Cognitive frailty as a predictor of adverse outcomes among older adults: A systematic review and meta‐analysis

Abstract Objective A systematic review and meta‐analysis basing on the prospective cohort studies were conducted to explore the risk of all‐cause mortality and dementia in cognitively frail older adults compared to robust older adults and to determine whether it was a predictor of adverse outcomes. Methods Pubmed, Web of Science, The Cochrane Library, EMBASE, and CINAHL databases were searched to retrieve studies on adverse outcomes of cognitive frailty. Two reviewers independently screened the literature, extracted data, and assessed the risk of bias of the included studies. Stata 15.0 Software was used to perform the meta‐analysis. The all‐cause mortality and dementia were observed to be the primary outcomes, while the other data were considered as the secondary outcome. Results A total of 14 studies were included in qualitative analysis and 12 studies were included in the meta‐analysis, with low risk of bias and moderate to good methodological quality. The results showed that cognitive frailty in older people had a higher risk of all‐cause mortality [HR = 1.93, 95%CI (1.67, 2.23), p < .001] and dementia [HR = 3.66, 95%CI (2.86, 4.70) as compared with robust. The subgroup analysis showed that the assessment tools were the main source of heterogeneity. Conclusion In older adults living in communities, the cognitive frailty was found to be a significant predictor of all‐cause mortality and dementia. Nonetheless, cognitive frailty was found to be a better predictor of all‐cause mortality and dementia than just frailty.

manifestation characterized by the simultaneous presence of both physical frailty and cognitive impairment. The key factors defining such a condition include (a) coexistence of physical frailty and cognitive impairment (Clinical Dementia Rating [CDR] = 0.5); and (b) exclusion of concurrent Alzheimer's disease or other dementias (Kelaiditi et al., 2013), that is, cognitive impairment caused by physical factors.
Existing research (Boyle et al., 2010;Canevelli et al., 2014;Solfrizzi et al., 2017) shows that physical frailty increases the risk of dementia in populations with normal cognition and accelerates cognitive decline in individuals. Current evidence has shown that cognitive functioning is the key to reversing the frailty state (Aprahamian et al., 2018).
Moreover, the annual growth rate from mild cognitive impairment to dementia is relatively very high (5%~15%), and extensive neuronal loss and irreversible neuronal damage are observed at this stage (Liu et al., 2018;Panza et al., 2015). Therefore, some scholars believe that mild cognitive impairment may not be the best intervention stage at which to prevent dementia.
As mentioned above, frailty and cognitive impairment were often viewed as two independent concepts in previous studies, which helps to respectively predict adverse outcomes (Jacobs et al., 2011). But when they coexist, cumulative negative effects are often detected, significantly increasing all-cause mortality or other adverse outcomes (St John et al., 2017), and the precursory signs need to be particularly identified that lead to adverse outcomes, and to target the specific interventions. Hence, cognitive frailty may be a physiological precursor to the degenerative nervous system diseases, which could play a key role in predicting the short-term and the long-term all-cause mortality, dementia, disability, and other adverse health outcomes. Additionally, it could also offer a new target for the prevention and intervention of the pathological aging and the adverse outcomes.
This study used meta-analysis to explore the risk of all-cause mortality and dementia in older adults with cognitive frailty compared to robust elderly participants. Furthermore, the understanding of the relationship between frailty and geriatric cognitive disorders help healthcare professionals become more informed of cognitive frailty and the associated risks. It will also help contribute to new interventions and management of geriatric cognitive disorders.

| ME THODS
This systematic review was drafted in accordance with the protocol of the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009) for transparent reporting of the systematic reviews and meta-analysis, registered on the PROSPERO website (CRD42020186200).

| Search strategy
A computer-based search of Web of Science, PubMed, Cochrane Library, CINAHL, and EMBASE was carried out from May 10, 2020.
At the same time, manual retrieval of cognitive decline-related reviews and a systematic evaluation of references was conducted to ensure the comprehensiveness of results. We used Medical Subject Headings (MeSH), free term, and word variants for two main themes, that is, "cognitive frailty" and "cohort study," and these were combined with Boolean operator OR/AND. The search terms included "cognitive frailty" OR "cognitive decline" OR "cognitive impairment" AND (frail* OR pre-frail* OR "frailty syndrome") AND ("cohort study" OR "longitudinal study") (Take PUBMED for example in Appendix S1).

| Inclusion criteria and exclusion criteria
Inclusion criteria: (a) a prospective cohort study or population-based longitudinal study; (b) the subjects included older people in the community aged 60 years old and above; (c) cognitive frailty needs to be defined as physical frailty and cognitive impairment simultaneously, while the internationally agreed-upon diagnostic criteria were used to define physical frailty (e.g., Frailty Phenotype, Frail Scale, Frailty Index, or others) and the cognitive impairment (Minimum Mental State Examination, Clinical Dementia Rating, or others), and (d) reported the hazard ratio (HR), the odds ratio (OR), or risk ratio (RR) of the primary outcomes (all-cause mortality and dementia) and other adverse outcomes or basic data that could facilitate the calculation of the above values.
Exclusion criteria: (a) If there was literature with identical data, studies with larger sample sizes were selected; (b) low-quality literature (NOS score < 5 points); (c) non-English language literature; and (d) full text could not be found.

| Study selection and data extraction
Two researchers independently searched for and selected literature, then extracted data for cross-checking. In the case of any differences, a third party was consulted to settle the issue. The screening process was carried out in strict accordance with the PRISMA guidance process. After reading the title and abstract of the article, literature that clearly did not meet the acceptance criteria was excluded, and articles that had initially been included were subsequently excluded after having read the full text. The basic information that was extracted included the author, publication time, country, sample size, average age, the definition of cognitive decline (assessment tools and diagnostic criteria), outcome indicators, follow-up time, adjustment factors, and the NOS score.

| Quality evaluation
Two researchers used the New Castle Ottawa Scale (NOS) (Stang, 2010) to independently evaluate the quality of the included literature, and a third party was consulted in the event of a disagreement.
The quality evaluation table consisted of eight items, which were divided into three categories as follows: selection, comparability, and results. The total score was nine points. When the score was ≥5 points, it was deemed high-quality literature.

| Statistical analysis
The primary outcomes of this systematic review were considered to be all-cause mortality and dementia. The other data were considered as secondary outcomes. The hazard ratios (HR) with 95% confidence intervals (CI) of the primary outcomes in the case of cognitive frailty compared robustly with the noncognitive frailty that were extracted from the studies for meta-analysis.
The subgroup analyses were conducted depending upon the different assessment tools and the different models of cognitive frailty. Since the physical frailty involved both prefrailty and frailty, the cognitive frailty involved both prefrailty cognitive impairment, along with frailty and the cognitive impairment (Avila-Funes et al., 2009;Montero-Odasso et al., 2016;Shimada et al., 2018;Zheng et al., 2020). Ruan (Ruan et al., 2015) divided cognitive frailty into two subtypes: the reversible and the potentially reversible. The cognitive impairment of reversible cognitive frailty is SCD and/or positive biomarkers resulting from physical factors. The cognitive impairment of potentially reversible cognitive frailty is MCI (CDR = 0.5). As mentioned above, and according to the description of each research study, the types of cognitive frailty were divided into four groups as follows: cognitive impairment (CI) + frail, CI + prefrail, CI + (pre)frail, and reversible cognitive frailty.
Meta-analysis was conducted by Stata 15.0 software. (a) Pass I 2 and the chi-square test (the test level was α = 0.1) were carried out to determine the heterogeneity among the included studies. If I 2 < 50%, p > .1, this indicated that the heterogeneity was small, and the fixed effect model was used; if I 2 ≥ 50%, p ≤ .1, this indicated that the heterogeneity was large. In this case, the random effect model was used; F I G U R E 1 Flow diagram of the selection process of the studies. CINHAL, Cumulative Index to Nursing and Allied Health Literature; HR, hazard ratio; OR, odds ratio

| RE SULTS
A total of 1,661 documents were retrieved, of which 967 were retrieved by importing the title into the endnote software.
Forty nine articles were left over after reading the title and the TA B L E 2 Quality assessment of included studies based on the Newcastle-Ottawa scale

Author/Year
Selection Representative-ness of the exposed cohort Selection of the non exposed cohort Ascertain-ment of exposure to implants Demonstration that outcome of interest was not present at start of study 1 1 1 1 F I G U R E 2 Forest plot of the association between different types of cognitive frailty and the all-cause mortality. CI: cognitive impairment; a: cognitive impairment + prefrail; b: cognitive impairment + frail; ①: followup 3.5 years; ②: follow-up 7 years; HR: hazard ratio; 95%CI: 95% confidence interval abstract, while 14 articles were included after reading the full text. Due to the limited number of studies on secondary outcome indicators, it was not possible to combine meta-analyses.
A descriptive analysis was carried out, and finally, 12 reports were included for meta-analysis. The specific screening process is shown in Figure 1.

| Characteristics of the included studies
In this study, 14 cohort studies were included. There were 57,559 subjects, and the average age ranged from 71.5 to 93.6 years old, with a prevalence of cognitive frailty between 2.5% and 50%. The subjects were followed up for a period of 2-14 years. Five assessment tools were utilized to diagnose the physical frailty, and three assessment tools were used to define the cognitive impairment.  Table 1.

| Methodological quality
The 14 cohort studies were evaluated for methodological quality using the NOS. The bias risk scores of all the reports ranged between 6 and 9 (total score of 9). The included studies indicated moderate to good methodological quality and exhibited low risk of bias. The highest risk of bias in this review was assessed for Outcome and Comparability. Most of the earlier studies had not followed up long enough for the outcomes to happen (n = 8), whereas certain other studies did not modify in order to indicate the specific control for the second important factor (n = 7), as indicated in Table 2.

| All-cause mortality
A total of 10 studies ( To explore the source of heterogeneity, this study further examined the relationship between cognitive frailty and all-cause mortality by carrying out a subgroup analysis of models of cognitive frailty and assessment tools. The results of the subgroup analysis are shown in Figure 2. Four models of cognitive frailty increased the incidence of mortality in older people. In addition, heterogeneity among the groups decreased, and heterogeneity disappeared in the CI + frail group and in the reversible cogni-  Figure 3.

| Sensitivity analysis
A sensitivity analysis of all-cause mortality related to the main adverse outcomes was carried out, and a small difference was found between the combined effect value and the total combined effect value after each study was eliminated, indicating that the results of this study showed high stability (see Figure 5 for details.)

| Publication bias
The bias analysis of all-cause mortality was carried out using Begg's test and Egger's test. The funnel plot showed that individual studies deviated from the confidence interval, as shown in Figure 6.
However, the results of Begg's test and Egger's test showed that p = .436 and p = .177, respectively, which was not statistically significant, and the possibility of publication bias was small. No published bias analysis was conducted because the number of studies on dementia included in the meta-analysis was <10.

| D ISCUSS I ON
In this research, 14 cohort studies were used to evaluate the risk of all-cause mortality, dementia, and other adverse outcomes of cognitive frailty, to identify the related risks and take corresponding measures to follow-up and intervene. The results showed that the risk of all-cause mortality, dementia, and other adverse health outcomes in older adults with cognitive frailty was higher than that in older adults without cognitive frailty. The results of the subgroup analysis, based on the types of cognitive frailty and assessment tools, did not change significantly, which was still statistically significant. It is suggested that cognitive frailty is a key risk factor for poor health outcomes among older people in the community, which has predictive value. As a heterogeneous geriatric syndrome, the pathogenesis of cognitive frailty has not been clarified, which may be related to hormones, inflammatory responses, cardiovascular risk factors, and so on (Panza et al., 2018). Cognitive impairment and physical frailty have similar time trajectories and pathological mechanisms (Furtado et al., 2019), and they interact with each other. (1.07, 1.67)] (Kojima et al., 2016). Our results illustrated that the community-dwelling older adults with cognitive frailty had a higher risk of developing dementia. Hence, it could be surmised that the risk of all-cause mortality and dementia of cognitive frailty was significantly higher than that of the physical frailty (alone), and its impact on the adverse outcomes seemed to be cumulative, with better predictive values (Especially dementia). In addition, timely intervention is also important for many negative health outcomes caused by cognitive frailty. A number of studies (Sink et al., 2015;Zhang et al., 2020) have shown that exercise can delay the progress of debilitation, prevent the occurrence of cognitive impairment, and a Mediterranean diet can also effectively prevent and improve cognitive impairment (Valls-Pedret et al., 2015). According to the definition of cognitive frailty, we can carry out more joint interventions in many fields and modes, such as physical exercise combined with diet changes, to find effective interventions that can reduce the incidence of adverse outcomes.
At present, there is no uniform standard for the assessment of cognitive frailty, so the prevalence of cognitive decline in this study varied from 2.5% to 50%, and the results of the subgroup analysis of all-cause mortality also showed that most of the heterogeneity derived from different assessment tools. In this study, FP is often used to evaluate physical frailty, including self-reported fatigue, muscle strength decline, slow gait, involuntary weight loss, and a low level of physical activity. Three or more of these factors indicate frailty, while one or two suggests prefrailty (Fried et al., 2001). At present, there is no unified judgment standard for each index of FP. In terms of its application process, it needs to be adjusted in combination with the heterogeneity of people in different countries (King-Kallimanis et al., 2014). Secondly, the results of the subgroup analysis showed that the all-cause mortality rate of the FP combined with the cognitive impairment assessment tool group was lower than that of other groups, which may be due to the fact that the FP did not account for comorbidity, psychosocial and other conditions, which resulted in the low number of patients detected, in addition to the low positive rate of the follow-up results. Solfrizzi (Solfrizzi et al., 2019) combined the physiological, the psychological, and the social domains to identify a new biopsychosocial frailty (BF) phenotype. The BF structure was found to be associated with the short-and the long-term risks of developing overall dementia. The frailty of the older adults to dementia risk was not fully captured relatively to the comparative physiological and defect accumulation approaches, while the BF model added significant value in the case of both the assessment and the target of intervention during frailty. Some of the studies employed the frailty index (FI) (Searle et al., 2008), which is based on the cumulative health defect method to evaluate the decline, including any symptoms and signs related to adverse outcomes.
The degree of decline is divided into frailty (FI > 0.25), prefrailty (FI = 0.20 ~ 0.25), and robust (FI < 0.25) (Fried et al., 2001). As the health defects in multiple dimensions are captured, the FI can facilitate more comprehensive and accurate classification criteria and risk predictions. Multiple system assessment (Kojima et al., 2018;Malmstrom et al., 2014;Theou et al., 2013) also confirms that the prediction effectiveness of FI on disability, mortality, and other adverse outcomes is better than that of other assessment instruments, which is consistent with the results of the subgroup analysis in this study. However, in practice, it is cumbersome to collect dozens of items related to health defects, which increases the workload of medical staff, and as such, the development of information technology (such as electronic medical records) can better reflect its advantages. Compared with the FI, the FP can identify debilitated patients more easily and quickly.
In conclusion, it is necessary to further explore the evaluation indicators with higher sensitivity to physical frailty, and to simplify the scale and ensure its effective prediction ability, so as to avoid situations in which false positives and false negatives are reported. Ruan (Ruan et al., 2015) divided cognitive frailty into two subtypes: Reversible cognitive frailty indicates physical frailty and subjective cognitive decline, while potential reversible cognitive decline involves physical frailty and cognitive impairment.
In this study, MMSE was used to diagnose cognitive impairment in cognitive frailty and to identify the criteria that were primarily adapted to study potential reversible cognitive frailty, while less attention was paid to recognizing reversible cognitive frailty. Only Solfrizzi et al. (2017) analyzed the predictive effect of reversible cognitive frailty on adverse outcomes, while subjective cognitive decline, as a precognitive impairment and form of dementia, is associated with better reversibility, which is an important stage in the secondary prevention of dementia. Therefore, in addition to the importance of reversible cognitive frailty, we should use the diagnostic criteria and screening tools of subjective cognitive decline in the early stage of mild cognitive impairment, combined with biomarkers to evaluate the cognitive decline.
Limitations of this study: (a) fewer outcome indicators that can be combined for meta-analysis, such as the hospitalization rate, the decline in the ability to engage in activities, and other adverse outcomes; only 1-2 of such indicators, which produces incomplete

| CON CLUS ION
This systematic assessment exhibited a high prevalence of the cognitive frailty in community-dwelling older adults, while there were no uniform assessment instruments to evaluate this syndrome. This meta-analysis had indicated that the cognitive frailty increased the risk of the all-cause mortality and dementia among older adults.
Moreover, the cognitive frailty was a better predictor of the allcause mortality and dementia than just frailty (alone).

| IMPLIC ATIONS
By shedding light on the impact of cognitive frailty on the occurrence of adverse outcomes in the older adults, this study believes that the medical staff could identify the risks of cognitive frailty and the high-risk groups as early as possible, besides taking corresponding prevention and intervention measures to reduce the occurrence of adverse outcomes. At the same time, a large-scale, multicenter clinical investigation was carried out, and standardized assessment tools with high sensitivity and convenience were developed according to different research sites to explore the risk factors of cognitive frailty and effective intervention measures, so as to delay any further progression of the disease.

ACK N OWLED G M ENTS
This study was supported by the National Natural Science Foundation of China (71673149). We thank China Finance Bureau for supporting the project.

CO N FLI C T O F I NTE R E S T
No conflict of interests.

AUTH O R CO NTR I B UTI O N
Zihan Bu designed the study, extracted and analyzed the data, performed a meta-analysis, and wrote the manuscript. Anle Huang performed data extraction and revised the manuscript. Mengting Xue and Qingyun Li collected the data. Yamei Bai and Guihua Xu supervised the project and revised the paper.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1926.