In total, 22 primary care physicians in Germany participated in this prospective, multicentre observational study. All of them were members of the EvaMed Network, which aims to evaluate CAM remedies in usual care with regard to prescribing patterns, efficacy, and safety [26
]. Physicians were recruited through the German National Association of Anthroposophic Physicians (Gesellschaft Anthroposophischer Ärzte in Deutschland
; GAÄD). A total of 362 physicians were contacted and informed about the EvaMed Network by standard mail and, in the event of non-response, 4 weeks later by telephone. For a physician to be eligible to participate in the study, his or her medical practice had to meet a number of technical requirements, including the presence of a special computerised patient documentation system (DocExpert, DocConcept, TurboMed, Duria, PDE-Top, Medistar), a local area network (LAN) connection, and Microsoft Windows and Internet Explorer (i.e. as client software). A total of 38 physicians (10.5%) fulfilled the technical requirements, gave informed consent, and agreed to participate in the EvaMed Network. Of these physicians, 16 specialised in paediatrics, dermatology, and gynaecology and were excluded from the study. Each of the remaining 22 physicians had practised for at least 5 years in addition to completing training in anthroposophic medicine.
The present study is based on secondary data provided by physicians for health insurance accounting. As such, the recommendations for good practice in secondary data analysis (e.g. anonymisation of data on prescriptions and diagnoses) developed by the German Working Group on the Collection and Use of Secondary Data [29
] were applied in full.
Data were included if patients had at least one diagnosis of dementia according to the 10th revision of the International Classification of Diseases (ICD10: F00-F03) during the 5-year study period (01.01.2004-01.01.2009). Data were excluded if the diagnosis dementia was only a suspected diagnosis and not confirmed during the study period and if a cross-validation of the ICD10 code and text of diagnosis indicated an incorrect coding. Figure shows the flow chart of the inclusion process.
Flow chart of the inclusion process.
During the study, physicians continued to follow their routine documentation procedures, recording diagnoses and all prescriptions for each consecutive patient using their existing, computerised patient documentation system. These data were exported to the QuaDoSta postgreSQL database hosted in each practice [30
]. Physicians used a browser-based interface to match individual diagnoses with the corresponding drugs or remedies that had been prescribed. Prescribed drugs were documented using the German National Drug Code. Diagnoses were coded according to the 10th revision of the International Classification of Diseases (ICD-10).
Dementia was classified as 'Dementia in Alzheimer's disease', 'Vascular dementia', 'Dementia in Parkinson's disease', and 'Unspecified dementia'. Co-morbidities were classified as hypertension, depression, heart failure, chronic ischemic heart disease, cardiac arrhythmias, diabetes mellitus, dyslipidemia, atherosclerosis, other cerebrovascular diseases, pressure ulcer, stroke, cerebral infarction, arthropathies, and dorsopathies. Multi-morbidity was considered if a patient had at least 2 co-morbidities.
Study investigators identified all drugs and remedies prescribed for dementia. Each substance was classified using the Anatomical Therapeutic Chemical Index (ATC) and anti-dementia drugs were clustered into cholinesterase inhibitors (i.e. donepezil, rivastigmine, galantamine), memantine, Ginkgo biloba, and older anti-dementia drugs (e.g. piracetam, nimodipine, and selegiline). Other medication was classified as antidepressants, anitpsychotics, benzodiazepine, anti-Parkinson drugs, antiepileptics, and sedatives.
Statistical analysis was performed with SPSS 18.0 for Windows. Descriptive analysis was used to determine prescription rates. Means and standard deviations (SD) were calculated for continuous, normal data. In cases where data were not normally distributed, medians and interquartile ranges (IQR) were reported. Subgroup analyses of prescribing rates were performed for patient age (under 60 years, 60-69 years, 70-79 years, 80-89 years, 90-99 years, 100 years and older), gender, and co-morbidities. The two-tailed Chi square test was used to analyse differences in prescription rates. A P value of less than 0.05 was regarded as indicating a statistically significant difference.
Adjusted odds ratios (AOR) and 95% confidence intervals (CI) were calculated using multiple logistic regressions to determine factors associated with a) a prescription of any anti-dementia drug and b) a prescription of Ginkgo biloba versus cholinesterase inhibitors and/or memantine. Age, sex, physician specialization (neurologist vs. GP), year of prescription, type of consultation (first vs. follow up), type of dementia, multi-morbidity, co-morbidities, neuroleptic, and antidepressant therapy were included as independent variables. Patient age was introduced in the model as a continuous variable. For the "best-fitting" logistic regression model, we employed the stepwise Akaike Information Criterion (AIC).