BACKGROUND
Computerized rules/alerts systems may reduce the frequency of adverse medical events. The design of the alert may impact the effectiveness of the alert. From the user perspective there are 3 types of alerts: carrots which offer help to the user, sticks which force the user to respond, and flies which users just swat away with a mouse click. In order to design effective alerts, we need to study the effects of existing alerts on the user.
OBJECTIVE
To determine the effects of an inpatient CPOE alert that reminds clinicians at the time of order entry that allergy information needs to be updated on the patient.
DESIGN/METHODS
We currently use CPOE to enter all inpatients orders. We developed an alert that triggers when the following occurs: 1) when a clinician attempts to enter orders in CPOE, the system checks to see if the allergy information is updated in the system fro the current admission; and 2) if not,, the alert is triggered and the clinician sees an alert showing the current allergy information and gives the physician 3 choices: accept as is, update, or close the alert. The alert fires for each order in a set unless it is addressed. All order and alert information is stored in a relational database. On4/1/04 the alert was turned on. We queried the database for all instances of the allergy alerts and all orders for the time period of 4/1/04 to 6/1/04
RESULTS
During the study time period there were 250,494 inpatient orders written by 343 physicians on 2745 inpatient visits (16 orders per patient per day). Of these orders 52,758 (21%) were part of order sets. There were 57 physicians (16%) whose orders triggered alert. They wrote a total of 60,951 orders during the study period. Of these orders 21,452 (35%) were part of an order set. A total of 873 (1.4%) triggered an allergy info alert. And 599 (69%) were part of an order set. Each order in an order set had the potential to trigger the alert. 87.7% of physicians triggering alerts did so in less than 3 patients. 92% of the physicians entered < 3 orders before addressing the alert.
CONCLUSIONS
The allergy alert was triggered by a small number of physicians. The alerts occurred only with 1.4% of orders. More than 2/3’s of the alerts were associated with order sets. The combination of the allergy alert and order sets resulted in our carrot alert becoming a stick alert.