1.32; 95% -CI: 1.08–1.63) or occupational therapy (OR = 1.24; 95% -CI: 1.02–1.50) significantly. Being a female decreased chances to utilize specialist physicians (OR = 0.57; 95% -CI: 0.37–0.87) and increased the chances to utilize no services (OR = 2.08; 95% -CI: 1.29–3.33).
Conclusion:
While health care acknowledges the importance and benefits of dementia care networks (i.e., in Germany, the results were considered in new German legislation (SGB XI)), further research is needed to define this kind of service delivery to facilitate comparison as well as promote evidence-based implementation.
INTRODUCTION
The increasing number of people with dementia (PwD) is one of the most challenging problems for health care systems worldwide [1, 2]. In Germany, there are approximately 1.5 million people affected, of whom about three-quarters are living at home [3]. Since there is no causal treatment of dementia available and most of the PwD want to live at home as long as possible, treatment and care need to focus on enabling PwD to live at home as independently and as long as possible. From the health care system perspective, there are also economic reasons to delay institutionalization.
Current evidence–based guidelines for the treatment of dementia recommend a wide variety of medication and non-pharmacological treatment options, in Germany and worldwide [4–6]. In principle, these treatment options are all available in the German Health Care System. However, utilization rates of most are low [7]. In primary care, treatment with antidementia drugs varies between 23% and 28% of PwD [8] according to health insurance data and are some 30% according to primary/survey data [9]. According to health insurance data, general practitioners are seen by 87.9% –94.9% of PwD in the course of 1 year, and medical specialists by 52.2% –56.4%. Non-physician services are utilized by 10.1–12.3%, any medical aids by 13.5–17.7%, home health care by 24.6–32.2%, and rehabilitation by 1.8–6.3% of PwD [10]. Studies with secondary data usually include only patients with a formal diagnosis of dementia and thus present a selection bias since dementia is underdiagnosed in primary care [11]. Consequently, figures derived from reimbursement data are often overestimating health care utilization of PwD in Germany.
The lack of utilization of health services has been attributed to the sectored German health care system by the expert council for the assessment of the development in health care already in 2012 [12]. According to the World Health Organization (WHO), models of integrated care can overcome this challenge [1]. In this line, among others, so called Dementia Care Networks (DCN) have been established in Germany to provide support to PwD [13]. The Federal Ministry of Family, Senior Citizens, Women and Youth (BMFSFJ) has actively initiated, funded, and supported 500 networks in the line of the “local alliances of dementia”; the Federal Ministry of Health (BMG) has supported networks in the funding scheme of the “lighthouse projects dementia”. DCN are primarily financed by membership fees, earnings of services provided, public funds, and payments by municipalities or a variety of other funds of health care providers. Funds to support the interpersonal and infrastructural development of a DCN, a mix of internal and external financing sources, the participation of the municipality as well as the availability of full-time employees are determinants of a sustainable financing [14]. The construction and operation is very heterogeneous. However, underlying types of knowledge management and governance have been described in more detail elsewhere [15–17]. A systematic approach to disentangle the heterogeneity of already existing DCN is needed.
There is sound scientific evidence that a DCN can be effective in improving treatment and care for PwD, especially in integrating people into the specialized dementia care system based on a single randomized-controlled trial [18]. Further analyses of 13 different DCN networks throughout Germany have recently confirmed that PwD served by these networks show higher utilization rates of antidementive drug treatment [19], receive specialized physician care more regularly, especially in regard to the consultations of neurologist/psychiatrists [20], and show higher utilization rates of non-pharmacological treatments and aids [21] than PwD in general. The overarching conclusion is that DCN provide better health care for PwD [19–21]. Furthermore, DCN seem to serve a diverse population of PwD in primary care [22], highlighting their importance as a health service infrastructure.
However, these analyses have been cross-sectional only, so far. While the results are promising, it is not clear whether and how utilization of services by PwD changes with time while being served by a DCN and, if so, what factors are associated with this.
Therefore, the objective of this paper is to 1) analyze changes in utilization of health care services in PwD served by DCN in a longitudinal perspective, 2) analyze differences in utilization over time associated with the orientation of the network as either medical or non-medical, and 3) analyze factors associated with changes in utilization of services over time such as person-oriented factors as well as network-oriented factors like governance.
METHODS
Study design
The present analysis is based on longitudinal data from the Dementia Networks in Germany (DemNet-D) study. In DemNet-D, 13 different dementia networks and their clients were analyzed with respect to structures, procedures, and outcomes. The ethics committee at the medical faculty of Greifswald approved the study protocols and the informed consent (BB 107/12). The procedures and outcomes of dementia networks were assessed using qualitative and quantitative methods. Inclusion criteria for DCN participating in this study were defined and evaluated by the Federal Ministry of Health, the funder of the study. The DCN applying for funding to the ministry had to document that they had already established a network and state their willingness to participate in a scientific evaluation. This procedure implies a selection bias in the DCN under analyses; representativeness could not be evaluated.
Inclusion criteria for PwD and their caregivers were the following: 1) be a patient in one of the 13 participating dementia networks and 2) provide written informed consent from the PwD and their caregiver. PwD were recruited by regional dementia network staff aiming to recruit 715 dyads of PwD and caregiver with each network providing 55 dyads. The number was chosen 1) after asking the DN how many users they serve and the estimated proportion that they could provide to the study, 2) to keep the number of participants equal, and 3) to be appropriate according to financial resources across DN. DCN were located in rural (n = 5) as well as urban areas (n = 8) spread across 4 out of 16 federal states of Germany. DCN provided between 17 and 56 dyads into the study (mean = 46.9, SD = 3.1). Out of the DCN, 5 were considered to be physician-oriented, while 8 “others” were less medically oriented (more networking, awareness, care, etc.). The categorization was based on the professional background of the head of the network. In five of the participating networks, the head was a neurologist/psychiatrist who was working in an outpatient dementia-specific institution. We assumed that the focus of the network was the provision of the physicians’ expertise and that the type of care in these networks may systematically differ from networks with a different focus. For more specific information of the DCN, see: https://demenznetzwerke.de/start/netzwerkforschung/projekt-demnet-d/ (in German).
Comprehensive data were assessed by both face-to-face interviews conducted by trained interviewers and paper-pencil questionnaires for PwD and their caregiver from February through September 2013 and 12 months thereafter (February through September 2014). Baseline assessment was successful in 560 dyads and a follow-up was conducted 1 year later successfully re-assessing 481 of the initial participants (response 85.9%; n = 13 had moved out of the areas, n = 62 withdrew informed consent, n = 4 with missing data). A total of 66 participants had died during the follow-up period. A detailed sample description has been provided elsewhere [22]
Data assessment: Procedure and instruments
All information regarding this analysis was provided by the caregiver of the PwD. We collected socio-demographic information from the participants, including sex, age, living situation, and region. To characterize the living situation, we asked for the PwD’s and the family caregiver’s place of residence. We dichotomized the living situation into 1) PwD living by themselves in their own household or 2) PwD not living alone. We categorized the region type by asking the caregiver whether the PwD is living in 1) an urban/suburban or 2) a rural region. To assess education and income of the PwD and his/her caregiver, we solicited all variables necessary to calculate the Scheuch-Winkler index (household income, years of education and professional status) [23]. This index yields a categorization into three social classes (high, middle, and low).
Comorbidities were assessed using a paper-pencil questionnaire. Participants could choose from a list of the most common geriatric diseases and were asked whether a formal diagnosis had been communicated by a physician. Non-listed diseases could be indicated in open text fields. For data analysis, we transformed the diagnoses into International Statistical Classification of Diseases and Related Health Problems (ICD-10) codes. At follow-up, we asked whether any comorbidity was additionally acquired since baseline and whether other changes had occurred in social and health matters. Functional status of participants was assessed using the IADL score of Lawton and Brody [24]. This instrument measures functionality with a score ranging from a maximum of 8 points, indicating intact function, and a minimum of 0 points, indicating no function in regards to daily activities. Depressive symptoms were operationalized using the Geriatric Depression Scale (GDS) [25, 26] with a score ranging from 0–15, the higher the score the more severe the symptoms.
We asked caregivers whether a formal diagnosis of dementia for their PwD existed. Severity of functional disability associated with cognitive impairment was assessed using the FAST scale [27]. The FAST is a specific hierarchical scale offering examples of difficulties with ADL. This specific hierarchy was found to be related to the cognitive decline within AD. In our study, we used the FAST as proposed by the authors as proxy ratings from the interviewed caregiver. Similar to other proxy ratings, the validity can be critically discussed. This instrument categorizes 7 dementia-related functional disability stages from 1 (normal) to 7 (very severe). Utilization of medical treatment is operationalized as “being treated by a physician” and “the frequency of contact”. We categorized physicians who are especially relevant for treating dementia in the German health care system into 1) primary care physicians (PCP) including general practitioners and general internists, 2) neurologists and psychiatrists (specialists), 3) other medical doctors (MDs), or 4) emergency care physicians. The other MDs category referred to answers from a free text field, which did not meet any of the categories (1), (2), or (4). In the paper-pencil questionnaire, the respondents could indicate whether they were treated by the physician as well as the number of visits in the 6 months preceding the interview.
To reduce the heterogeneity and account for different goals of the DN, we categorized them into two groups: 1) physician-oriented networks and 2) others. The categorization was based on the professional background of the head of the network. In five of the participating networks, the head is a neurologist/psychiatrist who is working in an outpatient dementia specific institution. We assume that the focus of the network is the provision of the physicians’ expertise and that the type of care in these networks may systematically differ from networks with a different focus (such as networks initiated by institutions focusing on improving quality of care, community-initiated networks, caregiver networks and others).
To assess drug treatment, drugs were recorded by name, dose, and frequency of use. The following drugs were considered: donepezil (N06AD02), rivastigmine (N06AD03), galantamine (N06AD04), memantine (N06AX01), and gingko biloba (N06DP01) [28]. In order to collect data about the utilization of therapies, we asked the caregiver to state whether the PwD had received physiotherapy, occupational therapy, logopedics, or outpatient rehabilitation in the last 6 months and how often the PwD had received the different therapy modalities. This list is not comprehensive for all possible treatments and thus we also asked the participants whether they had received no therapies at all. This variable was cross-checked with any specific utilization of therapies.
It is well known that DCN focus on different aims and thus networks structures differ. One factor to describe differences between networks is governance. Consequently, the participating DCN were categorized into four different types proposed by Schäfer-Walkmann et al. (2014) [15]. These types are based on evidence-based types of governance and divided into: Type A-Stakeholder (n = 3; focus on support through information and education by linking regional care providers and actively involving stakeholders), Type B-Organization (n = 3; high level of internal formal governance and a defined control center), Type C-Hybridization (n = 4; able to adapt their strategy quickly to changing requirements), and Type D-contract (n = 3; user-focused, specific care-related aims that are followed through consistently).
Statistical analysis
Data were expressed by means and standard deviations for metric variables and by proportions for nominal variables. For univariate changes between baseline and follow-up, we performed McNemar tests for binary variables and paired t-tests for metric variables. Before analyzing the variables of interest (utilization variables), drop out analyses were performed to test for systematic drop-out. In particular, mixed effect logistic regressions were carried out with the drop out variable as outcome variable and the baseline characteristics of the patients (ADL, GDS, rural living, gender, age, type of network) as predictors. We analyzed drop-out overall and drop out by death separately. In these models, the DCN was introduced into the model as random effect to account for the stochastic dependency of data sampled from the same DCN. As these analyses indicated a positive intra class correlation on drop out overall, we examined furthermore the drop-out rates per DCN revealing differences between the DCNs.
To explore baseline factors predicting the utilization at follow up a mixed effect logistic regression was fitted for each utilization variable using the baseline value of the utilization variable and sociodemographic variables (rural living, age and gender), GDS and ADL scores as well as the network type “governance” as predictors. Again, the network was used as random effect variable. The link function was chosen to be the logistic in the case of binary utilization variables and the identity function for metric variables. In all regression heteroscedastic robust standard errors were used. A p-value below 0.05 (two-sided) was considered to be statistically significant.
RESULTS
Loss at follow-up
The overall drop-out rate varied significantly (p < 0.001) between the DCNs over a range of 5.88% to 46.3% while the rate of deaths (0% –22.22%) was similar between the clusters. From 551 participants at baseline, 140 (76 women, 64 men) dropped out over the follow-up interval. 66 individuals died (33 men). Significant predictors for overall drop-out were a low ADL score (p < 0.001) and living in rural regions (p < 0.01). Predictors for drop-out by death were high age, male gender, a low ADL score and living in rural regions (all p < 0.05).
Changes over time
Comparing the clinical and utilization variables for the sample at baseline and at follow-up, there is a significant change in the instrumental activities of daily living, the severity of dementia and the utilization of no medical treatment (as defined by this study). The mean in daily functioning decreased from 2.17 to 1.54., indicating less function with regard to daily activities in this sample. Accordingly, the severity of functional disability associated with cognitive impairment has increased over time from 6.27 to 6.43. While there were no significant differences in utilization of medical services over time for the total sample, there has been a significant increase of the proportion of people having indicated to not use any medical treatment (from 41.3% to 53.5%).
Comparing the sub-sample of PwD served by medically-oriented DCN over time, there is a significant decrease in instrumental activities of daily living like in the total sample, but no significant decrease in functional disability associated with cognitive impairment or the utilization of no medical treatment. However, there is increase in the mean score for depressive symptoms (from 3.55 to 4.05). The sub-sample of PwD served by other DCN showed significant differences between baseline and follow-up like the total sample, but also a significantly smaller proportion of PwD receiving antidementia drug treatment at follow-up (drop from 44.3% to 37.3%). The results are provided in more detail in Table 1.
Sociodemographic information, clinical variables, and utilization of medical services in people with dementia served by dementia care networks at Baseline and 1-year follow-up
A) comparison between overall at baseline and overall at follow-up; B) comparison of other DCN types at baseline and other DCN types at follow-up; C) comparison between physician-oriented DCN at baseline and physician-oriented DCN at follow-up; *p-value < 0.05; **p-value < 0.01; ***p-value < 0.001; n.s., proportions not statistically significant; anew diagnoses since baseline; bno p-value available.
Analyzing the logistic regression models to predict the utilization of services at follow-up, all models were statistically significant, except for the models for utilization of primary care physicians and the utilization of emergency physicians. Hence, these latter models are neglected in the detailed analyses of all variables. The strongest predictor of utilization of any service was the use of this service at baseline (OR ranging from 3.23 in utilization of physiotherapy to 44.16 in utilization of antidementia drug treatment). However, higher scores of daily functioning increased the chances to utilize specialist physicians (OR = 1.32; 95% -CI: 1.08–1.63) or occupational therapy (OR = 1.24; 95% -CI: 1.02–1.50) significantly. Being a female decreased chances to utilize specialist physicians (OR = 0.57; 95% -CI: 0.37–0.87) and increased the chances to utilize no services (OR = 2.08; 95% -CI: 1.29–3.33) significantly. Regarding the network type, PwD in Type B-organization networks showed significantly higher rates of utilization of physiotherapy (OR = 2.24; 95% -CI 1.48–3.39) compared to PwD in Type A-stakeholder networks. The detailed results are presented in Table 2.
Predictive factors at baseline for utilization at follow-up
OR, odds ratio; *p-value < 0.05; **p-value < 0.01; ***p-value < 0.001. ALogistic regression (N = 300, 13 clusters), random effects model, Wald Test (9) = 6.26, p = 0.71; intra-class-correlation <0.001. BLogistic regression (N = 282, 13 clusters), random effects model, Wald Test (9) = 47.84, p < 0.001; intra-class-correlation = 0.15. CLogistic regression (N = 168, 13 clusters), random effects model, Wald Test (9) = 5.73, p = 0.77; intra-class-correlation <0.001. DLogistic regression (N = 315, 13 clusters), random effects model, Wald Test (9) = 28.75, p < 0.001; intra-class-correlation <0.001. ELogistic regression (N = 315, 13 clusters), random effects model, Wald Test (9) = 68.71, p < 0.001; intra-class-correlation <0.001. FLogistic regression (N = 330, 13 clusters), random effects model, Wald Test (9) = 35.70, p < 0.001; intra-class-correlation = 0.11. GLogistic regression (N = 258, 13 clusters), random effects model, Wald Test (9) = 70.62, p < 0.001; intra-class-correlation = 0.19.
DISCUSSION
This is the first longitudinal analysis of utilization of medical and other health services in PwD cared for in dementia care networks. It shows that utilization of all categories of care is constantly higher than comparable utilization rates in the general population. There is no significant decrease in utilization over the time of 1 year in general. However, the level of utilization is partly still lower than one would expect for some non-medical treatments including occupational therapy and outpatient rehabilitation. This might partly be due to the significant decrease in functional abilities over time, rendering a growing proportion of all PwD too severely impaired to be assigned to these therapies. However, there are still opportunities in DCN to increase utilization among their PwD.
DCN are a heterogeneous group of networks and categorizing them according to whether they are medically-oriented or not already yields considerable differences in outcomes. The medically-oriented DCN did not show a significant decrease in functional disability associated with cognitive impairment or a decrease in antidementive drug therapy like other DCN. Also, physician oriented DCN did not show a significant increase of PwD without any medical treatment. Our results imply that inclusion or involvement of physicians might be one crucial factor in improving medical therapy/utilization of medical therapies in PwD served by DCN when this is the outcome the DCN aims at. However, this is just one factor to improve outcome. The overall utilization of care is low in any type of DCN in this analysis.
The utilization of medical services after one year of participation in DCN is mostly dependent on the utilization of this service at baseline according to multivariate analyses. There seems to be a gender bias in utilization with being female predicting a lower utilization of physician specialists or a higher non-utilization in general. The other variables like age, living region, or functional ability have no predictive value for utilization. While it is re-assuring that there seems to be no age or region bias in utilization, there seems to be a gender issue for utilization in DCN. While this has been reported previously for the general population, it is surprising that the same tendency prevails even in a DCN setting aiming specifically to improve treatment and care for all PwD. One might assume that experts in the field of dementia are aware of this gender issue and would actively work against any gender-bias. However, these results show that gender equality is not fully achieved in our sample of DCN patients.
The type of network is not generally very predictive for utilization of medical services. The longitudinal analysis, however, revealed differences over time between physician led DCN and other DCN. Similarly, a more comprehensive categorization into X network types based on their governance yields only few significant results. Only PwD in DCN type B-organization have a higher chance to utilize physiotherapy after one year as compared to the network type A-stakeholder. This might indicate that DCN aiming at physiotherapy should be organized in type B rather than type A, but given the modest association and the complexity of medical therapies in PwD, this conclusion might be overstated.
The establishment of DCN generally improves utilization of health care services among PwD. Physicians should be involved in DCN, but the specific type of network is of less importance. In Germany, the results were considered in new German legislation (SGB XI), which took effect in January 2017. The §45c section 9 SGB XI 4 regulates that the Long-Term Care Insurance participates in the funding of self-organized networks for a structured collaboration in long-term care now.
Footnotes
ACKNOWLEDGMENTS
We gratefully thank the participating users of dementia networks for their time and effort to provide the data. Furthermore, we appreciate the cooperation and exchange with the members of each dementia network.
This study is funded by the Federal Ministry of Health (BMG) of Germany in the line of the funding scheme “Zukunftswerkstatt Demenz” (grant numbers IIA5-2512FSB031; IIA5-2512FSB032).
