The European Journal of Health Economics

, Volume 12, Issue 6, pp 533–539 | Cite as

Direct costs of Alzheimer’s disease in Germany

  • Peter Kiencke
  • Dietmar Daniel
  • Christine Grimm
  • Reinhard Rychlik
Original Paper



The aim of this analysis was to determine the direct costs for patients with Alzheimer’s disease (AD) based on data supplied by a large German statutory health insurance (BARMER). Focus of the present evaluation is the exposure of total direct costs, survival rates and the types of care distribution of patients with AD.


The analysis was based on anonymised data of patients with Alzheimer’s disease who were insured by a large German statutory health insurance (Barmer Ersatzkasse-BEK) in 2005 over 12 months (n = 35.684). The study population was classified into three treatment groups: patients who received memantine and no other antidementia drugs, psychotropic drugs or sedatives/hypnotics (memantine group); patients who neither received memantine nor other antidementia drugs but psychotropic drugs and/or sedatives/hypnotics (PHS group); and patients who received no antidementia drugs and no psychotropic drugs or sedatives/hypnotics (no AT group). A Markov model was designed to assess the costs of each treatment group depending on the type of care over a 5-year period.


The results obtained for the PHS group were the following: after 5 years, 25.1% were in inpatient health care, 1.1% in partial inpatient care, 5.4% in outpatient care, and 64.5% of patients died. Only 3.9% did not require care. On average, direct costs to the amount of 7.948 € incurred per patient and year and 13.099 € per surviving patient. In the group without pharmacotherapy, 20.6% were in inpatient care, 0.3% in partial inpatient care, 12.8% in outpatient care, and 52.5% of the patients died. For 13.8% of the patients, care was not necessary. The annual costs amounted to 6.760 € per patient and 9.926 € per surviving patient. The results obtained for the memantine group were more beneficial: annual costs per patient amounted to 6.100 € and to 8.376 € per surviving patient.


The results demonstrate that non-antidementive therapy for Alzheimer’s disease causes higher costs especially for care. The memantine group proved to be superior compared to PHS group and no AT group, despite higher costs in the specific drug category.


Costs Level of care Alzheimer’s disease Dementia Memantine 


Alzheimer’s disease (AD) is a progressive, neurodegenerative disease, which is associated with the loss of cognitive and intellectual function like memory, impaired judgement and abstract thinking, as well as loss in general everyday competence. Everyday competence includes cognitive, communicative, behavioural and practical skills. AD is the most common form of neurodegenerative dementia. It affects persons in the later period of life and is the most common type of dementia. The prevalence of AD is approximately 1% among those aged between 65 and 69, whereas up to 30% of persons aged 90 years suffer from dementia. The incidence of AD is estimated to be 120.000 per year [1]. In consideration of the demographic development, the number of persons who will be affected by 2050 is estimated at 2 million [2].

Especially for family member carers, dementia-associated conduct disorders, such as distrust, aggression, agitation, delusion, depression, cause the main problems and render the care in home settings almost impossible. This explains the four-time higher prevalence rate of dementia among nursing home residents [3].

Currently, there is no cure for Alzheimer’s disease, but drug and non-drug treatments may help with both cognitive and behavioural symptoms. There are three category groups for treatment: drug therapy, psychological and ecological interventions, as well as social interventions. For treatment of moderate to severe AD, memantine is the only antidementia drug available and recommended [4]. Antidepressants and antipsychotics are also used for non-cognitive impairments and conduct disorders in patients with Alzheimer’s disease [5].

The progression of Alzheimer’s disease is not only associated with an increase in physical and psychic symptoms but also with social and economical implications. The increasing number of demented patients represents a great challenge for the German health care system and compulsory long-term care insurance, in terms of quality of medical care, as well as financing.

Cost of dementia was calculated to the amount of 5.6 billion Euro in the year 2002 [6]. The highest cost is attributed to the moderate to severe stage of AD [2, 7]. At present, a comprehensive cost-of-illness calculation for Germany is not available.

The aim of this analysis was to determine the direct costs for patients with Alzheimer’s disease based on data supplied by a large German statutory health insurance (BARMER). Focus of the present evaluation is the evaluation of total direct costs and survival rates and the type of care distribution of patients with AD under different therapies. Three treatment groups were analysed: patients who received memantine (memantine group), treatment with psychotropic drugs and/or sedatives/hypnotics (PHS group), and patients with no antidementia drugs (no AT group).

Materials and methods

Data and patient selection

This evaluation is based on a retrospective analysis of routine data from 2005 supplied by a German statutory health and nursing care insurance (BARMER). The anonymised data included all persons with dementia insured from 2005, as well as all services provided for these patients in 2005.

Data of 48.122 patients with the diagnosis dementia were available. Since this analysis focused on Alzheimer’s disease, only 36.668 patients with the diagnosis Alzheimer’s disease coded with ICD-10 ‘F00’ or ‘G30’ by a medical specialist (neurologist, psychiatrist, internist, neurosurgeon) were selected. Patients who are insured by family members were excluded (n = 945) because the quality of these claims data were not on the same level of accuracy as for primary insured members. Thirty-nine patients were excluded because they left the BEK before the end of 2005 without the reason death. Accordingly, 35.684 patients were selected.

The analysis population was classified into three treatment groups:

Patients with at least one memantine prescription in 2005 who received no other antidementia drugs, psychotropics or sedatives/hypnotics (memantine group); patients who received neither memantine nor any other antidementia drugs but psychotropics or sedatives/hypnotics (PHS group) and patients with no antidementia drugs or symptomatic therapy (no AT group). The selection on the basis of the drug prescriptions add up to 1.448 patients in the memantine group, 12.561 patients in the PHS group and 7.503 patients in the no AT group. A total of 21.512 patients were included in the evaluation, 14.172 patients did not meet the inclusion criteria and were not considered in the analysis.


A matched-pairs-analysis was conducted to reveal a better comparability of the three groups. The following matching criteria were selected:
  • Age

  • Gender

  • Type of care (01.01.2005)

  • At least one prescription every quarter in 2005, up to the last quarter or before death, in the memantine and PHS groups.

The last criterion implicates that only patients, for whom memantine was continually prescribed, were included in the analysis. Therefore, 527 patients for each group could be analysed in the matched-pairs-analysis. The inclusion of the type of care was more relevant than the level of care because it allows the differentiation between ambulatory (outpatient) and inpatient care which has a great influence on cost. The German nursing system differentiate between three types of care: outpatient care, i.e. the patient received nursing care at home; partial inpatient care, i.e. the patient received nursing care half a day (by day or by night) at home and the rest of the time in a nursing home; inpatient care, i.e. the patient will be cared for in a nursing home.

The consideration of clinical parameters e.g. duration or severity of Alzheimer’s disease or social setting of the patient was not possible with the present data. Diabetes mellitus, cardiovascular disease and depression/psychoses were defined as relevant concomitant diseases. Diabetes mellitus and cardiovascular disease occurred with a similar frequency. In contrast, psychoses and depressive disorders were three times higher in the PHS group as in the other treatment groups, which of course correlate with the chosen pharmacotherapy. After matching, all services, such as non-inpatient diagnostics and therapy, pharmacotherapy, hospitalisation, costs of care, curatives and aids, as well as the development of the type of care were analysed.

Markov model

Based on these data, a Markov model was applied which represents the type of care (Markov states: no care, outpatient care, partial inpatient care, inpatient care, death) and the direct costs in the three groups.

The model was calculated with TreeAge Pro 2008 (TreeAge Software Inc., Williamstown MA). The assumptions made were as follows:
  • All three groups start with the same distribution of types of care.

  • In every cycle, patients can only remain in their state or change to a higher type of care. Death is an adsorbing state.

  • A cycle represents a quarter, the model was calculated for 20 cycles (5 years).

  • Costs depend on the type of care and treatment group, but not on time. Therefore, discounting was disregarded.

  • A benefit or utility was not defined. Only costs were calculated. Since death was the only state which did not generate costs and mortality rates varied in the three groups, a group with a higher mortality rate would cause less costs. Therefore, the costs per surviving patient were calculated additionally. Insofar, the survival rate is a clinical outcome parameter.

Transition probabilities per quarter for each group are presented in Table 1.
Table 1

Transition probabilities per quarter




No care


Partial inpatient



Memantine group

 No care











 Partial inpatient









PHS group

 No care











 Partial inpatient









No AT group

 No care











 Partial inpatient










Demographics and need of care

After matching, there were 62.0% women in each group. The average age was 79.85 years and the median 80 years. Three times more patients suffered from psychoses or depressions in the PHS group compared to the other groups (Table 2). At the beginning of 2005, 58.1% of the patients had no type of care, whereas 29.0% of patients received outpatient care, 0.9% partial inpatient care and 12.0% inpatient care.
Table 2

Relevant concomitant diseases


Memantine (%)

PHS (%)

No AT (%)

Diabetes mellitus




Cardiovascular disease








For 2005, the probabilities to remain in the same type of care or to switch to a higher (more severe) type were calculated (Table 3). Patients from the memantine group showed the highest rate and the lowest rate of in inpatient care. Poor results were obtained in the PHS group. The mortality rates were lower in the memantine (10.2%) group than in the no AT group (12.3%) and in the PHS group (15.4%).
Table 3

Distribution of types of care 31.12.2005





No care








Partial inpatient













The lowest averaged annual costs were generated by the memantine group with 6.962 € per patient. The highest costs occurred in the PHS group with 10.725 €, particularly caused by costs of care and hospitalisation. In the group without pharmacotherapy, the average costs added up to 7.839 € (Table 4). In the memantine group, drug costs contributed to the second highest costs of total costs; in the other two groups, hospitalisation costs were the second highest. The percentage of physician treatment costs was similar in all groups (6.7% no AT group and 6.4% in the PHS group). Remedies had the minor proportion in all groups.
Table 4

Average total direct costs (after matching)





Treatment costs

457.27 €

688.95 €

525.04 €

Drug acquisition (total)

1,765.99 €

1,445.04 €

772.28 €


1,028.86 €

464.17 €

0.00 €


737.13 €

980.87 €

772.28 €


990.57 €

3,087.73 €

2,179.12 €

Nursing expenses

3,138.42 €

4,716.22 €

3,671.91 €


242.23 €

206.29 €

172.23 €


367.74 €

580.73 €

518.35 €

Total direct costs

6,962.22 €

10,724.96 €

7,838.92 €

The above described calculation consisted of costs per patient incurred in 2005. A comparison of the three groups in terms of type of care was only given at the beginning of 2005; thereafter, changes in the type of care affects the costs for care, which concludes that the averaged costs for care differ. The Markov model was conducted with transition probabilities for the parameter type of care. Therefore, only costs for the same type of care were of interest (Table 5). The different costs of care in the outpatient sector were determined by different levels of care which were no matching criteria. Patients in the PHS group had the highest average cost in all types except for inpatient care. The results for partial inpatient care base on a small sample size. In the start distribution, only 0.9% of patients received this type of care. Therefore, an average of 12.281 € was used for all groups in the Markov model.
Table 5

Averaged total costs per type of care





No care

3,061 €

4,798 €

2,635 €


9,145 €

12,408 €

10,875 €

Partial inpatient

10,331 €

20,386 €

8,774 €


18,350 €

17,887 €

19,013 €

Markov Model

Basis for the model calculation was the results of the matched-pair analysis based on a unique distribution of the type of care, the calculated transition probabilities and the costs per type of care in the three treatment groups. The distribution of each type, including death was simulated, and costs were calculated in a period of 20 cycles (quarters).

After 5 years, more than half of the patients died (Table 6). The worst results by far were obtained in the PHS group: after 5 years, 64.5% of patients passed away, 25.1% received inpatient care, 1.1% partial inpatient care, and 5.4% outpatient care. Only 3.9% received no care. For the cost calculation, the proportion of surviving patients in the cycles was relevant, because the deceased did not generate any costs and therefore a higher mortality rate would reduce total costs. On average, costs to the amount of 7,954 € incurred per patient and year and 13,110 € per surviving patient in the PHS group.
Table 6

Final distribution of care types (after 20 cycles)


Memantine (%)

PHS (%)

No AT (%)

No care








Partial inpatient












In the group without pharmacotherapy, 52.5% of the patients died during the observation period, 20.6% received inpatient care, 0.3% partial inpatient care and 12.8% outpatient care. For 13.8% of the patients, care was not necessary. The annual costs amounted to 6.763 € per patient and 9.929 € per surviving patient.

The best results were obtained by patients treated with memantine. After 5 years, 47.0% of the patients in the memantine group passed away, 11.9% received inpatient nursing care, 4.4% partial inpatient care, and 19.8% outpatient care. In the memantine group, 16.9% did not receive any care services (Fig. 1). In the memantine group, averaged costs incurred to the amount of 6,117 € per patient and 8,400 € per surviving patient (Fig. 2). Figure 3 shows the costs per surviving patient and quarter. The costs increased continuously because switching back to a lower state of care was not possible due to the assumptions made.
Fig. 1

Final distribution of care types after 20 cycles

Fig. 2

Group comparison of annual costs

Fig. 3

Development of costs per survivor

The Markov model showed that a pharmacotherapy with memantine is cost-effective despite higher drug costs. Patients in the memantine group had higher survival rates; the surviving patients predominantly received outpatient care or no care services, less costs incurred than in the PHS group. The results of the cost calculation after 5 years are presented in Table 7. Despite the highest mortality rate, patients in the PHS group generated the highest total costs amounting to 39,770 €. This resulted from the rapid increase in the proportion of patients with cost-intensive inpatient care.
Table 7

Results of the Markov model (after 20 cycles)





Total costs (5 years)

30,584 €

39,770 €

33,813 €

Annual average

6,117 €

7,954 €

6,763 €

Per survivor/year

8,400 €

13,110 €

9,929 €

Sensitivity analysis

Table 8 contains the results of a Monte Carlo simulation based on 10,000 patients within the Markov model. This simulation confirmed the result of the Markov model. Significant differences in total costs between memantine group and the other groups could be shown.
Table 8

Results of the Monte Carlo simulation (after matching) (after 20 cycles)





Number of patients





29,946 €

39,908 €

33,948 €

Standard deviation

22,180 €

26,793 €

28,157 €


0 €

0 €

0 €

25% quartile

15,307 €

17,887 €

13,175 €


22,937 €

35,773 €

23,767 €

75% quartile

45,727 €

62,040 €

54,375 €


91,748 €

89,433 €

95,066 €

p value (Memantin vs.)




Discounting of the incurred costs was conducted by verifying the rate between annual 0 and 8%. The sensitivity analysis shows no influence of discounting on the differences of results between the three groups. With an increasing discounting rate, the total costs decreased in all groups equally.


The present model based on insurance data of a German statutory health care insurance in 2005. By means of these data, a cost calculation, as well as the distribution of care types at the beginning of 2005 and their changes during 2005, for patients with Alzheimer’s disease, was conducted. Three treatment strategies were classified on the basis of drug prescriptions. Neither the evaluation of the efficacy of the medication, duration nor severity of the disease was possible with the present routine data. These parameters can only be evaluated within clinical or non-interventional trials.

The time period of 1 year was sufficient for the cost calculation. For the presentation of the treatment strategies, the development of type of care and survival rates after initial diagnosis, a longer time period would be favourable. Thus, the classification of patients to the three treatment strategies was based on data from 2005; former treatment options were unknown and could not be considered. Due to the sample size, the chosen method was appropriate. In the cost calculation, all services paid by the BEK were included. Because of the elderly and multi-morbid population, services might be included which cannot be assigned to the Alzheimer’s disease, because a causal assignment on the basis of the data was mostly impossible. This problem existed in all three treatment groups. For this reason, this model differs from the internationally accepted AHEAD Model [8]. The AHEAD Model is validated for mild to moderate Alzheimer’s disease; it uses a variety of clinical parameters for the assessment of the severity of the disease and predicted transition probabilities and costs. The model in this analysis utilised the calculated transition probabilities between the care types and costs based on the BARMER.

Clinical parameters could not be derived from the routine data. Therefore, the type of care as an indicator for the severity of Alzheimer’s was chosen because of missing clinical data. It directly indicates the financial burden on the economy. Furthermore, dementia-associated symptoms, such as cognitive and emotional disorders or behavioural abnormalities, have only little impact on the classification for the level of care (step I, II or III) in Germany. Services to support everyday activities are currently the strongest predictors. This was shown in an evaluation of Grässel et al. [9]. In 390 patients with confirmed mild to moderate Alzheimer’s diagnosis, the Barthel-Index, which presents daily living skills of the patients, showed to be a highly significant predictor for the existence of a level of care, whereas the severity of the cognitive impairment measured with the Mini-Mental-State-Test had no influence.

Further, it has to be considered that the acceptance or modification of a level of care is an administrative procedure. Between the incidence of the necessary requirements and the access to a level of care might be a long period of time.

In the present analysis, only direct costs were calculated. Indirect costs due to productivity loss might be low in the observed population; however, indirect costs are generated by family members in terms of reduced work or exemption from work. In this analysis, the assessment of indirect costs, such as unpaid care services of family carers, was not possible.

However, this cost analysis showed that cost of care had a great impact on total costs. Also, Wimo et al. showed in their resource utilisation and cost analysis that 75% of total costs were generated by the treatment of patients with severe Alzheimer’s disease because of referral to a nursing home and the associated costs [10]. Results from another cost analysis by Wimo and co-worker showed that the application of memantine instead of placebo in patients with moderate to severe Alzheimer’s disease was cost saving [11]. The cost savings were due to fewer referrals to nursing homes (p = 0.04 versus placebo) and less expenditures on time of carers (~52 h, p = 0.02 versus placebo).

The memantine-treated group showed significant lower costs for patients with Alzheimer’s disease compared to treatments with psychotropics or compared to therapeutic nihilism (memantine versus PHS p < 0.001 and memantine versus no AT group p = 0.039). In Germany, memantine is the only antidementive licensed for patients with moderate to severe Alzheimer’s disease. The extent to which costs differ between these patients correlates with the severity of the disease. A prospective health economic study could indicate that the number of stationary treatments, costs of care and costs for aiding devices are less for memantine-treated patients in comparison with patients treated with other or no psychopharmaca. Also, the progress to increased care is delayed and mortality rates reduced for patients treated with memantine.



The study was financially supported by Merz Pharmaceuticals, Germany.


  1. 1.
    Bickel, H.: Dementia syndrome and Alzheimer disease: an assessment of morbidity and annual incidence in Germany. Gesundheitswesen 62(4), 211–218 (2000)PubMedCrossRefGoogle Scholar
  2. 2.
    Bickel, H.: Dementia in advanced age: estimating incidence and health care costs. Z. Gerontol. Geriatr. 34(2), 108–115 (2001)PubMedCrossRefGoogle Scholar
  3. 3.
    Jakob, A., Busse, A., Riedel-Heller, S.G., Pavlicek, M., Angermeyer, M.C.: Prevalence and incidence of dementia among nursing home residents and residents in homes for the aged in comparison to private homes. Z. Gerontol. Geriatr. 35(5), 474–481 (2002)PubMedCrossRefGoogle Scholar
  4. 4.
    Vollmar, C., Koch, M., Löscher, S., Butzlaff, M.: Demenz-Evidenzbasierte Leitlinie zur Diagnose und Therapie 2005 [cited 2008 Aug 26]Google Scholar
  5. 5.
    Deutsche Gesellschaft für Psychiatrie, Psychotherapie und Nervenheilkunde. Behandlungsleitlinie Demenz. Kurzversion 2008 [cited 2008 Sep 9]Google Scholar
  6. 6.
    Statistisches Bundesamt. Gesundheit. Krankheitskosten 2002. Statistisches Bundesamt 2004, Wiesbaden. 2008Google Scholar
  7. 7.
    Specht-Leible, N., Bender, M., Oster, P.: Causes for in-hospital treatment of nursing home residents. Z. Gerontol. Geriatr. 36(4), 274–279 (2003)PubMedCrossRefGoogle Scholar
  8. 8.
    Caro, J.J., Getsios, D., Migliaccio-Walle, K., Raggio, G., Ward, A.: Assessment of health economics in Alzheimer’s disease (AHEAD) based on need for full-time care. Neurology 57(6), 964–971 (2001)PubMedGoogle Scholar
  9. 9.
    Grassel, E., Donath, C., Lauterberg, J., Haag, C., Neubauer, S.: Dementia patients and levels of care: do symptoms of the illness affect the grading? Gesundheitswesen 70(3), 129–136 (2008)PubMedCrossRefGoogle Scholar
  10. 10.
    Wimo, A., Winblad, B.: Health economic aspects of Alzheimer’s disease and its treatment. Psychogeriatrics (1), 189–193 (2001)Google Scholar
  11. 11.
    Wimo, A., Winblad, B., Stoffler, A., Wirth, Y., Mobius, H.J.: Resource utilisation and cost analysis of memantine in patients with moderate to severe Alzheimer’s disease. Pharmacoeconomics 21(5), 327–340 (2003)PubMedCrossRefGoogle Scholar

Copyright information

© Springer-Verlag 2010

Authors and Affiliations

  • Peter Kiencke
    • 1
  • Dietmar Daniel
    • 1
  • Christine Grimm
    • 1
  • Reinhard Rychlik
    • 1
  1. 1.Institute of Empirical Health EconomicsBurscheidGermany

Personalised recommendations