Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

JSM 2013

Presented at a contributed SPEED session at the Joint Statistical Meetings, August 2013. Accompanying poster is available at http://z.umn.edu/jsm2013poster

Julian Wolfson

on 7 August 2013

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of JSM 2013

Censored Bayesian Networks
Predicting 5-year risk of cardiovascular events using electronic health record data.
Combines health information from clinic/hospital visits, prescription database, and vital records.
Bayesian network
A machine learning technique often used for classification problems.
2. Use assumptions on P(X|Y) to make life easier, e.g., independence = "Naive Bayes":
Electronic health record (EHR)
Censored Bayesian networks
Huge sample sizes
Incomplete follow-up and censoring
Missing data
Complex interactions
Opportunities and challenges
Our dataset provides 10 years of health information on >400,000 people.
BP, cholesterol, etc. measured irregularly (if at all), especially among young people.
CVD is multifactorial and several risk factors may interact.
Extending Naive Bayes to censored, time-to-event data
Naive Bayes assumption &
nonparametric smoothing
Kaplan-Meier estimator
Julian Wolfson, Assistant Professor, Division of Biostatistics
University of Minnesota School of Public Health
(in collaboration with the UMN/HP iPredict Research Group)
julianw@umn.edu z.umn.edu/julianw @DrJWolfson

1. Model class probabilities by applying Bayes' rule:
Yep, this assumption is pretty crazy. Amazingly, Naive Bayes often works pretty well even if the assumption is severely violated!
At the poster session this afternoon:
An app which uses these methods to calculate your five-year cardiovascular risk.
Full transcript