Problem to be Resolved:
In any given population, between 10% and 15% of the population account for more than 85% of the cost. Predictive modeling is a powerful statistical tool to help organizations find and manage these people early, before their health problems evolve to acute crisis. In this session, participants will learn how predictive modeling works and 10 settings in which it is being used to change the way healthcare is delivered in the United States.
Objectives of the Project:
The aims are the following:
* Discuss methods used to develop predictive modeling formulas to find those at highest risk for near-term, high care utilization.
* Describe 10 ways predictive modeling can be used to change the way healthcare is delivered, both in population-based and inpatient settings.
* Present powerful health and financial outcomes from a large government client.
Methods:
The sources of data and predictive modeling methods, in general, will be described along with the pros/cons of each method. Logistic regression was used for the predictive modeling that constitutes the prime example in this paper.
Outcomes:
Significant impact on process, health, and financial outcomes will be presented using a case study from a state government employer. Most of the emphasis will be on the way predictive modeling is and will be used to change healthcare delivery in the future.
Lessons Learned:
Predictive modeling provides a powerful way to detect and intervene with people before their symptoms evolve to acute crisis. As health information systems provide more and more data, predictive modeling will evolve more strongly as well in inpatient settings to detect those most at risk for iatrogenic complications and readmission.
Section Description
We are pleased to share the paper presentation abstracts from the Summer Institute in Nursing Informatics, Informatics at the Point of Care: A Barrier or a Bridge?, held at the University of Maryland School of Nursing, July 22 to 25, 2009. The program, chaired by Dr Judy Ozbolt, was a great success. Each of the following abstracts was selected for presentation by a peer-review committee.