To the best of our knowledge, this study is the first to develop and validate a diagnostic model for identifying community-dwelling elderly individuals with aMCI, and we have gathered numerous relevant factors. Finally, our model includes 11 significant variables: MoCA, MMSE, IADL, center, education, job, planting flowers/keeping pets, singing, Number of hobbies, UOB, and UP. The predictive model demonstrates good discrimination and calibration, as well as excellent clinical applicability.
It is widely recognized that there is no gold standard for aMCI diagnosis, and a detailed cognitive function assessment is a crucial diagnostic method. Consequently, cognitive function scores such as MMSE and MoCA are important predictive variables. However, it is not sufficient to solely rely on cognitive function scores as predictive variables. The results of this study, which attempt to analyze the independent predictive value of MoCA on aMCI using ROC, are unsatisfactory and unstable. Please refer to Supplementary Fig. 1 for further details. ADL, especially IADL, serves as a crucial predictor of aMCI32. This finding is consistent with Hyojin P's research11. While the ADL score is almost normal, IADL can be slightly impaired among elderly individuals with MCI.
Bai W et al.12reported that the presence of MCI among community dwellers is influenced by the region of study sites. Our research aligns with this conclusion. The prevalence of aMCI among older adults in the Hongshan center was 8.41times higher than that in the Wuchang center (95% CI: 3.99–17.71). Various complex factors contribute to the differences between the two regions, and the underlying reasons are beyond the scope of this study. We plan to explore the reasons for the disparity between the two centers in future research.
Education level is a recognized predictor of aMCI11,12. In this study, we took a tailored approach instead of using a "one size fits all" method that divides cognitive function by a certain score. We evaluated MMSE and MoCA scores based on educational level. Compared with the normal group, the MCI group had a higher proportion of lower education levels, but the difference was not statistically significant (see Supplementary Table 2). Engaging in manual work before retirement is a predictor of aMCI in this study. People who engaged in jobs with higher cognitive requirements tend to exhibit less cognitive decline. Manual labor before retirement poses a risk factor for cognitive impairment in the elderly. Literature suggests that the incidence rate of MCI of mental workers is far lower than that of manual workers33. A systematic review and meta-analysis conducted abroad suggest that increased physical activity can predict the transition from any type of MCI to all-cause dementia34. Our previous research has found that manual workers have a higher rate of MoCA damage compared to mental workers35
Previous studies have demonstrated a higher incidence of MCI among individuals with few or no hobbies, and different hobbies have varying impacts on cognitive function33,35,36. Our research aligns with this conclusion. Most studies have focused on investigating the general number of interests and hobbies, without exploring the relationship between specific hobbies and cognitive function, and there is a lack of data supporting the composition of different interests and hobbies in the MCI population. This survey conducted a classified survey on common hobbies and found that engaging in hobbies such as planting flowers, keeping pets and singing was associated with a decreased risk of aMCI.
Furthermore, this study examined the correlation between blood and urine biomarkers and aMCI. Previous studies37 have reported leukocytosis as an independent risk factor for Parkinson's disease, while other studies have found no significant difference in white blood cell (WBC) counts between the MCI group and the normal group22. It remains unclear whether the analysis of systemic inflammatory markers, represented by white blood cells, can serve as a useful indicator for predicting MCI. In this study, we observed that lower WBC counts appeared to be a predictor of aMCI. However, WBC counts did not ultimately enter the model. In the future, it is necessary to further control for confounding factors and actively explore the correlation between more microscopic inflammatory indicators and MCI.
Compared to other blood lipid indicators, HDL-C is considered important in the process of neurodegeneration and may be a stronger predictor of MCI and AD38. Liu Y et al.39reported a significant negative correlation between plasma HDL-C levels and the risk of MCI in elderly Chinese individuals, which has also been validated in the Japanese elderly population38. In this study, lower HDL-C levels appeared to be a predictor of aMCI. HDL-C is a protective anti-atherosclerosis plasma lipoprotein that can promote the reverse transport of cholesterol, exhibit antioxidant, anti-inflammatory, and other beneficial effects. Lower HDL-C levels have been associated with more severe white matter lesions, which can contribute to the development of MCI and even AD40. We will further examine the correlation between HDL and aMCI in a distinct article.
Urine serves as an important source of disease-related biomarkers, which can be effortlessly acquired, entirely non-invasive, continuously gathered and can reflect the overall state of the body41. There have been numerous studies exploring the correlation between urine and kidney diseases, as well as its use as an indicator for evaluating safety of treatment plans, there have been limited investigations into the correlation between routine urine tests and MCI. Bikbov B et al.42 have reported that an increase in proteinuria may play a role in the development of MCI or dementia. This study collected urine samples from individuals with aMCI and normal elderly individuals. The routine urine examination revealed that urinary occult blood (UOB) and urinary protein (UP) were predictors of aMCI, which warrants further exploration.
Studies predicting aMCI in community elderly people over 65 years old have not been reported. Recent researches have selected community middle-aged and older adults, Pu L, et al.6 used public databases to develop a predictive model for the risk of cognitive impairment, utilizing demographic information variables (AUC = 0.775). Mengli Huang et al.10 developed a nomogram for predicting MCI, which demonstrated good discrimination power (AUC = 0.870) and good calibration. As to older adults, Penfold R B, et al.43 developed a machine learning model to predict MCI using natural language processing in the absence of screening (AUC = 0.67). Li X, et al.44 reported the AUC values of RS models were between 0.64 and 0.785. Climent M T, et al.45 have developed a decision tree model to identify individuals at risk of MCI among non-institutionalized community elderly (AUC = 0.763). As previously mentioned, current researches are focused on predicting MCI using cerebrospinal fluid and imaging indicators. Prins S, et al.46 developed an algorithm based on biomarkers obtained through non-or minimally invasive procedures to identify healthy elderly subjects who have an increased risk of abnormal cerebrospinal fluid and amyloid beta 42 levels (AUC = 0.65). Pinaya W H L, et al.47 develop a model that performed on six queue data using magnetic resonance imaging results, and the model's performance in classifying normal aging and MCI was average (AUC was between 0.49 and 0.64). The AUC of our model was 0.787 (95% CI: 0.753–0.821) in the training set, and 0.780 in the validation set. In general, our model has easy collection of predictive indicators, good clinical applicability, and is suitable for promotion in community health services.
In summary, this study holds significant clinical importance as it aids in the identification of high-risk individuals with aMCI in the community through simple and feasible surveys as well as physical examinations.
Limitations and strengths
There are some limitations to this study. Firstly, the model is a diagnostic model that only utilizes one wave of data for modeling. Going forward, additional longitudinal data is necessary to monitor the fluctuations in cognitive function within this population. Secondly, this study population resided in an urban area of central China, thus it may not be representative of rural populations and other races, and our model was specifically developed for predicting aMCI among community-dwelling elders, so the results may not be generalized to hospitalized patients. Furthermore, our model has only undergone internal validation and lacks external validation, additional validation and model optimization will be conducted in other centers in the future.
Despite its limitations, this study still has several important strengths: This study has a rigorous design, in which the assessment of cognitive function was classified by experienced geriatric psychiatrists in accordance with strict criteria, and the diagnosis of MCI does not rely solely on one scale such as MMSE or MoCA; what’s more, the detailed questionnaire allows us to consider many important confounding factors for the found associations, in this study, we excluded the effects of diseases and vascular factors on MCI and included elderly individuals with aMCI who mainly experienced memory decline, ensuring the homogeneity of the study; Finally, the predictive indicators of this model are easy to obtain, our team has developed a cloud platform for data collection and storage, for screening, follow-up, and management of aMCI elderly in community, which is conducive to early identification and management of target populations, with good guiding implications in public health, clinical settings and health aging.