Abstract
Objective
Estimate the lifetime cost of dementia to Medicare and Medicaid.
Data Source
1997–2005 Medicare Current Beneficiary Survey.
Study Design
A multistage analysis was conducted to first predict the probability of developing dementia by age and then predict the annual Medicare/Medicaid expenditures conditional on dementia status. A cohort-based simulation was conducted to estimate the lifetime cost of dementia.
Principal Findings
The average lifetime cost of dementia per patient for Medicare is approximately $12,000 (2005 dollars) and for Medicaid about $11,000. Dementia onset at older age leads to shorter duration and lower lifetime cost. Increased educational level leads to longer longevity, more dementia cases per cohort, but shorter duration, and lower lifetime cost per patient, which could offset the cost increase induced by more dementia cases. Increased body mass index leads to more dementia cases per cohort and higher lifetime cost per patient.
Conclusion
Net cost of dementia is lower than the estimates from cross-sectional studies. Promoting healthy lifestyle to reverse the obesity epidemic is a short-term priority to confront the epidemic of dementia in the near future. Promoting higher education among the younger generation is a long-term priority to mitigate the effect of population aging on the dementia epidemic in the distant future.
Keywords: Health care costs, Medicare, aging/elderly/geriatrics
Dementia has become a major public health challenge for our aging society. Currently, about 5.2 million Americans live with Alzheimer's disease (AD) and other dementias, and the estimated societal cost of formal and informal care for these conditions is more than $140 billion annually (2008; Philip et al. 2002; Plassman et al. 2007; Gajital 2008; Goldman and Joyce 2008). In 2012, the first cohort of the 76 million “Baby Boomers” born between 1946 and 1964 will begin to turn 65. How this demographic trend will influence the future number of dementia cases and the societal cost of care for these cases is of great public health and policy significance. Based on the current estimates of the dementia prevalence rate and the cost of dementia care found in cross-sectional studies, the National Institutes of Health (NIH) and the Alzheimer's Association predict that increased population size and extended longevity will lead to significant increases in dementia cases among Baby Boomers—as many as double the cases—and escalating societal cost of dementia care that could top $20 trillion over the next 40 years (Alzheimer's Association 2010a,b).
However, the results from other studies suggest these prevalence numbers and cost estimates may not be directly applicable to the Baby Boomer generation and suggest longitudinal analysis regarding the future health and economic burden of dementia (Brookmeyer and Gray 2000; Brookmeyer et al. 2002; Langa et al. 2008). The major reason is like many other chronic conditions among the elderly, the total number of dementia cases and the cost of dementia care are determined not only by the percentage of elderly who will develop these conditions before death but also by the longitudinal course of illness of these conditions over the life span. This course of illness is related both to age and to proximity to death at disease onset. For example, using data in the Health and Retirement Study from multiple birth cohorts, Langa et al. (2008) found that despite increased longevity and higher probability of having dementia before death, younger cohorts have more compressed episodes of dementia shortly before death, characterized by later onset and shorter duration. This finding is attributed to special characteristics of the younger cohorts, especially higher education level, which is a known mitigating factor for dementia (Friedland 1993; Katzman 1993; Stern et al. 1994; Herbert et al. 2003; Gatz, Prescott, and Pedersen 2006; Plassman et al. 2007; Langa et al. 2008).
Such results suggest that the cost of care for dementia should be more precisely defined as a longitudinal cost over the course of illness. A patient with a shorter episode of dementia before death in an older age could cost either more or less to society than one with a more protracted course of illness that begins at a younger age. Existing studies that take the longitudinal perspective reached quite different conclusions regarding the effect of dementia on medical care costs from the cross-sectional studies. Using the National Long-Term Care Survey (NLCS) merged with Medicare claims, researchers found that patients with AD and related conditions do not have higher Medicare expenditures over the last 5 years of their life than the nondemented elderly, and that the total annual program payments of Medicare and Medicaid attributable to AD and related disorders even decreased slightly between 1994 and 1999 (Ayyagari, Salm, and Sloan 2008; Lamb, Sloan, and Nathan 2008).
However, these studies have several limitations that reduce their applicability to the general Medicare population and to future Medicare beneficiaries. First, these studies used comparatively less recent NLCS data from 1982 to 1999, which focus on a more severely ill group of elderly who needed assistance with at least one activity of daily living (ADL) and instrumental activity of daily living (IADL). Therefore, their results could overestimate the prevalence and underestimate the lifetime cost of dementia. Second, these measurements of dementia and related conditions are based exclusively on diagnosis codes from Medicare claims, the use of which could lead to underestimates of the prevalence of dementia (Taylor and Sloan 2000; Taylor et al. 2009). Third, NLCS has limited information regarding demographic and socioeconomic features of the respondents. In particular, 48.1 percent of the sample has missing data on education level, which is the most important mitigating factor of dementia and longevity; therefore, the absence of this information could limit the external validity of the results. Fourth, although both studies did estimate the longitudinal Medicare expenditures attributed specifically to dementia while controlling for the endogeneity of dementia status and aging via counter-factual simulation or propensity score matching, they were unable to control for cohort effects. Instead, they assumed that the aging process for future generations will be similar to that of the current elderly generation. Hence, they could not model future generations’ aging and dementia experience, which is predicted to feature extended longevity, but more compressed dementia episodes.
In this study, we addressed a number of these limitations in the previous literature to conduct a longitudinal analysis of the lifetime cost of dementia. Our study is designed to achieve three major goals: (1) to precisely measure the onset and duration of dementia after adjusting for longevity and education level, (2) to estimate lifetime Medicare and Medicaid expenditures related to dementia, and (3) to simulate the effect of changes in population character on the prevalence, course of illness, and longitudinal cost of care of dementia at population level. To achieve these goals, we first used data from a nationally representative sample of the entire elderly population—the Medicare Current Beneficiary Survey (MCBS) from 1997 to 2005. Second, we created an algorithm to measure dementia status by combining information from the MCBS survey, inpatient and outpatient claims, and prescription drug event files. Finally, we conducted a multistage analysis method to first estimate the probability of developing dementia before death based on age and then estimate Medicare and Medicaid expenditures based on dementia status, age, and time to death. Based on the estimation, we conducted a series of cohort-based simulations by imputing the parameters from the first two steps into a dynamic aging simulation dataset to incorporate the dementia onset and cost of care over the life span.
Methods
Data and Measurements
We used the 1997–2005 MCBS Cost and Use files to determine dementia status and estimate cost. Several features of this dataset make it ideal for our analysis. First, the MCBS is a longitudinal and multipurpose survey of a nationally representative sample of the entire Medicare population. Second, it facilitates identifying individuals with AD and other dementia. MCBS contains not only ample survey information that uniquely includes dementia history but also includes reliable claims data that have diagnosis codes for dementia from inpatient, outpatient, and physician service event files, as well as codes for AD drug prescription. Third, the dataset provides sufficient samples of both survivors and decedents that allow us to examine trends in health care expenditures among dementia and nondementia beneficiaries as they age toward death. Our study sample consisted of all Medicare beneficiaries aged 65 or older. Nonelderly and those older than age 101 were excluded due to small sample size. The final dataset for analysis has 113,811 person-year observations representing 53,244 unique individuals. Among these individuals, 34 percent were surveyed in only 1 year, 19 percent for 2 years, and 47 percent for 3 years.
The golden standard for defining dementia is clinical diagnoses using physical examinations and medical record review. But, with the exception of studies using the Aging, Demographics, and Memory Study (ADAMS) data (Plassman et al. 2007), this type of detailed clinical data is not included in most datasets. However, relying on single data source such as AD diagnosis from Medicare claims leads to underestimation of AD and potential selection bias (Taylor and Sloan 2000; Lin et al. 2009; Taylor et al. 2009). In this study, therefore, we used an inclusive case definition, which resulted in less underestimation of dementia prevalence and avoided sample selection bias from relying on a single data source, such as determining AD diagnosis from Medicare claims (Lin et al. 2009). Specifically, we categorized observations as having dementia if she/he met any of the following three criteria: (1) the respondents or proxy answered “yes” to the survey question “Has a doctor ever told you that you had Alzheimer's disease or dementia?” in the survey; or (2) had at least one of the following International Classification of Diseases-ninth revision-Clinical Modification (ICD-9-CM) diagnosis codes in his/her inpatient, outpatient, or physician services event files: 290 or 331.0; or (3) ever took any AD-targeted prescription drugs, which were identified by the reported drug names in the prescription drug event files. These drugs include donepezil (Aricept), rivastigmine (Exelon), galantamine (Reminyl or Razadyne), and memantine (Namenda). Because dementia is not reversible, respondents with 2 or 3 years of observations were coded as having dementia in all subsequent years if they were identified as a dementia patient in 1 year. We classified survey respondents without any of the three criteria in any year as not having dementia.
Total Medicare health care expenditures were defined in our analysis as the sum of expenditures paid by Medicare across all types of services events from the claims data, including inpatient, outpatient, physician visits, skilled nursing facility, home health, hospice, and other medical services. Similarly, total Medicaid health care expenditures were defined as the sum of expenditures paid by Medicaid across all types of services. All expenditures were converted to 2005 dollar value by the Consumer Price Index for medical care from the Bureau of Labor Statistics.
We examined the existing chronic diseases reported in the survey that are leading causes of death and disability among the older population, including cardiovascular diseases (heart diseases, stroke, and hypertension), respiratory system diseases, and cancer (Ferrucci and Guralnik 1997). We also measured the occurrence of acute medical care events related to these chronic conditions, as indicated by ICD-9 codes for primary diagnoses in the claims files. Demographic characteristics and socioeconomic status—most important, education level—were collected in the survey. Body mass index (BMI), which is an important indicator of general health and predictor of mortality, morbidity, and health care expenditures, was calculated from the self-reported weight and height reported in the survey and included as an independent variable.
Regression Analysis
We first estimated a Logit regression to predict the probability of dementia at a certain age. The dependent variable was whether a respondent is identified as having dementia by the algorithm we introduced above; the independent variables include the respondent's age, education level, the interaction of age and education level, BMI, existing chronic diseases, acute medical events, and other demographics, such as race and marital status. The equation is depicted below:
 |
We then used the Rand two-part model to predict annual Medicare and Medicaid expenditures conditional on dementia status, age, and time to death (Manning et al. 1987). Specifically, the first part of the model was a Logit model predicting the probability of having any Medicare and Medicaid expenditures, and the second part of the model was ordinary least square (OLS) predicting the natural log of annual Medicare and Medicaid expenditures among the observations with positive expenditures only. In these models, the key independent variables were dementia status, age, and death year. We added interactions between dementia status and each of age and death year. Besides its strong influence on longevity, education level is also a strong predictor of health care expenditures. We therefore included education level and its interaction term with age as independent variables. The equation is depicted below:
 |
Simulation
It is difficult to draw meaningful conclusions regarding the longitudinal course of illness and cost of care of dementia patients with the regression results alone. Thus, we conducted a simulation model to apply the parameters from our regression analyses to a virtual elderly cohort born between the 1920s and 1930s to approximate the longitudinal dementia course of illness and cost of care conditional on other population characteristics and the aging process. This simulation model adopts a similar theoretical framework and rationale as the Rand Future Expenditure Model (Goldman et al. 2004; Lakdawalla, Goldman, and Shang 2005) and addressed the dynamic nature of the aging process more precisely by including the development of other chronic diseases, fluctuation of body weight, and deterioration of functional status (Yang and Hall 2008).
Specifically, we first constructed and estimated a dynamic model using 1992–2004 MCBS data to quantify the relationship between several key factors that influence the aging process and Medicare/Medicaid expenditures. Such factors include, but are not limited to, BMI, development of chronic diseases, deterioration of functional status, enrollment into supplemental insurances, and mortality. Then, the aging process of a virtual cohort of elderly born between the 1920s and 1930s was simulated based on the dynamic model. Next, we incorporated the parameters obtained from the dementia status model in this study to simulate whether the respondent would develop dementia at a certain age. If the respondent was predicted to have dementia, such condition was considered permanent until death. Last, using the parameters from the two-part models, we simulated annual Medicare and Medicaid expenditures based on age, dementia status, time to death, education level, other chronic conditions, BMI, and demographics. Because the second part of the two-part model predicts the natural log of the real health care expenditures, we used smearing factors to transform the predicted log expenditures to level expenditures. The random error term was generated by the Monte Carlo process.
For the dementia patients, we calculated their lifetime Medicare and Medicaid expenditures from dementia onset until death. Previous research illustrated that dementia status is likely to be endogenous with higher education level, older age, and time to death because better educated people are more likely to live longer, and hence face higher risk of dementia before death. Not controlling for this issue will likely bias the estimate of the net cost of dementia. Therefore, instead of comparing the cost of demented elderly with those who never had dementia, we conducted a counter-factual simulation of lifetime Medicare/Medicare expenditures assuming the same aging process but without dementia for these observations. We then calculated the difference between the original and counter-factual simulations as the net longitudinal Medicare and Medicaid cost attributed to dementia condition over the life span exclusively.
Besides this baseline simulation among the elderly cohort as observed in MCBS, we conducted simulations evaluating the impact of changes in population character on dementia prevalence and cost. Specifically, we focused on increased educational level and increased BMI level. Education is a known mitigation factor for course of illness of dementia (Friedland 1993; Katzman 1993; Stern et al. 1994; Herbert et al. 2003; Gatz, Prescott, and Pedersen 2006; Plassman et al. 2007; Langa et al. 2008); therefore, increased education could possibly offset the additional cost of dementia due to increasing population size and extended longevity. Overweight and obesity, however, were found to increase late-life dementia risk in recent publications using longitudinal data of identical twins in Enrope (Xu et al. 2011). In the simulation, we first assumed that the 76 million Baby Boomers have similar characteristics as the cohort observed from MCBS and calculated the prevalence rate and cost of dementia per capita and at the population level. Second, we assumed that the Baby Boomer generation's high school graduation rate is 5 percent higher and their average BMI is 10 percent higher than the rates observed in the MCBS sample, and simulated the aging process, course of illness, and cost of dementia. Last, we compared the simulation results between the hypothetical scenarios with increased education and BMI with the baseline scenario to calculate the net effect of such changes on longevity, number of dementia cases, course of illness, and cost of care per patient and in total.
Results
The summary statistics of the entire MCBS 1997–2005 study sample, the dementia patient sample, and the sample of elderly without dementia are presented in Table 1. Approximately 6.4 percent of the sample was identified as dementia patients, who have significantly higher annual Medicare and Medicaid expenditures at $10,814 and $6,234 comparing to the nondemented elderly at $5,953 and $1,962, respectively (p < .05). However, the dementia patients are also older than the nondemented elderly by more than 10 years on average (82.5 versus 71.3, respectively, p < .05) and have a substantially higher mortality rate at 15.95 versus 4.76 percent (p < .05). The nondemented elderly are more educated, with about 5 percent more people having a high school or college education. There is no obvious difference in the prevalence rate of existing chronic diseases between the dementia and nondemented samples, but the rate of acute cardiovascular diseases is obviously higher in the dementia sample than in the nondemented sample (35.8 versus 27.1 percent, respectively, p < .05). Such results indicate that the onset of dementia closely associates with older age, proximity to death, and cardiovascular diseases; hence, cross-sectional analysis could easily overestimate the cost of dementia if such analysis does not control for proximity to death and the dynamic process of chronic disease development.
Table 1.
Summary Statistics of the Cost and Use Files MCBS 1997–2005 Study Sample
| Entire Sample N = 113,811 Mean | Dementia Sample N = 8,586 Mean | Nondementia Sample N = 105,225 Mean | |
|---|---|---|---|
| Dementia status | 6.4% | – | – |
| Health care expenditures | |||
| Total annual Medicare expenditures | $6,319 | $10,814 | $5,953 |
| Total annual Medicaid expenditures | $2,207 | $6,234 | $1,962 |
| Population characteristics | |||
| Age | 72.1 | 82.5 | 71.3 |
| Death year | 5.4% | 15.9% | 4.7% |
| BMI | 26.5 | 24.5 | 26.6 |
| Education | |||
| High school | 29.1% | 24.4% | 29.4% |
| Some college | 28.4% | 22.4% | 28.8% |
| College graduate | 5.5% | 3.8% | 5.6% |
| Existing chronic conditions | |||
| Cardiovascular diseases | 56.4% | 52.8% | 56.7% |
| Diabetes | 18.5% | 18.9% | 18.5% |
| Cancer | 16.7% | 12.5% | 17.1% |
| Respiratory system diseases | 15.4% | 13.2% | 15.6% |
| Acute medical events | |||
| Cardiovascular diseases (ICD-9) | 27.6% | 35.8% | 27.1% |
| Cancer (ICD-9) | 5.8% | 4.7% | 4.7% |
| Respiratory system diseases (ICD-9) | 5.8% | 6.9% | 6.9% |
| Other demographics | |||
| Race | |||
| Non-Hispanic white | 84.3% | 85.2% | 84.1% |
| Non-Hispanic black | 10.6% | 10.3% | 10.6% |
| Hispanics | 2.5% | 1.9% | 2.6% |
| Marital status | |||
| Married | 46.9% | 31.6% | 48.1% |
| Never married | 9.6% | 6.6% | 9.8% |
| Separated | 10.2% | 6.3% | 10.5% |
| Widowed | 33.2% | 55.0% | 31.4% |
The first column of Table 2 presents the results from our first Logit regression model, which predicts dementia status conditional on age and education level. Age and education are both strong predictors of dementia. Age has a statistically significant positive effect (coefficient = 0.53, p < .01), and education has a statistically significant negative effect (coefficient of high school education = −1.73, p < .01; coefficient of college = −1.76, p < .01) on the probability of having dementia. The coefficients of the interaction terms between age and education level are also significantly positive. BMI and existing chronic diseases are also strong predictors of dementia. Higher BMI has a negative effect on dementia (coefficient = −0.14, p < .001), but the interaction of BMI and age has a positive effect on dementia (coefficient = 0.11, p < .001). Patients with cardiovascular diseases and diabetes are more likely, but cancer patients are less likely, to develop dementia.
Table 2.
Regression Results from the Models Predicting AD Status and Health Care Cost
| Dementia Logit | Any Medicare Expenditures Logit | Ln (Medicare Expenditures) OLS | Any Medicaid Expenditures Logit | Ln (Medicaid Expenditures) OLS | |
|---|---|---|---|---|---|
| Demented | 0.82* | 1.18*** | 0.92** | −1.09** | |
| Age × dementia | −0.01 | 0.01*** | −0.01 | 0.03*** | |
| Death year | 0.17 | 3.65*** | 0.38 | −1.11*** | |
| Death year × dementia | 0.51*** | −0.41*** | −0.25*** | −0.34*** | |
| Age × death year | −0.01 | −0.03*** | 0.01 | 0.02*** | |
| Age | 0.53*** | 0.02*** | 0.03*** | 0.02*** | 0.02*** |
| Education: high school | −1.73*** | 0.08*** | 0.03* | −1.23*** | 1.44*** |
| Education: college | −1.76*** | 0.12*** | 0.09*** | −1.75*** | −1.66*** |
| Age × high school | 0.02*** | ||||
| Age × college | 0.02*** | ||||
| BMI | −0.14*** | −0.05** | 0.01 | 0.01 | 0.03 |
| BMI × age | 0.11*** | 0.07*** | −0.03* | −0.03 | −0.04 |
| Chronic conditions | |||||
| Cardiovascular diseases | 0.27*** | 0.03 | 0.29*** | 0.09*** | 0.01 |
| Cancer | −0.24*** | −0.12*** | 0.18*** | −0.36*** | −0.37*** |
| Diabetes | 0.15*** | 0.19*** | 0.47*** | 0.44*** | 0.51*** |
| Respiratory system disease | 0.05 | 0.05* | 0.22*** | 0.39*** | 0.38*** |
| Acute events | |||||
| Cardiovascular diseases | 4.94*** | 1.13*** | 0.51*** | 0.32*** | |
| Cancer | 5.25*** | 1.14*** | 0.16*** | 0.05 | |
| Respiratory system disease | 4.17*** | 0.99*** | 0.56*** | 0.59*** | |
| Moderately disabled | −0.01 | 0.29*** | −0.11*** | −0.36*** | |
| Severely disabled | 0.02 | 0.59*** | 0.20*** | −0.05 | |
| Male | 0.05 | −0.30*** | −0.05*** | −0.29*** | −0.15*** |
| Black | 0.21*** | −0.35*** | 0.05** | 0.98*** | 1.36*** |
| Hispanics | −0.17* | −0.50*** | 0.05 | 1.66*** | 2.50*** |
| Rural | 0.092*** | 1.11*** | −0.340*** | 0.09*** | 0.08*** |
Notes.
Statistically significant at 90% confidence interval (p < .1).
Statistically significant at 95% confidence interval (p < .05).
Statistically significant at 99% confidence interval (p < .01).
The second to fifth columns of Table 2 present the results from the two-part models used to predict the Medicare and Medicaid expenditures conditional on dementia, age, and death year. Having dementia increases the probability of having any Medicare expenditures (coefficient = 0.82, p < .1), as well as increases the log of annual expenditures for those with any Medicare expenditure (coefficient = 1.18, p < .01). Having dementia also has a positive effect on the probability of having any Medicaid expenditures (coefficient = 0.92, p < .05), but it has a negative effect on the log of annual expenditures for those with any Medicaid expenditure (coefficient = −1.09, p < .05). Death year does not influence the probability of having any Medicare or Medicaid expenditures. Instead, respondents have higher Medicare (coefficient = 3.65, p < .01) but lower Medicaid (coefficient = −1.11, p < .01) expenditures in their last year of life, conditional on having any Medicare/Medicaid expenditures. Both age and education level have a significant effect on Medicare and Medicaid expenditures. Older age was related to higher health care expenditures, while higher education level predicts higher Medicare expenditures, but lower Medicaid expenditures. Except for cancer, all existing chronic conditions and acute medical events significantly increase Medicare and Medicaid expenditures.
Table 3 summarizes the simulation results modeling the age of onset, duration, and lifetime Medicare and Medicaid expenditures for dementia patients, the counter-factual simulation of the dementia patients assuming equal longevity and other health conditions but no dementia, and the net effect of dementia on lifetime Medicare and Medicaid expenditures. About 14 percent of the simulated elderly cohort developed dementia before death. The simulated average dementia duration is 5.1 years before death, which is slightly less than the estimated duration between 5.5 and 6 years from longitudinal cohort-based studies among smaller populations in rural Pennsylvania, Baltimore, and Seattle (Brookmeyer et al. 2002; Larson et al. 2004; Ganguli et al. 2005). The average lifetime Medicare cost attributed to dementia is approximately $12,000 (p < .05), and the average lifetime Medicaid cost attributed to dementia is approximately $11,000 (p < .05). In addition to these results, we compared patients who developed dementia before age 80 to those who developed dementia after age 80. We found that the group with earlier onset of dementia, on average, has longer dementia duration at 6.4 years, higher Medicare costs of $15,607, and higher Medicaid costs of $14,336 over the course of illness, than the group with later onset of dementia, which has shorter dementia duration of 3.8 years, lower Medicare costs of $8,300, and lower Medicaid costs of $7,584 over the course of illness. Such results confirm the hypothesis that more compressed dementia episodes in older age before death lead to lower formal care expenditures reimbursed by Medicare and Medicaid.
Table 3.
Simulated Net Effect of Dementia on Lifetime Medicare and Medicaid Expenditures among the Cohort Born in the 1920s to 1930s
| Dementia | Counter Factual | Net Difference | |
|---|---|---|---|
| Entire sample | 14.2% | 0 | 13.9% |
| Percentage of onset | |||
| Age of onset | 80.2 | – | – |
| Duration (years) | 5.1 | – | – |
| Medicare | $101,810 | $89,835 | $11,975 |
| Medicaid | $27,028 | $15,955 | $11,073 |
| Early onset before 80 | |||
| Percentage of onset | 7.1% | 0 | 7.5% |
| Age of onset | 78.5 | – | – |
| Duration (years) | 6.4 | – | – |
| Medicare | $102,448 | $86,841 | $15,607 |
| Medicaid | $29,484 | $15,148 | $14,336 |
| Late onset after 80 | |||
| Percentage of onset | 7.1% | 0 | 7.1% |
| Age of onset | 83.2 | – | – |
| Duration (years) | 3.8 | – | – |
| Medicare | $102,771 | $94,450 | $8,321 |
| Medicaid | $24,401 | $16,817 | $7,584 |
Table 4 summarizes the simulation results comparing the baseline scenario for a cohort of 76 million with the same characteristics as the elderly cohort born in the 1920s to 1930s as observed in MCBS, with hypothetical scenarios of 5 percent increase in high school graduation rate and 10 percent increase in average BMI. Increased education level will lead to 0.7 years of increased longevity (from 85.3 to 86.0, p < .05), accompanied by a 0.5 percent (p < .05) increase in the percentage of the cohort that experiences dementia before death (from 13.9 to 14.4 percent). However, the average age of dementia onset will also increase by 0.9 years (p < .01), from 80.2 to 81.1 years, and the average duration of dementia will drop by 0.2 years from 5.1 to 4.9 (p < .05). In regard to cost, average lifetime Medicare expenditures will decrease $618 (from $11,975 to $11,357, p < .05) per patient, and the Medicaid lifetime expenditures will decrease $1,641 (from $11,073 to $9,432, p < .05) per patient. At population level, the total number of dementia cases will increase by 380,000 because of the 5 percent increase in high school graduation rate, but the net total Medicare cost of treating dementia will be $2.2 billion lower than that of the baseline scenario, and the total Medicaid cost will be $13.7 billion lower than that of the baseline scenario. Increase in BMI by 10 percent will lead to slight decrease in longevity (0.1 year, p < .05), 4.4 percent increase in dementia cases (from 13.9 to 18.3 percent, p < .05). Average Medicare expenditures per patient will increase $732 (p < .05), and Medicaid expenditures will decrease $376 (p < .05). At the population level of 76 million, a 10 percent increase in baseline BMI at age 65 will lead to more than 3 million more dementia cases, and $50.3 billion and $32.8 billion increases in Medicare and Medicaid cost.
Table 4.
Summary Statistics of the Simulated Net Effect of Improved Education Level or Increased BMI on Dementia Prevalence and Lifetime Dementia Care Cost (76 Million Baseline Population)
| Increased Education | Increased BMI | ||||||
|---|---|---|---|---|---|---|---|
| Baseline | Per Capita Estimate | Net Effect per Capita | Net Effect in Total | Per Capita Estimate | Net Effect per Capita | Net Effect in Total | |
| Longevity | 85.3 | 86.0 | 0.7 | – | 85.2 | −0.1 | – |
| Prevalence | 13.9% | 14.4% | 0.5% | 380,000 | 18.3% | 4.4% | 3,344,000 |
| Age of onset (years) | 80.2 | 81.1 | 0.9 | – | 80.1 | −0.1 | – |
| Duration (years) | 5.1 | 4.9 | −0.2 | – | 5.0 | −0.1 | – |
| Net Medicare cost | $11,975 | $11,357 | −$618 | −$2.2 billion | $12,707 | 732 | $50.3 billion |
| Net Medicaid cost | $11,073 | $9,432 | −$1,641 | −$13.7 billion | $10,697 | −376 | $31.8 billion |
Discussion
In this study, we used an innovative longitudinal multistage cohort-based simulation model to examine the lifetime cost of dementia to Medicare and Medicaid. Our results predict that roughly one in seven of today's elderly individuals will develop dementia before death. The average age of onset will be approximately 80; the typical patient will live approximately 5 years with dementia. The average per capita lifetime cost of dementia to Medicare is approximately $12,000 (in 2005 dollars) and to Medicaid about $11,000.
The magnitude of our cost estimates is much lower than the cost estimates from cross-sectional studies, especially for the Medicare costs (Hill et al. 2002). This is because the course of illness of dementia was found to have profound influence on the cost of care. Later dementia onset is associated with shorter duration of illness before death and lower health care costs. Our simulation predicts that increased educational levels could lead to longer life span and a higher percentage of future cohorts of elderly developing dementia before death. But this could also lead to shorter episodes of illness and overall lower lifetime Medicare and Medicaid costs per patient. However, increases in overweight and obesity prevalence will lead to both substantial increases in dementia cases and significant health care expenditure increases.
Our study demonstrates that it is crucial to take a longitudinal perspective in dementia research. Recently, NIH and Alzheimer's Association published the New Criteria and Guidelines for Alzheimer's Disease Diagnoses, which suggest that the entire course of illness of dementia could last two decades in three stages: preclinical, mild cognitive impairment, and Alzheimer's dementia. The guidelines advise treatment and prevention efforts be targeted at individuals in the preclinical stage (Clifford et al. 2011). With this type of intervention, the effect is inherently longitudinal, necessitating the use of longitudinal methods to understand the effect and/or cost-effectiveness of dementia screening and clinical interventions.
It is important to note that the algorithm we used to define dementia may not entirely precisely match the clinical definition of dementia. In our results, the point estimate of dementia prevalence in the MCBS sample is 6.4 percent, which is lower than found by most epidemiologic studies, which often report a prevalence rate of 10 percent or higher (Evans et al. 1990; Herbert et al. 2003; Plassman et al. 2007). Similarly, our simulation suggests 14 percent of the cohort born between the 1920s and 1930s will develop dementia, which is close to the point estimate from ADAMS of 13.9 percent, although the ADAMS sample only includes those age 70 or older. Both suggest underestimates of the true prevalence of dementia. Consequently, this could lead to underestimates of the cost of dementia. Possible reasons for the underestimates include a lack of knowledge of dementia diagnoses by the respondent that leads to under report of dementia history, and limitations of claims data of the events file that only the ICD-9 code of the top three reasons of hospital or outpatient care visits are recorded. It is also important to note that this study is a secondary data analysis to identify the relationship between population characteristics, dementia onset, and cost. Although our longitudinal simulation profiled the course of illness of dementia across the life span, we suggest caution in drawing a conclusion of causal relationship between education, obesity, dementia, and health care cost. Future studies with longitudinal data or randomized trial will answer such questions with better confidence.
Our study, however, does provide valuable insights into the policy debate regarding long- and short-term priorities to confront the challenge of the increasing trend of dementia. Over the long term, education is an important factor that alters the course of illness of dementia and reduces total costs. Therefore, the promotion of more education or other cognitively stimulating activities among American children and young adults may yield long-term public health benefits in the distant future. Over the shorter term, the promotion of healthy and active lifestyles among adults and older adults to reduce obesity rates may be the most viable strategy to reduce dementia costs. The U.S. health care system is going through rapid changes after the passage of the Patient Protection and Affordable Care Act. Some policy interventions targeting lifestyle change are under consideration. For example, the community-based YMCA version of Diabetes Prevention Program nationally may reduce Medicare spending by up to $50 billion across the entire Baby Boomer cohort by preventing chronic diseases related to obesity (Thorpe and Yang 2011). Added benefits from avoided dementia cases and shortened dementia episodes would further enhance the cost-effectiveness of such policies. In addition, proposals regarding Medicare and Medicaid financing reform and new service models are being discussed. Such suggestions include changing Medicare hospice care guidance to be based on patients' care needs instead of estimated life expectancy (Mitchell et al. 2010) and the proposal of Program for All-inclusive Care for the Elderly model to promote comprehensive care for the frail elderly (Chattopadhyay and Bindman 2010). Further research regarding the cost-effectiveness of these policy recommendations will provide valuable insights.
Acknowledgments
Joint Acknowledgment/Disclosure Statement: This study was supported by NIA Pilot Grant on Alzheimer's Disease research through Emory Alzheimer's Disease Research Center (P50 AG025688). We are grateful for the comments from the seminar participants at the Institute of Social Research at University of Michigan at Ann Arbor and two anonymous reviewers, who improved this paper significantly.
Disclosures: None.
Disclaimers: None.
SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article:
Appendix SA1: Author Matrix.
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.
References
- Alzheimer's Association. Incidence and Prevalence of the Major Causes of Brain Disorder. Chicago, IL: Alzheimer's Association; 2010a. [Google Scholar]
- Alzheimer's Association. Changing the Trajectory of Alzheimer's Disease: A National Imperative. Chicago, IL: Alzheimer's Association; 2010b. [Google Scholar]
- Ayyagari P, Salm M, Sloan F. “Effects of Diagnosed Dementia on Medicare and Medicaid Program Costs”. Inquiry. 2008;44:481–94. doi: 10.5034/inquiryjrnl_44.4.481. (Winter 2007/2008) [DOI] [PubMed] [Google Scholar]
- Brookmeyer R, Gray S. “Methods for Projecting the Incidence and Prevalence of Chronic Diseases in Aging Populations: Application to Alzheimer's Disease”. Statistics in Medicine. 2000;19(11–12):1481–93. doi: 10.1002/(sici)1097-0258(20000615/30)19:11/12<1481::aid-sim440>3.0.co;2-u. [DOI] [PubMed] [Google Scholar]
- Brookmeyer R, Corrada MM, Curriero FC, Kawas C. “Survival Following a Diagnosis of Alzheimer Disease”. Archives of Neurology. 2002;59(11):1764–7. doi: 10.1001/archneur.59.11.1764. [DOI] [PubMed] [Google Scholar]
- Chattopadhyay A, Bindman AB. “Linking a Comprehensive Payment Model to Comprehensive Care of Frail Elderly Patients”. Journal of the American Medical Association. 2010;304(17):1948–9. doi: 10.1001/jama.2010.1634. [DOI] [PubMed] [Google Scholar]
- Clifford R, Albert M, Knopman DS, McKhann G, Sperling R, Carrillo M, Thies B, Phelps C. “Introduction to the Recommendations from the National Institute on Aging and the Alzheimer's Association Workshop on Diagnostic Guidelines for Alzheimer's Disease”. Alzheimer & Dementia. 2011;7(2):1–6. doi: 10.1016/j.jalz.2011.03.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Evans D, Scherr P, Cook N, Albert M, Funkenstein H, Smith L, Hevert L, Wetle T, Branch L, Chown M, Hennekens C, Taylor J. “Estimated Prevalence of Alzheimer's Disease in the United States”. Milbank Quartorly. 1990;68:267–89. [PubMed] [Google Scholar]
- Ferrucci L, Guralnik J. “Hospital Diagnoses, Medicare Changes, and Nursing Admissions in the Year When Older Persons Become Severely Disabled”. Journal of the American Medical Association. 1997;277(9):728–34. [PubMed] [Google Scholar]
- Friedland R. “Epidemiolog, Education, and the Cology of Alzheimer's Disease”. Neurology. 1993;43:246–9. doi: 10.1212/wnl.43.2.246. [DOI] [PubMed] [Google Scholar]
- Gajital S. “One in Eight Baby Boomers Expected to Develop AD”. Neurology Today. 2008;20:20. [Google Scholar]
- Ganguli M, Hiroko H, Dodge H, Shen C, Pandav R, DeKosky S. “Alzheimer Disease and Mortality”. Archives of Neurology. 2005;62:779–84. doi: 10.1001/archneur.62.5.779. [DOI] [PubMed] [Google Scholar]
- Gatz M, Prescott C, Pedersen N. “Lifestyle Risk and Delaying Factors”. Alzheimer Disease and Associated Disorders. 2006;20:S84–8. doi: 10.1097/00002093-200607001-00013. [DOI] [PubMed] [Google Scholar]
- Goldman DP, Joyce GF. “Medicare Part D — A Successful Start with Room for Improvement”. Journal of the American Medical Association. 2008;299(16):1954–5. doi: 10.1001/jama.299.16.1954. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Goldman DP, Shekelle P, Bhattacharya J, Hurd M, Joyce G, Lakdawalla D, Matsui DH, Newberry SJ, Panis C, Shang B. Health Status and Medical Treatment of the Future Elderly. Santa Monica, CA: RAND Corporation; 2004. [Google Scholar]
- Herbert L, Scherr P, Bienias J, Bennet D, Evans D. “Alzheimer's Disease in the U.S. Population: Prevalence Estimates Using 2000 Census”. Archives of Neurology. 2003;60:1119–22. doi: 10.1001/archneur.60.8.1119. [DOI] [PubMed] [Google Scholar]
- Hill JW, Futterman R, Duttagupta S, Mastey V, Lloyd JR, Fillit I. “Alzheimer's Disease and Related Dementias Increase Costs of Comorbidities in Managed Medicare”. Neurology. 2002;58(1):62–70. doi: 10.1212/wnl.58.1.62. [DOI] [PubMed] [Google Scholar]
- Katzman R. “Education and the Prevalence of Dementia and Alzheimer's Disease”. Neurology. 1993;43:12–20. doi: 10.1212/wnl.43.1_part_1.13. [DOI] [PubMed] [Google Scholar]
- Lakdawalla DN, Goldman DP, Shang BP. “The Health and Cost Consequences of Obesity among the Future Elderly”. Health Affairs. 2005;24(6):W5R30–41. doi: 10.1377/hlthaff.w5.r30. [DOI] [PubMed] [Google Scholar]
- Lamb V, Sloan F, Nathan A. “Dementia and Medicare at Life's End”. Health Services Research. 2008;43(2):714–32. doi: 10.1111/j.1475-6773.2007.00787.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Langa K, Larson EB, Karlawish JH, Cutler DM, Kabeto MU, Kim HM, Rosen AB. “Trends in the Prevalence and Mortality of Cognitive Impairment in the United States: Is There Evidence of Compression of Cognitive Morbidity?”. Alzheimers & Dementia. 2008;4:134–44. doi: 10.1016/j.jalz.2008.01.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Larson EB, Shadlen M-F, Wang L, McCormick WC, Bowen JD, Teri L, Kukull WA. “Survival after Initial Diagnosis of Alzheimer Disease”. Annals of Internal Medicine. 2004;140(7):501–9. doi: 10.7326/0003-4819-140-7-200404060-00008. [DOI] [PubMed] [Google Scholar]
- Lin P, Kaufer D, Maciejewski M, Ganguly R, Paul J, Biddle A. “An Examination of Alzheimer's Case Defination Using Medicare Claims and Survey Data”. Alzheimer's and Dementia. 2009;6(4):334–41. doi: 10.1016/j.jalz.2009.09.001. [DOI] [PubMed] [Google Scholar]
- Manning W, Newhouse J, Duan N, Keeler EB, Leibowitz A, Marquis M. “Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment”. American Economic Review. 1987;77(3):251–77. [PubMed] [Google Scholar]
- Mitchell SL, Miller SC, Teno JM, Kiely DK, Davis RB, Shaffer ML. “Prediction of 6-Month Survival of Nursing Home Residents with Advanced Dementia Using ADEPT vs. Hospice Eligibility Guidelines”. Journal of the American Medical Association. 2010;304(17):1929–35. doi: 10.1001/jama.2010.1572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Philip D, Zimmerman S, Suchindran C, Reed P, Wang L, Boustani M, Sudha S. “The Public Health Impact of Alzheimer's Disease, 2000-2005: Potential Implication of Treatment Advances”. Annual Review of Public Health. 2002;23:213–31. doi: 10.1146/annurev.publhealth.23.100901.140525. [DOI] [PubMed] [Google Scholar]
- Plassman B, Langa K, Fisher G, Heeringa S. “Prevalence of Dementia in the United States: The Aging, Demographyics, and Memory Study”. Neuroepidemiology. 2007;29:125–32. doi: 10.1159/000109998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stern Y, Gurland B, Tatemichi T, Tang M, Wilder D, Mayeux R. “Influence of Education and Accupation on the Incidence of Alzheimer's Disease”. Journal of the American Medical Association. 1994;271:1004–10. [PubMed] [Google Scholar]
- Taylor D, Sloan F. “How Much Do Persons with Alzheimer's Disease Cost Medicare?”. Journal of the American Geriatrics Society. 2000;48:639–46. doi: 10.1111/j.1532-5415.2000.tb04721.x. [DOI] [PubMed] [Google Scholar]
- Taylor D, Ostbye T, Langa K, Plassman BL. “The Accuracy of Medicare Claims as an Epidemiological Tool: The Case of Dementia Revisited”. Journal of Alzheimer's Disease. 2009;17:807–15. doi: 10.3233/JAD-2009-1099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Thorpe K, Yang Z. “Enrolling People with Pre-Diabetes Ages 60-64 in a Proven Weight Loss Program Could Yield $7 Billion or More in Medicare Savings”. Health Affairs. 2011;30(9):1673–9. doi: 10.1377/hlthaff.2010.0944. [DOI] [PubMed] [Google Scholar]
- Xu WL, Atti AR, Gatz M, Pedersen N, Johansson B, Fratiglioni L. “Midlife Overweight and Obesity Increase Late-Life Dmentia Risk”. Neurology. 2011;76:1568–74. doi: 10.1212/WNL.0b013e3182190d09. (May 3) [DOI] [PMC free article] [PubMed] [Google Scholar]
- Yang Z, Hall AG. “Financial Burden of Overweight and Obesity among Elderly Americans: The Dynamics of Weight, Longevity, and Health Care Cost”. Health Services Research. 2008;43(3):849–67. doi: 10.1111/j.1475-6773.2007.00801.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.