Health Care Trails Project

Overview | People | Collaborators | Sponsors | Publications | Tools


Recommended strategies for improving health care quality, improving patient outcomes, and reducing costs include care coordination and optimization in the organization of the delivery of ambulatory and inpatient services. However, little is known about how to design and evaluate the effects of coordination programs in terms of what services are needed by whom and when. Coordinative care decisions could be improved if there was a better understanding of and ability to monitor: 1) how patients move between services; 2) how that patient flow changes with age, health care coverage, and medical condition; and 3) the differences in flow associated with different patient outcomes. We propose a management care approach based around health care trails. A health care trail is the time ordered sequence in which a patient obtains services such as dialysis, bloodwork or psychological counseling. This trail data can be analyzed to identify trails that are effective (improve patient outcomes) and/or are commonly used by patients with particular conditions (e.g. elderly patients with congestive heart failure). Such data is available through UPMC and is currently used in the Invision system developed by Augr for UPMC physicians. We propose during phase 1 to develop a prototype system for assessing a patient's health care trail and providing information on: likelihood of patient being readmitted given the trail they are following, expected next step in the trail. In addition, the physician would be able to see the trail that the patient has followed. We would assess the scalability of the technology, demonstrate the feasibility of using dynamic network analytics and machine learning to identify those optimal trails based on the patient's characteristics as available through the ADT data and identify conditions under which additional data might be needed. We would also work with the UPMC collaborators to determine a site for a physical trial in phase 2. Phase 1 will focus one at least one group of patients, those with heart issues. Depending on the scope of the data in the rolling and prior ADT data sets this might be expanded to another set of patients, most likely those with diabetes. Phase 2 would expand on the technology developed in phase 1, expand the approach to other types of patients, make the trail technology interoperable with the Invision system, and engage in a physical trial using a trial group of physicians already familiar with Invision. We plan in Phase 1 and 2 to focus on providing information to the physician. Later follow on applications include: a system to provide information and alerts to patients about their trails; and a system to be used internally to assess the relative cost of alternative trails. We plan in Phase 1 to focus on identification of trails that have better health outcomes, and in Phase 2, to bring in cost information if feasible to identify trails that can reduce cost yet achieve the same health outcome.