Prediction of leptospirosis cases using classification algorithms

• Published in: IET software Vol. 11; no. 3; pp. 93 - 99
• Main Authors: Nery, Nivison Ruy Rocha, Claro, Daniela Barreiro, Lindow, Janet C
• Format: Journal Article
• Language: English
• Published: The Institution of Engineering and Technology 01.06.2017
• Subjects: antibiotic treatment delivery; data mining; data mining classification algorithms; dengue cases; diseases; febrile illnesses; health care; health professionals assistance; JRIP algorithm; leptospirosis cases; life-threatening disease; patient diagnosis; pattern classification; Special Issue: Advances in Knowledge and Information Software Management; febrile illnesses; pattern classification; data mining; diseases; data mining classification algorithms; JRIP algorithm; leptospirosis cases; antibiotic treatment delivery; life-threatening disease; dengue cases; patient diagnosis; health care; health professionals assistance
• ISSN: 1751-8806; 1751-8814; 1751-8814
• DOI: 10.1049/iet-sen.2016.0193

Leptospirosis is a potentially life-threatening disease primarily affecting low-income populations, with an estimated annual incidence of 1.03 million infections worldwide. This disease has symptoms often confused with other febrile syndromes, such as dengue fever, influenza and viral hepatitis, often making diagnosis challenging. Improving the accuracy of early diagnosis of patients with leptospirosis will increase the speed of appropriate antibiotic treatment delivery, and both will improve clinical outcomes for this potentially fatal disease. The authors conducted an analysis of clinically and epidemiologically defined leptospirosis cases to predict disease using data mining classification algorithms. They conducted four sets of experiments to evaluate the performance of the algorithms, assessing their predictive accuracy of using different training and test datasets. The JRIP algorithm achieved 84% sensitivity using a dataset of only confirmed leptospirosis cases, and a specificity of 99% using a dataset of only confirmed dengue cases. Therefore, the approach successfully predicted leptospirosis cases, differentiated them from similar febrile illnesses, and may represent a new tool to assist health professionals, particularly in endemic areas for leptospirosis, accelerating targeted treatment and minimising disease exacerbation and mortality.