Estimating Life Expectancy and Lifetime Healthcare Costs for Alzheimer’s Disease in Taiwan: Does the Age of Disease Onset Matter?

Abstract

Background:

People with early onset Alzheimer’s disease (EOAD) seem to suffer greater impact. But there is a lack of population-based studies on loss of life expectancy (LE) and lifetime healthcare costs.

Objectives:

We conducted this study to estimate LE, expected years of life lost (EYLL), and lifetime healthcare costs for Alzheimer’s disease (AD) in Taiwan stratified by onset age and gender, using a method which integrates the product of the survival function and the mean cost function over a lifetime horizon.

Methods:

We linked the National Health Insurance datasets with the National Mortality Registry and extrapolated the survival to lifetime to estimate the mean cumulative costs since the date of the first AD diagnosis using medical claims between 2001 and 2012.

Results:

A total of 21,615 mild to moderate AD patients (including 20,358 late-onset (LOAD) and 1,257 EOAD) were recruited. The average onset age for EOAD was 61 years old, while that of LOAD was 78. Although the LE of EOAD was 4.8 years longer than that of LOAD due to younger age, the EYLL for the former was 8.7 years versus 1.7 years for the latter. EOAD also had higher lifetime healthcare costs than the LOAD group (USD$37,957±2,403 versus 33,809±786).

Conclusions:

Since EOAD patients had both higher EYLL and lifetime healthcare costs than LOAD, future studies should pay more attention to the needs of EOAD patients.

Keywords

Early-onset Alzheimer’s disease healthcare costs late-onset Alzheimer’s disease life expectancy

INTRODCTION

Alzheimer’s disease (AD), the most common cause of dementia, is one of the greatest health care challenges of the 21st century. It is a neurodegenerative disorder presented with progressive cognitive decline, functional dependence, and eventually death. With the global population aging rapidly, the prevalence of AD continues to rise [1]. It not only has great impact on the patients themselves, but the long-term disability caused by the disease also poses great burden to families, and the whole society.

Clinically, AD is categorized into early onset Alzheimer’s disease (EOAD) and late onset Alzheimer’s disease (LOAD) depending on the age of onset before or after 65. EOAD is distinct from LOAD in its pathophysiology and disease trajectory [2]. However, controversies exist on whether or not the age of disease onset is related to mortality. Although patients are usually at a better fitness level and with less comorbidity at the time of disease onset, many reported that EOAD is associated with a more malignant course [3 –5]. Other studies indicate that the survival in AD is independent of age of onset [6, 7]. As reported in previous literature, life expectancy (LE) for LOAD after starting cholinesterase inhibitor is 4.95 years in males and 6.05 years in females; while that of the EOAD group is 6.21 in males and 6.83 years in females [8]. Another study from a Korean medical center found that the median survival after diagnosis for EOAD and LOAD were 14.7 and 11.3 years, respectively [9]. In a propensity score matched study, Chang et al. demonstrated that EOAD patients have a two-time higher risk of mortality compared to LOAD patients [10]. However, data on the number of years of shortened LE among EOAD and LOAD patients compared with the general population are limited. A literature review by Brodaty et al. concluded that the loss of LE or expected years of life loss (EYLL) in EOAD and LOAD patients ranged from 9.6–19.4 and 1.3–9.2 years, respectively, with the relative loss of remaining LE for EOAD being 60–94% versus 16–73% for LOAD [11].

The economic burden of AD is another major concern of the disease. The cost of AD is huge and is increasing dramatically. In the US, while the annual cost per patient for direct medical care was estimated to be 7,929 USD [12], the average total cost (which includes formal and informal costs) per AD patient ranged from 27,700–47,000 USD per year in the 1990 s [13, 14] and rose to 41,000–56,000 USD in 2010 [15]. Another cohort of the Medicaid record suggested a similar increment in health care costs, with average Medicaid and informal care costs reaching 35,346 USD and 83,022 USD, respectively, between 2006 to 2010 for dementia patients [16]. In China, the annual direct medical cost per AD patient was 708 USD in 2007 [17] and the total cost (i.e., including direct medical, direct nonmedical, and indirect costs) per patient per year in 2015 was estimated to be 19,144 USD [18]. In Turkey, the annual direct medical cost per AD patient was 1,620–2,391 USD [19]. A detailed analysis of the disease-related costs also revealed gender differences. In a US study, the economic burden for female patients were higher not only in health care related costs, but also in informal costs as they tended to stay in nursing homes rather than in the community [20]. As AD prevalence increases over time, its medical costs approximately doubles every 2 years [21 –23]. To date, AD has been listed as the most expensive disease in America [24]. It is expected that the direct cost of AD care in America will increase to more than 1.2 trillion USD by 2050 [24]. In Asian countries, its costs are also on the rise. Our previous study showed that the annual total cost per person with dementia in Taiwan ranged from $NT 218,644 to $NT 439,972 (1NTD = 14.97 USD-PPP in 2014), and the costs increase with disease severity [25]. However, to our knowledge, no large scale population-based studies on the lifetime costs of AD have been conducted. Even less attention has been drawn to the costs of EOAD despite a seemingly greater health impact on these patients and their families [26, 27].

LE is important to clinicians, the patients, and their families when communicating prognosis and planning future needs. A detailed understanding of the lifetime costs for these patients is crucial for resource allocation and financial preparation. Therefore, we conducted this population-based retrospective cohort analysis to estimate LE, EYLL, and lifetime healthcare expenditures (HE) in EOAD and LOAD to provide insight into these important and distinct clinical subtypes of disease.

MATERIALS AND METHODS

Data and sample

This study was conducted using data from Taiwan’s National Health Insurance Research Database (NHIRD) from 2000 to 2012. The Institutional Review Board of National Cheng Kung University Hospital approved this study before commencement. The National Health Insurance (NHI) system of Taiwan was established in 1995 and has covered over 99% of Taiwan’s 23 million people since 2009 [28]. Figure 1 depicts the process of sample selection. First, patients with dementia were defined as those with at least two outpatient claims or one inpatient admission with ICD-9-CM codes 290, 294.1, or 331 in any year during 2002 and 2012. In order to identify incident cases of AD, we first excluded those diagnosed with dementia in the past two years prior to entering the sample since 2002. Next, we limited our AD sample to cholinesterase inhibitor (ChEI) users with the following two exclusion criteria: 1) Persons diagnosed with Parkinson’s disease (ICD-9-CM code 332) or vascular dementia (ICD-9-CM 290.4) in two outpatient or one inpatient claim; 2) Persons who did not use ChEIs within 365 days after their first diagnosis of dementia. ChEIs considered in our study included both rivastigmine and galantamine prescribed during 2002–2012 and donepezil prescribed during 2002–2010. According to the insurance reimbursement of Taiwan’s NHI, only mild to moderate AD patients with Mini-Mental State Examination scores between 10 and 26 as well as Clinical Dementia Rating scores from 1 to 2 were eligible for ChEI reimbursement during those periods. In other words, our sample of incident AD patients did not include persons who were diagnosed at a late or severe stage of the disease. Moreover, since the pre-approval of ChEIs prescription required clinical neurologists or psychiatrists to provide evidence of clinical symptoms and signs, blood tests, cognitive tests, and neuroimage work-ups, such as brain computed tomography (CT) or magnetic resonance imaging (MRI) of eligible AD patients, limiting our study sample to ChEI users would significantly reduce the possibility of miscoding common to other studies with claims data [29]. Because our National Health Insurance system have waived copayments for patients diagnosed with catastrophic illnesses including dementia, the likelihood that AD patients who are eligible for ChEI but do not use it due to economic constraints would be very low. The fourth criterion was to exclude those under 40 years old or whose sex or age information was missing. The above process yielded a final total sample of 21,615 persons, including 1,257 EOAD patients whose onset age was between 40 to 64 years and 20,358 LOAD patients who had their first diagnosis of AD at above 65 years old.

Fig.1

Flow diagram of sample selection of patients with Alzheimer’s disease (AD). ¹Dementia claims with the following ICD-9-CM diagnosis codes: 290, 294, 331. ²Parkinson’s disease: 332. ³Vascular dementia: 290. ChEI: cholinesterase inhibitors.

Since our data included incident AD cases that were diagnosed anytime during 2002 to 2012, the starting time of follow-up varied for each person, but all AD patients were followed up until either death or December 31, 2012, whichever came first. The mean follow-up time were 8.2 years for the EOAD group and 6.8 years for the LOAD group (Table 1). In addition to stratified analyses by onset age (EOAD versus LOAD), we also conducted stratified analyses to compare differences in survival and healthcare costs by gender.

Table 1

Demographic and clinical characteristics of patients with Alzheimer’s disease (AD) stratified by onset age

Characteristics	Early onset AD (N = 1,257)	Late onset AD (N = 20,358)
Age (mean, SE)	60.6 (4.4)	77.5 (6.4)
Number of deceased	139	5,554
Censored rate (%)	88.9	72.7
Mean survival in follow-up (SE)	8.2 (0.1)	6.8 (0.0)
Total healthcare expenditures in follow-up (USD$, SE)	20,787 (883)	24,073 (748)
Comorbidities (%)
Hypertension	30.2	51.3
Cardiovascular disease	12.2	19.7
Diabetes mellitus	20.7	23.5
Stroke	10.8	13.7
Cancer	3.1	6.0
Chronic kidney disease	2.4	5.0

HE measures

The HE in our study included direct medical costs but not direct non-medical, indirect costs, or long-term care spending. Direct medical costs incurred by patients with AD recorded in the National Health Insurance datasets contained cost of outpatient visits, emergency room visits, inpatient admission, and prescription drugs. Thus, our HE measures for AD patients were comprehensive and comparable to the literature in terms of direct medical costs [17].

Statistical analyses

Estimation of survival outcomes

We applied the Kaplan-Meier method to estimate the survival function for the follow-up period, and used Monte Carlo methods to generate survival data from an age-and-sex-matched reference population. To facilitate the estimation, we used the ISQoL2 software, a written package of the R program that can be downloaded freely from http://www.stat.sinica.edu.tw/isqol. The logit transformation of the relative survival was defined as the logit ratio between the survival functions of the AD cohort and the reference population, and because the rolling extrapolation of logit function was decreasing, we used the cubic splines model to extrapolate the survival function beyond the follow-up period until all patients in our sample died [30]. Thus, we estimated the lifetime survival and the area under the entire survival curve was the LE of the AD cohort. Next, we estimated the EYLL by subtracting the LE of the AD cases from that of the corresponding age-and-sex-matched reference population. In addition to a point estimate of LE, we obtained standard errors of LE and EYLL through a bootstrap method by implementing the extrapolation process with data simulated by 100 times of repeated sampling with replacements from the real dataset. For more details on the method used to estimate the lifetime survival and cost functions, interested readers can refer to another article published by Hwang and colleagues [31].

Estimation of HE outcomes

The lifetime HE of a patient refers to all direct healthcare costs paid by the NHI from the date of AD diagnosis until the date of death. The dates of death were retrieved from linkage with the National Mortality Registry of Taiwan. Since Taiwan’s National Health Insurance operates on a global budget system, we first converted the healthcare utilization points from the NHI database using quarterly point value tables to obtain actual healthcare cost estimates. Because medical expenditures are often higher during the period approaching the end of life, we used the rolling extrapolation survival-adjusted cost estimator that accounted for this aspect to construct a more accurate estimate of lifetime HE. Existing literature in health economics has shown that the mean aggregated costs in a time interval can be represented as a weighted average of the mean costs of subjects who died in the time interval and the expected costs of other subjects who remained alive in that interval [32].

To determine the appropriate time interval to examine increasing costs of end-of-life care, we drew a time series plot of the mean expenditure of subjects in K months prior to their death (red line) and the mean monthly costs for patients who would not die in the next 36 months (as shown in the black dashed line of Supplementary Figure 1). Because an increase in healthcare costs about 24 months before mortality was found in all four situations, we chose 24 months or the last two years of follow-up to estimate the monthly costs prior to death. After choosing these parameters, estimation of the expected lifetime costs involves two parts: (1) the survival function was estimated from the rolling extrapolation algorithm, and (2) the monthly mean cost function was estimated using a survival-adjusted cost estimator [31].

After estimating the lifetime HE for the AD cohort, we also obtained the estimates of HE per life year by dividing the total lifetime HE by the years of LE. All HE estimates were discounted at a rate of 3% following the recommendation of the US 2nd Panel on Cost Effectiveness [33] and converted to 2012 USD (Exchange rate 1 USD = 30 NTD). We used the SAS software, V.9.2 (SAS Institute Inc.) for data management and the iSQoL2 software for analyses involving rolling extrapolation.

RESULTS

Table 1 summarizes the demographic and clinical characteristics of 21,615 AD patients stratified by onset age. There was a 17-year difference in the average onset age between the two groups (61 versus 78 years old). The early-onset group had both higher survival and higher censored rates during follow-up than the late-onset group. The mean survival of the early-onset patients was 8.18 years after diagnosis, as compared with 6.81 years in the late-onset group. However, the average total HE during the 11 years of follow-up was higher in the late-onset group compared with that of the early-onset group ($24,073 versus $20,787). A comparison of the six common co-morbidities including hypertension and diabetes also showed that the late-onset group had a much higher prevalence of co-morbidities across all six diseases.

Table 2 illustrates the LE, EYLL, lifetime HE (in USD$), and HE per life year. Early onset cases had longer survival after the diagnosis of the disease than the late-onset cohort. However, considering the fact that the age of diagnosis was 17 years younger in the early onset group, they had an EYLL of 8.7 years compared with matched referents, which was over 5 times greater than the late-onset group. This implies that the early-onset cohort deserves more attention. The standard error of LE and EYLL of the two groups were no more than 1 year (1.0 and 0. 1), indicating relatively precise estimations. The lifetime HE estimates for the early-onset and the late-onset groups were $37,957 and $33,809, respectively. The early-onset group had a higher lifetime HE than the late-onset group, because the former had a longer LE. The HE per life year was still lower for the early-onset group compared with the late-onset cohort ($2,855 versus $3,814). The standard error in lifetime HE of the early-onset group ($2,403) was about three times higher than that of the late-onset group ($786), which may have resulted from a higher censored rate and lower number of deceased patients in the smaller cohort.

Table 2

Life expectancy (LE), expected years of life lost (EYLL), and healthcare expenditures (HE) of Alzheimer’s disease (AD) stratified by onset age or gender

Sample	Sample size	LE	EYLL	Lifetime HE with 3% discount rate	HE per life year with 3% discount rate
		Mean (SE) years	Mean (SE) years	Mean (95% CI) USD	Mean (95% CI) USD
Early onset AD	1,257	13.8 (0.9)	8.7 (1.0)	37,957 (35,554–40,360)	2,855 (2,714–2,966)
Late onset AD	20,358	9.0 (0.1)	1.7 (0.1)	33,809 (33,023–34,595)	3,814 (3,722–3,906)
Male	8,445	8.4 (0.1)	1.6 (0.2)	34,777 (33,663–35,891)	4,210 (4,088–4,332)
Female	13,170	10.0 (0.2)	2.1 (0.2)	33,631 (32,554–34,708)	3,416 (3,302–3,530)

Figure 2 provides a visual display of the EYLL by age of onset. The shaded area between the survival curves of EOAD patients and the age and sex-matched referents was much greater in size than that between the survival curves of the LOAD patients and the reference population, suggesting a greater loss of LE in the EOAD compared to the LOAD cohort. When stratified by gender, there was a greater loss of LE in female patients than male patients. Looking at the difference between the upper and lower panels, we can also conclude that among our AD population, the difference in EYLL was greater by onset age than by gender.

Fig.2

Lifetime survival curves compared with age- and sex-matched referents simulated from national vital statistics. The shadowed areas indicate the differences of life expectancies, or, expected years of life loss. Panel (a) represents subgroup of early-onset Alzheimer’s disease (AD); panel (b) represents subgroup of late-onset AD; panels (c) and (d) represent lifetime survival of AD patient males and females, respectively.

Table 2 also summarizes the differences in outcomes of survival and healthcare costs from stratified analyses by gender of AD patients. Females not only had higher LE (10.0 versus 8.4) but also had higher EYLL (2.1 versus 1.6) than their age-matched males. On the other hand, the lifetime HE were higher for males compared with females ($34,777 versus $33,631). Similarly, AD males had higher HE per life year than their female counterparts ($4,210 versus $3,416).

DISCUSSION

Before discussing the implications of our findings, we will first corroborate the validity of our estimations with the following arguments: First, as the LE of patients with AD would usually be less than 10 years except those with EOAD (Table 2), the bias toward the end of life for most AD patients would be mild or minimal given that we had 11 years of follow-up. Second, relative bias was calculated by entering the first 7 years of data to project another 4 years and by using Kaplan-Meier estimates of 11 years as the gold standard, we showed that the relative bias was below 4%. Namely, the relative bias was calculated by dividing the difference between the extrapolated estimate and the gold standard, using the latter as the denominator. We also conducted a sensitivity analysis and found that if we increased the use of actual data from 7 to 8 years, the estimated relative bias would be below 2%. Thus, we could reasonably anticipate that the bias would be even smaller when we applied 11 years of follow-up data to project lifetime survival. Third, since this study excluded all prevalent cases and patients with vascular dementia, and our AD sample was selected from ChEI users only, the accuracy of AD diagnoses among our study sample was increased so that our results could truly represent incident AD cases that were diagnosed at mild or moderate stage. Thus, the risk of mis-diagnosis of AD would be minimal in our study. Finally, our sample size of 1,257 people with EOAD was generally bigger than almost all of the previous studies [11]. The only one that also included more than 1,000 EOAD patients was conducted by Rait and colleagues in the U.K. Their EYLL estimate of 11.03 for EOAD was somewhat bigger than our EYLL estimate of 8.7 with mostly mild to moderate severity, but in both studies, greater EYLL was associated with EOAD rather than LOAD [11, 32]. Therefore, we can conclude that our estimates would be reasonably accurate.

As previous studies have found that the relative loss of remaining LE was greater in EOAD patients than that of LOAD [11], our study has come to the same conclusion with 8.7 years of EYLL (39% relative loss) for EOAD and 1.7 years (16% relative loss) for LOAD. However, our findings seem to be lower for EOAD (60–94%) and closer to the lower bound for LOAD (16–73%), because our selected sample of incident AD patients did not include persons who were diagnosed at a late or severe stage of the disease, which implies a possible under-estimation of overall EYLL and relative loss of LE. Moreover, we found that although EOAD patients had longer survival after disease onset due to their younger age, their EYLL was 5 times greater than that of LOAD. The LE for EOAD and LOAD in our study were relatively close to the LE reported in a Korean cohort: 13.8 versus 14.7 years for EOAD and 9.0 versus 11.3 years for LOAD [9].

In addition to a greater loss in survival, we also found that EOAD had higher lifetime healthcare costs than the LOAD group: the lifetime healthcare costs were 37,957 USD for EOAD and 33,809 USD for LOAD. While total healthcare costs were higher for LOAD than EOAD during the years between AD onset and the end of follow-up, after using the survival-adjusted cost estimator for extrapolation, lifetime healthcare costs became higher for EOAD than LOAD, and one plausible explanation could be due to EOAD’s longer LE, so even though annual healthcare costs was lower for EOAD, cumulative healthcare costs over a lifetime would be reversed for the two groups. The HE per life year per person were 2,855 USD for EOAD and 3,814 USD for LOAD in our cohort. These annual direct healthcare costs found in our study were lower than that reported in the US (7,929 USD, study participants mainly mild to moderate AD cases) [12], but higher than that reported in China (708 USD, study participants ranging from mild to severe AD cases) [17] and Turkey (ranging from 1,620 USD among the mild AD cases to 2,391 USD among the severe AD cases) [19]. Since the costs for AD increase with disease severity [12, 25], our results may be an underestimation because we only included incident AD cases that were diagnosed at mild or moderate stage.

Our finding that care costs for dementia were higher for male than female patients was consistent with a recent study of 19 major cancers in Taiwan, which showed that for most cancers including liver and lung cancer, the healthcare costs per life year were also higher among males than females [34]. However, our results were different from another US study that found women had higher lifetime costs of AD, but that study also included costs of nursing home and costs of informal caregiving that were not available in our data [20]. The US Medicaid data on dementia patients also revealed an increased Medicaid spending in non-married females, while the total social costs were generally higher in both married and non-married females [16]. We therefore postulate that such a difference in results may be related to different socio-economic and cultural environment between Taiwan and the US, and that our study did not account for long-term care spending which are highly associated with female longevity.

Compared to the large differences found in both survival and healthcare costs between the two groups stratified by their onset age, we found the differences to be much smaller among female and male AD patients. In summary, there seems to be a greater disparity by onset age than by gender in survival and healthcare costs among AD patients.

There are several limitations to this study. First, by selecting our study sample of AD patients from ChEI users only, we inherently chose internal validity over external validity. In other words, our results are only applicable to AD patients who are eligible for ChEI use but not to all AD patients. Due to the NHI’s reimbursement criteria for ChEI during the study period, severe AD patients were not included in our study sample. Moreover, in a sensitivity analysis, we compared two sets of results estimated by AD samples defined by ICD diagnosis only and by ChEI use, and found similar results in LE, EYLL, and lifetime expenditures across the two late onset AD samples as shown in Supplementary Table 1. Specifically, compared to the ChEI use sample, there was a 1.3-year reduction in LE and a 0.7-year increase in EYLL for patients with diagnosis only. The difference was likely due to our exclusion of severe AD patients with no prescription of ChEI and thus were found with lower LE and higher EYLL. There was also a small difference of USD $2,446 in lifetime HE across the two groups of AD patients. Nevertheless, this choice significantly increased the accuracy of AD diagnoses among our study sample so that our results could accurately represent incident AD cases that were diagnosed at mild or moderate stage. Second, our lifetime HE cost estimates could be considered as lower bounds since our study included only healthcare costs paid by the NHI, but previous studies have shown that social and informal care costs usually account for a large part of the total care costs, especially for EOAD patients [27]. Therefore, future studies are needed to collect out-of-pocket expenditures and care hours spent by family with dementia in order to obtain a more comprehensive estimate of the lifetime costs of dementia. Third, following the method recommended by the World Alzheimer Report, we used the total costs rather than the net costs approach [35], so our cost estimates are not directly attributable to dementia and include effects from other comorbid conditions.

In conclusion, the contribution of this study lies in using a large national sample to extrapolate LE, EYLL, and HE over the lifetime horizon for AD patients. EOAD and LOAD patients had a loss of 13.8 and 9.0 years of LE, and 8.7 and 1.7 EYLL, respectively. Given that previous studies on EOAD usually had very small sample sizes, we offer evidence from a large-scale population-based study on the lifetime costs of AD and highlight the differences in health outcomes between EOAD versus LOAD, with mean lifetime expenditures of USD 37,957 and 33,809, respectively.

Footnotes

ACKNOWLEDGMENTS

This work was supported by grant MOST 105-2628-B-006-011. The authors would like to thank Ms. Tsuey-Hwa Hu for her technical assistance and statistical consultation on the ISQoL2 software.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Livingston

, Sommerlad

, Orgeta

, Costafreda

, Huntley

, Ames

, Ballard

, Banerjee

, Burns

, Cohen-Mansfield

, Cooper

, Fox

, Gitlin

, Howard

, Kales

, Larson

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbaek

, Teri

, Mukadam

(2017) Dementia prevention, intervention, and care. Lancet 390, 2673–2734.

Mendez

(2017) Early-onset Alzheimer disease. Neurol Clin 35, 263–281.

Ueki

, Shinjo

, Shimode

, Nakajima

, Morita

(2001) Factors associated with mortality in patients with early-onset Alzheimer’s disease: A five-year longitudinal study. Int J Geriatr Psychiatry 16, 810–815.

Koedam

, Pijnenburg

, Deeg

, Baak

, van der Vlies

, Scheltens

, van der Flier

(2008) Early-onset dementia is associated with higher mortality. Dement Geriatr Cogn Disord 26, 147–152.

Barclay

, Zemcov

, Blass

, McDowell

(1985) Factors associated with duration of survival in Alzheimer’s disease. Biol Psychiatry 20, 86–93.

Newens

, Forster

, Kay

, Kirkup

, Bates

, Edwardson

(1993) Clinically diagnosed presenile dementia of the Alzheimer type in the Northern Health Region: Ascertainment, prevalence, incidence and survival. Psychol Med 23, 631–644.

Kay

, Forster

, Newens

(2000) Long-term survival, place of death, and death certification in clinically diagnosed pre-senile dementia in northern England. Follow-up after 8-12 years. Br J Psychiatry 177, 156–162.

Wattmo

, Londos

, Minthon

(2014) Risk factors that affect life expectancy in Alzheimer’s disease: A 15-year follow-up. Dement Geriatr Cogn Disord 38, 286–299.

, Lee

, Seo

, Chin

, Kang

, Moon

, Na

, Cheong

(2013) Survival of Alzheimer’s disease patients in Korea. Dement Geriatr Cogn Disord 35, 219–228.

10.

Chang

, Hong

, Lee

, Kang

, Lee

, Choi

, Kim

, Na

, Seo

, Kim

, Lee

, Chung

, Lim

, Noh

, Park

, Son

(2017) Mortality risk after diagnosis of early-onset Alzheimer’s disease versus late-onset Alzheimer’s disease: A propensity score matching analysis. J Alzheimers Dis 56, 1341–1348.

11.

Brodaty

, Seeher

, Gibson

(2012) Dementia time to death: A systematic literature review on survival time and years of life lost in people with dementia. Int Psychogeriatr 24, 1034–1045.

12.

Zhu

, Scarmeas

, Torgan

, Albert

, Brandt

, Blacker

, Sano

, Stern

(2006) Clinical features associated with costs in early AD: Baseline data from the Predictors Study. Neurology 66, 1021–1028.

13.

Leon

, Cheng

, Neumann

(1998) Alzheimer’s disease care: Costs and potential savings. Health Aff (Millwood) 17, 206–216.

14.

Rice

, Fox

, Max

, Webber

, Lindeman

, Hauck

, Segura

(1993) The economic burden of Alzheimer’s disease care. Health Aff (Millwood) 12, 164–176.

15.

Hurd

, Martorell

, Delavande

, Mullen

, Langa

(2013) Monetary costs of dementia in the United States. N Engl J Med 368, 1326–1334.

16.

Kelley

, McGarry

, Gorges

, Skinner

(2015) The burden of health care costs for patients with dementia in the last 5 years of life. Ann Intern Med 163, 729–736.

17.

Wang

, Cheng

, Zhang

, Bai

, Zeng

, Cui

, Zhang

, Sun

, Ren

, Deng

, Xu

, Wang

, Chen

(2008) Economic impact of dementia in developing countries: An evaluation of Alzheimer-type dementia in Shanghai, China. J Alzheimers Dis 15, 109–115.

18.

Jia

, Wei

, Chen

, Li

, Tang

, Qin

, Zhao

, Jin

, Xu

, Wang

, Zhou

, Zuo

, Wu

, Han

, Huang

, Wang

, Li

, Chu

, Shi

, Gong

, Du

, Zhang

, Zhou

, Lv

, Xie

, Ji

, Li

, Yu

, Luo

, Wang

, Yang

, Qu

, Guo

, Liang

, Zhang

, Tan

, Shen

, Zhang

, Peng

, Tang

, Lv

, Fang

, Chu

, Jia

, Gauthier

(2018) The cost of Alzheimer’s disease in China and re-estimation of costs worldwide. Alzheimers Dement 14, 483–491.

19.

Zencir

, Kuzu

, Beser

, Ergin

, Catak

, Sahiner

(2005) Cost of Alzheimer’s disease in a developing country setting. Int J Geriatr Psychiatry 20, 616–622.

20.

Yang

, Levey

(2015) Gender differences: A lifetime analysis of the economic burden of Alzheimer’s disease. Womens Health Issues 25, 436–440.

21.

Chan

, Cham

, Lin

(2009) Direct medical costs in patients with Alzheimer’s disease in Taiwan: A population-based study. Curr Ther Res Clin Exp 70, 10–18.

22.

Wimo

, Jonsson

, Winblad

(2006) An estimate of the worldwide prevalence and direct costs of dementia in 2003. Dement Geriatr Cogn Disord 21, 175–181.

23.

Wimo

, Winblad

, Jonsson

(2007) An estimate of the total worldwide societal costs of dementia in 2005. Alzheimers Dement 3, 81–91.

24.

Alzheimer’s Association (2017) Costs of Alzheimer’s to Medicare and Medicaid.

25.

, Pai

, Shih

(2016) Economic impact of dementia by disease severity: Exploring the relationship between stage of dementia and cost of care in Taiwan. PLoS One 11, e0148779.

26.

Werner

, Stein-Shvachman

, Korczyn

(2009) Early onset dementia: Clinical and social aspects. Int Psychogeriatr 21, 631–636.

27.

Sansoni

, Duncan

, Grootemaat

, Capell

, Samsa

, Westera

(2016) Younger onset dementia. Am J Alzheimers Dis Other Demen 31, 693–705.

28.

Lee

Y-C

, Huang

Y-T

, Tsai

Y-W

, Huang

S-M

, Kuo

, McKee

, Nolte

(2010) The impact of universal National Health Insurance on population health: The experience of Taiwan. BMC Health Serv Res 10, 225.

29.

Sun

, Lai

, Lu

, Chen

(2008) How long can patients with mild or moderate Alzheimer’s dementia maintain both the cognition and the therapy of cholinesterase inhibitors: A national population-based study. Eur J Neurol 15, 278–283.

30.

Andersson

TML

, Dickman

, Eloranta

, Lambe

, Lambert

(2013) Estimating the loss in expectation of life due to cancer using flexible parametric survival models. Stat Med 32, 5286–5300.

31.

Hwang

, Hu

, Lee

, Wang

(2017) Estimating lifetime medical costs from censored claims data. Health Econ 26, e332–e344.

32.

Basu

(2010) Estimating lifetime or episode-of-illness costs under censoring. Health Econ 19, 1010–1028.

33.

Sanders

, Neumann

, Basu

, Brock

, Feeny

, Krahn

, Kuntz

, Meltzer

, Owens

, Prosser

, Salomon

, Sculpher

, Trikalinos

, Russell

, Siegel

, Ganiats

(2016) Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: Second panel on cost-effectiveness in health and medicine. JAMA 316, 1093–1103.

34.

T-Y

, Chung

C-H

, Lin

C-N

, Hwang

J-S

, Wang

J-D

(2018) Lifetime risks, loss of life expectancy, and health care expenditures for 19 types of cancer in Taiwan. Clin Epidemiol 10, 581–591.

35.

Martin

, Anders

, Maëlenn

(2015) World Alzheimer Report 2015: The Global Impact of Dementia. Alzheimer’s Disease International, London.