Managing the alert process at NewYork-Presbyterian Hospital

Introduction: This paper describes the approach to the development of alerts within the CDS for the clinical information system at New York-Presbyterian Hospital. Following implementation of their CIS, request for alerts were numerous and not responded to in a timely fashion due to several process issues.

Objective: To improve the process of development of alerts to be used in the CDS system.

Methods: Review of the previous process with identification of process issues followed by an improvement initiative. Classic plan, do, check, act process improvement was not utilized. Results: With initial implementation of the CIS, an alerts committee was formed with the goal of developing and managing alerts. The committee was multidisciplinary but did readily respond to outside request for alert rules and had essentially no tracking or follow-up of such request. In addition, the committee could not address the need to update the current alerts. Four problems were identified: 1) no data were being collected regarding how often alerts were firing, 2) there was no record of the functioning of many of the alerts, 3) it was not clear who had responsibility for managing request for changes to the alerts and 4) there was a backlog of request for alerts. This prompted a redesign of the process around two guiding principles. First, alerts could come from outside the committee and second, request for alerts should come from limited sources, those with the authority to develop medical policy. The new model required four steps: 1) a sponsoring body (hospital committee or department) must sponsor the alert, 2) there must be a named requestor for the alert, 3) the sponsoring committee and named requestor were tasked with the responsibility of reapproving the alert every two years and 4) the alert must be requested using an approved form. The alert committee chair was tasked with reviewing the request and the committee prioritized it based on its importance and the availability of IT resources. In the 9 months following implementation of the new process, 10 requests were received.

Comments: While it appears that the new process should lead to improvement, this paper has essentially no data to support the proposition that the outlined process improvement resulted in measurable improvement. Larry Dean