Week 6: Some preliminary results

This week I attended another coronary artery bypass (CABG) surgery, but this time there were three bypasses. The patient was very morbidly obese so it was challenging to access the heart. I also continued to work on my project and finally have some preliminary results.

I wrote a script to analyse my data efficiently using Matlab. The type of data I’ve encountered is very different from the one I typically collect in my lab. From a mathematical standpoint, medical researchers are interested in the understanding how a collection of independent variables (predictors) affect one or multiple outcomes of surgery. In cardiac surgery, the major outcome of interest is described by dichotomous variables that indicate the occurrence of events such as death, infection, stroke, or re-operation. Independent predictors of those events can also be dichotomous and describe patient characteristics such as sex (male/female), a binary diagnostic result (positive/negative), or a medical condition (symptomatic/asymptomatic). Therefore, to study the relationship between dichotomous variables, I had to perform multiples logistic regression.

I am interested to model which factors are predictors of left atrial thrombosis (LAT) and/or spontaneous echo contrast (SEC). LAT and SEC, which are respectively a blood clot in the left atrium and the manifestation of a “smoke-like” signal on an echocardiogram, are frequently found associated post-operatively with arterial fibrillation, stroke and valvular hear disease in patients that have undergone transcatheter aortic valve implantation (TAVI). So far, multivariate regression have demonstrated that pre-operative fibrillation is the strongest predictor of LAT and SEC. Similarly, I also modeled which factors can predict 30 day mortality as well as 1 year mortality. My model identified that pre-operative fibrillation, transient ischemic attack, post-operative stroke, were predictors of 30 day mortality. In addition, high creatine levels (a marker of kidney function),  high B-type natriuretic peptide (BNP) levels (a marker of heart failure) were significant predictors of 1 year mortality. I plan to discuss the clinical significance of these results with my mentor next Monday.