Logistic regression modelling for STHR analysis
2014 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE credits
Student thesisAlternative title
Logistisk regression för STHR-analys (Swedish)
Abstract [en]
Coronary artery heart disease (CAD) is a common condition which can impair the quality of life and lead to cardiac infarctions. Traditional criteria during exercise tests are good but far from perfect. A lot of patients with inconclusive tests are referred to radiological examinations. By finding better evaluation criteria during the exercise test we can save a lot of money and let the patients avoid unnecessary examinations.
Computers record amounts of numerical data during the exercise test. In this retrospective study 267 patients with inconclusive exercise test and performed radiological examinations were included. The purpose was to use clinical considerations as-well as mathematical statistics to be able to find new diagnostic criteria.
We created a few new parameters and evaluated them together with previously used parameters. For women we found some interesting univariable results where new parameters discriminated better than the formerly used. However, the number of females with observed CAD was small (14) which made it impossible to obtain strong significance. For men we computed a multivariable model, using logistic regression, which discriminates way better than the traditional parameters for these patients. The area under the ROC curve was 0:90 (95 % CI: 0.83-0.97) which is excellent to outstanding discrimination in a group initially included due to their inconclusive results.
If the model can be proved to hold for another population it could contribute a lot to the diagnostics of this common medical conditions
Place, publisher, year, edition, pages
2014.
Series
TRITA-MAT-E ; 2014:44
National Category
Probability Theory and Statistics
Identifiers
URN: urn:nbn:se:kth:diva-148971OAI: oai:DiVA.org:kth-148971DiVA, id: diva2:742118
Subject / course
Mathematical Statistics
Educational program
Master of Science in Engineering -Engineering Physics
Supervisors
Examiners
2014-08-312014-08-162022-06-23Bibliographically approved