TY - JOUR T1 - Development of a decision tree based on real life primary care data: Can we predict the diagnoses asthma, COPD and asthma/COPD overlap syndrome based on automatic data analysis? JF - European Respiratory Journal JO - Eur Respir J VL - 44 IS - Suppl 58 SP - 4418 AU - Esther Metting AU - Niels Chavannes AU - Richard Dekhuizen AU - Daniel Dresden AU - Frederiek Swart AU - Hans In 't Veen AU - Thys van der Molen Y1 - 2014/09/01 UR - http://erj.ersjournals.com/content/44/Suppl_58/4418.abstract N2 - Introduction The Asthma/COPD(AC) tele-medicine service in the North of the Netherlands supports general practitioner(GP)s in managing their patients by cooperating with pulmonologists. GPs (n=352) can refer patients (current n=17,000) with respiratory complaints to the service where patients are assessed according to a strict protocol. The pulmonologists (n=13) evaluate the outcomes online and send diagnosis (asthma, COPD, asthma/COPD-overlap syndrome (ACOS) or diagnosis uncertain) and management advice to the GP. Data from 12,0000 patients is stored and available for research.Aim To explore if a decision tree could support the evaluation of individual assessments automatically.Method A decision tree was developed using exhaustive Chi-squared Automatic Interaction Detection(CHAID) analysis. The database was randomly divided in a training file (n=4,697) and test file (n=4,623). The post-bronchodilator FEV1/FVC ratio was used as first predictor, other predictors in order of significance were age of onset, smoking, allergy, hyperreactivity, reversibility, post-FEV1 , wheezing, family history and age.Result The current decision tree (50% men, mean age 54±17 years, 49% smoker) was able to correctly predict 80% of the asthma patients, 81% of the COPD patients and 31% of the ACOS patients.Discussion The diagnosis asthma and COPD could be correctly predicted in most patients by using this instrument. Our decision tree is considered to be promising because it was based on real life primary care data and used diagnoses made by pulmonologists. The clinical implication of the decision tree needs further debate. ER -