Save Time by Eliminating False Positives

Issue link:

Contents of this Issue


Page 1 of 1

Frequently Asked Questions About Eliminating False Positives with FairWarning As humans continue to grade alerts, the machine will be retrained for better data decision making Confidence analyzed – FairWarning standard confidence level is 99.5% Alerts are categorized based off predictive probability calculation (inbox, clutter, junk) Machine learns how humans grade alerts from a risk perspective Past human decisions analyzed by machine How Does It Work? Behavioral Correlations FairWarning's Behavioral Correlations will automatically document and dismiss an alert when our system determines a user has a business reason for accessing a patient record. The filtering algorithms conduct an analysis to determine the following clinical scenarios within the event and user data: 1. Has the user modified the patient? 2. Has the user modified the patient record in the last month? 3. Is the user part of the care team? The most common question we receive is related to the care team. Due to the inconsistent use of care team members within the source application, FairWarning has found more accurate results by defining "care team" as "two other users from the same department have modified or edited the patient record in the last month." Machine Learning FairWarning Machine Learning technology learns how to respond to alerts in your application by prioritizing and scoring each future alert based on how your end users have previously reviewed and closed alerts manually. You can leverage this technology to automatically document and close alerts that meet a certain score or risk tolerance. We start by teaching the technology to score future alerts based on the following data: THE MACHINE LEARNING PROCESS Name of the Enforced Policy User flagged in the alert Patient flagged in the alert Access types Other attributes Status of similar alerts Event types Event names What's Required to Begin Eliminating False Positives in FairWarning? FairWarning eliminates false positives by layering Enforced Policies, Behavioral Correlations, and Machine Learning technologies. Behavioral Correlation Algorithms FairWarning will apply our latest Behavioral Correlation algorithms, which include clinical context and identification (e.g., existing members of the patient's care team) to automatically close and document alerts that were business-related. To enable Behavioral Correlations, you must have: How Do I Get Started? FairWarning's Behavioral Correlations and Machine Learning technology has been implemented with over 150 in-production customers. While the results vary based on workflow, applications, and risk tolerance, customers are seeing false positives reduced by 40 to 90 percent. This has allowed them to expand their monitoring programs, save time through efficiencies, and spend more time with staff and patients. Step 1 Contact your Customer Success Manager or create a support case in the online Community. Step 2 Review the Enforced Policies on which you would like to reduce the volume of alerts. This review will be performed during a call with our team. FairWarning will provide initial recommendations based on best practices. Step 3 FairWarning will implement Behavioral Correlations on the active Enforced Policies. We will review the alert volume change with you and share the resulting Governance Report to track for future use. Step 4 (Machine Learning Only) For those eligible, FairWarning will implement Machine Learning technology and begin training the system. This technology learns the difference between appropriate and inappropriate access based on alerts manually reviewed by your end users. FairWarning predicts which alerts should be documented and dismissed as "appropriate access" and review the results with the customer. When an alert is closed, it's not gone forever, but rather archived in case a record of the alert is ever needed. This will move to production if the customer is satisfied with the results. FairWarning will review the learned behaviors every six months, for accuracy. Proactive Monitoring (Enforced Policies) enabled or planned for coworker, household, or manager snooping 10 or more alerts per day Epic, Cerner, or Meditech Machine Learning Machine Learning will automatically document and dismiss alerts based on actions taken by your end users on similar alerts in the past. To enable Machine Learning, you must: Have proactive monitoring enabled (Enforced Policies) Have three months of alert history to draw upon Implement FairWarning Behavioral Correlations for at least three months Determine your risk tolerance Be willing to accept some level of false negatives Follow the FairWarning best practice workflow For more information, please visit | 727-576-6700 | © Copyright 2004-2019 FairWarning | All rights reserved. Various trademarks held by their respective owners. For more information, please visit | 727-576-6700 | © Copyright 2004-2019 FairWarning | All rights reserved. Various trademarks held by their respective owners. ?

Articles in this issue

Archives of this issue

view archives of Datasheets - Save Time by Eliminating False Positives