Machine Learning-Based Routine Laboratory Tests Predict One-Year Cognitive and Functional Decline in a Population Aged 75+ Years

BACKGROUND

Cognitive and functional decline are common problems in older adults, especially in those 75+ years old. Currently, there is no specific plasma biomarker able to predict this decline in healthy old-age people. Machine learning (ML) is a subarea of artificial intelligence (AI), which can be used to predict outcomes Aim: This study aimed to evaluate routine laboratory variables able to predict cognitive and functional impairment, using ML algorithms, in a cohort aged 75+ years, in a one-year follow-up study.

METHOD

One hundred and thirty-two older adults aged 75+ years were selected through a community-health public program or from long-term-care institutions. Their functional and cognitive performances were evaluated at baseline and one year later using a functional activities questionnaire, Mini-Mental State Examination, and the Brief Cognitive Screening Battery. Routine laboratory tests were performed at baseline. ML algorithms-random forest, support vector machine (SVM), and XGBoost-were applied in order to describe the best model able to predict cognitive and functional decline using routine tests as features.

RESULTS

The random forest model showed better accuracy than other algorithms and included triglycerides, glucose, hematocrit, red cell distribution width (RDW), albumin, hemoglobin, globulin, high-density lipoprotein cholesterol (HDL-c), thyroid-stimulating hormone (TSH), creatinine, lymphocyte, erythrocyte, platelet/leucocyte (PLR), and neutrophil/leucocyte (NLR) ratios, and alanine transaminase (ALT), leukocyte, low-density lipoprotein cholesterol (LDL-c), cortisol, gamma-glutamyl transferase (GGT), and eosinophil as features to predict cognitive decline (accuracy = 0.79). For functional decline, the most important features were platelet, PLR and NLR, hemoglobin, globulin, cortisol, RDW, glucose, basophil, B12 vitamin, creatinine, GGT, ALT, aspartate transferase (AST), eosinophil, hematocrit, erythrocyte, triglycerides, HDL-c, and monocyte (accuracy = 0.92).

CONCLUSIONS

Routine laboratory variables could be applied to predict cognitive and functional decline in oldest-old populations using ML algorithms.

Building the National Database of Health Centred on the Individual: Administrative and Epidemiological Record Linkage - Brazil, 2000-2015

Introduction

In Brazil, the National Health System (SUS) provides healthcare to the public. The system has multiple administrative databases; the major databases record hospital (SIH) and outpatient (SIA) procedures. Epidemiological information is collected for all populations in subsystems, such as mortality (SIM), live births (SINASC) and diseases of compulsory declaration (SINAN). Each subsystem has its own information system, which is able to provide information about consultations, clinical information and medicines dispensed. However, these systems are not linked, thereby preventing individual-centred analysis.

Objective

To describe the methods and results of parameter setting that are needed to execute the probabilistic deduplication of large administrative and epidemiological databases in Brazil and to create a National Health Database Centred on the individual.

Methods

This paper shows the results of a record linkage model to integrate data from SIH, SIA, SIM, and SINAN, which have different formats and attributes between them and over time. These data consist of 1.3 billion records from 2000-2015. Probabilistic and deterministic record linkages were used to deduplicate these data. The Kappa statistic and clerical review were used to ensure the quality of the linkage. The graph algorithm and depth-first search were used to generate the identifiers.

Results

The deterministic deduplication process resulted in a database with 403,113,527 possible unique individuals. After the probabilistic deduplication process of the former database was performed, 159,703,805 unique individuals were identified. This result had an estimated a false positive error rate of 3.3%, and the false negative error was estimated at 12.3%.

Conclusions

The National Health Database centred on the individual was generated and will allow researchers to use real-world evidence to conduct clinical, epidemiological, economic and other studies. This database represents a significant cohort, spanning 15 years of historical data and preserving patient privacy. The success of the process described will allow repeating and appending the data for future years and enable important studies to promote SUS efficiency and provide better treatments for patients.