Importance of Serum Albumin in Deep Learning-Based Prediction of Cognitive Function Data in the Aged Using a Basic Blood Test.

BACKGROUND

Recently, a method using deep learning has been developed to estimate the risk of developing dementia. This method uses general blood test data from routine health examinations that reveal lifestyle-related diseases, which can lead to vascular cognitive impairment via arteriosclerosis, as well as systemic metabolic disorders that are unrelated to lifestyle, such as nutritional disorders. In this study, we investigated the differences in the accuracy of estimating the risk of dementia based on the presence or the absence of blood test parameters reflecting nutritional disorders while focusing on the association between malnutrition and the risk of dementia in frail, elderly individuals.

OBJECTIVES

The objective of this study was to evaluate the impact of including or excluding serum albumin, which reflects nutritional status, on the accuracy of predicting cognitive function in older adults using blood test data.

METHODS

We estimated cognitive function, as measured by the Mini-Mental State Examination (MMSE), using the deep learning model (DLM). The estimation was performed based on general blood test data, including complete blood tests and basic metabolic panels, obtained from a selection of 1287 patients admitted to Osaka Medical and Pharmaceutical University Hospital. The data were divided into two groups: individuals aged 65 and above and those aged below 65. The impact of including or excluding serum albumin on the predictive performance of MMSE was examined within each group.

RESULTS

In those aged below 65, the mean squared error (MSE) of the DLM was 5.33 without albumin and 4.62 with albumin, showing a -0.71 improvement with albumin. In those aged 65 and above, the MSE of the DLM was 6.38 without albumin and 6.28 with albumin, showing a -0.1 improvement with albumin.

DISCUSSION

The present study demonstrated that including serum albumin in the input data resulted in lower estimation errors for MMSE across all applied algorithms in the group aged 65 and above. This is consistent with previously reported studies that have shown the adverse effects of malnutrition on cognitive function in older adults.

CONCLUSIONS

This study highlighted the significance of serum albumin, which reflects nutritional status, as an important assessment variable for estimating MMSE from blood test data, particularly in individuals aged 65 and above.