Sample and data collection
The analysis is based on cross-sectional data of the ongoing study “Dementia Networks in Germany (DemNet-D)”. DemNet-D was conducted to analyse structures, procedures, and outcomes of DN in Germany using qualitative and quantitative methods. Ethical approval was obtained from the Committee of Ethics at the University of Greifswald (register number BB 107/12). Thirteen DN throughout Germany were included in the study. These networks applied for funding to participate and were chosen by the funding agency—the Federal Ministry of Health (BMG). Inclusion criteria were (a) having been previously evaluated and (b) being considered a sustainable network. Currently the total number of DN in Germany is unknown, however the funding opportunity was made public and all existing DN in Germany were eligible to participate.
For the analyses, we collected data from 560 pairs of PWD and their caregivers. Inclusion criteria for each pair were (a) being served by one of the 13 networks and (b) issuing of written informed consent by both the PWD and the caregiver. Participants were randomly selected by the DN according to these inclusion criteria during a timeframe of 6 months. Data were collected through in-person interviews and paper questionnaires by trained interviewers. The interviewer were employed by the DN, the professional background was heterogeneous. All were experienced in communicating and dealing with PWD and caregivers by employment criteria. Specific qualification for conducting the assessments was provided by two group trainings and a written manual, both specifically designed for this study. The interview period lasted from the 1st of February until the end of September 2013, and recruitment was conducted by regional DN staff. Except for the geriatric depression scale (GDS)  all information was given by the interviewed caregiver. Information of the GDS was based on the answers given by the PWD.
The collected socio-demographic information included gender, age, living situation and region. Living situation was categorised as either (a) living alone in own household or (b) living together with others. Type of region was classified as either urban/suburban or rural. Socio-economic status (SES) was operationalized using the Scheuch-Winkler index , a combination of household income, years of education and profession and categorised as high, middle and low SES. Information about comorbidity was collected using a paper questionnaire. To identify which diseases affect each PWD, participants were provided with a list of the most common diseases in geriatrics (hypertension, hyperlipidemia, adipositas, diabetes, coronary heart disease, heart attack, cardiac insufficiency, stroke, asthma, chronic bronchitis, renal insufficiency, hepatic insufficiency, enteritis, gastritis, stomach ulcer, duodenal ulcer, arthrosis, rheumatoide arthritis, osteoporosis, chronic back pain, cancer, deafness, visual impairment) from that they could check off. More specific and rare diseases could be indicated in open text fields. Diagnoses were transferred into ICD 10 codes. Functional status was measured using the Instrumental Activities of Daily Living (IADL) score according to Lawton and Brody . This index ranges from 8 (no activity restrictions in daily living) to 0 (comprehensive activity restrictions in daily living). Depression was measured using the short-form geriatric depression scale (GDS) , with a score greater than 5 indicating depression. The caregiver was also asked about the presence of a diagnosis and the type of the dementia from the PWD.
The DN was categorised into either (a) physician associated networks, or networks led by a specialist (neurologist/psychiatrist) or (b) others. The other networks where community oriented networks focused on care providers. They aim to improve the management, evaluation and service integration for PWD in the sector of nursing care providers. 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 ginkgo biloba (N06DP01) .
We assessed the daily target dose according to the national guidelines and recommendations for acetylcholinesterase inhibitors and N-Methyl-D-Aspartate . The daily dosage was categorised as “target dose as recommended” or “others,” including: 20 mg/d for memantine, 10 mg/d for donepezil, 6–12 mg/d (9.5 mg/d – transdermal therapeutic system) for rivastigmine, and 16–24 mg/d for galantamine . We defined the period of intake as the beginning of the antidementia drug treatment until time of the interview.
We used descriptive statistics to summarize the demographic characteristics of the sample. To evaluate the associations of the prescription of antidementia drugs, we performed multiple logistic regression analyses with the prescription of the antidementia drug as the outcome variable. The model was adjusted for age, sex, comorbidities, functional status, depression, living situation (dichotomous: alone vs. not alone), diagnosis of dementia, and the network association (medical vs. others). Additionally, we accounted for the correlated nature of our data because observations from PWD in the same network were unlikely to be independent. Thus, we included the network as a random effect in our model. Prior to fitting the final regression model, we checked for non-linear associations of the covariates with the outcome by using the multivariate fractional polynomial approach . As expected, the association with age showed a departure from linearity and age was therefore categorised into four age groups using the quartiles of the birth years as cut-off values (group 1: 1910–1927; group 2: 1928–1932; group 3: 1933–1937; group 4: 1938–1969). For graphical analysis, we modelled age with restricted cubic splines using three equally spaced knots .
We listed the name of the drug and the prevalence of usage by the PWD in the cohort. To show the drug dosage, we used the mean frequencies of the intake per day and the dose of each antidementia drug in the cohort. Two-sided p-values were calculated with a two-sided significance level (p-value = 0.05). The statistical package used for the analysis was STATA 11 (StataCorp LP, Texas, USA).