Abstract
Objective:
The main objective of this analysis was to assess the medical and economic differences between patients with and without diagnosed Alzheimer’s Disease (AD). Analysis included co-morbidities, patterns of drug use, and clinical course, as well as the magnitude of these differences attributable to AD.
Methods:
This evaluation is based on retrospective analyses of anonymized claims data from 2005–2008 provided by a large German Statutory Health Insurance (SHI). Cross-sectional analyses were performed with the following data: demographics, number of hospitalizations, number of in-patient days, number of ambulatory physician visits, number of drugs used, total number of defined daily doses (DDD) of prescribed drugs, and costs. A propensity score model was used to assess costs attributable to AD.
Results:
Patients with AD caused substantially higher costs from the perspective of a SHI. The differences in long-term care costs explained 70–75% of the total cost difference, with being responsible for about half of the total costs in AD patients. Comparing matched AD patients with controls resulted in 2.2–2.4-fold higher costs in the cross-sectional analysis, €7413–€9207 for AD patients vs €3378–€3850 for controls. The propensity score model resulted in a difference of €3771 attributable to AD. These costs are mainly caused by differences in hospitalization and long-term care costs.
Conclusion:
This analysis is one of the largest health economic studies of AD in Germany. The limitations of this study include the fact that reported diagnosis couldn’t be validated and disease severity was not taken into account. Despite these methodological constraints, it can be concluded that AD is a substantial cost driver from the SHI payer perspective in Germany.
Introduction
Dementia is the generic term used for different forms of cognitive degeneration. According to the International Classification of Diseases (ICD), dementia is a syndrome resulting from chronic or progressive disease of the brain with impairment of several high cortical functions including memory, orientation, thinking, perception, calculating, learning, language, and judgmentCitation1.
Alzheimer’s disease (AD), which is caused by neurodegenerative processes in the brain, is the most common type of dementia in the elderly, accounting for two-thirds of dementia casesCitation2. Vascular dementia, caused by circulatory disturbances, accounts for 15–20% of dementia cases, with the remaining cases being mixed forms of the degenerative-vascular type or other forms secondary to diseases like HIV, primary Parkinson’s syndrome, chronic intoxication, and othersCitation1,Citation2.
Currently ∼1.3 million people suffer from dementia of any type in GermanyCitation3. Prevalence of AD depends on age and rises from ∼1% in the 65–69-year-old age category to more than 20% in the 90–94-year-old age categoryCitation2,Citation4,Citation5. Prevalence is higher among women than menCitation2,Citation4,Citation5. Like prevalence, incidence also increases with ageCitation5,Citation6. In Germany, it is estimated that between 120,000–160,000 new AD cases are arising each yearCitation2.
AD has a discrete disease onset, progresses slowly, and is currently incurable. The course of the disease is characterized by increasing loss of cognitive function and reductions in activities of daily living (ADL). As AD progresses, patients come to completely rely on daily care. Current therapeutic options with medications and psycho-social interventions are limited to reducing disease symptoms, preserving cognitive and non-cognitive functions, and improving quality-of-lifeCitation2,Citation7.
In 2008, dementia of any type (ICD F00-F03, G30) in Germany cost 10.3 billion Euros, most of which was attributable to long-term care. For comparison, total healthcare expenses in Germany totaled to 254 billion EurosCitation8. Taking into account the high proportion of patients in need of care and the unsalaried time for supporting the patients provided by families, AD is likely one of the most expensive diseases in older patientsCitation5. Due to demographic changes, the number of patients and related financial and social burden are expected to increaseCitation5.
This analysis covers anonymized data from 2005–2008 from ∼90,000 insurants with a diagnosis of any kind of dementia and thereof 20,000 insurants with specific AD diagnosis. This makes this analysis one of the largest studies of AD. With cooperation from the DAK-Gesundheit, a German SHI covering 6 million insurees, an assessment of medical and economic differences between patients with and without diagnosed AD was conducted. Analysis included co-morbidities, patterns of drug use, and clinical course, as well as the magnitude of these differences attributable to AD.
Patients and methods
Study setting and patient selection
The evaluation is based on retrospective analyses of routine data sets from 2005–2008 from a large German SHI company (DAK-Gesundheit—ca. 6 million insured persons). The anonymized data included information on hospitalizations, drug prescriptions, outpatient visits, and level of care provided by the health insurance for care related to AD.
The definition of dementia was based on the International Classification of Disease (ICD-10 GM), 10th revision for GermanyCitation1. A sub-group of patients with AD was defined as having at least one diagnosis of AD without any diagnoses indicating other forms of dementia. Although quality of coding is improving in previous years, diagnoses in routine data are based on average physician’s coding skills without a mandatory validation by epidemiologic instrumentsCitation9.
gives an overview of all ICD-10 codes considered during the selection process.
Table 1. Numbers and proportions of ambulatory diagnoses (ICD-10 GM) of dementia during the study period.Citation1.
Cross-sectional study
Patients were included if they were insured and alive on December 31st of the respective calendar year, continuously insured during each respective calendar year, and documented as having at least one ambulatory or in-patient diagnosis. An equal number of age- and sex-matched insurants who were continuously insured during the respective calendar year but had no diagnosis of dementia were included in the comparator group. Both patient groups were characterized for each 1-year period with the following characteristics: Demographics, number of hospitalizations, number of in-patient days, number of ambulatory physician visits, number of different drugs used, total number of defined daily doses (DDD) of drugs prescribed, and costs. Prevalence odds ratios (POR) of dementia including 95% confidence intervals were calculated for all categorical prevalence variables under study. The Wilcoxon signed rank test was used to investigate differences in the number of certain characteristics.
The cross-sectional analysis is able to identify changes of time and shows the characteristics and distributions of ambulatory diagnoses of patients with dementia in each year ().
Propensity score matched cohort study
In order to assess medical and economic consequences attributable to dementia, a propensity-score matched cohort design was used comparing patients with incident AD and patients without dementia. We applied a modified high-dimensional propensity score techniqueCitation10 to identify the most important variables for the propensity score model. As potential variables, we considered all ICD-10 diagnoses (4-digit level) and drugs (ATC substance level) in the year prior to the quarter of first dementia diagnosis or respective matching quarter in subjects without dementia.
Insurants were eligible to be in the cohort after at least 1 year of available person time. Potential quarter of cohort entry was the first quarter with at least one diagnosis of AD after becoming eligible. To be classified as an incident case of AD, insurants were required to have had no diagnosis of dementia and no prescriptions for anti-dementia drugs in the four quarters prior to the quarter of the first diagnosis of AD (index quarter). To clearly separate the person time before first diagnosis of dementia from the person time after this diagnosis, baseline characteristics were assessed in the four quarters prior to the index quarter, while follow-up began only on the first day of the next quarter. All cohort members were followed-up from the first day of the quarter following the index quarter until the end of the study period (November 30, 2008), death, or end of insurance with DAK ().
Figure 1. Study design of propensity score matched cohort.
Patients with incident AD differ from patients without, not only with respect to the presence of dementia, but also with respect to comorbidities. As the intention of this part of the study was to study the consequences attributable to AD, a propensity-score matching approach was used. Usually, propensity score is defined as the conditional probability of receiving the treatment given the observed covariatesCitation11. In our study, the propensity score was defined as the conditional probability of having a diagnosis of AD given the observed covariates. Matching by this propensity score leads to balanced patient characteristics between patients with and without AD.
A modified, high-dimensional propensity score technique, as suggested by Schneeweiss et al.Citation10, was applied to identify and consider the most important variables for the propensity score model. All ICD-10 diagnoses (4-digit level) and drugs (ATC substance level) in the year prior to first AD diagnosis (or respective matching quarter in subjects without AD) were considered potential variables. Each diagnosis’ and each drug’s ability to cause confounding was calculated. The 250 diagnoses and 250 drugs with the highest potential for causing bias were included in the propensity score model together with the factors of age, gender, number of hospitalizations, drugs prescribed, physician visits, and level of ambulatory nursing care. Based on their propensity score and using a greedy matching algorithm, the cohorts were matched 1:1Citation12.
Absolute standardized differences for all covariates were estimated to assess pre-match imbalances and post-match balances achieved and results for the main categories were presented as Love plots (). An absolute standardized differences of 0% indicates no bias, and values <10% suggest inconsequential biasCitation13.
Figure 2. Standardized differences (Love plot, difference in percent) of diagnoses and prescriptions before and after propensity score matching (Alzheimer’s dementia).
Outcomes of interest in this cohort study were event rates (hospitalizations, days in hospital, prescriptions, and physician visits) and costs before and after the index quarter. Event rates and costs were referenced to person time under observation (i.e., per person-year).
Costs
Costs analysis referred to direct costs from the SHI payer perspective. Hospitalization, rehabilitation, and drug prescription costs were calculated based on the costs documented in the database, which consist of financial expenditures reimbursed by the SHI for the patient.
Costs for outpatient services (i.e., the value of doctor–patient contacts) were based on a measurement proposed by Krauth et al.Citation14 in 2005. This involved multiplying the number of contacts per patient, which was available in the present dataset, by a rate of €23.07, €23.12, €23.16, and €23.21 per physician visit in 2005, 2006, 2007, and 2008, respectively. In 2008, the number of ambulatory service claims dropped by ∼18%, likely due to the introduction of lump sums in Germany. Lump sums do not reflect real resource use because not every physician visit causes deductible services. Therefore, it was assumed that the number of service claims remained fairly constant, and a correction factor of 1.2 was used, resulting in €27.85 per physician visit in 2008.
In Germany, the level of care needed (‘Pflegestufe’) represents the extent of daily care needed by the patient and is classified into three severity levels based on the care time needed: Level 1 = at least 1.5 h per day; Level 2 = 3 h per day; and Level 3 = 5 h per day. Depending on the year and the nature of the care provided, costs borne by the SHI vary. In 2008, for care provided by family members the SHI paid €215, €420, or €675 per month for care at levels 1, 2, or 3, respectively, to the caregiver. For ambulatory care administered by a nurse, SHI paid €420, €980, or €1470 per month for care at levels 1, 2, or 3, respectively. Long-term care in a nursing home amounts to €1023, €1279, or €1470 per month for care at levels 1, 2, or 3, respectivelyCitation15.
As only data on the level of care and not on the type of care were available, mean costs for each severity level were calculated by weighting with the official number of patients in need of care per kind of maintenance, resulting in €320 for level 1, €802 for level 2, and €1271 for level 3Citation16.
All calculations were performed with SAS 9.2 (SAS Institute, Cary, NC).
Results
Cross-sectional study
Between 86,000–96,000 insurants with at least one diagnosis of any type of dementia and an equal number of control subjects were identified. Approximately 20% of insurants with dementia had a diagnosis specific for AD without diagnoses indicating other forms of dementia. Approximately 70% of patients with any type of dementia and AD specifically were female. The proportion of insurants with AD increased with age, with mean age being ∼80 years during all 4 calendar years ().
Table 2. Characteristics of the study and control population (SD) before propensity-score matching.
There were substantial differences in the level of care needed between insurants with AD and the control group. During all years, the proportion of AD insurants needing the highest level of care (level 3) is at least 15-fold higher than the proportion of insurants from the control group ().
Number of hospitalizations, number of in-patient days, number of ambulatory physician visits, number of different drugs used, and total number of defined daily doses (DDD) of drugs prescribed were analyzed as measures of burden of disease. During all 4 years, these measures were substantially higher in insurants with AD as compared to controls ().
Regarding costs for hospitalization, rehabilitation services, outpatient services, ambulatory drug prescriptions, and long-term care during all years, patients with AD caused substantially higher costs than patients without AD. For instance, total costs were €7413, €8534, €8629, and €9207 in 2005, 2006, 2007, and 2008, respectively, for patients with AD, whereas total costs were €3378, €3540, €3707, and €3850 for controls. Differences in long-term care costs explained 70–75% of the total cost difference, with long-term care costs being responsible for about half of the total costs in AD patients ().
Table 3. Annual healthcare costs in Euros (SD) before propensity-score matching.
Propensity score matched cohort study
A total of 14,249 patients with incident AD and 2,328,331 potential controls (without AD) met the inclusion criteria for the propensity score analysis. For 14,116 patients with incident AD (99.1%), one matched control could be identified. After propensity score matching, both groups were well balanced with respect to the baseline variables which, including age, gender, number of hospitalizations, number of different drugs prescribed, number of physician visits, and the level of ambulatory nursing care, as well as leading diagnoses and drugs ().
Table 4. Leading diagnoses and drugs after propensity score matching.
While the event rates and costs in AD patients and controls were well balanced at baseline, substantial differences were noted after the index quarter between both groups. Total healthcare costs for patients with AD accrued to €10,266 after the index quarter and to €6495 for controls. The difference of €3771 is attributable to AD and mostly caused by differences in hospitalization and long-term care costs ().
Table 5. Event rates and costs in propensity-score matched cohort before and after diagnosis of dementia (POR 95% CI).
AD patients have a 42% and 54% higher rate of hospitalizations and number of in-patient days than control patients. However, the differences in the number of prescriptions and number of physician visits were smaller between both patient groups (22% and 28% higher rates for AD patients, respectively).
Discussion
AD syndromes are frequent in the elderly and, due to the substantial care needed, result in a considerable financial burdenCitation5. With an aging population, the number of patients with AD will increase. Therefore, a better understanding of the economic implications of AD is relevant for policymakers, as well as the general public.
The focus of this analysis was the assessment of differences between subjects with and without AD using routine data from a large German SHI between 2005–2008. Four annual cross-sectional studies and one propensity score matched cohort model were analyzed.
In our cross-sectional analyses, patients with AD scored higher than patients without AD on the following measures: hospitalization rates, number of in-patient days, number of prescriptions, and number of physician visits. The considerable differences in the total annual healthcare costs are mainly due to the higher costs for long-term care.
A propensity score matched cohort study was used to assess the medical and economic consequences attributable to AD. In this analysis, AD patients cost €10,266 after the index quarter of diagnosis, while the matched controls cost €6495. The difference of €3771 can be attributed to AD alone.
The results for AD patients fit well with the results of another recent claim-based study by Kiencke et al.Citation17 based on 35,684 insured AD patients. Depending on the AD treatment (memantine, psychotropic drugs, or no anti-dementia drugs), patients in 2005 caused average total direct costs of between €7000–€11,000. Reese et al.Citation18 performed an interview-based analysis of 395 AD patients. From the societal perspective, it has been estimated that the average annual cost of an AD patient is ∼€13,080 per patient, with €25,500 for patients living in long-term care facilities and €7450 for patients living at homeCitation18. Leicht et al.Citation19 obtained data for Germany from structured interviews with both patients and caregivers as well as controls with a difference of €2149 for medical care and €9856 for nursing care, including direct costs borne by the patients. Despite this study being based on 349 patients and controls only, due to their strong time-based resource cost calculation approach it leads to the assumption that nursing care cost is under-estimated if calculated on claims only and, hence, is under-financed from a statutory insurance point of view.
The present study used data from 20,000 AD patients is the largest study to our knowledge using data over a time period of 4 years. The DAK covers persons in all regions of Germany, but, because of the historically evolved social-demographic structure, the results are not fully transferable to the general German populationCitation20. A higher social status and better general health of the investigated population could lead to an under-estimation of prevalence and incidence, also at the same time to an over-estimation of usage of healthcare resources and costs. Nevertheless, health claims data from health insurance funds provides most valid resources for analysis of healthcare utilization, compared to alternatives like survey estimates based on self-reportsCitation21.
Another limitation of this study was uncertainty with regard to the reported diagnosis. AD is the most common form of dementia. According to the German Alzheimer SocietyCitation22, clinical diagnosis is correct in more than 80% of patients, while a validated diagnosis is only available after death via examination of the brain. In this dataset, only 20% of dementia patients had a specific AD diagnosis, so this clinical diagnosis seems to be under-represented in SHI routine data. As family members often miss the signs of cognitive failure, the dementia is not brought to the attention of a physicianCitation23. Also in many cases, the early symptoms are mistaken for normal aging or senilityCitation24. A recent study showed the relevance of primary care physicians for diagnosis and treatment of patients with dementia in Germany. This could raise the proportion of undiagnosed dementia or not valid AD-specific diagnosed patients and possibly reduce the share of patients with appropriate drug-treatmentCitation25.
The study only reflects direct cost from a SHI perspective. All direct costs borne by the patients (co-payments) and other insurances (pension insurance paying for rehabilitation services) as well as indirect costs (work loss) were not included. Also the differentiation with regard to disease severity was not possible with the dataset. Finally, data reflecting rehabilitative inpatient care, physiotherapy, and technical aids were not available.
Conclusions
From the perspective of the German statutory health insurance system, AD is a substantial cost driver. Comparing matched AD patients with controls in a German SHI setting results in 2.2–2.4-fold higher costs for insured people with AD in a cross-sectional analysis. Long-term care costs in the expenses of the SHI account for about half of the total costs for AD patients.
Transparency
Declaration of funding
This study was sponsored by Janssen Alzheimer Immunotherapy Research & Development, LLC, and Pfizer Inc.
Declaration of financial/other relationships
BB and NE are employees of Boston Healthcare Associates International GmbH, who were paid consultants to Janssen Alzheimer Immunotherapy Research & Development, LLC, and Pfizer Inc. in connection with the development of this manuscript. LL was an employee of Janssen Alzheimer Immunotherapy Research & Development LLC at the time the work was conducted. FA and SW received no payment in connection with the development of this manuscript.
Acknowledgements
We would like to thank Dr. Jörg Burkowitz for review and support.
References
- Deutsches Institut für Medizinische Dokumentation und Information. ICD-10-GM. Altersdemenz. Berlin: Robert Koch-Institut, 2005. http://www.dimdi.de/static/de/klassi/icd-10-gm/index.htm. Accessed 2012
- Weyerer S. Alterdemenz - Gesundheitsberichterstattung des Bundes Heft 28. Robert Koch-Institut, 2005
- Sütterlin S, Hoβmann I, Klingholz R. Demenz-Report. Berlin: Berlin-Institut für Bevölkerung und Entwicklung, 2011
- Lobo A, Launer LJ, Fratiglioni L, et al. Prevalence of dementia and major subtypes in Europe: a collaborative study of population-based cohorts. Neurologic Diseases in the Elderly Research Group. Neurology 2000;54(11 Suppl 5):S4-S9
- Bickel H. Epidemiologie und Gesundheitsökonomie. In: Wallesch CW, Förstl H, eds. Demenzen. Referenzreihe Neurologie. Stuttgart: Thieme Verlag, 2005. pp. 1-15
- Fratiglioni L, Launer LJ, Andersen K, et al. Incidence of dementia and major subtypes in Europe: a collaborative study of population-based cohorts. Neurologic Diseases in the Elderly Research Group. Neurology 2000;54(11 Suppl 5):S10-S15
- S3-Leitlinie ‘Demenzen’ Langversion. Deutsche Gesellschaft für Psychiatrie PuN, Deutsche Gesellschaft für Neurologie, Deutsche Alzheimer Gesellschaft, Berlin (location of the society: Deutsche Gesellschaft für Psychiatrie und Psychotherapie, Psychosomatik und Nervenheilkunde e.V. (DGPPN)) 2009
- Gesundheitsberichterstattung des Bundes. Krankheitskosten in Mio. € für Deutschland. Gliederungsmerkmale: Jahre, Alter, Geschlecht, ICD 10. http://www.gbe-bund.de/oowa921-install/servlet/oowa/aw92/WS0100/_XWD_PROC?_XWD_110/6/XWD_CUBE.DRILL/_XWD_136/D.946/14332. Accessed May 1, 2012
- Schubert I, Koster I, Kupper-Nybelen J, et al. [Health services research based on routine data generated by the SHI. Potential uses of health insurance fund data in health services research]. Bundesgesundheitsbla Gesundheitsforschung Gesundheitsschutz 2008;51:1095-105
- Schneeweiss S, Rassen JA, Glynn RJ, et al. High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology 2009;20:512-22
- Rosenbaum P. The Propensity Score. Observational studies, 2nd edn. New York: Springer-Verlag, 2002. pp. 296
- Parsons LS. Performing a 1:N Case-Control Match on Propensity Score. Montreal: SAS Institute Inc. 2004. http://www2.sas.com/proceedings/sugi29/165-29.pdf. Accessed 2004
- Austin P. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Statist Simula Computa 2009;38:1228-34
- Krauth C, Hessel F, Hansmeier T, et al. Empirical standard costs for health economics evaluation in Germany - a proposal by the working group methods in health economic evaluation. Gesundheitswesen 2005;67:736-46
- Bundesministerium der Justiz. Sozialgesetzbuch (SGB) - XI - Soziale Pflegeversicherung. Berlin: Bundesministerium der Justiz, 2012
- Statistisches Bundesamt. Pflegestatistik 2005 - Pflege im Rahmen der Pflegeversicherung. Wiesbaden: 2007
- Kiencke P, Daniel D, Grimm C, et al. Direct costs of Alzheimer's disease in Germany. Eur J Health Econ 2011;12:533-9
- Reese JP, Hessmann P, Seeberg G, et al. Cost and care of patients with Alzheimer's disease: clinical predictors in German health care settings. J Alzheimers Dis 2011;27:723-36
- Leicht H, Heinrich S, Heider D, et al. Net costs of dementia by disease stage. Acta Psychiatr Scand 2011;124:384-95
- Hoffmann F, Icks A. [Structural differences between health insurance funds and their impact on health services research: results from the Bertelsmann Health-Care Monitor]. Gesundheitswesen 2012;74:291-7
- Hunger M, Schwarzkopf L, Heier M, et al. Official statistics and claims data records indicate non-response and recall bias within survey-based estimates of health care utilization in the older population. BMC Health Serv Res 2013;13:1
- Haupt M. Die Diagnose der Alzheimer-Krankheit. Deutsche Alzheimer Gesellschaft, editor. Das Wichtigste - Informationsblätter. Berlin 1999
- Ross GW, Abbott RD, Petrovitch H, et al. Frequency and characteristics of silent dementia among elderly Japanese-American men. The Honolulu-Asia Aging Study. JAMA 1997;277:800-5
- Small GW, Rabins PV, Barry PP, et al. Diagnosis and treatment of Alzheimer disease and related disorders. Consensus statement of the American Association for Geriatric Psychiatry, the Alzheimer's Association, and the American Geriatrics Society. JAMA 1997;278:1363-71
- IGES. Neurologische und psychiatrische Versorgung aus sektorenübergreifender Perspektive. Berlin: IGES Institut GmbH 2014. http://www.iges.de/publikationen/gutachten__berichte/neurologisch_psychiatrische_versorgung/e15931/infoboxContent15932/IGESInstitut_StudieNeurologischpsychiatrischeVersorgung_2014_ger.pdf. Accessed October 9, 2014