Direct costs associated with depressive symptoms in late life: A 4.5-year prospective study

Abstract

Background:
Depression in old age is common. Only few studies examined the association of depressive symptoms and direct costs in the elderly in a cross-sectional way. This study aims to investigate prospectively health service use and direct costs over a course of 4.5 years considering also different courses of depressive symptomatology.

Methods:
305 primary care patients aged 75+ were assessed face-to-face regarding depressive symptoms (Geriatric Depression Scale), and service use and costs at baseline and 4.5 years later. Resource utilization was monetarily valued using 2004/2005 prices. The association of baseline factors and direct costs after 4.5 years was analyzed by multivariate linear regression.

Results:
Mean annual direct costs of depressed individuals at baseline and follow-up were almost one-third higher than of non-depressed, and highest for individuals with chronic depressive symptoms. Most relevant cost drivers were costs for inpatient care, pharmaceuticals, and home care. Costs for home care increased at most in individuals with chronic depressive symptoms. Baseline variables that were associated with direct costs after 4.5 years were number of medications as a measure of comorbidity, age, gender, and depressive symptoms.

Conclusions:
Presence and persistence of depressive symptoms in old age seems to be associated with future direct costs even after adjustment for comorbidity. The findings deign a look to the potential economic consequences of depressive symptoms in the elderly for the healthcare system in the future.

Full text

International Psychogeriatrics: page 1 of 11 C
International Psychogeriatric Association 2012
doi:10.1017/S1041610212001688
Direct costs associated with depressive symptoms in late life: a
4.5-year prospective study
.........................................................................................................................................................................................................................................................................................................................................................................
Melanie Luppa,1Hans-Helmut König,2Dirk Heider,1,2 Hanna Leicht,2Tom Motzek,1
Georg Schomerus3and Steffi G. Riedel-Heller1
1Institute of Social Medicine, Occupational Health and Public Health, University of Leipzig, Leipzig, Germany
2Department of Medical Sociology and Health Economics, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
3Department of Psychiatry, Ernst-Mor itz-Arndt University Greifswald, Greifswald, Germany
ABSTRACT
Background: Depression in old age is common. Only few studies examined the association of depressive
symptoms and direct costs in the elderly in a cross-sectional way. This study aims to investigate prospectively
health service use and direct costs over a course of 4.5 years considering also different courses of depressive
symptomatology.
Methods: 305 primary care patients aged 75+were assessed face-to-face regarding depressive symptoms
(Geriatric Depression Scale), and service use and costs at baseline and 4.5 years later. Resource utilization
was monetarily valued using 2004/2005 prices. The association of baseline factors and direct costs after 4.5
years was analyzed by multivariate linear regression.
Results: Mean annual direct costs of depressed individuals at baseline and follow-up were almost one-third
higher than of non-depressed, and highest for individuals with chronic depressive symptoms. Most relevant
cost drivers were costs for inpatient care, pharmaceuticals, and home care. Costs for home care increased
at most in individuals with chronic depressive symptoms. Baseline variables that were associated with direct
costs after 4.5 years were number of medications as a measure of comorbidity, age, gender, and depressive
symptoms.
Conclusions: Presence and persistence of depressive symptoms in old age seems to be associated with future
direct costs even after adjustment for comorbidity. The findings deign a look to the potential economic
consequences of depressive symptoms in the elderly for the healthcare system in the future.
Key words: costs, economic burden, service use, depression, depressive symptoms
Introduction
Depressive symptoms are common in late life.
Prevalence is estimated with 17% by a recently
published meta-analysis (Luppa et al., 2012a).
Elderly individuals with depressive symptoms are
faced with a number of negative consequences
including functional decline, marked disability,
decreased quality of life, and higher mortality
from co-morbid medical conditions (Fiske et al.,
2009). Furthermore, it has been shown that
depressive symptoms in old age are associated with
an increased presentation of unexplained physical
Correspondence should be addressed to: Melanie Luppa, Institute of Social
Medicine, Occupational Health and Public Health, University of Leipzig,
Philipp-Rosenthal-Straße 55, D-04103 Leipzig, Germany. Phone: +49-
341-9724534; Fax: +49-341-9715409. Email: Melanie.Luppa@medizin.uni-
leipzig.de. Received 24 Apr 2012; revision requested 26 Sep 2012; revised
version received 12 Sep 2012; accepted 19 Sep 2012.
symptoms such as headache, gastrointestinal
symptoms, and dizziness [8,9], which partly may
contribute to higher health service utilization and
costs of depressed individuals in the healthcare
system. Findings of systematic reviews examining
health service utilization and costs of depression
suggest that depression in late life contributes
substantially to an increased service use and costs
(Luppa et al., 2007; 2012b; König et al., 2010).
However, so far only five studies examined the
impact of depressive symptoms on direct costs in
elderly individuals (Livingston et al., 1997; Unutzer
et al., 1997; Luber et al., 2001; Katon et al., 2003;
Luppa et al., 2008). Only one study examined costs
in a longitudinal perspective (Unutzer et al., 1997);
yet, this study did not analyze the impact of different
types of course of depressive symptomatology on
direct cost (i.e. chronic, remittent, and incident
depressive symptoms).

2M. Luppa et al.
EXCLUSION
n = 29
Refused 10
Deceased 15
Not traceable 2
Severe illness 0
Incomplete
assessment 2
Participants at baseline
n = 451
Depressive
individuals
n = 63
Non-depressive
individuals
n = 388
EXCLUSION
n = 117
Refused 60
Deceased 44
Not traceable 5
Severe illness 6
Incomplete
assessment 2
Depressive
individuals
n = 29
Non-depressive
individuals
n = 276
Baseline
Follow-up
n = 13
n = 21
n = 16
n = 255
Participants at follow-up
n = 305
Figure 1. Sample attrition and course of depressive symptoms.
Thus, the aim of this prevalence-based bottom-
up cost-of-illness study with incremental cost
approach is to investigate health service use and
direct costs from a societal perspective in a sample
of depressed and non-depressed German primary
care patients aged 75 years and older over a period
of 4.5 years considering also different courses of
depressive symptomatology.
Methods
Sample
The study is a part of the longitudinal German
study on Ageing, Cognition and Dementia in
Primary Care Patients (AgeCoDe study), a study
on early detection of mild cognitive impairment
and dementia funded by the German Research
Network on Degenerative Dementia (Kom-
petenznetz Degenerative Demenzen, KNDD),
a study that was conducted at six centers
(Bonn, Düsseldorf, Hamburg, Leipzig, Mannheim,
Munich), representing an urban area of cities with
a total population ranging between about 300,000
(Mannheim) and 1.8 million (Hamburg). Our
analyses refer to the data of the study center Leipzig,
where a total of 1,133 randomly selected individuals
from 20 general medical practices in Leipzig
meeting the inclusion criteria received a postal
invitation from their General Practitioner (GP) to
participate in the study. Inclusion criteria were
being 75 years of age or older and at least one visit
of the patient to the GP within the last 12 months.
Exclusion criteria were GP consultations only
through home visits, residence in a nursing home,
severe illness that, according to the GP, would
prove fatal within three months, insufficient ability
to speak German, deafness or blindness, lacking
ability to consent, and being only an irregular
patient of the participating practice. Of the 451
patients who consented to participate in the baseline
assessment between May 2004 and December
2005 (for detailed information on sampling frame,
eligible subjects, and respondents, see Luppa et al.,
2008), 305 were re-assessed at follow-up after
4.5 years. Seventy patients (15.5%) refused to
participate, 59 (13.1%) had died before the follow-
up interview, seven patients were not traceable
(1.6%), six patients (1.3%) were not consultable
due to very severe illness, and four patients (0.9%)
had incomplete assessments (see Figure 1).
Comparison of baseline characteristics of
participants in the prospective study (n=305) and
non-participants at follow-up (n=146) showed

Direct costs and depressive symptoms 3
no significant differences regarding age, gender,
and marital status between both groups, but
non-participants showed a higher mean Geriatric
Depression Scale (GDS)-score at baseline than
participants (3.1 vs. 2.3, p<0.001), whereby 29
(19.9%) of them showed a GDS-score of 6 and
more.
Instruments
Depressive symptoms were measured using a short
version of the GDS-15, a 15-item screening
instrument developed by Sheikh and Yesavage
(Sheikh and Yesavage, 1986) to identify elderly
patients with depressive symptoms. We defined
significant depressive symptoms with a score of 6 or
greater on the GDS (range 0–15) as recommended
for the German version (Gauggel and Birkner,
1999; Allgaier et al., 2011).
Comorbidity was measured by the number of
medications prescribed at baseline and follow-up:
the higher the number of prescriptions, the higher
the comorbidity (Farley et al., 2006).
Service use was assessed retrospectively over
different time periods for services to control for
recall difficulties, i.e. use of more memorable
and consistent services was recorded over longer
periods (Johnston et al., 1999). Time periods
were: three months for outpatient services, use of
outpatient non-physician provider, medical supply
and dentures, pharmaceuticals and transport to
medical treatments; six months for inpatient care
and assisted living; and 12 months for use of home
care. Measures of use were recorded depending on
the type of service, i.e. frequency or mean duration
of use (see also Luppa et al., 2008). Unit costs from
a societal perspective were calculated for all services
used, and for all pharmaceuticals and medical
supply prescribed or privately purchased. Costs
were extrapolated to annual costs and calculated
in EURO at 2004/2005 price levels. For detailed
information regarding determination of service
use and monetary valuation, see Luppa et al.
(2008).
Statistical analyses
Statistical analyses were performed using PASW
for Windows, version 20.0 (SPSS Inc.). Missing
values for healthcare utilization were imputed by
mean values of respective variables. Differences
in characteristics between participants and non-
participants at follow-up (age, gender, marital
status, and GDS score) and between depressed
and non-depressed individuals as well as between
individuals with no, chronic, remittent, and incident
depressive symptoms over study course (age,
gender, marital status, education, comorbidity,
GDS, and costs) were investigated using two-
sided t-test, one-way ANOVA, Mann-Whitney
U-test, Kruskal-Wallis test, and χ2-analyses as
appropriate.
A common problem for health economics studies
is a non-normal distribution of cost variables
skewed to the right. For that reason, a multiple
linear regression model with non-parametric
bootstrapped standard errors (3,000 replications)
and bias-corrected confidence intervals (Efron and
Tibshirani, 1986) with mean annual direct costs at
follow-up after 4.5 years as dependent variable was
developed. In the fully adjusted model the following
baseline variables were included as explanatory
variables: age in years, gender, the number of
medications as measure of comorbidity, and the
GDS score (different cut-offs) as measurement of
depression status.
Ethical approval
The ethics committee approved the study.
Written informed consent was obtained from all
participants.
Results
Sample characteristics
DEPRESSIVE AND NON-DEPRESSIVE
SYMPTOMS
Sixty-three (14.0%) individuals of the 451 patients
at baseline, and 29 (9.5%) individuals out of 305 at
follow-up were classified as depressed (GDS-Score
≥6). Individuals with depressive symptoms did not
differ significantly from non-depressed individuals
regarding age, gender, education, and marital status
(Table 1).
DIFFERENT COURSES OF DEPRESSIVE
SYMPTOMS
A comparison of characteristics of individuals that
were neither depressed at baseline (BL) nor at
follow-up (FUP) with individuals showing chronic
(depressive symptoms at baseline and follow-up,
n=13, thus 38% of followed-up depressed
individuals), remittent (depressive symptoms only
at baseline, n=21, thus 62% of followed-
up depressed individuals), or incident depressive
symptoms (depressive symptoms only at follow-up,
n=16, thus 6% of followed-up non-depressed
individuals) revealed no significant differences
regarding age, gender, educational level, and marital
status at baseline or follow-up (Table 1).

4M. Luppa et al.
Table 1. Baseline sociodemographic characteristics of depressed and non-depressed at baseline (n=451) as well as individuals without depressive symptoms
at baseline and follow-up, and with chronic, remittent, and incident depressive symptoms (n=305)
VARIABLE
DEPRESSED
INDIVIDUALS
AT BASELINE
(n=63)
NON-
DEPRESSED
INDIVIDUALS
AT BASELINE
(n=388)
t-TEST/
χ2-TEST
P
-VALUE
NON-
DEPRESSED
INDIVIDUALS
AT BASELINE
AND
FOLLOW-UP
(n=255)
INDIVIDUALS
WITH
CHRONIC
DEPRESSIVE
SYMPTOMS
(n=13)
INDIVIDUALS
WITH
REMITTENT
DEPRESSIVE
SYMPTOMS
(n=21)
INDIVIDUALS
WITH
INCIDENT
DEPRESSIVE
SYMPTOMS
(n=16)
ONE-WAY
ANOVA/
χ2-TEST
P
-VALUE
................................................................................................................................................................................................................................................................................................................................................................................................................................................................
Age in years: mean
(SD)
80.98 (4.04) 80.88 (3.63) t=−0.215 0.830 81.20 (3.23) 80.34 (2.14) 81.99 (4.52) 83.27 (5.33) F=2.443 0.064
Gender: n(%)
Female 47 (74.6) 240 (61.9) χ2=3.806 0.051 159 (62.4) 8 (61.5) 16 (76.2) 10 (62.5) χ2=1.622 0.654
Education: n(%)
Low 44 (69.8) 239 (61.6) χ2=3.849 0.146 151 (59.2) 10 (76.9) 15 (71.4) 11 (68.8) χ2=6.153 0.406
Middle 13 (20.6) 72 (18.6) 47 (18.4) 2 (15.4) 5 (23.8) 3 (18.8)
High 6 (9.5) 77 (19.8) 57 (22.4) 1 (7.7) 1 (4.8) 2 (12.5)
Marital status: n(%)
Single 7 (11.1) 31 (8.0) χ2=7.131 0.068 21 (8.2) 0 (0) 2 (9.5) 2 (12.5) χ2=7.921 0.542
Married 24 (38.1) 168 (43.3) 109 (42.7) 7 (53.8) 6 (28.6) 7 (43.8)
Widowed 25 (39.7) 174 (44.8) 114 (44.7) 5 (38.5) 10 (47.6) 7 (43.8)
Divorced 7 (11.1) 15 (3.9) 11 (4.3) 1 (7.7) 3 (14.3) 0 (0)

Direct costs and depressive symptoms 5
Service use
DEPRESSIVE AND NON-DEPRESSIVE
SYMPTOMS
A comparison of service utilization of depressed
and non-depressed individuals at baseline and
follow-up showed significant differences between
depressed and non-depressed individuals for mean
total number of medications only at baseline (BL
5.8 vs. 4.9, p<0.05, FUP 7.1 vs. 5.6, p=0.068)
and for use of antidepressants (AD) including St.
John’s wort (as verbal AD, which is commonly
prescribed in Germany for mild depression), as
well as tri- and tetracyclic AD, and SSRIs (20.6
vs. 4.9%, p<0.001) also only at baseline as well
as for use of medical supply (BL 25.4 vs. 14.2%,
p<0.05, FUP 48.3 vs. 25.8%, p<0.05), and
home care (BL 38.1 vs. 20.4%, p<0.01, FUP
55.2 vs. 32.4%, p<0.05) at baseline and follow-up.
Also, 14% of depressed individuals visited a mental
health specialist (i.e. psychiatrist or neurologist)
at baseline and follow-up compared to 8% and
10% of non-depressed individuals. From baseline
to follow-up, an increase in inpatient care (15.9 vs.
20.7%), medical supply (25.4 vs. 48.3%), and use of
home care (38.1 vs. 55.2%), and a decrease in use
of antidepressants (20.6 vs. 6.9%) and specialists
visits (79.4 vs. 69.0%) in depressed individuals were
found, which was not found in that degree in non-
depressed individuals.
DIFFERENT COURSES OF DEPRESSIVE
SYMPTOMS
A comparison of individuals without depressive
symptoms over the study course with individuals
with chronic, remittent, or incident depressive
symptoms showed only a significant difference
between individuals with no depressive symptoms
and with chronic depressive symptoms regarding
mean number of medications at baseline (mean
(SD) =4.8 (0.2) vs. 6.8 (0.9) medications, p<
0.05). We found the highest mean number of
physician contacts and medications at baseline
in individuals with chronic depressive symptoms
(mean (SD) =6.8 (1.1) contacts, and 6.8
(0.9) medications). At follow-up, individuals with
incident depressive symptoms showed the highest
mean number of physician contacts and of
medications (mean (SD) =4.9 (0.6) contacts,
and 7.3 (1.3) medications). For incident depressed
individuals, an increase in use of outpatient non-
physician services, medical supply, and home care
was found. For individuals with chronic depressive
symptoms, use of inpatient care, medical supply,
and home care increased, and use of outpatient
non-physician services and use of specialists
decreased. For individuals with remittent depressive
symptoms, the use of specialists, medical supply,
and antidepressants decreased from baseline to
follow-up. Mental health specialists were used
most often by individuals with chronic depressive
symptoms at baseline, followed by incident,
remittent, and non-depressed individuals (23%,
14%, 13%, 9%), and by individuals with remittent
depressive symptoms at follow-up followed by
incident, chronic, and non-depressed individuals
(29%, 25%, 0%, 8%).
Direct costs
DEPRESSIVE AND NON-DEPRESSIVE
SYMPTOMS
Table 2 shows the mean annual direct costs for
different cost components and total mean annual
direct costs as well as median values and ranges
for depressed and non-depressed participants at
baseline and follow-up. The mean annual direct
costs per patient were €5,241 for depressed and
€3,648 for non-depressed individuals at baseline,
and €6,491 for depressed and €4,537 for non-
depressed at follow-up. Thus, the mean annual
direct costs of depressed individuals exceeded
the costs of those without depressive symptoms
by €1,593 and €1,954, or 30.4% and 30.1%,
respectively. Mean costs of depressed individuals
were higher in each of the cost components except
for outpatient costs at baseline, and for inpatient
care costs, medical supply, and assisted living
at follow-up. Statistically significant differences
between depressed and non-depressed individuals
were found for pharmaceutical costs, costs for
medical supply, and costs for home care at baseline,
and for costs for medical supply as well as for home
care at follow-up. The most relevant cost drivers
of the direct costs of the depressed elderly were
inpatient costs at baseline (40% of total annual
direct costs), followed by costs of pharmaceuticals
(27%) and home care (15%). At follow-up, the most
relevant cost drivers for depressed individuals were
costs for home care (50%), for pharmaceuticals
(18%), and inpatient care (17%).
DIFFERENT COURSES OF DEPRESSIVE
SYMPTOMS
Mean annual total costs at baseline and follow-up
were highest for individuals with chronic depressive
symptoms (see Table 3). Mean annual direct costs
increased from baseline to follow-up for individuals
with no depressive symptoms by 16%. For chronic,
remittent, and incident depressed participants, costs
increased by 39%, 17%, and 27%, respectively.
We found no significant differences in mean
annual direct costs between these four groups at
baseline and follow-up.

6M. Luppa et al.
Table 2. Mean annual direct costs of depressed and non-depressed individuals at baseline (n=451) and follow-up (n=305)
BASELINE FOLLOW-UP
COST COMPONENTS
DEPRESSED:
N
=63
NON-
DEPRESSED:
N=388
MANN-
WHITNEY
U
-TEST
P
-VALUE
DEPRESSED:
N
=29
NON-
DEPRESSED:
N
=276
MANN-
WHITNEY
U
-TEST
P
-VALUE
.......................................................................................................................................................................................................................................................................................................................................................................................................................................................
Outpatient physician care: mean (SD) 385 (388) 573 (2,085) −1.388 0.165 430 (378) 405 (435) −0.526 0.599
Median (range) 230 (0–1,884) 301 (0–40,311) 328 (17–1,270) 270 (0–3,386)
Outpatient non-physician care: mean (SD) 186 (354) 167 (385) −1.075 0.282 188 (363) 168 (332) −0.113 0.910
Median (range) 0 (0–2,062) 0 (0–3,195) 0 (0–1,512) 0 (0–2,142)
Inpatient care: mean (SD) 2,087 (5,714) 1,291 (3,692) −0.107 0.915 1,115 (3,244) 1,388 (4,256) −0.124 0.901
Median (range) 0 (0–23,259) 0 (0–25,261) 0 (0–13,968) 0 (0–48,522)
Pharmaceuticals: mean (SD) 1,410 (1,359) 1,017 (2,048) −3.705 0.000 1,192 (1,026) 959 (897) −0.947 0.343
Median (range) 923 (0–7,178) 648 (0–26,421) 774 (0–3,667) 719 (0–8,405)
Medical supply/dentures: mean (SD) 225 (696) 216 (969) −2.205 0.027 113 (182) 402 (3,598) −2.319 0.020
Median (range) 0 (0–4,492) 0 (0–8,485) 0 (0–702) 0 (0–56,479)
Home care: mean (SD) 790 (2,042) 292 (1,232) −3.279 0.001 3,214 (6,575) 1,080 (4,216) −2.692 0.007
Median (range) 0 (0–13,008) 0 (0–14,053) 133 (0–27,360) 0 (0–32,476)
Assisted living: mean (SD) 48 (189) 32 (175) −1.417 0.156 34 (184) 59 (224) −0.792 0.428
Median (range) 0 (0–993) 0 (0–1,534) 0 (0–993) 0 (0–1,534)
Transportation: mean (SD) 110 (273) 60 (80) −1.132 0.258 205 (435) 76 (144) −0.871 0.384
Median (range) 12 (0–1,416) 6 (0–2,275) 42 (0–2,169) 28 (0–1,344)
Total direct costs: mean (SD) 5,241 (6,670) 3,648 (5,854) −2.806 0.005 6,491 (7,761) 4,537 (7,606) −1.857 0.063
Median (range) 2,412 (281–25,872) 1,625 (0–65,302) 3,404 (0–29,251) 1,981 (124–70,283)

Direct costs and depressive symptoms 7
Table 3. Mean annual direct costs of individuals with no,chronic,remittent,and incident depressive symptoms at baseline and follow-up (n=305)
BASELINE FOLLOW-UP
COST COMPONENTS
NO
DEPRESSIVE
SYMPTOMS
(
N
=255)
CHRONIC
DEPRESSIVE
SYMPTOMS
(
N
=13)
REMITTENT
DEPRESSIVE
SYMPTOMS
(
N
=21)
INCIDENT
DEPRESSIVE
SYMPTOMS
(
N
=16)
KRUSKAL-
WALLIS
TEST
P
-VALUE
NO DEPRESSIVE
SYMPTOMS
(
N
=255)
CHRONIC
DEPRESSIVE
SYMPTOMS
(
N
=13)
REMITTENT
DEPRESSIVE
SYMPTOMS
(
N
=21)
INCIDENT
DEPRESSIVE
SYMPTOMS
(
N
=16)
KRUSKAL-
WALLIS
TEST
P
-VALUE
......................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
Outpatient physician care:
mean (SD)
465 (32) 605 (123) 459 (101) 537 (148) χ2=2.381 0.497 418 (28) 310 (91) 250 (58) 527 (100) χ2=8.760 0.033
Median (range) 301 (0–4,718) 535 (50–1,484) 270 (66–1,884) 306 (0–2,331) 283 (17–3,386) 202 (34–1,211) 149 (0–972) 477 (17–1,270)
Inpatient care: mean (SD) 1,489 (248) 1,131 (1,131) 3,424 (1,655) 0 (0) χ2=4.728 0.193 1,248 (202) 2,092 (1,259) 3,092 (2,324) 322 (322) χ2=4.319 0.229
Median (range) 0 (0–24,933) 0 (0–14,704) 0 (0–23,259) 0 (0–0) 0 (0–21,555) 0 (0–13,968) 0 (0–48,522) 0 (0–5,146)
Pharmaceuticals: mean (SD) 938 (112) 1,633 (333) 982 (133) 909 (219) χ2=9.329 0.025 964 (57) 1,179 (236) 905 (157) 1,202 (295) χ2=1.417 0.701
Median (range) 670 (0–26,421) 972 (363–3,759) 904 (206–2,620) 615 (0–3,324) 719 (0–8,405) 1,045 (292–2,915) 639 (80–2,434) 767 (0–3,667)
Home care: mean (SD) 257 (76) 1,538 (1,008) 119 (92) 557 (318) χ2=8.552 0.036 1,018 (259) 5,460 (2,543) 1,831 (1,122) 1,389 (588) χ2=8.273 0.041
Median (range) 0 (0–14,053) 0 (0–13,008) 0 (0–1,920) 0 (0–4,920) 0 (0–32,476) 1,109 (0–27,360) 0 (0–23,264) 67 (0–6,059)
Total direct costs: mean (SD) 3,656 (331) 5,912 (1,802) 5,431 (1,782) 2,843 (621) χ2=4.586 0.187 4,375 (454) 9,707 (2,881) 6,508 (2,439) 3,878 (785) χ2=4.816 0.186
Median (range) 1,625 (0–38,653) 3,202 (513–21,507) 1,911 (520–25,755) 2,170 (555–8,950) 1,906 (124–70,283) 4,393 (446–29,251) 3,011 (379–49,544) 2,296 (368–9,723)
Table 3 shows also costs for different cost
categories according to course of depressive
symptomatology. Costs for outpatient care de-
creased in all subgroups, and were strongest in
individuals with chronic depressive symptoms.
Costs for pharmaceuticals increased in individuals
with incident and no depressive symptoms, and
decreased for individuals with chronic and remittent
depressive symptoms. Costs for inpatient care
increased only for individuals with chronic and
incident depressive symptoms. Costs for home
care increased in all subgroups, but were strongest
in individuals with chronic depressive symptoms.
Statistical significant differences were found
between individuals with remittent and incident
depressive symptoms for outpatient costs at follow-
up (Z =–2.193, p<0.05), between individuals
with chronic and no depressive symptoms (Z =
–2.658, p<0.01), and with chronic and incident
depressive symptoms (Z =–2.017, p<0.045)
for costs for pharmaceuticals at baseline, and for
costs for home care at baseline between individuals
with chronic and no depressive symptoms (Z =
–2.135, p<0.05) as well as incident depressive
symptoms (Z =–2.069, p<0.05) and at follow-up
between individuals with chronic and no depressive
symptoms (Z =–2.426, p<0.05).
Multivariate analyses
The multivariate linear regression model (see
Table 4) investigated the impact of depressive
symptoms, number of medications, age, and gender
at baseline on mean annual direct costs after 4.5
years, and revealed three variables predicting future
direct costs: number of medications, age, and
gender. With each additional medication, the mean
annual costs increased by €275. With every year
of age, the direct costs increased by €245. Women
caused €2,336 higher mean annual direct costs than
men. Depressive symptoms, defined with a score
of 6 or more on the GDS, showed no significant
impact on future directs costs. But if the cut-off
score was lowered to 5, 4, or 3, depressive symptoms
at baseline showed a significant association with
mean annual direct costs in the future (p<0.05).
The explained variance of mean annual direct costs
in the linear regression model was with 7% quite
low indicating that other factors than these included
contribute to the occurrence of direct costs.
Discussion
The aim of the study was to investigate health
service use and direct costs of a German
primary care sample over a 4.5-year study course

8M. Luppa et al.
Table 4. Linear regression of annual mean direct costs (n=305)
COEFFICIENT (95% CIa)STANDARD ERRORS P-VALUE
........................................................................................................................................................................................................................................................................
Age (years) 244.72 (67.64–446.85) 100.44 0.023
Gender 2,335.56 (1,137.96–3,652.14) 685.44 0.006
Number of medications 274.95 (19.00–513.91) 121.24 0.031
Depressive symptoms 2,845.66 (–235.34 to 6,472.36) 1,876.65 0.126
Constant –20,683.57 (–38.322.17 to –5,753.82) 8,253.75 0.022
Rb0.069
N305
aBias-corrected confidence intervals.
bBootstrap standard errors with 3,000 replications.
considering also the impact of different courses of
depressive symptoms on direct costs.
We found that mean annual direct costs
of individuals with depressive symptomatology
exceeded the total direct costs of non-depressed
individuals by almost one-third at baseline and
follow-up. This is in line with the few studies on
direct costs of depression in old age conducted in
the USA (Livingston et al., 1997; Unutzer et al.,
1997; Katon et al., 2003), and with the findings
of a recently published review of health service use
and costs of depressive symptoms in old age (Luppa
et al., 2012b).
A novel finding is that presence of depressive
symptoms at baseline shows a significant association
with the mean annual costs after 4.5 years, and
that this effect sustained after controlling for
comorbidity as measured by number of medications
at baseline (if a cut-off score between 3 and
5 for depressive symptomatology was chosen).
This finding suggests an independent association
of presence of depressive symptoms with future
direct costs that is not only explained by a higher
degree of comorbidity, but also that it apparently
needs only few depressive symptoms to trigger
an increase of direct costs. It has been discussed
that the amplification of present symptoms of
diseases and the increased presentation of non-
specific medical complaints such as headache,
gastrointestinal symptoms, and dizziness, which
leads to an increase in medical examinations and
prescriptions cause the higher direct costs (Waxman
et al., 1985; Luber et al., 2001; Katon et al.,
2003). Furthermore, it has been suggested that the
reason why individuals with depressive symptoms
amplify somatic complaints and use more health
services than their non-depressed counterparts is
not due to the desire of emotional support through
heavy contacts to healthcare professionals, but
rather due to a distorted perception of severity
of somatic symptoms and diseases: depressed
individuals seem to appraise themselves worse
and symptoms of diseases more severe than non-
depressed individuals (Kirmayer, 2001; Levinson
and Druss, 2005). The new aspect that depressive
symptoms are also associated with direct costs in
distant future may be explained by the common
unfavorable course of depressive symptoms in old
age: intermittency and chronicity rates in old
age were reported with almost 50–60% (Cole
et al., 1999; Harris et al., 2006). Furthermore,
higher baseline depression level is considered as
prognostic indicator for unfavorable course of
depressive symptoms (Licht-Strunk et al., 2007).
Hence, individuals with higher depression score
at baseline are at high risk of still suffering from
depressive symptoms 4.5 years later. In our primary
care-based sample, we found a somewhat lower
rate of chronicity of 38% over the study course
of 4.5 years than population-based representative
studies in old age (Cole et al., 1999; Harris
et al., 2006). Conceivable explanations may be:
the higher age of our study population (75+)
connected with a higher mortality rate of especially
depressed individuals (Blazer, 2003); the exclusion
of elderly individuals at higher risk of depressive
symptoms, e.g. nursing home residents, patients
with a dementia disorder; or the higher refusal
rate at follow-up of patients with more depressive
symptoms at baseline. Since patients with chronic
depressive symptoms showed the highest direct
costs over the study course, our findings of direct
costs may be rather underestimated (see also,
strength and limitations).
However, the association of depressive symp-
toms with future direct costs may also be affected
by a higher comorbidity at follow-up in individuals
with depressive symptoms. We found a considerable
difference of costs for home care between depressed
and non-depressed individuals at both time points,
and a substantial increase of costs for home care
for depressed individuals from baseline to follow-
up, which may be due to a substantial increase in
use of benefits from the nursing care insurance by
depressed individuals: at baseline 13% of depressed
individuals received benefits from nursing care

Direct costs and depressive symptoms 9
insurance (non-depressed: 4%); at follow-up, the
percentage increased to 34% (non-depressed: 9%).
It is a fact that the need of care increases
substantially in the age group of 80+(Federal
Statistical Office, 2011) and is closely related to
increasing morbidity and functional impairment.
Since in our sample an increase in need of care
occurred mainly in depressed individuals, a close
relationship of depressive symptoms and morbidity
or functional impairment was demonstrated as
previously reported (Blazer, 2003; Fiske et al.,
2009). The higher mean number of medications of
depressed compared to non-depressed individuals
at baseline supports the postulated association of
depressive symptoms and morbidity. However, in
this study one could not say which one caused the
other. In the literature, on the one hand higher
somatic comorbidity and functional impairment
were reported as prognostic indicators for an
unfavorable course of depressive symptoms in old
age (Licht-Strunk et al., 2007). On the other hand,
depressive symptoms were also reported to have
an impact on development and deterioration of
somatic diseases and functional impairment (Fiske
et al., 2009). A view to direct costs associated
with the course of depressive symptoms showed the
highest costs for individuals with chronic depressive
symptoms, and an increase of costs of chronic
depressed individuals by 39% over the study course
compared to 16% in non-depressed individuals.
This increase was mainly due to an increase in
costs for home care, so that in our sample especially
chronic depressive symptoms seem to be highly
associated with need of care. This finding was
emphasized by the results that chronic depressed
individuals showed the highest mean number of
physician contacts and the highest mean number
of medications at baseline. They were followed
by individuals with incident depressive symptoms,
which is, besides individuals with chronic depressive
symptoms, a further important subgroup. For these
individuals, the highest mean number of physician
contact and medications was found at follow-up
and an increase in use of outpatient non-physician
services, medical supply, and home care from
baseline to follow-up, which could be attributed (1)
to the distorted perception of medical complaints
due to the depressive perspective, (2) to a new severe
illness that may have triggered the development of
depressive symptoms, or (3) to a combination of
both.
Mental healthcare was not often used by
individuals with depressive symptoms. Treatment
with antidepressants only occurred for 21% and
7% of depressed individuals at baseline and follow-
up compared to 5% of non-depressed individual
(whereby reasons of antidepressant prescription for
non-depressed individual may be due to physician’s
diagnosis of depression or due to anxiety/panic
or sleeping disorders (Gardarsdottir et al., 2007)).
Moreover, individuals with remittent depressive
symptoms showed a decrease in antidepressants
use from baseline to follow-up, which may be
due to a termination of the treatment by the
physician or by the patient him/herself due to
an improvement of symptomatology. Only 14%
of depressed individuals visited mental health
specialists. This indicated that more unrecognized
than recognized depressed cases were contained in
the sample, and this is emphasized by the findings
that depression in old age in primary care is often
under-diagnosed (Callahan, 2001). However, one
should keep in mind that clinical relevance and
need of treatment of minor depressive symptoms
remain unclear. Mean GDS score of depressed
individuals at baseline and follow-up was 7.5 and
7.6, respectively, thus exceeding the threshold of 6
only slightly. On the other hand the general increase
in health service use and costs (Luppa et al., 2012b),
the often unfavorable course (Cole et al., 1999), and
the number of well-investigated negative outcomes
of depressive symptoms in old age (Blazer, 2003)
may justify their treatment.
Furthermore, in our study increasing age was
significantly associated with higher direct costs.
This is in line with findings from another study
conducted in Germany before (Linden et al., 1997),
and may be explained by the close association of
advanced age with illness, frailty, and higher need
for care (Linden et al., 1997), and also by the
influence of death-related costs (Raitano, 2006).
Strengths and limitations
Our study used a bottom-up approach, collecting
data retrospectively by applying a questionnaire
on service utilization in face-to-face interviews
with the individuals. This approach allows a
comprehensive assessment of resource use data
of each participant, and individual data such as
out-of-pocket expenses or costs of social services
can be assessed, too. However, recall bias cannot
be ruled out completely. Therefore, we aimed at
minimizing the bias applying different time periods
for cost components following recommendations by
Johnston et al. (1999). Furthermore, cost estimates
reported should be seen as rather conservative
estimates. Via exclusion criteria, nursing home
residents and patients with dementia disorders,
known to have higher depression rates and
supposedly higher healthcare costs, were not
included in this sample. Also, non-participants at
follow-up showed a higher mean depression score
and a higher prevalence of depressive symptoms at

10 M. Luppa et al.
baseline than participants. Moreover, sub-samples
of depressed and non-depressed participants as
well as of different courses of depressive symp-
tomatology were rather small limiting statistical
power, and a comparison of results in different
cost categories between sub-samples should be done
with caution. Lastly, we only have information
about depressive symptomatology at baseline and
4.5 years later, excluding information about
fluctuation of depressive symptomatology during
the time between both time points. Nonetheless, our
results convey a first impression of impact of course
of depressive symptomatology on direct costs, and
call for further in-depth investigations.
Conclusions
Depressive symptoms in old age are common,
highly persistent, and lead to enormous personal,
social, and economic consequences for individuals,
families, and societies. The manifold research
insights in late-life depression achieved in the
past decades urgently demand for an optimization
of its early detection and adequate treatment.
Our findings emphasize this requirement as they
illustrate the economic consequences of depressive
symptoms for the healthcare system over the
course of time, which appeared in a substantial
increase in health service use and direct costs with
presence and persistence of depressive symptoms.
Potential reasons for the increased service use
of depressed elderly individuals may be first an
increase of comorbidity with rising age, and second
the amplification of existing medical complaints
as well as the increased presentation of unspecific
medical symptoms. Strategies and interventions for
improvement of diagnostic validity and treatment
success of depressive symptoms in late life would
not only lead to preserve and improve quality of life
for people with dementia and their families, but may
also have a potential effect on economic burden for
societies (Wang et al., 2003; Barrett et al., 2005;
Donohue and Pincus, 2007).
Conflict of interest
None.
Description of authors’ roles
M. Luppa formulated the research questions,
designed and carried out the study, analyzed the
data, and wrote the first draft of the paper;
H.-H. König formulated the research questions,
designed the study, and commented on and revised
the paper; D. Heider and H. Leicht carried out
the study, analyzed the data, and commented
on and revised the paper; T. Motzek formulated
the research questions, designed the study, and
analyzed the data; G. Schomerus formulated
the research questions, designed the study, and
commented on and revised the paper; and S.G.
Riedel-Heller formulated the research questions,
designed and carried out the study, and commented
on and revised the paper.
Acknowledgments
This work is part of German Research Network
on Dementia (KND) and the German Research
Network on Degenerative Dementia (KNDD) and
was funded by the German Federal Ministry of
Education and Research (AgeCoDe study, grant:
01GI431 and 01GI0714) and is published in
affiliation with the Study on Late-Life Depression
in Primary Care (AgeMooDe study, grant:
01GY1155A).
References
Allgaier, A. K., Kramer, D., Mergl, R., Fejtkova, S. and
Hegerl, U. (2011). Validity of the geriatric depression
scale in nursing home residents: comparison of GDS-15,
GDS-8, and GDS-4. Psychiatrische Praxis, 38, 280–286.
Barrett, B., Byford, S. and Knapp, M. (2005). Evidence of
cost-effective treatments for depression: a systematic
review. Journal of Affective Disorders, 84, 1–13.
Blazer, D. G. (2003). Depression in late life: review and
commentary. Journals of Gerontology, Series A: Biological and
Medical Sciences, 58, 249–265.
Callahan, C. M. (2001). Quality improvement research on
late life depression in primary care. Medical Care, 39,
772–784.
Cole, M. G., Bellavance, F. and Mansour, A. (1999).
Prognosis of depression in elderly community and primary
care populations: a systematic review and meta-analysis.
American Journal of Psychiatry, 156, 1182–1189.
Donohue, J. M. and Pincus, H. A. (2007). Reducing the
societal burden of depression: a review of economic costs,
quality of care and effects of treatment. Pharmacoeconomics,
25, 7–24.
Efron, B. and Tibshirani, R. (1986). Bootstrap methods for
standard error, confidence interval, and other measures of
statistical accuracy. Statistical Science, 1, 54–77.
Farley,J.F.,Harley,C.R.andDevine,J.W.(2006). A
comparison of comorbidity measurements to predict
healthcare expenditures. American Journal of Managed Care,
12, 110–119.
Federal Statistical Office (2011). Pflegestatistik 2009: Pflege
im Rahmen der Pflegeversicherung – Deutschlandergebnisse.
Wiesbaden: Statistisches Bundesamt.
Fiske, A., Wetherell, J. L. and Gatz, M. (2009).
Depression in older adults. Annual Review of Clinical
Psychology, 5, 363–389.

Direct costs and depressive symptoms 11
Gardarsdottir, H., Heerdink, E. R., van Dijk, L. and
Egberts, A. C. G. (2007). Indications for antidepressant
drug prescribing in general practices in the Netherlands.
Journal of Affective Disorders, 98, 109–115.
Gauggel, S. and Birkner, B. (1999). Validity and reliability
of a German version of the Geriatric Depression Scale
(GDS). Zeitschrift fuer Klinische Psychologie-Forschung und
Praxis, 28, 18–27.
Harris, T., Cook, D. G., Victor, C., DeWilde, S. and
Beighton, C. (2006). Onset and persistence of depression
in older people: results from a 2-year community follow-up
study. Age Ageing, 35, 25–32.
Johnston, K., Buxton, M. J., Jones, D. R. and Fitzpatrick,
R. (1999). Assessing the costs of healthcare technologies in
clinical trials. Health Technology Assessment, 3, 1–76.
Katon, W. J., Lin, E., Russo, J. and Unutzer, J. (2003).
Increased medical costs of a population-based sample of
depressed elderly patients. Archives of General Psychiatry,
60, 897–903.
Kirmayer, L. J. (2001). Cultural variations in the clinical
presentation of depression and anxiety: implications for
diagnosis and treatment. Journal of Clinical Psychiatry,
62, 22–28.
König, H.-H., Luppa, M. and Riedel-Heller, S. (2010).
Die Kosten von Depression und die Wirtschaftlichkeit ihrer
Behandlung. Psychiatrische Praxis, 37, 213–215.
Levinson, C. M. and Druss, B. G. (2005). Health beliefs
and depression in a group of elderly high utilizers of
medical services. General Hospital Psychiatry, 27, 97–
99.
Licht-Strunk, E., van der Windt, D. A., van Marwijk,
H. W., de Haan, M. and Beekman, A. T. (2007). The
prognosis of depression in older patients in general practice
and the community: a systematic review. Family Practice,
24, 168–180.
Linden, M., Horgas, A. L., Gilberg, R. and
Steinhagen-Thiessen, E. (1997). Predicting health care
utilization in the very old. Journal of Aging and Health,9,
3–27.
Livingston, G., Manela, M. and Katona, C. (1997). Cost
of community care for older people. British Journal of
Psychiatry, 171, 56–59.
Luber, M. P. et al. (2001). Depression and service utilization
in elderly primary care patients. American Journal of
Geriatric Psychiatry, 9, 169–176.
Luppa, M., Heinrich, S., Angermeyer, M. C., König,
H.-H. and Riedel-Heller, S. G. (2007). Cost-of-illness
studies of depression: a systematic review. Journal of
Affective Disorders, 98, 29–43.
Luppa, M. et al. (2008). Direct costs associated with
depression in old age in Germany. Journal of Affective
Disorders, 105, 195–204.
Luppa, M. et al. (2012a). Age- and gender-specific
prevalence of depression in latest-life: systematic review and
meta-analysis. Journal of Affective Disorders, 136, 212–221.
Luppa, M., Sikorski, C., Motzek, T., Konnopka, A.,
König, H.-H. and Riedel-Heller, S. (2012b). Health
service utilization and costs of depressive symptoms in late
life: a systematic review. Current Pharmaceutical Design,
Epub Jun 6.
Raitano, M. (2006). The impact of death-related costs on
health-care expenditures: a survey. European Network of
Economic Policy Research Institutes Research Report No.
17. Available at: http://ssrn.com/abstract=2012337.
Sheikh,J.I.andYesavage,J.A.(1986). Geriatric
Depression Scale (GDS): recent evidence and development
of a shorter version. Clinical Gerontologist, 5, 165–173.
Unutzer, J. et al. (1997). Depressive symptoms and the cost
of health services in HMO patients aged 65 years and older:
a 4-year prospective study. JAMA, 277, 1618–1623.
Wang,P.S.,Simon,G.andKessler,R.C.(2003). The
economic burden of depression and the cost-effectiveness
of treatment. International Journal of Methods in Psychiatric
Research, 12, 22–33.
Waxman,H.M.,McCreary,G.,Weinrit,R.M.and
Carner,E.A.(1985). A comparison of somatic
complaints among depressed and non-depressed older
persons. Gerontologist, 25, 501–507.