A Study of Chronic Fatigue and Its Relation to Absenteeism using Stepwise and Elastic-net, 2019, Neisse et al

[Dolphin]

Mainly posting this for completeness

Free full text
ftp://venus.est.ufmg.br/pub/fcruz/pub/iwsm-chr.pdf

A Study of Chronic Fatigue and Its Relation to Absenteeism using Stepwise and
Elastic-net

Anderson Cristiano Neisse1, Fernando Luiz Pereira de Oliveira2, Anderson Castro Soares Oliveira3, Frederico Rodrigues Borges da Cruz4, Fausto Aloisio Pedrosa Pimenta2

1 Universidade Federal de Vicosa (UFV), Brazil
2 Universidade Federal de Ouro Preto (UFOP), Brazil
3 Universidade Federal de Mato Grosso (UFMT), Brazil
4 Universidade Federal de Minas Gerais (UFMG), Brazil
E-mail for correspondence: a.neisse@gmail.com

Abstract:

Chronic Fatigue Syndrome (CFS) is an illness that has been commonly present in clinical practice in the last decades. It is characterized by persistent fatigue, pain, cognitive impairment, and sleep difficulties. The factors that contribute to the CFS development, as studies suggest, are: poor sleep, psychological stress, hormonal dysfunction, nutrient deficiencies, among others. Its development can increase in poor work conditions, such as in shift work in mines, therefore increasing the risk of fatal accidents.

A possibly effective tool to prevent the development of CFS is predictive modeling. This study aims to assess the risk of CFS and its relation to absenteeism by means of biochemical and anthropometric variables.

A cross-sectional study collected data on 621 shift workers in a mine, measuring 19 variables. After imputing missing data, logistic regression was fitted by four approaches: stepwise, lasso, ridge and elastic-net. Each model was compared between imputed and complete-cases datasets as well as with each other. The stepwise model was chosen for further exploration since the other three approaches did not show performance improvements.

Results suggest a lack of discrimination power due to noise that is inherent to the dependent variable's nature. However, significative effect was observed for the LDL, total cholesterol, triglycerides, and sodium on the risk of skipping work.

Keywords: Chronic Fatigue; Bioinformatics; Elastic-Net; Logistic Regression.