Purpose: This study aimed to develop a fall prediction model for community-dwelling older adults using machine learning. Methods: The present study was conducted with a secondary data analysis that used data from the 2020 national survey of older Koreans. Among 10,097 participants, data 177 were excluded due to incompleteness and 9,920 were included in the final analysis. Because of data imbalance, upsampling was performed to increase the number of individuals who fell. Forty-five independent variables for fall prediction were selected based on the fall risk factors from previous studies and univariate statistical analysis. The data were split into training and testing sets at an 80:20 ratio. Three machine learning algorithms—logistic regression, random forest, and artificial neural network—were used to develop a fall prediction model. Results: The random forest model outperformed the others, with an area under the curve of .91, accuracy of .94, precision of .94, recall of .74, and F1 score of .83. An analysis of feature importance revealed that satisfaction with health condition, visual difficulty, instrumental activities of daily living, performance of 400m walk, and cognitive ability were the top five features for fall prediction. Conclusion The fall prediction model developed using machine learning demonstrated high model performance, implying its suitability for use as a primary screening tool for fall risk. Subjective satisfaction with one’s health should be considered as an important factor in predicting falls in community-dwelling older adults. It is necessary for community health nurses to reinforce positive health awareness by continuous disease management and physical function improvement for older adults to prevent falls.

Purpose: This study aimed to develop a fall prediction model for community-dwelling older adults using machine learning. Methods: The present study was conducted with a secondary data analysis that used data from the 2020 national survey of older Koreans. Among 10,097 participants, data 177 were excluded due to incompleteness and 9,920 were included in the final analysis. Because of data imbalance, upsampling was performed to increase the number of individuals who fell. Forty-five independent variables for fall prediction were selected based on the fall risk factors from previous studies and univariate statistical analysis. The data were split into training and testing sets at an 80:20 ratio. Three machine learning algorithms—logistic regression, random forest, and artificial neural network—were used to develop a fall prediction model. Results: The random forest model outperformed the others, with an area under the curve of .91, accuracy of .94, precision of .94, recall of .74, and F1 score of .83. An analysis of feature importance revealed that satisfaction with health condition, visual difficulty, instrumental activities of daily living, performance of 400m walk, and cognitive ability were the top five features for fall prediction. Conclusion The fall prediction model developed using machine learning demonstrated high model performance, implying its suitability for use as a primary screening tool for fall risk. Subjective satisfaction with one’s health should be considered as an important factor in predicting falls in community-dwelling older adults. It is necessary for community health nurses to reinforce positive health awareness by continuous disease management and physical function improvement for older adults to prevent falls.

Purpose: This study aimed to develop a fall prediction model for community-dwelling older adults using machine learning. Methods: The present study was conducted with a secondary data analysis that used data from the 2020 national survey of older Koreans. Among 10,097 participants, data 177 were excluded due to incompleteness and 9,920 were included in the final analysis. Because of data imbalance, upsampling was performed to increase the number of individuals who fell. Forty-five independent variables for fall prediction were selected based on the fall risk factors from previous studies and univariate statistical analysis. The data were split into training and testing sets at an 80:20 ratio. Three machine learning algorithms—logistic regression, random forest, and artificial neural network—were used to develop a fall prediction model. Results: The random forest model outperformed the others, with an area under the curve of .91, accuracy of .94, precision of .94, recall of .74, and F1 score of .83. An analysis of feature importance revealed that satisfaction with health condition, visual difficulty, instrumental activities of daily living, performance of 400m walk, and cognitive ability were the top five features for fall prediction. Conclusion The fall prediction model developed using machine learning demonstrated high model performance, implying its suitability for use as a primary screening tool for fall risk. Subjective satisfaction with one’s health should be considered as an important factor in predicting falls in community-dwelling older adults. It is necessary for community health nurses to reinforce positive health awareness by continuous disease management and physical function improvement for older adults to prevent falls.

Purpose: This study aimed to develop a fall prediction model for community-dwelling older adults using machine learning. Methods: The present study was conducted with a secondary data analysis that used data from the 2020 national survey of older Koreans. Among 10,097 participants, data 177 were excluded due to incompleteness and 9,920 were included in the final analysis. Because of data imbalance, upsampling was performed to increase the number of individuals who fell. Forty-five independent variables for fall prediction were selected based on the fall risk factors from previous studies and univariate statistical analysis. The data were split into training and testing sets at an 80:20 ratio. Three machine learning algorithms—logistic regression, random forest, and artificial neural network—were used to develop a fall prediction model. Results: The random forest model outperformed the others, with an area under the curve of .91, accuracy of .94, precision of .94, recall of .74, and F1 score of .83. An analysis of feature importance revealed that satisfaction with health condition, visual difficulty, instrumental activities of daily living, performance of 400m walk, and cognitive ability were the top five features for fall prediction. Conclusion The fall prediction model developed using machine learning demonstrated high model performance, implying its suitability for use as a primary screening tool for fall risk. Subjective satisfaction with one’s health should be considered as an important factor in predicting falls in community-dwelling older adults. It is necessary for community health nurses to reinforce positive health awareness by continuous disease management and physical function improvement for older adults to prevent falls.

Primary afferents sensitive to capsaicin and noxious heat express vanilloid receptor 1(VR1) in both their peripheral and central fibers and terminals. We used multiple immunofluorescence and confocal microscopy to characterize their pattern of termination in rat spinal cord, colocalization of neurochemical markers of primary afferents and other presynaptic receptors. VR1-positive unmyelinated fibers mainly terminate in lamina I, where they co-stain for CGRP, and to a limited extent for SP, and in lamina II, especially its medial half, where they co-stain for IB4. VR1 positive thin myelinated fibers terminate in lamina I and co-stain for the neurochemical tracer CTB, injected in the sciatic nerve. As revealed by simultaneous staining for the synaptic marker synaptophysin, VR1-positive terminals are abundant in lamina I and sparse in lamina II. In L6-S1 spinal cord, VR1-positive fibers and terminals were abundant in Lissauer's tract, lamina I-V, medial collateral path to lamina X, and lateral collateral path to sacral parasympathetic nucleus. Most of VR1 positive fibers in Lissuer's tract and LCP are colocalized with SP. In conclusion, it is suggested that VR1 positive fibers in spinal cord are both peptidergic and non-peptidergic, IB4 positive fibers, mediating both somatic and visceral sensations, and that peptidergic VR1 positive fibers are mainly related with visceral sense.
Animals ; Capsaicin ; Fluorescent Antibody Technique ; Ganglia, Spinal ; Hot Temperature ; Microscopy, Confocal ; Myelin Sheath ; Negotiating ; Rats* ; Receptors, Presynaptic ; Sciatic Nerve ; Sensation ; Spinal Cord ; Synaptophysin

Various probiotic strains have been reported to affect emotional behavior. However, the underlying mechanisms by which specific probiotic strains change brain function are not clearly understood. Here, we report that extracellular vesicles derived from Lactobacillus paracasei (Lpc-EV) have an ability to produce genome-wide changes against glucocorticoid (GC)-induced transcriptional responses in HT22 hippocampal neuronal cells. Genome-wide analysis using microarray assay followed by Rank-Rank Hypergeometric Overlap (RRHO) method leads to identify the top 20%-ranked 1,754 genes up- or down-regulated following GC treatment and their altered expressions are reversed by Lpc-EV in HT22 cells. Serial k-means clustering combined with Gene Ontology enrichment analyses indicate that the identified genes can be grouped into multiple functional clusters that contain functional modules of “responses to stress or steroid hormones”, “histone modification”, and “regulating MAPK signaling pathways”. While all the selected genes respond to GC and Lpc-EV at certain levels, the present study focuses on the clusters that contain Mkp-1, Fkbp5, and Mecp2, the genes characterized to respond to GC and Lpc-EV in opposite directions in HT22 cells. A translational study indicates that the expression levels of Mkp-1, Fkbp5, and Mecp2 are changed in the hippocampus of mice exposed to chronic stress in the same directions as those following GC treatment in HT22 cells, whereas Lpc-EV treatment restored stress-induced changes of those factors, and alleviated stress-induced depressive-like behavior. These results suggest that Lpc-EV cargo contains bioactive components that directly induce genome-wide transcriptional responses against GC-induced transcriptional and behavioral changes.

Animal models of depression are used to study pathophysiology of depression and to advance therapeutic strategies. Stress-induced depression models in rodents are widely used. However, amenable behavioral criteria and experimental procedures that are suitable for animal models have not been established. Given that depression is clinically diagnosed by multiple symptomatic criteria and stress effects are imposed to the brain non-specifically in stress-induced depression models, analyses of depression states in rodents using multiple symptomatic criteria may provide more power than any methods relying on a single symptomatic criterion. To address this, C57BL/6 inbred mice were restrained for 2 h daily for 14 d, and depression states of individual mice were assessed using the U-field test, behavioral assessment developed to measure animal's sociability, and the tail suspension test and/or forced swim test, which are the typical methods that measure psychomotor withdrawal states. Although the majority of these mice showed severe depressive behaviors in both tests, a significant proportion of them, which were all inbred mice and received the same amount of restraints, expressed differential depression states in the sociability test and psychomotor withdrawal tests. To easily read-out differential depression states of individuals in two different tests, a standard method and basic parameters required to construct two-way behavior matrix were introduced. The utility and features of this two-way behavior analysis method for studies of different depressive states of individuals were discussed.
Animals ; Brain ; Depression* ; Hindlimb Suspension ; Mice ; Models, Animal ; Rodentia*

Emerging evidence has suggested that the gut microbiota contribute to brain dysfunction, including pathological symptoms of Alzheimer disease (AD). Microbiota secrete membrane vesicles, also called extracellular vesicles (EVs), which contain bacterial genomic DNA fragments and other molecules and are distributed throughout the host body, including blood. In the present study, we investigated whether bacteria-derived EVs in blood are useful for metagenome analysis in an AD mouse model. Sequence readings of variable regions of 16S rRNA genes prepared from blood EVs in Tg-APP/PS1 mice allowed us to identify over 3,200 operational taxonomic units corresponding to gut microbiota reported in previous studies. Further analysis revealed a distinctive microbiota landscape in Tg-APP/PS1 mice, with a dramatic alteration in specific microbiota at all taxonomy levels examined. Specifically, at the phylum level, the occupancy of p_Firmicutes increased, while the occupancy of p_Proteobacteria and p_Bacteroidetes moderately decreased in Tg-APP/PS1 mice. At the genus level, the occupancy of g_Aerococcus, g_Jeotgalicoccus, g_Blautia, g_Pseudomonas and unclassified members of f_Clostridiale and f_Ruminococcaceae increased, while the occupancy of g_Lactobacillus, unclassified members of f_S24-7, and g_Corynebacterium decreased in Tg-APP/PS1 mice. A number of genus members were detected in Tg-APP/PS1 mice, but not in wild-type mice, while other genus members were detected in wild-type mice, but lost in Tg-APP/PS1 mice. The results of the present study suggest that the bodily microbiota profile is altered in Tg-APP/PS1 mice, and that blood EVs are useful for the metagenome analysis of bodily microbiota in AD.
Alzheimer Disease* ; Animals ; Brain ; Classification ; DNA ; Extracellular Vesicles ; Gastrointestinal Microbiome ; Genes, rRNA ; Membranes* ; Metagenome* ; Metagenomics ; Mice* ; Microbiota* ; Reading