Data were derived from the Odense study, an epidemiological survey in which the objective was to estimate the prevalence and incidence of dementia in Denmark [10
] In this study, a total of 244 patients with dementia agreed to participate in an interview accompanied by a relative or caregiver. The study was approved by the Scientific-Ethical Committee of the Counties of Funen and Vejle, Denmark, and by the Danish Data Protection Agency.
Demented patients were classified by type of dementia and by severity of dementia. Alzheimer's disease (AD) was diagnosed according to the NINCDS-ADRDA criteria for probable dementia [12
]. Vascular dementia and other types of dementia were diagnosed according to the DSM-IIIR criteria [13
]. Severity of dementia was diagnosed according to the Clinical Dementia Rating (CDR) scale [14
] and the Mini Mental State Examination (MMSE) [15
]. The complete examination programme is described in Andersen, Lolk et al, 1997 [10
All interviews were conducted by a certified nurse in the patient's home., Patient's and caregiver's socio-economic and socio-demographic status as well patients' health status and ADL were recorded. In the event that a relative was not present during the interview, a professional caregiver verified information provided by the patient.
Each interview included the following information:
- sociodemographic questions (age, gender, setting).
- activity of daily living (ADL) questionnaire using 7 items describing patients' ability to perform physical activities (personal care, dressing, mobility and personal toiletry) and psychosocial activities (activities in the home and hobbies inside and outside of the home). Each activity was scored using a four-point Likert scale anchored at the ends with 1 = "Unable to perform the activity" and 4 = "Perform the activity without help from others". The physical ADL ranged between 4 (worst state) and 16 (best state), while the psychosocial ADL scored between 3 (worst state) and 12 (best state) [16
- Questions mapped into each of the five dimensions of the EQ-5D: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression as presented in the Table [17
Mapping questions from the Odense Study into the EQ-5D
For the mobility dimension of the EuroQol instrument, we assumed that patients that were able to walk without assistance from others had no problems in performing this activity, whereas patients who were unable to walk unassisted were classified as confined to bed. Patients that needed help from others were classified as having some mobility problems.
Two questions from the ADL instrument in the Odense Study were used to classify patients on the EuroQol personal care dimension. Patients that performed both activities without help from others were classified as having no problems on this dimension. Patients in need of help with either personal care or dressing or both were classified as having some problems. Only patients unable to wash and dress without help from others were classified as such on the EuroQol instrument.
The patient's ability to carry out hobbies in the home was used as a proxy for their ability to perform usual activities.
For the pain/discomfort dimension of the EuroQol instrument, it was assumed that patient assessment of their own health status covered this dimension. Therefore, if they found their health status to be very good they were classified as having no pain or discomfort. A good or fair assessment was categorised as having moderate pain or discomfort, whereas a poor assessment was assumed to correspond to extreme pain or discomfort.
In the Odense Study, patients stated how often they experienced emotional problems, whereas the health state being described by the EuroQol instrument refers to the patient's health at the time of filling in the instrument. Thus, the questions in the Odense Study included an aspect of time that the EuroQol does not cover. To overcome this, it was assumed that the occurrence of emotional problems converts to the degree of anxiety or depressions. That is, patients that never experience emotional problems were assumed not to be anxious or depressed, whereas patients that sometimes or often experience problems converted to moderate or extreme anxiety or depression, respectively.
The procedure of mapping returned a five-digit code, where the first digit referred to the patient's mobility level; the second to the patient's level on personal care; and so forth. This five-digit code described a health state for which we looked up the HRQoL utility weight in a table of EQ-5D tariffs. The EQ-5D tariffs take values between zero and one, where zero is the worst imaginable health status, and one is the best imaginable health status. We used Danish EQ-5D tariffs from a survey based on the time trade-off technique [18
Based on their cognitive and functional scores patients were classified by severity and dependency level.
The severity of a patient's dementia was defined by score intervals on the MMSE [15
]. Those scoring ≥20 were considered as having mild dementia, while patients scoring between 10 and 19 were classified as having moderate dementia. Patients scoring ≤9 were classified as suffering from severe dementia. MMSE scores were not available for 30 patients (21 AD patients and 9 patients suffering from vascular dementia). In order to determine these patients degree of dementia we used the CDR score to classify them into the above three severity groups. Patients with a CDR score of 0.5 were classified as mild, patients with a CDR of 1 were classified as moderate and patients with a CDR of 2 to 3 were classified as severe [19
Patients were classified by their ability to perform physical and psychosocial activities of daily living (ADL) This resulted in a classification of either dependent or independent [3
]. A binary variable was based on a non-hierarchical cluster analysis [20
]. Firstly, we identified possible initial seeds for the analysis. The seeds were identified from a cross table of the physical ADL score and the psychosocial ADL score. Combinations of the physical and psychosocial ADL scores with five or more observations were included in the cluster analysis as possible seeds. Secondly, we carried out the cluster analysis using the PROC FASTCLUS procedure in SAS 8.2 (SAS Institute Inc., Cary, NC, USA) in order to identify two clusters.
One cluster included patients with low scores on both the physical and psychosocial ADL scales. As low scores on both instruments meant that patients required help from others in performing the activities in question, they were classified as "Dependent". The other cluster included patients with high scores on both ADL scales and they were classified as "Independent". As these were composite criteria, the characteristics of the two groups of dependency were analysed.
After analysing the descriptive results of the EQ-5D scores, we performed a regression analysis of EQ-5D scores on sociodemographic and clinical characteristics. Sociodemographic characteristics included gender, age and setting (living in the community or institutionalised) and clinical characteristics took into account the level of severity (Mild, Moderate, and Severe), the type of dementia (AD, vascular or other) and ADL status (independent or dependent).
The regression analysis was performed with backward selection (level 5%) in order to determine the main factors influencing QoL.
Because of heteroscedasticity, we estimated the heteroscedasticity consistent covariance matrix, which was used to calculate test statistics for the coefficients.
Observations with missing data were automatically excluded from the analyses. That is, observations with insufficient information to establish a EQ-5D weight, e.g. that information was lacking to determine a patient's mobility level on the EQ-5D instrument.