Determinação de intervalos de referência de exames laboratoriais bioquímicos na pessoa idosa por métodos direto e indireto

Abstract

Laboratory analyzes are essential for clinical decision making. From laboratory results it is possible to confirm a diagnosis, exclude a hypothesis and evaluate the prognosis of a pathology. The use of reference intervals estimated for adults in the interpretation of laboratory results in elderly people can lead to inadequate interpretation and generate unnecessary treatments as aspects of senescence and senility can considerably affect the results of biomarker measurements. There are different methodologies for determining the reference range. In the direct approach, healthy individuals are selected based on a clinical assessment. In the indirect approach, laboratory data used to establish the reference interval are obtained from the laboratory database. The general objective of this work was to determine the reference intervals of biochemical laboratory tests frequently used in the evaluation of elderly people by the direct and indirect method using data mining algorithms. This is an analytical, quantitative, observational, cross-sectional epidemiological study. This work was approved by the departments involved and the Research Ethics Committee. The population of this study was made up of patients over the age of 60, of both sexes, treated at the Geriatrics and Gerontology Center at HC-UFMG between 2010 and 2019 and who collected a biological sample for laboratory analysis. Data collection for this research can be subdivided into two main stages: (a) data collection in the Care Plan and (b) extraction of reports from the Laboratory Information System. The direct method approach included identification of outliers, carrying out descriptive statistics, checking data normality, determining reference intervals estimated by the 2.5% and 97.5% percentiles adopting 95% CI. For the indirect method approach, robust statistical models were used applied from Data Mining algorithms developed in programming languages: R and RStudio, JAVA, Python and C. In this study, the reference range for 24 biomarkers was determined from from the direct approach and 3 biomarkers from a data mining algorithm frequently used in the laboratory assessment of elderly people. The sampling procedure and robust statistical analysis resulted in reference intervals suitable for the Brazilian population. From the comparison between the reference intervals determined for the elderly person from the direct approach with the indirect method using different data mining tools and algorithms, it was possible to observe agreement between the different algorithms. The methodological approach is replicable and the indirect method is an important tool for laboratories to update the IR systematically.