As there is an increasing trend in rates of dementia when aging population has been increasing but awareness is low, considering an effective and efficient
community based dementia screening programs is of paramount important. This has been upheld by the current findings on low awareness of dementia has been
ascertained in routine health insurance health care system as the P/I ratio of community-based survey (active detection method) was greater than that of health insurance heath care system (passive detection method). The results regarding the validity of the accuracy in the early diagnosis of AD8 and MMSE shows AD8 had higher sensitivity but lower specificity in detection of dementia than MMSE when both tools were considered in population- and community-based screening. The optimal sensitivity and specificity of MMSE could be achieved using the optimal cut-off based on empirical data. Compared with the use of AD8 or MMSE alone, the
combination of AD8 with MMSE with the parallel mode enhanced the sensitivity whereas that with the serial mode enhanced the specificity. Community-based screening for dementia with AD8 and MMSE is more cost-effective and almost near cost-saving compared with no screening program. The most economic screening strategy is the parallel mode of combining AD8 with MMSE. There are two three novelty of the methodological development including (1) a P/I-ratio-based regression model under the framework of generalized linear regression model to estimate the P/I ratio of active detection method in comparison with that of passive detection method after controlling for confounding factors; (2) the application of Bayesian model for multiple detection modalities in the absence of gold standard for estimating posterior distributions of sensitivity and specificity by combing prior information on the use of AD8 and MMSE with the likelihood based on the current empirical data; and (3) the development Markov cycle decision tree model for economic evaluation of
population-based screening program AD8 in combination with MMSE.
a Table 3.2.1 Distribution assigned for parameters in probabilistic cost-utility analysis
Parameters Base case estimate Distribution
Prevalence90
Mild dementia 5.88%
Moderate dementia 1.13%
Severe dementia 1.13%
Screening Interval Every 5 year
AD8
Sensitivity 84.1% Beta(14.3, 2.7)
Specificity 77.1 Beta(128.0, 38.0)
MMSE
Sensitivity 82.2% Beta(14.8, 3.2)
Specificity 86.1% Beta(142.9, 23.1)
Parallel Test
Sensitivity 97.2% Beta(18.5, 0.5)
Specificity 66.4% Beta(110.2, 55.8)
Sequential Test
Sensitivity 69.1% Beta(13.8, 6.2)
Specificity 96.8% Beta(160.7, 5.3)
Parameters Base case estimate Distribution
Transition probability2 ( Nature history )
Normal to mild114 0.047
Mild to moderate 0.25387
Mild to severe 0.0774
Mild to death 0.021
Moderate to mild 0.09773 Moderate to severe 0.44886
Moderate to death 0.053
Severe to death 0.153
Transition probability2, 115 (Post treatment)
Normal to mild2, 114 0.047
Mild to moderate 0.22054
Mild to severe 0.0774
Mild to death 0.021
Moderate to mild 0.09773 Moderate to severe 0.2467
Parameters Base case estimate Distribution
Moderate to death 0.053
Severe to death 0.153
Costs116
Medical cost of mild* 1,418 Triangular (706, 1418, 2836) Care cost of mild* 11,644 Triangular (5822, 11644, 23288) Medical cost of
moderate*
1,765
Triangular (883, 1765, 3530)
Care cost of moderate* 23,626 Triangular (11813, 23626, 47252) Medical cost of severe* 2,113 Triangular (1056, 2113, 4226) Care cost of severe* 37,262 Triangular (18613, 37262, 74524) Screening Cost
AD8 11.87
MMSE 10.68
Parallel 12.42
Sequential 10.97 QALY97
Mild 0.68 Beta (26.98, 12.69)
Moderate 0.54 Beta (24.45, 20.83)
Parameters Base case estimate Distribution
Severe 0.37 Beta (8.84, 15.06)
*All cost were in USD
b Table 4.1.1: Age and gender specific prevalence and incidence rate of dementia
Gender Age Prevalent
Case Subtotal 27965 910285 3.07 49865 2571999 1.94 1.58 All 65-69 5470 659042 0.83 11560 1943978 0.59 1.41
70-74 9450 582622 1.62 19325 1691100 1.14 1.42 75-79 13198 378169 3.49 25317 1056982 2.4 1.45 80-84 13045 188721 6.91 22107 493459 4.48 1.54 85-89 9142 84265 10.85 14009 203434 6.89 1.57 90+ 5577 28489 19.58 7291 56635 12.87 1.52 Total 55882 1921308 2.91 99609 5445586 1.83 1.59
Active Survey Data
65-79 11 135 8.15* 56202 4692060 1.2 6.79 80-90+ 6 51 11.76* 43407 753528 5.76 2.04 total 17 183 9.29* 99609 5445586 1.83 6.59
#P/I ratio= prevalence/incidence ratio
*Prevalence was estimated from other active survey study
c Table 4.1.2 Effects of Age, gender, geographic on the risk of incidence rate of dementia by Poisson regression model
Variable Univariate Multivariate
RR (95%CI) P-value RR (95%CI) P-value
Age <.0001 <.0001
65-69 1.00 1.00
70-74 1.02 (0.99-1.04) 1.01(0.99-1.04) 75-79 1.03 (1.01-1.06) 1.03(1.01-1.05) 80-84 1.05 (1.03-1.08) 1.05(1.03-1.07) 85-89 1.09 (1.06-1.12) 1.09(1.06-1.12) 90+ 1.14( 1.11-1.18) 1.14(1.11-1.18)
Gender 1.01 (1.00-1.02) 0.2003 1.01(1.00-1.03) 0.0450
Area 0.0057 0.0146
Central 1.00 1.00
North 1.02(1.00-1.04) 1.01(1.00-1.04 ) South 1.00(0.98-1.01) 0.99(0.98-1.02) East 1.03(0.99-1.07) 1.02(0.97-1.06)
d Table 4.1.3 Adjusted P/I ratios of dementia measured by passive survey
Intercept 0.197(0.145-0.248)
Age
65-69 baseline 1.35(1.29-1.41)
70-74 -0.12(-0.176-0.063) 1.20(1.15-1.24) 75-79 -0.091(-0.145-0.037) 1.23(1.19-1.27) 80-84 0.141(0.087-0.195) 1.55(1.51-1.60) 85-89 0.36(0.304-0.417) 1.93(1.86-2.00) 90+ 0.886(0.823-0.949) 3.27(3.13-3.41) Area
Central baseline 1.20(1.17-1.24)
North 0.186(0.15-0.222) 1.45(1.41-1.49) South 0.096(0.056-0.136) 1.32(1.29-1.37) East -0.322(-0.417-0.228) 0.87(0.80-0.96)
Female
Intercept 0.251(0.2-0.302)
Age
65-69 baseline 1.47(1.41-1.54)
70-74 0.137(0.082-0.191) 1.69(1.63-1.74) 75-79 0.152(0.1-0.205) 1.71(1.66-1.76) 80-84 0.032(-0.021-0.085) 1.52(1.47-1.57) 85-89 -0.148(-0.204-0.091) 1.27(1.22-1.32) 90+ -0.492(-0.557-0.428) 0.90(0.86-0.94) Area
Central baseline 1.37(1.32-1.41)
North 0.211(0.175-0.247) 1.69(1.64-1.73) South 0.025(-0.015-0.066) 1.40(1.60-1.44) East 0.889(0.807-0.97) 3.32(3.07-3.59)
e Table 4.1.4 Adjusted P/I ratios of dementia in comparison between passive and active survey
Variables Regression Coefficient
(2.5%-97.5%)
Adjusted P/I ratios (2.5%-97.5%)
Intercept 0.252(0.220-0.283) Age 65-79 vs. Age 80+ 0.120(0.100-0.140) Active Survey vs. Passive
Survey 1.071(0.614-1.53)
Passive Survey
Age 65-79 1.45(1.43-1.47)
Age 80+ 1.64(1.61-1.66)
Active Survey
Age 65-79 4.23(2.68-6.69)
Age 80+ 4.77(3.02-7.54)
f Table 4.1.5 The incidence rate (per 1000 person years) of dementia by gender
Area Age Range
Male Female All
Taiwan 65+ 17.3 19.4 18.3
Japan 102 65+ 19.3 20.9 20.0
European 40 65+ 14.4 21.3 18.4
Sweden 103 75+ 22.9 45.7 40.0
Spain 105 65+ 10.8 14.3 12.8
Italy 104 65-84 12.9 13.8 13.3
USA 106 65+ 19.2 21.0 20.3
g Table 4.1.6 The Age-standardized incidence rate of dementia (per 1000 person years) in European, Sweden, Spain, Italy, USA, and Taiwan
Country Age Range Incidence rate
Age-standardized incidence rate
Taiwan 65+ 18.3 17.5
European 40 65+ 18.4 13.1
Sweden 103 75+ 40.0 33.8
Spain 105 65+ 12.8 10.2
Italy104 65-84 13.3 11.5
USA 106 65+ 20.3 14.6
h Table 4.1.7 Prevalence, incidence, and ratio of dementia in Taiwan and
Taiwan (NHI) 65+ cohort Rural/urban 2.91 1.83 1.59 Taiwan (Active survey) 65+ screening Rural 9.29 1.83 5.08 Taiwan(Chen)117 65+ screening Rural/urban 10.55 1.83 5.77 Taiwan(Sun)96 65+ screening Rural/urban 8.14 1.83 4.45 Denmark 101 65-84 cohort Urban 7.10 2.95 2.40 Japan102 65+ cohort subrural 8.50 2.00 4.25
China30 65+ screening Urban 6.19 0.99 6.25 Sweden103 75+ cohort Urban 19.3 4.00 4.83 Italy104 65-84 cohort Rural/urban 8.00 1.25 6.40
Spain105 65+ cohort Rural/urban 5.79 1.06 5.47
USA106 65+ cohort Urban 9.60 2.03 4.73
i Table 4.2.1 Demographic characteristics and average scores in AD8 and MMSE tests of 282 samples in dementia screening program
Characteristic Numbers Percentage(%) AD8 MMSE
50-59 60 21.28 0.17±0.59 27.91±2.20
60-69 82 28.72 0.28±0.64 25.39±3.84
70-79 90 31.91 0.78±1.24 21.89±4.15
80+ 52 18.09 1.10±1.50 20.51±4.88
Gender
Male 144 51.06 0.59±1.23 22.59±5.06
Female 138 48.94 0.54±1.06 25.23±3.99
Education level
Illiteracy 84 29.79 0.83±1.31 19.77±3.90
< 6 years 123 43.62 0.56±1.09 24.02±3.90
>6 years 75 26.60 0.22±0.65 28.43±1.66
j Table 4.2.2 The findings of AD8, MMSE and clinical diagnosis
Characteristic Numbers % Clinical diagnosis
Normal MCI Dementia
Abnormal 73 26.9 34(19.5%) 26(30.9%) 13(72.2%)
AD8 and MMSE*
Normal 260 94.2 170(97.7%) 77(91.7%) 13(72.2%)
Abnormal 16 5.8 4(2.3%) 7(8.3%) 5(27.8%)
MMSE**
Normal(>21) 198 71.7 140(80.5%) 54(64.3%) 4(22.2%) Abnormal(≤21) 78 28.3 34(19.5%) 30(35.7%) 14(77.8%) AD8 or MMSE**
Normal 183 66.3 133(76.4%) 48(57.1%) 2(11.1%)
Abnormal 93 33.7 41(23.6%) 36(42.9%) 16(88.9%)
AD8 and MMSE**
Normal 249 90.2 165(95.8%) 75(89.3%) 9(50.0%)
Abnormal 27 10.7 9(5.2%) 9(10.9%) 9(50.0%)
*with education adjustment
** With optimal cut-off score of 21
k Table 4.2.3 The relationship between tests and dementia
Tests Pearson Correlation
Coefficients
P-value＊
AD8 0.27 P<0.0001
MMSE -0.26 P=0.0004
＊P value by Pearson Correlation Coefficients
l Table 4.2.4 The relationship between AD8 and MMSE
Tests Pearson Correlation
Coefficients
P-value＊
AD8 with MMSE -0.37 P<0.0001
＊P value by Pearson Correlation Coefficients
mTable 4.2.5 Performance of individual tests and the combination of tests for early detection of dementia
Tests Sensitivity(%) Specificity(%) Positive predictive
MMSE(cut off=26) 94.44 (69.36-99.22 )
* With cut-off score of 2 ** With education adjustment # with optimal cut-off score of 21
¥
Parallel 1 positive: AD8 (+) MMSE(+) and AD8(+) MMSE(-)and AD8 (-) MMSE(+) Parallel 2 positive: AD8 (+) MMSE (+) and AD8 (+) MMSE (-)
n Table 4.2.6 Performance of individual tests and the combination of tests for early detection of memory impairment (MCI plus Dementia)
Tests Sensitivity(%) Specificity(%) Positive predictive
value (%)
Negative predictive value (%)
AD8* 25.74 90.70 61.9 67.5
MMSE** 24.75 87.36 53.2 66.7
AD8 or MMSE** 38.24 80.46 53.4 68.9
AD8 and MMSE** 11.8 97.70 75.0 65.4
MMSE# 43.14 80.46 56.4 70.7
AD8 or MMSE# 50.98 76.44 55.9 72.7
AD8 and MMSE# 17.7 94.8 66.7 66.3
* with cut-off score of 2
** with education adjustment
# with optimal cut-off score of 21
o Table 4.2.7 Performance of individual tests and the combination of tests for early detection of mild impairment impairment (MCI)
Tests Sensitivity(%) Specificity(%) Positive predictive
MMSE(cut off=26) 79.76 (69.82-87.03 )
* With cut-off score of 2 ** With education adjustment # with optimal cut-off score of 21
¥
Parallel 1 positive: AD8 (+) MMSE (+) and AD8 (+) MMSE (-) and AD8 (-) MMSE (+) Parallel 2 positive: AD8 (+) MMSE (+) and AD8 (+) MMSE (-)
p Table 4.2.8 Bayesian analysis in prevalence estimation and screening test evaluation in the absence of a gold-standard test
Outcome Tests
Prior Prevalcne
Prior Probability Estimated Prevalence
Parallel testing Beta(18, 264)
Serial testing Beta(18, 264)
MMSE Sen~Beta(298,177) 23
Spe~Beta(875,505) 23
Parallel testing Sen~Beta(155,37) 23
Spe~Beta(96,16) 23
Serial testing 50.5%
(45.6%-55.3%)
95.1%
(93.5%-96.6%)
q Table 4.2.9 Logistic regression of clinical diagnosis dementia status against MMSE and AD8 for sample (n=282)
Variable -2logL AIC P
None 133.88 133.88
MMSE 110.35 114.35 <0.0001
AD8 114.95 118.95 <0.0001
Education 125.76 129.76 0.0044
MMSE+AD8 105.38 111.38 0.0258
MMSE + education 110.24 116.24 0.7401
AD8+Education 110.84 116.84 0.0426
MMSE+AD8+education 105.14 113.14 0.8869
MMSE+AD8+MMSE*AD8+Education 103.04 113.04 0.1473
r Table 4.2.10 Effects of MMSE, AD8, Age, gender and education, on the risk of dementia by Logistic regression model (Univariate analysis)
Variable Regression coefficient
Standard Error
OR 95%CI Wald Chi-Square
P-Value
AD8 0.6067 0.1618 1.834 1.336- 2.519 14.0518 0.0002
MMSE -0.2187 0.0553 0.804
0.721-0.896 15.6480 <.0001
Education -1.4138 0.5027 0.243 0.091-0.651 7.9099 0.0049
Gender -0.2838 0.490 0.75 0.288-1.967 0.3353 0.5626
Age 0.1104 0.0328 1.117 1.047-1.191 11.3194 0.0008
s Table 4.2.11 Effects of MMSE, AD8, and other risk factor on dementia by Logistic regression model (Multivariate analysis, model 1-3)
Variable Regression
t Table 4.3.1 The distribution of simulated cohort of dementia by stage and death in the end of simulation (year 10)
Screening strategy
Status at year 10 RR
Normal Mild Moderate Severe Death Severe Death No Screen 0.4591 0.0875 0.0414 0.1310 0.2809 1.0000 1.0000
AD8 0.4591 0.0978 0.0618 0.1133 0.2680 0.8646 0.9540
MMSE 0.4591 0.0976 0.0614 0.1136 0.2683 0.8674 0.9549 Parallel test 0.4591 0.0993 0.0647 0.1108 0.2661 0.8457 0.9472 Serial test 0.4591 0.0960 0.0584 0.1162 0.2702 0.8871 0.9618
u Table 4.3.2 Cost-utility analysis for different screening strategies compared with no screening over 10-year span, considering both direct and indirect cost
Strategy Cost ($US) Incremental
Costs
QALY Incremental QALYs
ICURc
No Screen 46285.9 (Reference) 6.0847 (Reference) (Reference) (33488.6, 62241.8) (6.0068, 6.1709)
AD8 44683.5 -1602.4 6.1286 0.0439 -36505.7
(32014.7, 58746.3) (6.0480, 6.2104)
MMSE 44719.9 -1566.1 6.1276 0.0429 -36466.4
(32353.5 ,58886.6) (6.0502, 6.2103)
Parallel test 44450.0 -1836.0 6.1351 0.0505 -36377.1
(31836.3, 58397.6) (6.0575, 6.2179)
Serial test 44960.2 -1325.7 6.1209 0.0363 -36543.1
(32231.3, 59492.7) (6.0391, 6.2068) a ICUR: incremental cost-utility ratio
v Table 4.3.3 Cost-utility analysis for different screening strategies compared with no screening over 10-year span, considering direct cost only
Strategy Cost ($US) Incremental Costs
QALY Incremental QALYs
ICURa
No Screen 3264.3 (Reference) 6.0847 (Reference) (2300.1, 4368.4) (6.0068, 6.1709)
AD8 3281.9 17.6 6.1286 0.0439 401.4
(2335.5, 4346.9) (6.0480, 6.2104)
MMSE 3283.9 19.7 6.1276 0.0429 457.7
(2343.2, 4347.6) (6.0502, 6.2103)
Parallel test 3284.9 20.7 6.1351 0.0505 409.8
(2334.4, 4340.2) (6.0575, 6.2179)
Serial 3282.4 18.1 6.1209 0.0363 499.2
(2335.7, 4346.6) (6.0391, 6.2068) ICURa: incremental cost-utility ratio
AD8
[+]
MMSE [+]
Serial Test
[+]
Screening Strategy
No Screening [+]
Parallel Test
[+]
a Figure 3.2.1 Strategies for community-based dementia screening
b Figure 3.2.2 Five-state model for dementia progression for
cost-effectiveness analysis
Mild
Moderate
Severe
Death
Normal
c Figure 3.2.3 Markov decision tree with dementia screening program using Parallel Test
d Figure 4.2.1 ROC curve of AD8, MMSE and parallel test
e Figure 4.3.1 Distribution of disease status of dementia by stage and
death by time in a naïve cohort
0 0.1 0.2 0.3
0 2 4 6 8 10
Cumulative risk
Time (years) Mild
Moderate Severe Death
(a) AD8 vs no screening (b) MMSE vs no screening
* AD8 was 92.2% being cost-effective over no screening * MMSE was 92% being cost-effective over no screening
(c) Parallel test vs no screening (d) Serial test vs no screening
* Parallel test was 93.8% being cost-effective over no screening * Sequential test was 88.6% being cost-effective over no screening
(e) Parallel test vs AD8 (f) Parallel test vs MMSE
* Parallel test was 88.8% being cost-effective over AD8 * Parallel test was 90.0% being cost-effective over MMSE
(g) Serial test vs AD8 (h) Serial test vs MMSE
* Serial test was 11.4% being cost-effective over AD8 * Serial test was 12.8% being cost-effective over MMSE
f Figure 4.3.2 Scattered incremental cost-effectiveness analysis for different screening strategies for dementia compared with no screening over 10-year span, considering both direct and indirect cost
(a) AD8 vs no screening (b) MMSE vs no screening
* AD8 was 97.6% being cost-effective over no screening * MMSE was 97.2% being cost-effective over no screening
(c) Parallel test vs no screening (d) Serial test vs no screening
* Parallel test was 99.0% being cost-effective over no screening * Sequential test was 93.4% being cost-effective over no screening
(e) Parallel test vs AD8 (f) Parallel test vs MMSE
* Parallel test was 95.4% being cost-effective over AD8 * Parallel test was 95.2% being cost-effective over MMSE
(g) Serial test vs AD8 (h) Serial test vs MMSE
* Serial test was 13.2% being cost-effective over AD8 * Serial test was 15.2% being cost-effective over MMSE
g Figure 4.3.3 Scattered incremental cost-effectiveness analysis for different screening strategies for dementia compared with no screening over 10-year span, considering only direct cost
h Figure 4.3.4 Acceptability curve for cost-effectiveness analysis
dementia screening over 10-year span, considering both direct and indirect cost
i Figure 4.3.5 Acceptability curve for cost-effectiveness analysis
dementia screening over 10-year span, considering only direct cost
References
revisited. American Journal of Public Health 1994;84:1261‐1264.
integration of the literature. Acta psychiatrica scandinavica 1987;76:465‐479.
the silent epidemic. Annals of Internal Medicine 1982;97:231‐241.
58. Law S, Wolfson C. Validation of a French version of an informant‐based
progression rate from preclinical to clinical phase, sensitivity and positive predictive
87. Chiu L, Tang K‐Y, Liu Y‐H, Shyu W‐C, Chang T‐P. Willingness of families caring for
Study. Neurology 1999;52:85‐85.
events in Alzheimer's disease. Brain 2009:awp062.
115. Fuh JL, Wang SJ. Cost‐effectiveness analysis of donepezil for mild to moderate Alzheimer's disease in Taiwan. International journal of geriatric psychiatry
2008;23:73‐78.
116. Fuh J. Dementia in Taiwan: current status. Tai Wan Lao Nian Yi Xue Ji Lao Nian Xue Za Zhi 2008;3:169‐181.
117. Ta‐Fu Chen M‐JC, Allen GM. Institution Type‐Dependent High Prevalence of Dementia in Long‐Term Care Units Neuroepidemiology 2007;28:142–149