Trends in the prevalence of cognitive impairment at old age in China, 2002–2018

Abstract

INTRODUCTION

China has the world's largest number of older adults with cognitive impairment (CI). We aimed to examine secular trends in the prevalence of CI in China from 2002 to 2018.

METHODS

Generalized estimating equations (GEE) was used to assess changes in CI trend in 44,154 individuals (72,027 observations) aged 65 to 105 years old.

RESULTS

The prevalence of CI increased from 2002 to 2008 and then decreased until 2018. The age‐standardized prevalence increased from 25.7% in 2002, 26.1% in 2005, to 28.2% in 2008, then decreased to 26.0% in 2011, 25.3% in 2014, and 24.9% in 2018. Females and those ≥ 80 years old had greater CI prevalence.

DISCUSSION

The prevalence of CI showed an inverted U shape from early 2000s to late 2010s with a peak in 2008. Follow‐up studies are needed to confirm the decreasing trend after 2008 and examine the contributing factors and underlying mechanisms of this trend.

Highlights

• Generalized estimating equations (GEE) were used to assess trends of changes in cognitive impairment (CI).

• CI prevalence in China increased from 2002 to 2008 and then decreased until 2018.

• Females and those ≥ 80 years old had greater CI prevalence.

• Stroke, diabetes, and cigarette smoking were risk factors for CI.

1. INTRODUCTION

There is ongoing debate on whether there is a decrease in the prevalence of cognitive impairment (CI) in the older population. Several studies from high‐income countries have shown evidence that incidence and prevalence rate of dementia, the severe clinical stage of CI, have been decreasing in the past decades. ¹ , ² , ³ , ⁴ , ⁵ For example, in the United States, the Health and Retirement Study revealed a decreasing trend over time, ⁶ with an annual decline of 3.4% from 1993 to 2004. ⁷ Cautious optimism appears justified. ⁸ It is argued that the favorable trends in population brain health may be limited to those countries that had major investments in population health over the life course of those now in older ages. ² As the second largest economy in the world, China has made significant progress in improving public health, especially in the past three decades, with its gross domestic product almost tripled. How the rapid social development affected cognitive health of the Chinese older adults remains largely unknown.

Multiple factors can determine a person's cognitive function in late life. Epidemiological studies have established a number of protective (e.g., higher education, physical exercise) and risk (e.g., increasing age, poor cardiovascular health) factors. ⁹ , ¹⁰ The fast growth of China in wealth has been accompanied by the rising tides of non‐communicable diseases (NCD), ¹¹ or the so‐called “rich man's diseases.” The prevalence of Type II diabetes, for example, has increased from < 1% in 1980 to 12.4% in 2018. ¹² Given the increase in cardiovascular and cerebrovascular disease burden in China, ¹³ , ¹⁴ , ¹⁵ it is hypothesized that the prevalence of CI has increased as a consequence of worsening vascular health, on top of influences from changing demographics and lifestyles.

To date, no study has examined secular trends in the prevalence of CI in China using comparable data from multiple time points. This gap in scholarship has important social and economic implications as governmental agencies need such information for proper planning and allocation of resources for the assessment, management, and long‐term care of cognitively impaired older adults. Identifying major drives behind potential change will inform policy makers and practitioners on the best preventive strategies that may work at a population level. To bridge the existing knowledge gaps, the secular trends of CI among Chinese older adults was investigated by analyzing six waves of cross‐sectional data from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) over a 16‐year period from 2002 to 2018. Mechanisms of these observed trends were explored through examining a series of demographic, socioeconomic, behavioral, and health factors.

2. METHODS

2.1. Study design and participants

Data were extracted from the CLHLS survey conducted as follow‐ups in 2002, 2005, 2008, 2011, 2014, and 2018. The CLHLS is a national, representative survey conducted in a randomly selected sample that covered approximately half of the counties and cities in 22 of the 31 provinces and municipalities of China. ¹⁶ The study adopted a targeted random ‐sample design to insure representativeness, even distribution across age and sex, and an adequate subsample size of the oldest old (aged 80 to 105 years), comparable with young old (aged 65 to 79 years). All the CLHLS surveys included re‐interviews to survival participants, and information on the date and cause of death and health status of older adults who died before the subsequent survey was collected by interviewing a close family member. In the surveys in 2002, 2005, and 2008, CLHLS included new recruitments to replace deceased and older adults lost to follow‐up with the same sex and age. However, due to the limited budget of the survey, new recruitments after 2018 were conducted only within 8 longevity counties/cities (instead of 745 counties/cities). These counties/cities were randomly selected to minimize regional differences. Out of the six statistical regions of China, the eight longevity counties/cities covered the South Central China, East China, Central China, and the East China regions. Meilande Consulting conducted the surveys in 2002 and 2005. From 2008 onward, the China Centre for Disease Control and Prevention (CCDC) conducted all the surveys.

2.2. Procedures

In CLHLS, trained interviewers conducted face‐to‐face interviews to collect cognitive data and other information. A Chinese version of the Mini‐Mental State Examination (MMSE) originally developed by Folstein et al. ¹⁷ was used to assess global cognitive function. The total score ranges from 0 to 30 points with lower scores representing poorer cognitive function. The same MMSE version was used across all CLHLS data waves to ensure data comparability. Education‐specific cut‐offs were used to define CI, based on the latest normative and validation study of MMSE from the Chinese population ¹⁸ ≤ 16 points for those without schooling, ≤ 19 points for those with 1 to 6 years of education, and ≤ 23 points for those with > 6 years education. For sensitivity analyses, we also defined CI using higher education specific cut‐offs: ¹⁹ (1) ≤ 17 points for those without schooling, ≤ 20 points for those with 1 to 6 years of education, and ≤ 24 points for those with > 6 years education; (2) and a single cut‐off of < 18 points.

RESEARCH IN CONTEXT

1. Systematic review: Several studies from high‐income countries, such as Sweden, the United States, and England, have shown that incidence and prevalence rate of cognitive impairment (CI) have been decreasing. However, to date, no study before ours has examined secular trends in the prevalence of CI in China using comparable data from multiple time points. We used data from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) conducted in 2002, 2005, 2008, 2011, 2014, and 2018. The CLHLS is a national, representative survey conducted in a randomly selected sample that covered approximately half of the counties and cities in 22 of the 31 provinces and municipalities of China. The study adopted a targeted random sample design to insure representativeness, even distribution across age and sex, and large enough subsample size of the oldest old (aged 80 to 105), compatible with young old (aged 65 to 79). There were 81,485 interviews in total from CLHLS database 2002, 2005, 2008, 2011, 2014, and 2018. We removed participants aged < 65 years old or > 105 years old. We also dropped the cases with missing or irregular values. The final sample size for the 2002, 2005, 2008, 2011, 2014, and 2018 wave was 15,059, 14,337, 14,943, 8043, 5756, and 13,889, respectively, with a total of 72,027 cases of 44,154 participants.

2. Interpretation Although a decreasing trend in occurrence of CI in older adults has been reported from several high‐income countries except Japan, studies have been rare from low‐ and middle‐income countries such as China. Our findings on changes of cognitive health among elderly Chinese in the recent decades are new and important for the field. Different from meta‐analyses in previous literatures, our study took advantage of multiple waves of a national survey with a representative sample of older adults in China to reveal a non‐linear trend of CI in China, which reached a peak from 2002 to 2008 and then gradually declined to 2018.

3. Future directions Our findings provide the first evidence of changing trend of CI from China. The data are critical for policy making for aging and health care and deserve attention from clinicians, public health professionals, the government, and relevant non‐governmental organizations. With our research findings, we urge the health care and social welfare system of China to be prepared for the large number of cognitively impaired elderly in the near future. Nationwide initiatives on dementia prevention and brain health promotion should be carefully planned and implemented.

At each wave of survey, the CLHLS collected extensive information on demographics, lifestyle, and health conditions. For this study, multiple data variables were selected to look for potential explanatory variables for the hypothesized trends of increasing CI in the study population. These variables includes age, sex, education, income per capita, residence status (urban vs. rural), marital status, diagnosis of major comorbidity (hypertension, diabetes mellitus, heart diseases, stroke, etc.), cigarette smoking (ever vs. never), alcohol consumption, physical exercise, and usual diet based on a food frequency questionnaire. ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ These factors will act as covariates in the statistical models.

Details about CLHLS, including sampling design, follow‐up interviews with surviving participants, data quality, and the variables collected have been previously published, ²⁹ and can be found in Text S1‐S4 in supporting information.

2.3. Statistical analysis

Standardized prevalence rates (per 100 persons) of CI at each wave of surveys in 2002, 2005, 2008, 2011, 2014, and 2018 from CLHLS by sex and age group were reported.

Generalized estimating equations (GEE) were used to assess changes in prevalence across the six time points. As an expansion of the generalized linear model (GLM), GEE can effectively model repeated categorical variable from longitudinal studies. Logit link function and specified within‐subject covariance were used as first order autoregressive (AR1) in all analyses. Four models were built to examine potential explanatory variables. In model 1, only intercept, data wave, age, sex, education, income, residence status, and marital status were included; in model 2, four cardiometabolic health‐related medical conditions: hypertension, diabetes mellitus, heart disease, and stroke were added. In model 3, lifestyle factors such as cigarette smoking, alcohol consumption, physical exercise, consumption of vegetables, and consumption of fruits were added. In model 4, the logarithm of real family income per capita was considered. Thereafter, subgroup analyses were conducted for young old (≤ 79), the oldest old (≥ 80), women and men, respectively.

We also expand the time trend of CI of elderly population to the whole 20‐year period of CLHLS from 1998 to 2018. Considering that there were only oldest‐old subjects aged 80+ in the 1998 and 2000 waves, we removed the young old aged 65 to 79 from the 20‐year trend analyses of CI. Family income per capita was not included as a covariate in the GEE models because information about the economic condition of respondents was not collected in the 1998 and 2000 waves.

To examine the mechanisms of changes in prevalence of CI, the changes in the variables mentioned above were systematically examined in the investigation period. Statistical analyses were performed using STATA version 14.1. A two‐tailed P value of 0.05 was used as the cut‐off value for statistical significance.

2.4. Role of funding source

The funders provided financial support for data collection and analysis but had no role in the writing of the report, interpretation of the results, or submission for consideration of publication. The first and corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

3. RESULTS

After excluding older adults < 65 or > 105 years old, and observations with missing or repeated data, a study sample of 72,027 observations, contributed from 44,154 CLHLS individuals, was obtained. The sample sizes for 2002, 2005, 2008, 2011, 2014, and 2018 data waved were 15,059, 14,337, 14,943, 8043, 5756, and 13,889, respectively. The data cleaning process can be seen in Figure S1 in supporting information.

Various characteristics of the study samples are presented in Table S1 in supporting information. The mean age varied from 84.17 years to 86.91 years across the six waves. There were more females than males, and the older adults had low education levels in all waves. The proportion of older adults with a diagnosis of hypertension, diabetes, heart disease, and stroke were higher in recent data waves. Figure 1 shows the age‐standardized prevalence rates of CI. The prevalence of CI was higher in women than in men in all six waves. The age‐standardized prevalence increased during the period 2002 to 2008 and then decreased during 2008 to 2018.

FIGURE 1.

Age‐standardized prevalence of cognitive impairment (CI) per 100 persons, standardized based on 2002 age distribution in 5‐year intervals: men versus women. Cognitive impairment uses cut‐offs 16/17, 19/20, 23/24 for those without schooling, 1 to 6 years of education, and more than 6 years of education, respectively.

Table 1 further shows the prevalence rates of CI by sex and age groups. In general, the prevalence of CI among Chinese older adults increased with age at all waves of the surveys. For example, in the 2002 survey, the prevalence of CI was 1.5% in those aged 65 to 69, 12.1% in those aged 80 to 84, 34.7% in those aged 90 to 94, and 58.8% in those aged 100 to 105 year. Concerning the time trend of prevalence of CI, a similar trend was also observed for most age groups, in both men and women.

TABLE 1.

Prevalence rate of cognitive impairment (per 100 persons) in CLHLS cohorts in 2002 to 2018, by sex and age groups.

	2002			2005			2008			2011			2014			2018
Total	No. of subjects	No. of cases	P (%)	No. of subjects	No. of cases	P (%)	No. of subjects	No. of cases	P (%)	No. of subjects	No. of cases	P (%)	No. of subjects	No. of cases	P (%)	No. of subjects	No. of cases	P (%)
All Subjects
65–69	1544	23	1.5	1625	24	1.5	1378	21	1.5	615	9	1.5	205	10	4.9	1430	27	1.9
70–74	1577	35	2.2	1557	48	3.1	1420	60	4.2	1211	41	3.4	862	36	4.2	1626	38	2.3
75–79	1466	69	4.7	1517	82	5.4	1281	72	5.6	1103	51	4.6	1079	51	4.7	1881	96	5.1
80–84	1968	238	12.1	1222	150	12.3	1776	197	11.1	1153	131	11.4	926	103	11.1	2005	169	8.4
85–89	1988	402	20.2	2340	497	21.2	2077	460	22.1	1088	226	20.8	916	156	17.0	1602	279	17.4
90–94	2225	773	34.7	2252	684	30.4	2739	1002	36.6	1172	396	33.8	762	232	30.4	1903	547	28.7
95–99	1375	628	45.7	1393	632	45.4	1433	719	50.2	768	353	46.0	514	224	43.6	1197	566	47.3
100–105	2916	1697	58.2	2431	1500	61.7	2839	1861	65.6	933	582	62.4	492	299	60.8	2245	1360	60.6
Women
65–69	750	12	1.6	806	14	1.7	646	9	1.4	259	6	2.3	75	5	6.7	726	16	2.2
70–74	781	19	2.4	754	19	2.5	646	24	3.7	552	17	3.1	398	13	3.3	754	10	1.3
75–79	728	32	4.4	739	48	6.5	613	36	5.9	503	30	6.0	498	22	4.4	963	57	5.9
80–84	928	127	13.7	599	72	12.0	846	107	12.6	571	82	14.4	468	54	11.5	1041	97	9.3
85–89	1013	217	21.4	1180	286	24.2	1030	270	26.2	530	126	23.8	453	90	19.9	857	165	19.3
90–94	1221	458	37.5	1227	431	35.1	1513	616	40.7	646	253	39.2	411	140	34.1	1021	345	33.8
95–99	846	422	49.9	865	434	50.2	918	487	53.1	474	240	50.6	304	144	47.4	690	348	50.4
100–105	2291	1364	59.5	1912	1218	63.7	2261	1545	68.3	728	482	66.2	390	253	64.9	1673	1061	63.4
Men
65–69	794	11	1.4	819	10	1.2	732	12	1.6	356	3	0.8	130	5	3.8	704	11	1.6
70–74	796	16	2.0	803	29	3.6	774	36	4.7	659	24	3.6	464	23	5.0	872	28	3.2
75–79	738	37	5.0	778	34	4.4	668	36	5.4	600	21	3.5	581	29	5.0	918	39	4.2
80–84	1040	111	10.7	623	78	12.5	930	90	9.7	582	49	8.4	458	49	10.7	964	72	7.5
85–89	975	185	19.0	1160	211	18.2	1047	190	18.1	558	100	17.9	463	66	14.3	745	114	15.3
90–94	1004	315	31.4	1025	253	24.7	1226	386	31.5	526	143	27.2	351	92	26.2	882	202	22.9
95–99	529	206	38.9	528	198	37.5	515	232	45.0	294	113	38.4	210	80	38.1	507	218	43.0
100–105	625	333	53.3	519	282	54.3	578	316	54.7	205	100	48.8	102	46	45.1	572	299	52.3

Note: Cognitive impairment uses cut‐offs 16/17, 19/20, 23/24 for those without schooling, 1 to 6 years of education, and more than 6 years of education, respectively.

Abbreviation: CLHLS, Chinese Longitudinal Healthy Longevity Survey.

Table S2 in supporting information shows results from GEE models. Analysis based on the entire study sample revealed that the prevalence of CI increased until 2008 for non‐stratified analysis or 2011 in stratified analysis for women. The trend then decreased until the last wave in 2018. The observed trend also held when demographic, medical, and lifestyle variables were sequentially controlled for in analyses. Stratified analysis with full adjustments revealed that this pattern generally applied except that the degree of dynamics varied by sex and age group (see Figure 2). Sensitivity analysis with CI defined using higher education‐specific MMSE cut‐off values and a single cut‐off of 17/18 points revealed a similar trend of CI in the entire study sample between 2002 and 2018 (results available in Figures S2 and S3 in supporting information). When we conducted GEE analyses on pre‐2008 and post‐2008 separately, we observed that the time trend of CI on pre‐2008 (2002 to 2008) is positive while the time trend on post‐2008 (2008 to 2018) is negative, which shows that an inverted U shape of the prevalence of CI existed during 2002 to 2018 (Table S3 in supporting information). In the 20‐year time trend (1998 to 2018) analysis, which only included those aged ≥ 80 years, similar trends of CI were observed (Tables S4 and S5 in supporting information). Figures S4 and S5 in supporting information provide graphical representations of the trend that is similar to the inverted U shape trend observed in the young olds.

FIGURE 2.

Time trend of cognitive impairment from 2002 to 2018: odds ratios from hierarchical generalized estimating equations models with 2002 as reference. Cognitive impairment uses cut‐offs 16/17, 19/20, 23/24 for those without schooling, 1 to 6 years of education, and more than 6 years of education, respectively.

The odds ratios of the independent factors were examined to look at the mechanisms of the observed trends. As shown in Table 2, being married, drinking alcohol, having physical exercise, and vegetable consumption were found to be protective against CI, while increasing age, being female, having had a stroke, and being diagnosed with Type II diabetes were found to be risk factors for CI. Although we were not able to calculate joint risk with additive or multiplicative interactions in the GEE models, we could expect that older adults with the joint characteristics of higher age being female, having chronic diseases and low family income, being widowed/divorced, seldom performing physical exercise, and seldom consuming vegetables or fruits to have the highest risk of CI, and those older adults with the joint characteristics of lower age, being male, having no chronic disease, having high family income, being married, regularly performing physical exercise, and consuming vegetables or fruits daily to have the lowest risk of CI. When investigating how these independent factors changed with time from 2002 to 2018, as shown in Figure 3, it was found that the prevalence of major chronic diseases, such as hypertension and Type II diabetes increased, while smoking and drinking both declined rapidly from 2002 to 2018. The economic situation of the older adults measured by income also increased over the period of time.

TABLE 2.

Risk and protective factors of cognitive impairment in the CLHLS cohorts: odds ratios from hierarchical GEE models.

Factors	OR (95% CI)	P value
Protective factors
Income per capita	0.90 (0.88, 0.91)	<0.001
Married and living with spouse	0.75 (0.70, 0.80)	<0.001
Drink alcohol	0.75 (0.70, 0.81)	<0.001
Regular physical exercise	0.39 (0.37, 0.42)	<0.001
Consume vegetables daily	0.75 (0.71, 0.79)	<0.001
Consume fruits daily	0.92 (0.85, 0.99)	0.022
Risk factors
Age (in years)	1.13 (1.13, 1.13)	<0.001
Female	1.23 (1.16, 1.30)	<0.001
Diagnosed stroke	2.50 (2.28, 2.74)	<0.001
Diagnosed diabetes	1.38 (1.20, 1.59)	<0.001
Diagnosed hypertension	0.88 (0.83, 0.94)	<0.001

Note: Cognitive Impairment, as dependent variable, uses cut‐offs 16/17, 19/20, 23/24 for those without schooling, 1 to 6 years of education, and more than 6 years of education, respectively.

Abbreviations: CI, confidence interval; CLHLS, Chinese Longitudinal Healthy Longevity Survey; GEE, generalized estimating equation; OR, odds ratio.

[Correction added on December 13, 2023, after first online publication: Error in Table 2 has been fixed. The row with 1.02 (1.01, 1.03)… has been taken out.]

FIGURE 3.

Time trend of risk and protective factors of cognitive impairment from 2002 to 2018: odds ratios (OR) from hierarchical generalized estimating equations models with 2002 as reference. Cognitive impairment, as dependent variable, uses cut‐offs 16/17, 19/20, 23/24 for those without schooling, 1 to 6 years of education, and more than 6 years of education, respectively.

4. DISCUSSION

It was hypothesized that the prevalence of CI would have increased between 2002 and 2018 in China because of increasing vascular health burden, on top of influences from demographics and lifestyles factors. However, the current study showed that the prevalence of CI increased from 2002 to 2008, and then decreased from 2008 to 2018, when adjusted for demographic, lifestyle, and health factors. The observed pattern of change in CI prevalence was consistent across sex and age groups. The pattern of change could be associated with changes in several modifiable factors including presence of a chronic disease, which increased from 2002 to 2018, and being a smoker or an alcohol drinker, which decreased over the period. The change in prevalence of these factors could be contributing to the increasing and then decreasing trend of CI prevalence.

Findings on changes of cognitive health among elderly Chinese in recent decades are novel and important for the field. In a systematic review, Chan et al. concluded that the prevalence of dementia was increasing in China from 1990 to 2010. ³⁰ Another review by Zhang et al., however, found that the prevalence increased from 1.3% in 1985 to 1990 to 3.9% in 2001 to 2005, and then decreased slightly in 2006 to 2010 (3.6%). ³¹ Different from these meta‐analyses, the current study took advantage of multiple waves of a survey of a national, representative sample of older adults in China to reveal a non‐linear trend of CI in China, which reached a peak from 2002 to 2008 and then gradually declined to 2018.

The observed pattern of changes in CI of older adults in China reflects the profound interactions of compression and expansion of morbidity in population aging, two major opposite forces jointly shaping health changes of the older population. As proposed by Zeng et al., ²⁹ these interactions make up a balance between the so‐called benefits of success and costs of success for aging societies. As a result of such dynamics, the health trends of aged individuals may not always be linear and could even change to the opposite direction with time. ³² , ³³ What this study observed in China is thus not exceptional.

To better understand the non‐linear scenario in China, this study has provided clues of specific mechanisms. As shown in results, less smoking and drinking may reduce risk of CI for Chinese older adults, while the rise of “rich‐man diseases” could simultaneously lead to increased risk of CI. This is in consensus with previous literature, which found that smoking was associated with worse cognitive performance, ³⁴ and heavy drinking was associated with significant impairment in a number of neurocognitive domains. ³⁵ Although between 2002 and 2018 there was a decreasing trend of smoking and drinking among older adults, the prevalence of major chronic diseases, such as hypertension and Type II diabetes, was on an increasing trend. Therefore, we cannot adequately explain the observed inverse U shape trend just by using the individual‐level factors as examined. In this regard, the changes in CI could also be due to certain macro‐historical and social changes, which is beyond these individual‐level factors in the current analysis. The rapid epidemiological transition in China and its related changes in the Chinese medical system, for example, could be an important factor. From 1998 to 2008, there was an increase in number of doctors and hospital facilities. ³⁶ It is thus likely that the recent investment in public health of the Chinese government helped to reverse the increasing trend of CI since 2008. From the perspective of a life‐course approach to age‐related diseases, ³⁷ cognitive status in late life is the consequence of exposure to various risk and protective factors through an entire life and the timing and duration (absolute and relative) of such exposures may make a difference. A major event distinguishing the pre‐2008 waves and post‐2008 waves was the establishment of the People's Republic of China (PRC) in the year 1949. For the waves of 2002, 2005, and 2008, older adults were born before 1949; in contrast, cohorts of 2011, 2014, and 2018 were born after the establishment of PRC. The end of war and rapid improvement of socioeconomic condition such as nutrition, hygiene, medical services, and education could have helped improve the cognitive health of these specific cohorts, and thus reduced the prevalence of CI of later years, as observed in this study. However, more studies are needed to clarify these cohort effects.

Although the future change in prevalence of CI, based on the trend observed in this study, is unclear, we would expect the absolute number of cognitively impaired individuals in China to increase in the coming decades, due to rapid population aging. Proper planning of health care and the social welfare system should be put into place to accommodate the needs of those individuals and their families. Moreover, long‐term monitoring of population cognitive health is necessary to track this trend closely. The 2018 wave marks the eighth wave, which is the wave with the latest available data before the COVID‐19 pandemic hit. The ninth wave began in 2021 and the data will provide further updates on the CI trend after the pandemic. Cognitive screening should be considered for front‐line health‐care practitioners who provide services to the elderly population. The rate of cognitive decline in normal and abnormal aging and related brain changes among the Chinese population should be established as reference for both research and practice purposes.

The criteria used here in defining CI are good surrogates for dementia diagnosis. Based on the latest study, the sensitivity and specificity of the cut‐offs ranged from 87.6% to 94.3% and 80.8% to 94.3%, respectively. ¹⁸ However, in the studies by Wu et al., the increase in prevalence was attenuated when variation in methods was taken into account, suggesting that introduction of new and more inclusive diagnostic criteria may be instrumental in the noted increase in dementia case identification. ³⁸ , ³⁹ In this regard, there are objective imaging markers of brain health available. In the Rotterdam Study, the researchers demonstrated that the 2010 subcohort had larger brain volumes and less cerebral small vessel disease than the 1990 subcohort, which supported their findings on declining dementia incidence. ⁵ Future cohort studies in China should consider such markers of brain aging and brain health.

The strengths of our study are the use of identical assessment tools and criteria of CI over a relatively long study period at multiple time points, the large sample sizes, and the considerations of major risk and protective factors in statistical analyses. In our analysis, the same cognitive assessment tool and the same set of criteria are used across six time points over a 12‐year period, which guaranteed comparability and excluded the possibility of artificial increase or decrease due to more restrictive or inclusive criteria.

The study was not without limitations. First, after 2008, new recruitments were conducted only in 8 longevity counties/cities compared to 745 counties/cities in previous waves, increasing the possibility of a selection bias. Second, the study included relatively small sample sizes in the 2011 and 2014 waves due to no new recruitment conducted to replace the losses compared to other waves. Third, there were no objective measures of blood pressure and fasting glucose, and health conditions such as hypertension, diabetes, heart disease, and stroke were self‐reported. Underdiagnosis of those medical conditions are highly possible and may introduce unknown influences on the analysis. Last, CI was defined based on the total score from MMSE, which may not be as accurate as operational definitions based on detailed neuropsychological assessment and structured clinical interview.

Cognitive impairment and dementia are complex clinical conditions that are affected by aging and genetics and modified by multiple factors, and it is necessary for the biological mechanisms to be further elucidated in future studies. Due to the complex pathophysiological pathways, it would be useful to identify biomarkers in association with cognitive changes underlying CI and dementia.

Nationwide initiatives on dementia prevention and brain health promotion should be carefully planned and implemented to slow the effect of a rapidly aging population. Randomized controlled trials from the developed world have produced promising results on potential measures. ⁴⁰ , ⁴¹ Findings from the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability, a large, long‐term, randomized controlled trial, suggest that a multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring could improve or maintain cognitive functioning in at‐risk older adults from the general population. ⁴⁰ Another study also suggested that cognitive training could result in improved cognitive abilities specific to the abilities trained, and reasoning training could result in less functional decline. ⁴¹ An early dementia prevention program also has been initiated in comparable populations such as Singapore. ⁴² , ⁴³ Existing approaches and models would be useful, but research evidence derived from China's own population must be produced to support major public health measures. Moreover, it is known that population‐based epidemiological research with use of consistent methods can provide robust evidence for policy making and dementia care planning and a comprehensive understanding of health in old age. ³⁹

Overall, the data are critical for policy making for the aging population and health care and deserve attention from clinicians, public health professionals, the government, and relevant non‐governmental organizations. Based on the findings and the rapidly aging population, we urge the health care and social welfare system of China to be prepared for the large number of cognitively impaired older adults.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

The CLHLS study was approved by the Research Ethics Committee of Peking University (IRB00001052‐13074). Written informed consent was obtained from all study participants and their proxy respondents.

Supporting information

Supporting Information

ICMJE Disclosure Forms

Chen H, Ye KX, Feng Q, et al. Trends in the prevalence of cognitive impairment at old age in China, 2002–2018. Alzheimer's Dement. 2024;20:1387–1396. 10.1002/alz.13545