INTERNATIONAL JOURNAL OF GERIATRIC PSYCHIATRY

Int J Geriatr Psychiatry 2006; 21: 449–459.

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/gps.1489

Determinants of costs of care for patients with

Alzheimer’s disease
Linus Jönsson1*, Maria Eriksdotter Jönhagen1, Lena Kilander2, Hilkka Soininen3,
Merja Hallikainen3, Gunhild Waldemar4, Harald Nygaard5, Niels Andreasen1,
Bengt Winblad1 and Anders Wimo1
1
Division of Geriatric Epidemiology, the Neurotec Department, Karolinska Institutet, Stockholm, Sweden
2
Department of Public Health and Caring Sciences, Uppsala University, Uppsala, Sweden
3
Department of Neurology, Kuopio University Hospital and Clinical Research Center, Mediteknia,
University of Kuopio, Kuopio, Finland
4
Department of Neurology, Copenhagen University Hospital, Copenhagen, Denmark
5
Section for Geriatric Medicine, Department of Public Health and Primary Health Care, University of Bergen

SUMMARY
Background Alzheimer’s disease (AD), the most common cause of dementia, is a major cause of disability and care bur-
den in the elderly. This study aims to estimate the costs of formal and informal care and identity determinants of care costs.
Materials and methods Two hundred and seventy-two (AD) patients and their caregivers were recruited among patients
attending regular visits at six memory clinic in Sweden, Denmark, Norway and Finland. Patients with a diagnosis of AD and
with an identifiable primary caregiver were eligible for inclusion. Data was collected by questionnaires at baseline, and at
scheduled follow-up visits after 6 months and again after 12 months. Cognitive function was assessed with the Mini Mental
State Examination (MMSE) and behavioural disturbances were measured using a brief version of the neuropsychiatric
inventory (NPI).
Results Total annual costs were on average 172 000 SEK, ranging from 60 700 SEK in mild dementia to 375 000 SEK in
severe dementia. Costs for community care (special accommodation, home help, etc.) constituted about half of total costs of
care and increase sharply with increasing cognitive impairment. Informal care costs, valued at the opportunity cost of the
caregiver’s time, make up about a third of total costs and also increased significantly with disease severity. Medical care
costs (inpatient care, outpatient care, pharmaceuticals), on the other hand, were not significantly related to disease severity.
Regression analysis confirmed a strong association between costs and cognitive function, between patients as well as within
patients over time. There was also a significant influence on costs from behavioural disturbances. Sensitivity analysis
showed that the method chosen to value informal care can have considerable impact on results.
Conclusions Costs of care in patient with AD are high and related to dementia severity as well as presence of behavioural
disturbances. The cost estimates presented have implications for future economic evaluation of treatments for Alzheimer’s
disease. Copyright # 2006 John Wiley & Sons, Ltd.

key words — dementia; cognition; economics; informal care

INTRODUCTION mortality in the elderly (Henderson and Jorm, 2000).

Alzheimer’s disease (AD) is the most common cause
of dementia and a major cause of morbidity and
*Correspondence to: Dr L. Jönsson, European Health Economics,
London SW1Y4JS, UK. Tel: þ44(0)20 7661 8128 þ44(0) 7789
763256. E-mail: linus.j@healtheconomics.se
Contract/grant sponsor: Merck & Co., Inc. Whitehouse Station, NJ,
USA.
Copyright # 2006 John Wiley & Sons, Ltd.
To optimize the use of limited resources available for
dementia care, new treatments should be assessed and
compared with existing options and priority should be
given to cost-effective strategies, i.e. interventions that
generate large health gains in relation to the incremen-
tal cost (Karlsson et al., 1998). An understanding of
the resource utilization and costs in different stages
of AD is a central part of any economic evaluation
in this field (Jönsson et al., 2000).
Received 1 December 2003
Accepted 11 October 2005
450 l. jönsson
When assessing the economic burden of AD it is
important to distinguish between costs for patients
with the disease, and costs for AD. It is only the
costs for patients with AD that can be observed; the
costs ‘due’ to AD is a theoretical construct requiring
assumptions regarding the attributability of different
resources to AD, alternatively comparison with non-
demented controls or with subjects in the very early
stage of the disease.
Studies from several countries with different health
care systems have demonstrated higher total costs of
care for patients with AD compared with matched
controls (Østbye and Crosse, 1994; Wimo et al.,
1997; Kronborg Andersen et al., 1999; Souetre et al.,
1999; Taylor and Sloan 2000). Differences have been
shown also in costs of care for other somatic conditions
in AD patients with no direct relation to AD (e.g. hip
fracture, stroke, pneumonia), indicating a high degree
of co-morbidity, and that the management of co-morbid
conditions is complicated by the AD diagnosis (Sloan
and Taylor, 2002). For example, balance disturbances
and dysphagia due to late-stage AD predisposes for
infections and traumatic injuries.
As the disease progresses there is a loss of cogni-
tive and physical functions which ultimately leads
to complete dependency. Several studies have demon-
strated a strong relationship between total costs of
care and the level of dementia severity. Severity has
been assessed by the level of cognitive impairment
(e.g. MMSE scores) (Rice et al., 1993; Ernst et al.,
1997; Hux et al., 1998; Schulenberg et al., 1998;
Jönsson et al., 1999b; Souetre et al., 1999; Kavanagh
and Knapp, 2002; Small et al., 2002; Wolstenholme
et al., 2002), the occurrence of behavioural distur-
bances (Kavanagh and Knapp, 2002; Murman et al.,
2002; Small et al., 2002; Wolstenholme et al.,
2002), the physical abilities of the patient in terms
of activities of daily living (ADL) (Chiu et al.,
1999; Kavanagh and Knapp 2002; Small et al.,
2002; Wolstenholme et al., 2002), and by global
scales such as the clinical dementia rating (CDR)
(Østbye and Crosse 1994; Kronborg Andersen et al.,
1999; Leon and Neumann, 1999; Trabucchi, 1999;
Fillenbaum 2001a,b;).
The costs for patients with AD fall to a large extent
outside the health care sector (Wimo et al., 1997). Costs
for special accommodation and other community care
services dominate in previous cost-of-illness studies,
while costs for medical care are relatively low. Compar-
ison of previous studies is hampered by difficulties in
determining which resources have been included and
how resources are valued. This is particularly important
in relation to informal care, i.e. unpaid care-giving pro-
Copyright # 2006 John Wiley & Sons, Ltd.
ET AL.
vided by families or friends of the disabled. Quantify-
ing time spent on informal care activities is a
challenging task, and the RUD (Resource Utilization
in Dementia) instrument used in this study has been
developed for this purpose (Wimo et al., 2000b). Dif-
ferent principles have been proposed for the valuation
of informal care (Brouwer et al., 1999; McDaid, 2001).
In this study we employ a method based on the oppor-
tunity cost of the caregiver’s time.
The aim of this study was to estimate the cost of
medical care, community care and informal care in
different stages of AD in Sweden, Norway, Denmark
and Finland. The second aim was to estimate the rela-
tionship between costs and disease severity and to
identify important cost drivers. This is accomplished
by cross-sectional analysis comparing costs in differ-
ent stages of AD, and also by repeated observations to
relate costs to disease stage within individual patients.
METHODS
Study design
Two hundred and seventy-two AD patients and their
primary caregiver (spouse or child to the patient in
most cases) were enrolled in a prospective observa-
tional study. Subjects were recruited among patients
attending regular visits at memory clinics at six cen-
ters: Stockholm, Uppsala and Piteå (Sweden), Copen-
hagen (Denmark), Kuopio (Finland) and Bergen
(Norway). Patients with a diagnosis of Alzheimer’s
disease according to the diagnostic criteria utilized
in clinical practice at the participating centres and
with an identifiable primary caregiver were eligible
for inclusion. The study was approved by the local
ethics committees and informed consent was obtained
from caregivers and from patients (when possible)
prior to inclusion. Data were collected by question-
naires at baseline, and at scheduled follow-up visits
after 6 months and again after 12 months.
Data collection procedure
At each data collection point, questionnaires were
administered individually to patients, caregivers and
the responsible physician. Physicians answered ques-
tions about the occurrence of specific diagnoses
including cardiovascular disease, diabetes, autoim-
mune disease, cancer, pulmonary disorders and neu-
rological disease, as well as hospitalizations,
outpatient care and pharmaceuticals. Cognitive func-
tion was assessed with the Mini Mental State Exam-
ination (MMSE) (Folstein et al., 1975). The MMSE
Int J Geriatr Psychiatry 2006; 21: 449–459.
 determinants of cost of care for ad patients
score ranges from 0 (severe cognitive impairment) to
30 (no cognitive impairment) and is known to depend
on age and educational level (Crum et al., 1993).
Behavioural disturbances were measured using a brief
version of the neuropsychiatric inventory (NPI),
based on presence and severity (mild, moderate,
severe) of ten important symptoms (Cummings
et al., 1994; Kaufer et al., 2000). The maximum score
is 30, higher scores indicating more disturbances.
Patients were asked to respond to questions regard-
ing their use of community care services, accommoda-
tion and employment status (when relevant). In cases
when the patient was unable to respond reliably to the
questions, this was noted and the caregiver or physi-
cian completed the questionnaire on behalf of the
patient. Caregivers were asked about their relationship
to the patient, employment status, the amount of infor-
mal care provided and also questions about their own
health and use of health care services. Patients and
caregivers also completed the EuroQoL (EQ-5D)
health status questionnaire (EuroQoL-group, 1990).
Cost estimates
Resources were quantified in units defined as to be
relevant for all four countries, based on the Resource
Utilization in Dementia (RUD) instrument (Wimo
et al., 1998). It is thus possible to aggregate resource
use for all patients and apply unit costs for a specific
country. Costs were estimated from the societal per-
spective by combining resource usage data with unit
costs for Sweden (national average costs when avail-
able and otherwise local cost data), presented in
Table 1. Forty-five percent of the study sample was
from Sweden. Unit costs were estimated as to reflect
the opportunity cost of the resource, including over-
heads. All costs are presented in 2003 Swedish
Kronor (SEK), inflated using the standard consumer
price index (SCB, 2003b). Costs for inpatient care
were estimated by multiplying the length of stay by
the daily cost at the specific clinic, plus costs for
intensive care (if any). Costs for prescription drugs
were calculated from the current price list (Apoteket
AB, 2003).
Informal care was assessed by asking caregivers
about the time spent during an average day on differ-
ent care-giving tasks (feeding, continence, hygiene,
clothing, medication, indoor and outdoor transporta-
tion, supervision). For validation purposes caregivers
were also asked to state the amount of time the patient
was left unattended each day. The opportunity cost of
the caregiver’s time was estimated separately for lost
production time and lost leisure time. Caregivers with
Copyright # 2006 John Wiley & Sons, Ltd.
451
gainful employment were asked for the amount of
time they had been absent from work to care for the
patient. The value of lost production was valued at the
average hourly salary including social insurance
taxes, 196 SEK per hour (SCB, 2003a). Remaining
informal care was valued as lost leisure time at 28
SEK per hour, according to a previous contingent
valuation study performed by the Swedish Road
Authority (Vägverket, 1997). Since there is contro-
versy regarding methods for valuing informal care,
different approaches were explored and compared in
a sensitivity analysis.
Analysis
For analysis and exposition patients were divided into
states according to the MMSE score. This avoids
choosing a particular functional form for the relation-
ship between MMSE scores and costs (which is often
non-linear) and is also in line with the methods used
in simulation modelling (e.g. Markov models). Cut-
off points were set to create five sub-states with clini-
cally meaningful differences: very mild dementia
(MMSE 26–30), mild dementia (MMSE 21–25),
moderate dementia (MMSE 15–20), moderately
severe dementia (MMSE 10–14) and severe dementia
(MMSE 0–9) (Jönsson et al., 1999a).
Co-morbidity was included in the analyses as a sin-
gle variable: the number of co-morbid conditions
(ranging from 0 to 13). Physical dependency was
derived from EQ-5D responses. Patients for which
caregivers indicated severe impairment of mobility,
self-care or the ability to perform usual activities were
classified as dependent.
The Kruskal-Wallis test, a non-parametric rank
analogue of one-way analysis of variance, was used
to test for differences in costs of care between
patients in different MMSE states (StataCorp,
2003). As the cost data were skewed, confidence
intervals were estimated using the bias-corrected
accelerated (BCa) percentile bootstrap method
(Briggs et al., 1997). Ordinary linear regression as
well as fixed-effects panel data regression methods
were used to estimate the relationship between total
costs of care and patient characteristics. In the fixed
effects model individual patient effects are esti-
mated as parameters in the model, and the remain-
ing variation comes from changes over time within
each subject. This is of particular interest for the
study of how costs change as dementia progresses
in AD patients, and also has the advantage of requir-
ing less distributional assumptions than the random
effects model.
Int J Geriatr Psychiatry 2006; 21: 449–459.
452 l. jönsson
All calculations were performed in STATA 8.0 for
Windows.
RESULTS
Two hundred and eight patients (76.5%) completed
all three interviews. One center (20 patients) only par-
ticipated in the baseline and first follow-up visits
since the clinic was closed down. Other reasons for
discontinuing the study were the following: 25
patients were lost to follow-up (moved or were other-
wise lost), seven died, consent was withdrawn by
patient or caregiver in six cases, and six patients dis-
continued due to somatic illness. Patients who did not
complete all interviews were not significantly differ-
ent at baseline from patients who completed all three
interviews with respect to age, gender, education,
MMSE scores, brief NPI scores, co-morbidity or costs
of care. Thirteen percent of patients were able to com-
plete the questionnaire unassisted, 42% were able to
respond but required some prompting by an inter-
viewer, 40% required help by the interviewer to fill
out the questionnaire and 5% were not able to partici-
pate. Patients with more severe cognitive impairment
required more assistance with the questionnaire.
Table 2 shows demographic and disease charac-
teristics of the cohort. The average age was 75.9
years, and patients had been diagnosed with AD
on average 1.6 years prior to inclusion in the study
(range 0 years to 8.8 years). Of the patient, 71.1%
were on treatment with a cholinesterase inhibitor
at baseline. The distribution of patients across
MMSE states changed during the study as patients
progressed towards more severe states. The average
MMSE score was 19.2 at baseline, 17.8 at 6 months
and 16.3 at 12 months for patients completing all
three visits (p < 0.001 for trend). The share of
patients in special accommodation (service flats,
home for the elderly, group living or nursing home)
ranged from 3.5% for patients with MMSE
26 to
40% for patients with MMSE < 10.
Utilization of medical care resources is presented
in Table 2 for patients in different MMSE groups
(pooled observations). The average annual number
of days in hospital was similar across states ( 4
days/year) except for the most severe state where
the average was 24 days in hospital per year. The dis-
tribution was highly skewed with the hospital days
concentrated in a few patients, particularly in the most
severe state where four patients were hospitalized for
longer than 3 months. The ‘other’ category includes
several different types of wards: ophthalmology, reha-
bilitation, specialized ward for dementia care and
Copyright # 2006 John Wiley & Sons, Ltd.
ET AL.
respite care. The total annual number of outpatient
visits ranged from 4.9 to 10.7 with a tendency towards
more frequent visits in more severe disease states.
Table 3 shows the utilization of community care
resources. The distribution of patients over different
forms of accommodation is related to the MMSE
score. For example, the share of patients in group liv-
ing, home for the elderly or nursing home was 0% and
33.3% for patients with scores
26 and < 10, respec-
tively. As expected, patients in ordinary living or ser-
vice houses required more home help and other
services than those in special accommodation with a
higher level of service.
The average number of hours per day of informal
care is presented in Figure 1, where supervision is
separated from more active care tasks. The average
amount of care provided was 9.3 hours per day in
the most severe state, indicating practically constant
need for supervision. Fifty-six percent of these
patients could not be left unattended by the caregiver
at any time.
Table 4 shows annual costs for medical, community
and informal care by MMSE state, calculated by pool-
ing data from all three measurements. Costs for
cholinesterase inhibitors were highest in mild to mod-
erate dementia. Costs for other drugs acting on the
central nervous system increase consistently with
increasing disease severity, while total drug costs
were similar across MMSE states.
Community care and informal care costs varied
strongly across states (p < 0.001). Costs for medical
care, however, did not significantly differ between
MMSE states (p ¼ 0.45). Informal care costs consti-
tute on average 27% of total care costs, and the share
is as expected lower for patients in institutions (5%)
compared with patients living at home (37%).
Supervision Specific activities
12
10
Hours per day
8
6
4
2
0
MMSE 26-30 MMSE 21-25 MMSE 15-20 MMSE 10-15 MMSE 0-9
Figure 1. Informal care in hours per day for patients living in the
community in different MMSE levels (95% confidence intervals)
MMSE ¼ Mini-Mental State Examination.
Int J Geriatr Psychiatry 2006; 21: 449–459.
 determinants of cost of care for ad patients
To analyse the independent effects of time since
diagnosis, cognitive function, behavioural distur-
bances and co-morbidities on total costs of care, a lin-
ear regression analysis was conducted on data from
the three visits separately, and also for pooled data
from all visits, adjusting standard errors for multiple
observations per patient (Table 5).
After testing alternative specifications and excluding
non-significant covariates, a log-lin model was selected
including dummy variables for MMSE states, brief NPI
score, time since diagnosis and co-morbidities, since
this model produced normally distributed, homoskedas-
tic residuals and passed the Ramsey RESET test for
model misspecification (Gujarati, 1988). In this model
the natural logarithm of total costs are regressed on the
(untransformed) independent variables. Physical
dependency was also included in an alternative model.
A homogeneity test was carried out to check the appro-
priateness of pooling data from the four countries par-
ticipating in the study (Gujarati, 1988); the p-value for
this test was fairly low but not significant at the 5%
level (p ¼ 0.08).
For the pooled analysis, using the least severe
MMSE state (> 26) as reference, costs were 2.4 times
higher for patients with MMSE 15–20, 3.1 times
higher for MMSE 10–14 and 4.2 times higher for
MMSE 0–9. There was no significant difference
between the two mildest states (MMSE > 20). Costs
increased by 8% for each point on the brief NPI scale,
10% for each extra year since diagnosis and 11% for
each additional co-morbidity (not significant).
When individual co-morbidities were introduced in
the model rather than the total count of co-morbid-
ities, only heart failure was associated with signifi-
cantly higher costs. Weighting co-morbidities with
the same weights used by the Charlson Index
(Charlson et al., 1987) did not alter the results.
In an alternative model including a dummy variable
for physical dependency, coefficients for MMSE and
brief NPI scores were reduced but remained signifi-
cant. Dependent patients incurred 2.8 times higher
costs, controlling for other factors.
In the fixed-effects model, coefficients for the
MMSE states were jointly significant (p ¼ 0.004).
Costs in the mildest state were significantly lower
than in the two most severe states. There was no sig-
nificant difference between any of the three mildest
states, or between the two most severe states
(p> 0.09). Costs increased substantially with time
since diagnosis but the coefficients for brief NPI
and co-morbidities were not significant. One explana-
tion could be that there was insufficient variation over
time in co-morbidities and behavioural disturbances
Copyright # 2006 John Wiley & Sons, Ltd.
453
to separate these effects from the overall increase in
costs with time.
Since the estimation and valuation of informal care
is associated with uncertainty, the consequences of
applying different valuation principles are investi-
gated in a sensitivity analysis. An alternative way of
valuing leisure time is at the average after-tax salary
(Gold et al., 1996), about 90 SEK per hour [the pre-
tax salary is 140 SEK (SCB, 2003c), and the marginal
tax rate is 35.2% (RSV, 2003)]. Another alternative is
the ‘replacement cost method’ applying the cost of a
formal caregiver to informal care (227 SEK per hour,
using the cost of home help in Table 1). Average total
care costs in different states are presented in Table 6
under different valuation principles, as well as results
from pooled linear regressions with robust standard
errors. Coefficients for MMSE states are jointly sig-
nificant under all scenarios ( p < 0.05), but coeffi-
cients are smaller in absolute values in the model
with zero costs for informal care indicating a smaller
cost difference between states.
A number of independent variables were initially
included in the regression models but later excluded
due to non-significance at the 10% level. This
includes age, gender and cholinesterase drug use.
DISCUSSION
This study presents data on costs of care for 272 AD
patients and their caregivers in Sweden, Denmark,
Norway and Finland. Total annual costs were on
average 172 000 SEK, ranging from 60 700 SEK in
mild dementia to 375 000 SEK in severe dementia.
These estimates are comparable to previous findings
by cost-of-illness studies in the Nordic countries. In
Sweden, Wimo and Jönsson (2000) estimated formal
care costs to 247 000 SEK and total cost (including
also informal care) to 289 000 SEK per demented
person in 2000. Jönsson et al. (1999b) calculated
annual formal care cost to 81 700 SEK in mild
dementia, 277 800 SEK in moderate dementia
and 446 700 SEK in severe dementia in 1999. Formal
care costs of AD in Denmark have been estimated to
DKK1 85 000 in mild dementia, DKK 152 000 in
moderate dementia and DKK 207 000 in severe
dementia (Kronborg Andersen et al., 1999). This
study used a low cost for nursing home care, exclud-
ing costs for food, heating, etc. In Finland, annual
costs of special accommodation and inpatient care
has been estimated to on average US$2 16 000 per
demented person in 1999 (Viramo et al., 2003).
There are to our knowledge no previous cost esti-
mates for Norway.
Int J Geriatr Psychiatry 2006; 21: 449–459.
454 l. jönsson ET AL.

Table 1. Unit costs (per day unless otherwise stated), SEK (2003)
Resource Cost Reference

Accommodation
Block of service flats 573 (Wimo and Sandman, 1999)
Home for the elderly 709 (Wimo and Sandman, 1999)
Group living 1076 (Wimo and Sandman, 1999)
Nursing home 1473 (Wimo and Sandman, 1999)
Community services
Home help, per visit 227 (Henriksson and Akeson, 2000)
Home-delivered meals 84 (Jönsson, Gurner, 2001)
Day center visit 235 (Jönsson, Gurner, 2001)
Transportation, per hour 171 (Vägverket, 1997)
Personal alarm 9 (Jönsson, 2001)
Inpatient care
Geriatrics and respite care 2993 (Samverkansnämnden-Uppsala, 2003)
Internal medicine 3300 (Landstingsförbundet, 1999)
Cardiology 5642 (Södra-Samverkansnämnden, 2003)
Surgery 4846 (Landstingsförbundet, 1999)
Orthopaedics 4770 (Landstingsförbundet, 1999)
Intensive care (ICU) 21 634 (Samverkansnämnden-Uppsala, 2003)
Outpatient care
General practicioner visit 752 (Jönsson and Gurner, 2001)
Specialist visit 1616 (Jönsson and Gurner, 2001)
Nurse visit 367 (Jönsson and Gurner, 2001)
Emergency room visit 2057 (Jönsson and Gurner, 2001)
Day care 403 (Samverkansnämnden-Uppsala, 2003)
Informal care
Lost production (per hour) 196 (SCB, 2003a)
Lost leisure time (per hour) 28 (Vägverket, 1997)

Dementia care organization and delivery is quite
similar across Nordic countries, with a relatively high
availability of formal dementia care resources includ-
ing institutional living, and similar financing arrange-
ments with high coverage of tax-based social
insurances (Gerdtham and Jönsson 1994). Although
no significant difference in resource utilization
between the four included countries was found, the
study was not designed or powered to detect such dif-
ferences so these finding should be interpreted with
caution. We believe, however, that the principal
results of this study are applicable to each of the coun-
tries, and that in future studies it can be defended to
pool resource use data collected in a comparable way
(i.e. with the RUD instrument) across the Nordic
countries.
A limitation of this study is the selected sample on
which the analysis was based; patients were recruited
from memory clinics only, although it is only a pro-
portion of AD patients who are diagnosed and treated
at memory clinics. Patients managed by general prac-
titioners, or not recognized at all by the health care
system, may have different resource utilization pat-
terns. The findings from this study therefore need to
be compared and validated against population-based
data.
Another potential limitation is the patients lost to
follow-up (23.5%), which is known to be a source
of bias if the drop-out is informative. The patients
who were lost to follow-up were, however, not dif-
ferent from the rest of the sample at baseline. Also,
similar relationships between costs and MMSE scores
were observed in all three data points as well as in the
panel data analysis.
Costs for community care (special accommoda-
tion, home help, etc.) constitute about half of total
costs of care and increase sharply with increasing
cognitive impairment. Medical care costs (inpatient
care, outpatient care, pharmaceuticals), on the other
hand, were not significantly related to disease sever-
ity. A possible explanation is that in the early stages
of AD there are medical costs for establishing the
diagnosis, while there are costs for somatic compli-
cations and co-morbidities in the later stages of the
disease. Previous studies have also found non-linear
relationships between time since diagnosis and
medical care costs (Taylor et al., 2001), however
the present study is not powered for a detailed
analysis of this functional form. Administrative
databases and similar large data sets may provide
materials of sufficient size for addressing these
issues (Gutterman et al., 1999).

Copyright # 2006 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2006; 21: 449–459.
 determinants of cost of care for ad patients 455

Table 2. Characteristics of the study population and medical care resource utilization
MMSE MMSE MMSE MMSE MMSE ALL P for difference
26–30 21–25 15–20 10–14 0–9 between groupsy

Subjects (visit 1) 47 (17.4%) 89 (33%) 85 (31.5%) 26 (9.6%) 23 (8.5%) 270 (100%)

Subjects (visit 2) 27 (11.5%) 76 (32.5%) 76 (32.5%) 22 (9.4%) 33 (14.1%) 234 (100%)
Subjects (visit 3) 14 (6.9%) 60 (29.4%) 66 (32.4%) 30 (14.7%) 34 (16.7%) 204 (100%)
% on cholinesterase inhibitor* 68.1% 71.9% 77.6% 57.7% 65.2% 71.1% < 0.01
Time since diagnosis (years)* 1.3 1.2 1.6 2.8 2.8 1.7 < 0.01
Age (years)* 76.0 75.9 77.0 74.6 73.2 75.9 n.s.
Gender (% male)* 37.8% 48.3% 27.1% 42.3% 30.4% 37.7% 0.058
Mean hospital days per year
Geriatrics 2.95 0.43 1.32 0.67 18.09 3.30 < 0.01
Internal medicine 0.14 0.25 0.70 0.67 0.49 0.46 n.s
Cardiology 0.00 0.05 0.04 0.00 0.00 0.03 n.s
Psychiatric 0.00 0.00 0.53 0.00 0.00 0.17 n.s
Orthopaedics 0.00 0.07 0.00 0.56 0.00 0.08 n.s
Surgery 0.23 0.20 0.39 0.00 0.00 0.21 n.s
Other 0.05 0.66 1.02 2.41 5.38 1.49 n.s
All departments 3.36 1.65 4.00 4.31 23.96 5.75 < 0.01
Mean health care visits per year
General practicioner 1.95 1.53 1.64 1.90 2.00 1.72 0.053
Specialist physician 1.82 1.19 1.02 0.77 0.44 1.07 0.043
Emergency room 0.00 0.21 0.12 0.10 0.18 0.14 n.s
Day care 0.00 0.28 0.70 0.26 0.31 0.39 n.s
Nurse visit 0.50 2.22 2.31 6.77 4.49 2.82 0.052
Physiotherapist 0.50 0.84 1.04 0.82 0.84 0.86 n.s
Other visits 0.09 0.32 0.05 0.05 0.00 0.14 n.s
All visits 4.86 6.60 6.89 10.67 8.27 7.14 < 0.01

MMSE ¼ Mini-Mental State Examination;

*At baseline (visit 1)
y
one-way ANOVA
n.s ¼ not significant (p > 0.1)

Table 3. Type of accommodation (%) and community care resource utilization (mean number of occations per day)
Ordinary, alone Ordinary, together* Service flat Group living Home for the elderly Nursing home

MMSE 26–30 16 (18.8%) 66 (77.6%) 3 (3.5%) 0 (0%) 0 (0%) 0 (0%)

MMSE 21–25 63 (28%) 151 (67.1%) 6 (2.7%) 0 (0%) 0 (0%) 5 (2.2%)
MMSE 15–20 62 (27.3%) 123 (54.2%) 15 (6.6%) 4 (1.8%) 5 (2.2%) 18 (7.9%)
MMSE 10–15 12 (15.4%) 45 (57.7%) 5 (6.4%) 3 (3.8%) 2 (2.6%) 11 (14.1%)
MMSE 0–9 10 (11.1%) 44 (48.9%) 6 (6.7%) 4 (4.4%) 8 (8.9%) 18 (20%)
Type of service
Home help 2.68 0.44 3.25 0.99 0.00 0.00
Nurse assistance 0.61 0.06 1.20 0.95 0.00 0.00
Transportation 0.11 0.13 0.03 0.13 0.00 0.00
Other assistance 0.05 0.39 0.45 0.00 0.00 0.00
Home-delivered meals 1.54 0.12 3.57 4.20 0.45 0.00
Day center visits 0.58 0.23 1.09 0.13 0.00 0.00
Alarm 6% 2% 29% 13% 0% 4%

*Patient living together with caregiver.

MMSE ¼ Mini-Mental State Examination.

Informal care costs, valued at the opportunity cost
of the caregiver’s time, make up about a fourth of total
costs and also increased significantly with disease
severity. Informal care is an important factor in the
management of patients with AD still living in the
community. As is shown in Table 3, patients living
together with a caregiver have much lower utilization
of community care resources. The valuation of infor-
mal care is controversial and cost estimates depend
highly on the choice of valuation method. We have
used a standard method for valuing lost production
(the human capital approach), and an estimate of

Copyright # 2006 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2006; 21: 449–459.
456 l. jönsson ET AL.

Table 4. Mean annual costs per patient (SEK 2003) by category and MMSE state
MMSE 26–30 MMSE 21–25 MMSE 15–20 MMSE 10–14 MMSE 0–9 All patients

Medical care
Cholinesterase inhibitors 7631 8478 8276 6958 6102 7838
95% confidence interval (6728–8535) (7954–9049) (7790–8819) (5842–7948) (5182–7289) (7416–8092)
Other CNS drugs 633 874 1292 1542 1919 1185
(ATC class N except N06D)
95% confidence interval (430–952) (669–1149) (1006–1603) (922–2256) (1369–2611) (1022–1357)
All pharmaceuticals* 10 148 10 720 11 189 10 322 9814 10 640
95% confidence interval (8865–11 261) (10 015–11 419) (10 323–11 872) (9018–11 974) (8277–10 927) (10 196–11 132)
Outpatient care 4715 6419 8085 7029 6208 6925
95% confidence interval (3382–6633) (4696–8024) (5875–11 244) (4470–12 756) (3704–11 008) (5742–7907)
Inpatient care 13 903 5930 13 603 15 017 71 729 20 605
95% confidence interval (6846–26 313) (3261–12 135) (7436–24 845) (4371–41 861) (36 197–133 004) (14 289–28 176)
Community care
Special accommodation 7214 17 131 68 105 106 814 164 136 61 222
95% confidence interval (2222–20 699) (8380–34 234) (50 034–90 049) (74 309–156 740) (122 384–215 008) (51 410–74 739)
Other community care 6049 18 765 35 441 20 985 46 131 26 662
95% confidence interval (3454–14 195) (13 913–24 425) (26 757–49 908) (11 351–40 025) (28 535–74 656) (22 404–31 746)
Informal care
Lost production 0 6394 6720 7317 11 698 6338
95% confidence interval (–) (2795–18 516) (3631–11 260) (877–36 509) (3276–25 539) (4091–10 617)
Lost leisure time 18 702 28 601 40 939 59 393 65 247 39 733
95% confidence interval (11 701–27 963) (22 138–36 997) (32 752–52 053) (45 242–82 612) (49 241–85 823) (34 776–45 147)
TOTAL 60 730 93 959 184 081 226 876 374 962 172 121
95% confidence interval (50 203–80 149) (79 284–115 097) (159 262–210 577) (180 938–287 141) (314 115–472 000) (152 059–188 109)

MMSE ¼ Mini-Mental State Examination; CNS ¼ Central Nervous System; ATC ¼ Anatomical Therapeutical Classification.
*Including cholinesterase inhibitors.

Table 5. Regression analysis on (log) total costs of care

Linear regression analysis Fixed-effects
regression analysis

Baseline 6 months 12 months All observations* All observations*

with dependency

exp(
) p exp(
) p exp(
) p exp(
) p exp(
) p exp(
) p

MMSE
MMSE 26–30 (reference category)
MMSE 21–25 1.121 0.712 1.761 0.115 1.347 0.513 1.363 0.091 1.255 0.194 0.958 0.891
MMSE 15–20 1.774 0.071 2.643 0.008 3.349 0.008 2.421 < 0.001 1.954 0.001 1.331 0.440
MMSE 10–14 1.243 0.605 4.453 0.002 6.481 < 0.001 3.123 < 0.001 2.224 0.001 3.271 0.009
MMSE 0–9 3.663 0.004 6.201 < 0.001 4.185 0.005 4.246 < 0.001 2.385 0.001 2.808 0.056
Brief NPI 1.118 0.001 1.070 0.019 1.039 0.180 1.077 < 0.001 1.046 0.013 1.008 0.731
Years since diagnosis 1.171 0.030 1.062 0.421 0.986 0.856 1.104 0.035 1.089 0.066 1.637 < 0.001
Co-morbitities 1.099 0.160 1.157 0.213 1.265 0.023 1.114 0.088 1.106 0.050 1.149 0.203
Dependency — — — — — — — — 2.770 < 0.001 —
(Constant) 9.809 < 0.001 9.842 < 0.001 10.266 < 0.001 9.910 < 0.001 9.943 < 0.001 9.656 < 0.001
2
R 0.1628 0.1908 0.2354 0.1749 0.2683 0.0998

MMSE ¼ Mini-Mental State Examination; NPI ¼ Neuropsychiatric Inventory; Dependency: estimated from EQ-5D scores (see text);
*Robust standard errors, adjusted for clustering.

the cost of lost leisure time that is low by comparison,
only 14% of the cost of lost production. Valuing lei-
sure time at the after-tax salary or applying the repla-
cement cost method leads to substantially higher cost
estimates but has little effect on the relative cost dif-
ference between MMSE states.
Regression analysis confirmed a strong associa-
tion between costs and cognitive function, and also
a significant influence on costs from behavioural
disturbances. Costs were log-transformed prior to
analysis to reduce skewness (Andersen et al.,
2000). This may lead to problems when attempting

Copyright # 2006 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2006; 21: 449–459.
 determinants of cost of care for ad patients 457

Table 6. Sensitivity analysis on different principles for valuation of informal care

Valuation principle Base-case Zero cost of informal care After-tax salary Replacement cost method

Value of leisure time (SEK/h) 28 0 90 227

Value of working time (SEK/h) 196 0 196 227
Total cost of care (SEK/year)
MMSE 26–30 60 730 42 029 102 143 193 649
MMSE 21–25 93 959 58 965 157 291 298 243
MMSE 15–20 184 081 136 423 274 733 476 104
MMSE 10–14 226 876 160 167 358 390 650 149
MMSE 0–9 374 962 298 018 519 439 840 533
Linear regression analysis exp(
) p exp(
) p exp(
) p exp(
) p
MMSE — — — —
MMSE 26–30 (reference category)
MMSE 21–25 1.363 0.091 1.011 0.954 1.427 0.107 1.502 0.126
MMSE 15–20 2.421 < 0.001 1.634 0.035 2.616 < 0.001 2.829 < 0.001
MMSE 10–14 3.123 < 0.001 1.923 0.072 3.301 < 0.001 3.475 < 0.001
MMSE 0–9 4.246 < 0.001 2.791 0.017 4.231 < 0.001 4.260 < 0.001
Brief NPI 1.077 < 0.001 1.094 0.001 1.077 < 0.001 1.077 0.001
Years since diagnosis 1.104 0.035 1.023 0.768 1.171 0.003 1.233 0.001
Co-morbitities 1.114 0.088 1.099 0.203 1.143 0.041 1.169 0.026
(Constant) 9.910 < 0.001 9.604 < 0.001 10.121 < 0.001 10.355 < 0.001
R2 0.1749 0.0708 0.1828 0.1763

MMSE ¼ Mini-Mental State Examination; NPI ¼ Neuropsychiatric Inventory.

to re-transform predicted values to (non-logged)
costs; however, the aim in this study was to examine
the link between disease indicators and costs rather
than prediction.
Costs were higher for patients with several co-mor-
bidities, but the coefficient was only borderline signif-
icant. The regression analysis also showed that costs
increase with time since diagnosis after controlling
for MMSE and brief NPI scores. This may simply
be an effect of increasing age leading to higher care
expenditures, however, there may also be other
important cost drivers not presently included in the
model. Physical function could be one such cost dri-
ver; indeed the coefficient for time since diagnosis
was non-significant in the model including physical
dependency. As cognitive and physical function are
highly correlated, including both in the same model
leads to problems with multicollinearity. Reduced
cognitive function is likely responsible for an impor-
tant share of the physical dependency, particularly
regarding more complex tasks, although this may
KEY POINTS
* Costs of care in patient with Alzheimer’s
disease are high and increase dramatically with
increasing dementia severity.
* Costs are higher for patients with behavioural
disturbances, assessed with the brief NPI.
not be true for some aspects of physical function
(e.g. continence, mobility). The cost estimates pro-
vided by this study can be useful in future economic
evaluations of interventions aimed at reducing the
progression of AD in the Nordic countries.
Notes
1
1 DKK ¼ 1.22 SEK (September 2003)
2
1 US$ ¼ 8.09 SEK (September 2003)
ACKNOWLEDGEMENTS
This study was supported by an unrestricted grant
from Merck & Co., Inc. Whitehouse Station, NJ,
USA.
REFERENCES
Andersen CK, Andersen K, Kragh-Sorensen P. 2000. Cost function
estimation: the choice of a model to apply to dementia. Health
Econ 9(5): 397–409.
Apoteket AB. 2003. ‘Apotekets varuregister NLTS för datajournal-
system.’ 2003.
Briggs AH, Wonderling DE, Mooney CZ. 1997. Pulling cost-
effectiveness analysis up by its bootstraps: a non-parametric
approach to confidence interval estimation. Health Econ 6(4):
327–340.
Brouwer WB, van Exel NJ, Koopmanschap MA, Rutten FF. 1999.
The valuation of informal care in economic appraisal. A consid-
eration of individual choice and societal costs of time. Int J Tech-
nol Assess Health Care 15(1): 147–160.

Copyright # 2006 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2006; 21: 449–459.
458 l. jönsson
Charlson ME, Pompei P, Ales KL, MacKenzie CR. 1987. A new
method of classifying prognostic comorbidity in longitudinal
studies: development and validation. J Chronic Dis 40(5):
373–383.
Chiu L, Tang KY, Liu YH, Shyu WC, Chang TP. 1999. Cost com-
parisons between family-based care and nursing home care for
dementia. J Adv Nurs 29(4): 1005–1012.
Crum RM, Anthony JC, Bassett SS, Folstein MF. 1993. Population-
based norms for the Mini-Mental State Examination by age and
educational level. JAMA 269(18): 2386–2391.
Cummings JL, Mega M, Gray K, Rosenberg-Thompson S, Carusi
DA, Gornbein J. 1994. The Neuropsychiatric Inventory: compre-
hensive assessment of psychopathology in dementia. Neurology
44(12): 2308–2314.
Ernst RL, Hay JW, Fenn C, Tinklenberg J, Yesavage JA. 1997.
Cognitive function and the costs of Alzheimer disease. An
exploratory study. Arch Neurol 54(6): 687–693.
EuroQoL-group. 1990. EuroQol—a new facility for the measure-
ment of health-related quality of life. The EuroQol Group.
Health Policy 16(3): 199–208.
Fillenbaum G, Heyman A, Peterson BL, Pieper CF, Weiman AL.
2001a. Use and cost of hospitalization of patients with AD by
stage and living arrangement: CERAD XXI. Neurology 56(2):
201–206.
Fillenbaum G, Heyman A, Peterson BL, Pieper CF, Weiman AL.
2001b. Use and cost of outpatient visits of AD patients: CERAD
XXII. Neurology 56(12): 1706–1711.
Folstein M, Folstein S, McHugh P. 1975. ‘Mini-Mental State: A
practical method for grading the cognitive state of patients for
the clinician. J Psychiat Res 12: 189–198.
Gerdtham UG, Jonsson B. 1994. Health care expenditure in the
Nordic countries. Health Policy 26(3): 207–220.
Gold M, Siegel J, Russel L, Weinstein M. 1996. Cost-Effectiveness
in Health and Medicine. Oxford University Press: New York.
Gujarati D. 1988. Basic Econometrics. McGraw-Hill: New York.
Gutterman EM, Markowitz JS, Lewis B, Fillit H. 1999. Cost of
Alzheimer’s disease and related dementia in managed-medicare.
J Am Geriatr Soc 47(9): 1065–1071.
Henderson AS, Jorm AF. 2000. Definition and epidemiology of
dementia: a review. In Dementia, Maj M, Sartorius N (eds). John
Wiley & Sons, Ltd: Chichester
Henriksson S, Akeson A. 2000. Styrning av äldreomsorg. Järfälla,
Kommun.
Hux MJ, O’Brien BJ, Iskedjian M, Goeree R, Gagnon M,
Gauthier S. 1998. Relation between severity of Alzheimer’s
disease and costs of caring. CMAJ 159(5): 457–465.
Jönsson L, Gurner U. 2001. Kostnader för omsorg och vård om
äldre multisjuka. Stockholm, Stockholm Gerontology Research
Center.
Jönsson L, Jönsson B, Wimo A, Whitehouse P, Winblad B. 2000.
Second International Pharmacoeconomic Conference on
Alzheimer’ s Disease. Alzheimer Dis Assoc Disord 14(3): 137–
140.
Jönsson L, Lindgren P, Wimo A, Jönsson B, Winblad B. 1999a. The
Cost-Effectiveness of Donepezil Therapy in Swedish Patients
with Alzheimer’s Disease: a Markov Model. Clin Therapeut
21(7): 1230–1240.
Jönsson L, Lindgren P, Wimo A, Jönsson B, Winblad B. 1999b.
Costs of MMSE-related cognitive impairment. Pharmacoeco-
nom 16(4): 409–416.
Karlsson G, Wimo A, Jönsson B, Winblad B. 1998. Methodological
issues in health economic studies of dementia. In The Health
Economis of dementia, Wimo A, Karlsson G, Jonsson B,
Winblad B (eds). John Wiley & Sons Ltd: Chichester.
Copyright # 2006 John Wiley & Sons, Ltd.
ET AL.
Kaufer DI, Cummings JL, Ketchel P, et al. 2000. Validation of the
NPI-Q, a brief clinical form of the Neuropsychiatric Inventory.
J Neuropsychiatry Clin Neurosci 12(2): 233–239.
Kavanagh S, Knapp M. 2002. Costs and cognitive disability:
modelling the underlying associations. Br J Psychiatry 180:
120–125.
Kronborg Andersen C, Sogaard J, Hansen E, et al. 1999. The cost of
dementia in Denmark: the Odense Study. Dement Geriatr Cogn
Disord 10(4): 295–304.
Landstingsförbundet. 1999. Kostnader per intagen patient (cost per
admitted patient) 1996. Stockholm, Sweden, Cooperation of
County Councils (Landstingsförbundet).
Leon J, Neumann PJ. 1999. The cost of Alzheimer’s disease in
managed care: a cross-sectional study. Am J Manag Care 5(7):
867–877.
McDaid D. 2001. Estimating the costs of informal care for people
with Alzheimer’s disease: methodological and practical chal-
lenges. Int J Geriatr Psychiatry 16(4): 400–405.
Murman DL, Chen Q, Powell MC, Kuo SB, Bradley CJ, Colenda
CC. 2002. The incremental direct costs associated with beha-
vioral symptoms in AD. Neurology 59(11): 1721–1729.
Østbye T, Crosse E. 1994. Net economic costs of dementia in
Canada. J Can Med Assoc 151: 1457–1464.
Rice D, Fox P, Max W, et al. 1993. The economic burden of
Alzheimeŕs disease care. Health Affairs 12: 164–176.
RSV. 2003. Skattestatistisk årsbok 2002. Stockholm, Swedish
National Tax Authority (Riksskatteverket).
Samverkansnämnden-Uppsala. 2003. Regional price list 2003.
Samverkansnämnden Uppsala-Örebroregionen: Uppsala.
SCB. 2003a. Average hourly wages, private sector. Statistiska
Centralbyrån (Statistics Sweden): Stockholm.
SCB. 2003b. Konsumentprisindex (Consumer Price Index) 1980–
2003. Statistiska Centralbyrån (Statistics Sweden): Stockholm.
SCB. 2003c. Labour Costs in Private Sector, March 2003. Statis-
tiska Centralbyrån (Statistics Sweden): Stockholm.
Schulenberg J, Schulenberg I, Horn R. 1998. Cost of treatment and
cost of care for Alzheimer’s disease in Germany. In The Health
Economis of dementia, Wimo A, Karlsson G, Jönsson B, Winblad
B (eds). John Wiley & Sons Ltd: Chichester; 217–230.
Sloan FA, Taylor DH, Jr. 2002. Effect of Alzheimer disease on the
cost of treating other diseases. Alzheimer Dis Assoc Disord
16(3): 137–143.
Small GW, McDonnell DD, Brooks RL, Papadopoulos G. 2002.
The impact of symptom severity on the cost of Alzheimer’s dis-
ease. J Am Geriatr Soc 50(2): 321–327.
Södra-Samverkansnämnden. 2003. Regional price list 2003. Södra
Samverkansnämnden: Malmö.
Souetre E, Thwaites RM, Yeardley HL. 1999. Economic impact
of Alzheimer’s disease in the United Kingdom. Cost of
care and disease severity for non-institutionalised patients with
Alzheimer’s disease. Br J Psychiatry 174: 51–55.
StataCorp. 2003. Stata 8.0 user’s manual. Stata Corporation. Texas.
Taylor DH Jr, Schenkman M, Zhou J, Sloan FA. 2001. The
relative effect of Alzheimer’s disease and related dementias,
disability, and comorbidities on cost of care for elderly persons.
J Gerontol B Psychol Sci Soc Sci 56(5): S285–S293.
Taylor DH, Jr, Sloan FA. 2000. How much do persons with
Alzheimer’s disease cost Medicare? J Am Geriatr Soc 48(6):
639–646.
Trabucchi M. 1999. An economic perspective on Alzheimer’s dis-
ease. J Geriatr Psychiatry Neurol 12(1): 29–38.
Vägverket. 1997. Vägverkets samhällsekonomiska kalkylmodell.
Ekonomisk teori och värderingar. Vägverket (Swedish Road
Authority): Stockholm.
Int J Geriatr Psychiatry 2006; 21: 449–459.
 determinants of cost of care for ad patients
Viramo P, Nikiforov O, Mikko T, et al. 2003. The care of dementia
patients in Finland in 1999 - use and costs of services (abstract).
Wimo A, Jönsson L. 2000. Demenssjukdomarnas samhällskostna-
der. Socialstyrelsen: Stockholm.
Wimo A, Karlsson G, Sandman P, Corder L, Winblad B. 1997. Cost
of illness due to dementia in Sweden. Int J Geriatric Psychiatry
12: 857–861.
Wimo A, Nordberg G, Jansson W, Grafström M. 2000b. Assess-
ment of informal services to demented people with the
RUD instrument. Int J Geriatric Psychiatry 15: 969–
971.
Copyright # 2006 John Wiley & Sons, Ltd.
459
Wimo A, Sandman P. 1999. Demensvård för nästa sekel. Svenska
Kommunförbundet (Swedish Cooperation of Municipalities):
Stockholm.
Wimo A, Wetterholm A, Mastey V, Winblad B. 1998. Evaluation of
the healthcare resource utilisation and caregiver time in anti-
dementia drug trials. In The Health Economis of Dementia.
Wimo A, Karlsson G, Jönsson B, Winblad B. John Wiley & Sons
Ltd: Chichester; 217–230.
Wolstenholme J, Fenn P, Gray A, Keene J, Jacoby R, Hope T. 2002.
Estimating the relationship between disease progression and cost
of care in dementia. Br J Psychiatry 181: 36–42.
Int J Geriatr Psychiatry 2006; 21: 449–459.