Multivariate analysis and data mining help predict asthma exacerbations
Authors:
- Stefan MihaicutaSMStefan Mihaicuta
- Lucretia UdrescuLucretia Udrescu
- Adrian MilitaruAMAdrian Militaru
- Valentin NadasanVNValentin Nadasan
- Angelica TiotiuATAngelica Tiotiu
- Andras BikovAndras Bikov
- Sorin UrsoniuSUSorin Ursoniu
- Romina BirzaRBRomina Birza
- Alina Mirela PopaAPAlina Mirela Popa
- Stefan FrentStefan Frent
- 1 January 2024
- journal article
- research article
- Published by Taylor & Francis in Journal of Asthma
- Vol. 61 (6) , 608-618
Abstract
Background. Work-related asthma has become a highly prevalent occupational lung disorder. Objective. Our study aims to evaluate occupational exposure as a predictor for asthma exacerbation. Method. We performed a retrospective evaluation of 584 consecutive patients diagnosed and treated for asthma between October 2017 and December 2019 in 4 clinics from Western Romania. We evaluated the enrolled patients for their asthma control level by employing the Asthma Control Test (ACT < 20 represents uncontrolled asthma), the medical record of asthma exacerbations, occupational exposure, and lung function (i.e., spirometry). Then, we used statistical and data mining methods to explore the most important predictors for asthma exacerbations. Results. We identified essential predictors by calculating the odds ratios (OR) for the exacerbation in a logistic regression model. The average age was 45.42 ± 11.74 years (19-85 years), and 422 (72.26%) participants were females. 42.97% of participants had exacerbations in the past year, and 31.16% had a history of occupational exposure. In a multivariate model analysis adjusted for age and gender, the most important predictors for exacerbation were uncontrolled asthma (OR 4.79, p < 0.001), occupational exposure (OR 4.65, p < 0.001), and lung function impairment (FEV1 < 80%) (OR 1.15, p = 0.011). The ensemble machine learning experiments on combined patient features harnessed by our data mining approach reveal that the best predictor is professional exposure, followed by ACT. Conclusions. Machine learning ensemble methods and statistical analysis concordantly indicate that occupational exposure and ACT < 20 are strong predictors for asthma exacerbation.Keywords
This publication has 60 references indexed in Scilit:
- Lung function reduction and chronic respiratory symptoms among workers in the cement industry: a follow up studyBMC Pulmonary Medicine, 2011
- Work-related respiratory symptoms and lung function among solderers in the electronics industry: a meta-analysisEnvironmental Health and Preventive Medicine, 2011
- Predicting asthma control and exacerbations: chronic asthma as a complex dynamic modelCurrent Opinion in Allergy and Clinical Immunology, 2007
- Assessment of asthma control and its impact on optimal treatment strategyAllergy, 2007
- Asthma exacerbations {middle dot} 3: PathogenesisThorax, 2006
- Occupational Asthma and Occupational Rhinitis in HairdressersChest, 2005
- Comparison of self-reported occupational exposure with a job exposure matrix in an international community-based study on asthmaAmerican Journal of Industrial Medicine, 2005
- The Association between Asthma and Allergic Symptoms in Children and Phthalates in House Dust: A Nested Case–Control StudyEnvironmental Health Perspectives, 2004
- The relationship of sex to asthma prevalence, health care utilization, and medications in a large managed care organizationAnnals of Allergy, Asthma & Immunology, 2003
- Baker's asthmaOccupational and Environmental Medicine, 2002