Open AccessLogit and fuzzy models in data analysis: estimation of risk in cardiac patients
Author(s) -
Petr Honzík,
Lubomír Křivan,
Petr Lokaj,
Růžena Lábrová,
Zuzaováková,
Bohumil Fišer,
Nataša Honzíková
Publication year - 2010
Publication title -
physiological research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.647
H-Index - 70
eISSN - 1802-9973
pISSN - 0862-8408
DOI - 10.33549/physiolres.932008
Subject(s) - logistic regression , receiver operating characteristic , medicine , logit , myocardial infarction , risk stratification , cardiology , statistics , mathematics
The aim of this study was a comparison of risk stratification fordeath in patients after myocardial infarction (MI) and of riskstratification for malignant arrhythmias in patients withimplantable cardioverter-defibrillator (ICD). The individual riskfactors and more complex approaches were used, which take intoaccount that a borderline between a risky and non-risky value ofeach predictor is not clear-cut (fuzzification of a critical value)and that individual risk factors have different weight (area underreceiver operating curve – AUC or Sommers´ D – Dxy). The riskfactors were baroreflex sensitivity, ejection fraction and thenumber of ventricular premature complexes/hour on Holtermonitoring. Those factors were evaluated separately and theywere involved into logit model and fuzzy models (Fuzzy, FuzzyAUC, and Fuzzy-Dxy). Two groups of patients were examined:a) 308 patients 7-21 days after MI (23 patients died within periodof 24 month); b) 53 patients with left ventricular dysfunctionexamined before implantation of ICD (7 patients with malignantarrhythmia and electric discharge within 11 month afterimplantation). Our results obtained in MI patients demonstratedthat the application of logit and fuzzy models was superior overthe risk stratification based on algorithm where the decisionmaking is dependent on one parameter. In patients withimplanted defibrillator only logit method yielded statisticallysignificant result, but its reliability was doubtful because all othertests were statistically insignificant. We recommend evaluatingthe data not only by tests based on logit model but also by testsbased on fuzzy models.