Health care costs are rapidly outstripping society's ability to pay. Practice guidelines have been proposed as a way to reduce costs while improving health care quality, and many kinds of organizations (e.g., American College of Physicians, AHCPR, Mayo Clinics, etc.) are developing such guidelines. Some (Lamas 1992, Lamas 1989, Kosecoff 1987), however, argue that guidelines will have little effect on practice patterns if they are merely published without employing reminders or incentives of some kind. We have shown that computer reminders based on simple guidelines can change patterns (McDonald 1976, McDonald 1984, Tierney 1986, Litzelman 1993). However, we have not dealt with the rich practice guidelines being developed by AHCPR and others because the advice of such guidelines depends upon 'first order' data (collected from patients about their history, physical, and current symptoms) as well as the 'second order' data (from hospital services such as the pharmacy, laboratory, radiology, etc.) that the Regenstrief system now contains. In the proposed work, we will rigorously measure the real effect of automated guidelines on physician practice problems and patient outcome. Specifically, using our current network of ordering workstations as the platform, we propose to: 1) Define, test and refine detailed computer executable guidelines for the management (and/or prevention) of a number of medical problems including congestive heart failure, pneumonia, and urinary tract infection. 2) To define, test and refine instruments for capturing the historical, physical and symptom (first order) data needed by these guidelines. 3) To determine the reliability of these instruments when used by research assistants and 4) to refine and perfect mechanisms for delivering reminders produced by the computer executed guidelines to care providers. 5) To assess provider agreement with the guideline logic and their attitudes about the computer system. 6) To build statistical models that predict adverse effects and extended hospital stays based on the collected 'first order' data and incorporate the model's predictions into reminders. 7) To perform a randomized controlled trial of the effect of automated guideline reminders on a number of outcomes including provider compliance with the guideline advise, patients health status, length of stay, problem related costs, 30 and 90 day re-admission rates and adverse events. As American medicine attempts to halt the reckless growth of health care costs, proposed solutions should be rigorously studied. This project will shed light on the practicality, costs, and benefits of both practice guidelines automatically applied by a CBPR.
Showing the most recent 10 out of 22 publications
