Development and validation of a nomogram for predicting the six-year risk of cognitive impairment among Chinese older adults

Abstract

Objective:

Although some people with mild cognitive impairment may not suffer from dementia lifelong, about 5% of them will progress to dementia within one year in community settings. However, a general tool for predicting the risk of cognitive impairment was not adequately studied among older adults.

Design:

Prospective cohort study.

Setting:

Community-living, older adults from 22 provinces in China.

Participants:

We included 10,066 older adults aged 65 and above (mean age, 83.2±11.1 years), without cognitive impairment at baseline in the 2002–2008 cohort and 9354 older adults (mean age, 83.5±10.8 years) in the 2008–2014 cohort of the Chinese Longitudinal Healthy Longevity Survey.

Methods:

We measured cognitive function using the Chinese version of the Mini-Mental State Examination (MMSE). Demographic, medical, and lifestyle information was used to develop the nomogram via a Lasso selection procedure using a Cox proportional hazards regression model. We validated the nomogram internally with 2000 bootstrap resamples and externally in a later cohort. The predictive accuracy and discriminative ability of the nomogram were measured by area-under-the-curves (AUCs) and calibration curves, respectively.

Results:

Eight factors were identified with which to construct the nomogram: age, baseline MMSE, activities of daily living and instrumental activities of daily living score, chewing ability, visual function, history of stroke, watching TV or listening to the radio, and growing flowers or raising pets. The AUCs for internal and external validation were 0.891 and 0.867, respectively, for predicting incident cognitive impairment. The calibration curves showed good consistency between nomogram-based predictions and observations.

Conclusion and Implications:

The nomogram-based prediction yielded consistent results in two separate large cohorts. This feasible prognostic nomogram constructed using readily ascertained information may assist public health practitioners or physicians to provide preventive interventions of cognitive impairment.

Summary

A nomogram for assessing future risk of cognitive impairment was developed and validated with good performance via readily ascertained information including age, baseline MMSE, stroke, and other five modifiable variables.

Introduction

Cognitive impairment, ranging from mild cognitive impairment to dementia, is a major public health problem all over the world.^1,2 Modest declines in cognitive function are reported in some in younger adults but cognitive function rapidly worsens in the sixth decade of age and thereafter throughout the remaining adult lifespan.^1,2 The disease burden of cognitive impairment in China is dramatically growing with the aging population; by the end of 2019 in China, 176 million people were aged 65 and above, consisting of 12.6% of the Chinese population.⁴ However, drug therapy for cognitive impairment, dementia, or Alzheimer’s disease remains relatively inefficient.^5,6 Therefore, it is important to inform older adults about their future risk of cognitive impairment and provide individualized advice on interventions.⁷

Previous studies focused little on predicting cognitive impairment in cognitively healthy older adults. A risk nomogram based on age, family history of dementia, and memory has been explored to calculate relative risk (RR) rather than predict the incidence of cognitive impairment;⁸ another study focused on the prediction of two-year risk of 390 patients with Parkinson’s disease without external validation;⁸ other studies mainly concentrated on the risk prediction of dementia or Alzheimer’s disease using personal demographic information, disease, and lifestyle factors.^9–13 Genetic information (APOE-ε4), brain magnetic resonance imaging (MRI), fasting plasma glucose (FPG), cerebrospinal fluid (CSF) biomarkers, and HbA1c provide modest predictive power for cognitive impairment or dementia as measured by the concordance index (C-index) or area-under-the-curve (AUC).^9–13 However, collection and examination of these predictors are difficult among older adults: models constructed with these predictors would not be easily applied in general populations; most of the previous studies have limited sample size, especially lack for the oldest old; thus, models developed in these populations may not work well for older adults, who are more likely to progress to cognitive impairment. Therefore, the goal of this study is to provide a feasible and easy to deploy tool for predicting the six-year risk of cognitive impairment among older adults aged 65 and above.

Methods

Study participants

Two separate and independent cohorts of the Chinese Longitudinal Healthy Longevity Survey (CLHLS) were analyzed. The CLHLS was conducted face-to-face using internationally compatible questionnaires, was the first national longitudinal survey on determinants of healthy aging and has the largest sample of the oldest-old individuals in China. The baseline survey and the follow-up surveys at a three-year interval were conducted in a randomly selected half of the counties and cities in 22 of China’s 31 provinces; more information about this prospective cohort has been provided elsewhere.^15,16 Briefly, all centenarians were invited to be interviewed, two nearby nonagenarians, two octogenarians, and approximately three older adults aged 65–79 of pre-designated age and sex were randomly chosen to be interviewed in conjunction with every two centenarians, which ensured comparable numbers of randomly selected male and female younger elders aged 65 years and above. Overall, we identified 16,064 participants investigated at the 2002 baseline who were potential candidates for the cohort used to develop the predictive algorithm. Of these, we excluded 44 individuals younger than 65 years, 4338 who suffered from cognitive impairment, 1501 who were lost-to-follow-up before the first follow-up survey, and 115 with missing values of key variables, leaving 10,066 older adults aged 65 and above in the final cohort used for developing the predictive algorithm (Appendix 1 Figure A1). Similarly, we identified 16,954 participants investigated at the 2008 baseline in the external validation cohort. Of these, we excluded 7600 participants through the same criteria (described above), leaving 9354 older adults aged 65 and above in the cohort used to validate the predictive algorithm (Appendix 1 Figure A2). The biomedical ethics committee of XX University approved the study (IRB00001052–13074), all participants or their proxy respondents provided written informed consent.

Assessment of cognitive impairment

Cognitive function was assessed for all participants at the baseline survey by trained public health practitioners from local centers for disease control and prevention using the traditional Chinese version of the Mini-Mental State Examination (MMSE), which has been widely used as a screening test for cognitive impairment in epidemiological studies and proven to be applicable to China.^17,18 The MMSE is usually used to screen populations for cognitive impairment, track changes in cognitive function over time, and oftentimes to assess the effects of interventions on cognitive function. It shows high specificity and sensitivity in defining dementia and is recognized as a relatively perfect tool to detect cognitive impairment, especially among those with advanced age who are not able to go through complex clinical diagnostic testing.¹⁹ The Chinese version of the MMSE addressed the following aspects of cognitive functioning: orientation to time, orientation to place, three object registration, calculation, word recall, naming, repeating, comprehension, three-step tasks, sentence construction, and copying figures. It has been tested for reliability and validity in China and adaptation to China’s cultural and socioeconomic characteristics.²⁰ The MMSE scores range from 0 to 30; a higher score represents a better cognitive function. It has been proven rational to use education-based cutoff points to determine cognitive function among Chinese.^20,21 Therefore, we defined cognitive impairment as follows: for participants without formal education, an MMSE score <18 was considered cognitive impairment; for participants with one to six years of education, an MMSE score <21 was considered cognitive impairment; for participants with more than six years of education, an MMSE score <25 was considered cognitive impairment; not meeting these criteria was considered as not cognitive impairment. As described above, individuals meeting criteria for cognitive impairment at baseline were excluded from this analysis. Cognitive function was reassessed by the MMSE in the later follow-up surveys to serve as the standard for determining whether incident cognitive impairment had occurred during two three-year follow-up intervals.

Candidate predictors

Information about demographics and lifestyle were acquired through a structured face to face questionnaire by trained public health practitioners from local centers for disease control and prevention. History of diseases was ascertained using self-reported questionnaire. Blood pressure was measured by trained internists with at least three years of work experience using a mercury sphygmomanometer. The following were defined continuous variables: age, systolic pressure (SBP), body mass index (BMI), years of education, activities of daily living (ADL) and instrumental activities of daily living (ADL) score, baseline MMSE, daily amount and duration of smoking, daily dose and duration of drinking. ADL and IADL score ranges from 14 to 42 with 14 activities considered overall, including: six ADLs -- dressing, feeding, bathing, continence, toileting and indoor transfer;²² and eight IADLs -- going outside to visit neighbors, shopping, preparing meals, doing laundry, walking 2 kilometer, lifting a 5-kilogram weight, doing 3 squats, and taking public transport. For each activity, 1 point was assigned if the participant could easily finish it successfully without any assistance; 2 points were assigned if the participant could finish it successfully with difficulty or with assistance in specific part; 3 points were assigned if the participant could only finish it successfully with assistance in more than one part, or was unable to finish it (Detailed scoring rubric was provided in Appendix 1 Table A1). The total score of ADLs and IADLs was calculated for analysis.

Other variables were categorized as follows: sex was defined as male or female; race was defined as the Han nationality or minority; occupation was defined as farmer or other; marital status was defined as married if the participant was currently married and living with a spouse, formerly married if the participant was separated or divorced from a spouse, widowed, or single if the participant was never married; residence was defined as urban or rural; living arrangement was defined as alone, with family members or in a nursing home; self-reported heart disease, history of stroke, diabetes mellitus, and respiratory disease were defined as yes or no; chewing ability, visual function, and dietary diversity were defined as good or poor (more than 20 teeth or use of artificial teeth were defined as having good chewing ability);²³ self-reported frequencies of food consumption, including vegetables, fruits, legumes and soy products, meat, eggs, fish, and tea were defined as never, sometimes, and often; leisure activities including reading books or newspapers, playing cards or mahjong, watching TV or listening to the radio, attending social activities, growing flowers or raising pets, and personal outdoor activities were defined as never, sometimes, or often. All interactions between age and other predictors were also considered.

Statistical analysis

We used complete datasets for model development and validation because less than 0.3% of all key variables were missing values. We computed person-years of follow-up from the investigation date of the baseline survey until either cognitive impairment, death, or the end of follow-up (September 1, 2008 for the development cohort, September 1, 2014 for the validation cohort), whichever came first. Results are presented as mean ± standard deviation (SD) or median (interquartile range, IQR) for continuous variables, and number (percentage) for categorical variables.

To address the potential bias, participants who were lost-to-follow-up at the first follow-up survey were excluded. All participants in the development cohort were used for model establishment. Multivariable time-to-event analysis was performed using Cox proportional hazards regression model. Continuous variables were additionally explored for their potential non-linear association with the risk of cognitive impairment using restricted cubic splines. Variables in the final model were obtained using a Lasso selection procedure, which incorporates different penalties for different coefficients to retain the most important variables. Lasso is useful for preventing model overfitting and model misspecification.^24,25 The nomogram was constructed using variables included in the final model. A score was assigned to each predicting variable in the nomogram such that total points could be easily ascertained to estimate the probability of cognitive impairment. The estimated risk for the predictors was presented through the hazard ratio (HR).

In accordance with the validation rule described in the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) statement,²⁶ the nomogram was subjected to 2000 bootstrap resampling for internal validation of the development cohort; external validation was performed using a separate independent CLHLS cohort from 2008 to 2014. Discrimination was tested using AUC. The value of the AUC ranges from 0.5 to 1; a value of 0.5 indicates no ability to discriminate and a value of 1.0 indicates perfect separation.²⁷ The calibration was examined using a calibration curve derived by the regression analysis to test the consistency of predicted probabilities and observed frequencies. In detail, we ordered all participants by the value of predicted risk probabilities, and then divided them into ten sample-balanced subgroup; the predicted probabilities and observed frequencies of each group were combined to draw the calibration curve.

To evaluate the efficacy of our nomogram, we applied different thresholds of predicted risk to calculate in the validation cohort the corresponding sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), counts of true positive, false positive, true negative and false negative. In detail, all participants were arranged in decreasing order according to their nomogram predicted risk of cognitive impairment. We selected thresholds corresponding to the top 1%, 3%, 5%, 10%, 20%, 30%, 40%, 50%, and 60% of older adults with highest risk to assess the performance of classification.

To evaluate the net benefits of the nomogram, we performed a decision curve analysis in the validation cohort.^28,29 This method assesses the benefits of correctly detecting people who will have cognitive impairment compared with the harms from a false positive classification (which could lead to unnecessary interventions). Net benefits were computed across a range of thresholds of risk probabilities to compare with alternative strategies of “intervention in all participants” and “intervention in none”.

All statistical analyses were performed using R 3.6.2 for Windows (R Foundation for Statistical Computing, Vienna, Austria) with the “rms”, “Hmisc”, “glmnet”, “survival”, “survivalROC” and “ggplot2” packages.

Results

Baseline Characteristics

Table 1 displays the baseline characteristics of the development and validation cohorts. In the development cohort, we included 10,066 older adults with a mean age of 83.2±11.1, ranging from 65 to 114 years. Median follow-up time was 4.7 years (IQR 2.4–6.1). The average baseline MMSE was 26.8± 3.2 points, and summed ADL and IADL score was 18.7± 6.1 points. Overall 53.8% (5412) had good chewing ability, 79.1% (7959) had good visual function, 4.7% (478) had history of stroke, 47.4% (4768) often watch TV or listen to the radio, 10.1% (1013) often grow flowers or raise pets.

Table 1.

Baseline characteristics of older adults aged 65 years of age and above without cognitive impairment at study entry in development and validation cohorts.

Characteristics	Development cohort (n=10,066)	Validation cohort (n=9354)
Age (y)
Mean (SD)	83.2 (11.1)	83.5 (10.8)
Age group
65–79	3972 (39.5)	3481 (37.2)
80–89	3001 (29.8)	2817 (30.1)
90–99	1939 (19.3)	2130 (22.8)
≥100	1154 (11.5)	926 (9.9)
Baseline MMSE (points)
Mean (SD)	26.8 (3.2)	26.8 (3.3)
ADL and IADL score (points)
Mean (SD)	18.7 (6.10)	17.9 (5.52)
Chewing ability
Good	5412 (53.8)	5177 (55.3)
Poor	4654 (46.2)	4177 (44.7)
Visual function
Good	7959 (79.1)	6861 (73.3)
Poor	2107 (20.9)	2493 (26.7)
History of stroke
No	9588 (95.3)	8856 (94.7)
Yes	478 (4.7)	498 (5.3)
Watching TV or listening to the radio
Never	2782 (27.6)	2047 (21.9)
Sometimes	2516 (25.0)	2019 (21.6)
Often	4768 (47.4)	5288 (56.5)
Growing flowers or raising pets
Never	8314 (82.6)	7898 (84.4)
Sometimes	739 (7.3)	544 (5.8)
Often	1013 (10.1)	912 (9.7)
Median follow-up duration (IQR; years)	4.7(2.4–6.1)	4.8(2.6–5.8)

Abbreviations: MMSE= the Mini-Mental Status Examination; ADL = activities of daily living; IADL = instrumental activities of daily living; IQR = interquartile range. Values are presented as numbers (percentages) of participants unless stated otherwise.

In the validation cohort, we included 9354 older adults with a mean age of 83.5±10.8, ranging from 65 to 116 years. The distribution of characteristics was similar to the development cohort. Detailed information is provided in Table 1.

Incidence of cognitive impairment

Appendix 1 Table A2 displays the incidence of cognitive impairment stratified by age and sex. In the development cohort, 18.2% (1828) of participants suffered from cognitive impairment during a total follow-up time of 41,645 person-years. In the validation cohort, 20.7% (1932) participants suffered from cognitive impairment during a total follow-up time of 38,039 person-years. The cognitive impairment incidence tended to be higher among females and older adults in both cohorts.

Predictor variables

Appendix 1 Table A3 displays the HR (95% confidence interval [CI]) for cognitive impairment in different subgroups. Eight predictor variables were included in the final model, including age, baseline MMSE, ADL and IADL score, chewing ability, visual function, history of stroke, watching TV or listening to the radio, and growing flowers or raising pets. As shown in the nomogram (Figure 1), age served as the most important predictor for cognitive impairment. Age had a non-linear relationship with predicted risk probability of cognitive impairment, which appeared to increase at a slower rate among older adults with advanced age, especially those after 90 years old (Appendix 1 Figure A3; Figure 1). Interactions between age and other predictors were not observed.

Figure 1. Nomogram for predicting the six-year cognitive impairment probability.

The scores corresponding to each characteristic are listed on the top. To determine the six-year probability of cognitive impairment, compute the sum of scores for each characteristic and locate this sum on the “Total Points” axis, then read the corresponding probability from the lowest line.

Abbreviations: ADL= activities of daily living; IADL = instrumental activities of daily living; MMSE= the Mini-Mental Status Examination.

Nomogram for predicting cognitive impairment

Figure 1 displays the nomogram established based on regression analysis. The nomogram illustrates that age made the biggest contribution to the prediction of incident cognitive impairment, followed by the ADL and IADL score, and baseline MMSE. A score was assigned on the point scale for each subtype within these variables. By counting the total score and locating it on the total points scale, we could easily draw a straight line down to determine the estimated probability of cognitive impairment over six years. For instance, an age of 85 was allocated 55 points; the ADL and IADL score of 30 was allocated 18 points; baseline MMSE score of 24 was allocated 11 points; with history of stroke was allocated 9 points; poor chewing ability was allocated 4 points; poor visual function was allocated 7 points; never watching TV or listening to the radio was allocated 7 points; never growing flowers or raising pets were allocated 12 points. In this example, a total point of 123 is achieved by an older adult aged 85 with the characteristics listed above corresponds to a six-year risk probability of cognitive impairment of about 53.8%.

Besides, we also developed a risk calculation tool using Excel spreadsheets on the basis of nomogram, via which future risks of cognitive impairment can be easily accertained by imputing personal information required. Appendix 2.

Discrimination and calibration of the nomogram

The performance of the nomogram was evaluated by AUC. The AUCs for internal and external validation were 0.891 and 0.867, respectively (Figure 3). These results indicate a good and highly consistent performance of the nomogram for predicting a six-year risk of cognitive impairment among older adults without cognitive impairment at baseline. Overall, the AUC was higher among females, indicating that the model performed better among females than males. However, this model had poor performance in different age groups as age served as the most important predictors (Appendix 1 Table A4; Figure 1). We recalculated the AUCs excluding age; this led to a decline in the AUCs to 0.821 for the development cohort and 0.781 for the external validation cohort.

Figure 3. ROC curves for development and validation cohort.

The ROC curves were illustrated in red and blue for the development and validation cohort on the basis of nomogram-predicted probability of cognitive impairment over six years.

Abbreviations: ROC= receiver operating characteristic curve; AUC= area-under-the-curve.

Figure 2 depicts the calibration curves for internal development (A) and external validation (B). The plot showed good agreement between the predicted and observed six-year probability of cognitive impairment.

Figure 2. The internal (A) and external (B) calibration curves of the nomogram.

Nomogram-predicted probability and observed frequency over six years for cognitive impairment among participants without cognitive impairment at baseline were plotted in the x- and y-axis, respectively. The gray line indicates the ideal plot for the calibration curve, where the nomogram-predicted probabilities perfectly match the observed probabilities in all subgroups.

Sensitivity and specificity

Table 2 displays the sensitivity, specificity, PPV, NPV, counts of true positive, false positive, true negative and false negative at select thresholds of the top 1%, 3%, 5%, 10%, 20%, 30%, 40%, 50%, and 60% of older adults with highest risk in the validation cohort. The corresponding risk threshold of cognitive impairment for a cut-off at the top 5% was 89.6%, for the top 10% was 82.1%, and for the top 60% was 34.6%. With a risk threshold of 34.6% to identify the 60% of participants with the highest risk of cognitive impairment over six years, the sensitivity was 96.0%, specificity was 60.8%, positive predictive value was 38.9%, and negative predictive value was 98.3%.

Table 2.

Sensitivity, specificity, PPV and NPV for cognitive impairment at different thresholds of predicted risk of cognitive impairment over six years in the validation cohort.

Threshold	Risk threshold (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	True positive (count)	False positive (count)	True negative (count)	False negative (count)
Top 1%	95.9	2.4	99.8	79.66	79.7	47	12	7410	1885
Top 3%	92.5	6.7	99.2	67.7	80.3	130	62	7360	1802
Top 5%	89.6	11.2	98.6	68.0	81.0	217	102	7320	1715
Top 10%	82.1	23	97.6	71.1	82.96	445	181	7241	1487
Top 20%	71.5	42	92.7	60.1	86.0	811	539	6883	1121
Top 30%	61.8	60	87.4	55.3	89.4	1159	936	6486	773
Top 40%	52.2	75.4	80.5	50.17	92.6	1457	1447	5975	475
Top 50%	43.5	85.9	70.8	43.4	95.1	1659	2165	5257	273
Top 60%	34.6	96	60.8	38.9	98.3	1854	2907	4515	78

Abbreviations: PPV= positive predictive value; NPV= negative predictive value.

Decision curve analysis

The net benefit curves for cognitive impairment over six years shows that the predictive nomogram had higher net benefit than strategies based on considering either no participants or all participants for intervention at risk thresholds up to approximately 70% (Appendix 1 Figure A4).

Discussion

To the best of our knowledge, this is the first study using easily accessible personal information to predict the risk of cognitive impairment among Chinese older adults aged 65 and above. We used data from the CLHLS study, which is a nationally representative prospective cohort scattered over 806 counties in 23 provinces around China, enabling us to provide a general tool to predict the risk of cognitive impairment for Chinese older adults. The restricted cubic spline was explored to better explain the non-linear correlation between age and cognitive impairment.^30,31 In this prospective cohort of 10,066 older adults without cognitive impairment at baseline, we developed and validated an easy-to-use nomogram-illustrated model that proved to be robust for predicting six-year incident cognitive impairment among Chinese older adults. The nomogram includes an extensive set of personal risk factors that are easy to ascertain. Thus, the nomogram may be feasible for use by medical practitioners to inform the future risk of cognitive impairment at six years for older adults. In addition, the results of the nomogram might be used to guide preventive interventions; for instance, older adults may maintain better cognitive function via watching TV or listening to the radio, growing flowers or raising pets, and undergoing treatment to maintain ADL or IADL independence, good chewing ability, and good visual function.

Validation of the nomogram is essential to avoid overfitting of the model and to determine its generalizability.³² In our study, a 2000 bootstrap resampling strategy was used for internal validation and a separate cohort of CLHLS from 2008 to 2014 was used for external validation; both efforts provided validation that the nomogram as a feasible tool with good performance with respect to discrimination and calibration. A good discriminative ability was demonstrated by AUC of 0.891 for the internal development and 0.867 for the external validation cohorts. These values are comparable to previous studies based on easily accessible personal information with C-indices or AUCs from 0.66 to 0.78 for predicting dementia or Alzheimer’s disease, and up to around 0.86 with CSF, genetic and imaging information.^1,9–13,33 Calibration plots in our study showed good agreement between predicted and actual observed risk in ten subgroups, which further verified the repeatability and reliability of the nomogram.

The major risk factors we identified are generally consistent with previous studies. Age, ADL and IADL score, baseline MMSE, and growing flowers or raising pets are the most important predictors for cognitive impairment. We observed a negative quadratic relationship between age and cognitive impairment (Appendix 1 Table A3 and Appendix 1 Figure A3), indicating that the risk is higher among older adults with advanced age.^3,30,34 However, the risk magnitude of cognitive impairment tends to increase at a slower rate especially among those older than 90 years old who were more likely to be examples of successful aging, which may support cognitive reserve^35,36. Inability to perform daily activities (ADL or IADLs) has been found to be closely related to cognitive impairment.^37,38 Chewing ability, which declined with age, may influence the intake of food and increase the risk of cognitive impairment.³⁹ In this study, growing flowers or raising pets rather than reading books or newspapers was found to be strongly associated with lower risk of cognitive impairment; in general, participants in this study were poorly-educated with limited reading ability, as reported in previous studies using CLHLS data, and therefore unlikely to have engaged in activities like the reading of books.⁴⁰

This study also has some limitations. First, cognitive impairment was assessed in three-year intervals, which may cause bias as cognitive function may decrease largely within three years if any adverse event happened. Second, participants with older age may suffer from a higher risk of mortality leading to more missing cognitive function data for these individuals that may shrink the risk prediction. Third, data related to other potential risk factors like the apolipoprotein APOE ε4 allele, FPG, CSF, HbA1c, and vitamin use are not included in our study, but the nomogram provided relatively good performance in predicting six-year risk of cognitive impairment. Fourth, there may be a limitation in using the nomogram in developed countries where older adults have been well-educated. Finally, the nomogram was not subjected to an external validation in other studies due to the lack of research data including MMSE in populations with similar age structure. However, we used a 2000 bootstrap resampling strategy for internal validation and a separate independent cohort of the CLHLS for external validation. We showed that the nomogram has good performance in terms of discrimination (AUC 0.891 and 0.867, respectively) and calibration for predicting incident cognitive impairment in older adults.

Conclusions and Implications

In summary, the nomogram herein developed can be deployed as an easy-to-use and general tool for assessing the risk of cognitive impairment over six years. Through this nomogram, people can estimate the risk of cognitive impairment for older adults aged 65 and above. Public health practitioners can use it to evaluate future risk of cognitive impairment in later surveys. Physicians may also use it as an instructional tool for preventive interventions to potentially maintain better cognitive function in the clinic. For instance, older adults may maintain better cognitive function via watching TV or listening to the radio, growing flowers or raising pets, and undergoing treatment to maintain ADL or IADL independence, good chewing ability, and good visual function. However, whether it is applicable to other countries remains to be validated by future studies.