Risk of Wrong-Patient Errors With Multiple Records Open

Overview

This study is designed to achieve the following aims: 1. Assess the relationship between the number of records open at the time of placing an order, and the risk of placing an order on the wrong patient. 2. Compare the incidence of wrong-patient orders in a "restricted environment" that limits its providers to only one record open at a time to an "unrestricted environment" where users can open a maximum of four records at once. 3. The results of this study will help inform decisions on how to safely implement EHR systems. 4. The results of this study will inform a larger scale health IT implementation research project evaluating the balance between the wrong-patient error risks and potential efficiency gains of having multiple records open at once, with rigorous research methodologies.

Currently, at least 70,000 U.S. physicians use computerized provider order entry (CPOE) to place orders. This number is expected to rise sharply as hospitals continue to take advantage of federal incentives and adopt electronic health record (EHR) technology. Although CPOE is associated with a reduction in medical errors, when orders are placed electronically certain types of errors, including placing orders on the wrong patient, may occur more frequently. Although there have been no studies quantifying (or even establishing) an increased risk of wrong-patient errors when providers have multiple records open at once, there have been several articles and expert opinions that warn of this potential risk. The mechanism by which multiple patient records opened simultaneously can lead to a wrong-patient error may be related to the ease with which users can toggle between patient records and the similar looking computer screens. The magnitude of this risk needs to be established to help Information Technology (IT) leadership decide on how to safely implement CPOE systems. There have been no studies demonstrating whether multiple records increase the risk of wrong-patient errors, by how much, and if any increase is dependent on the number of records open. This research project is an important first step in quantifying this risk. In a randomized controlled trial conducted at Montefiore Medical Center, investigators propose to randomize inpatient and outpatient providers to a maximum of one record open at the time of ordering (restricted mode) or a maximum of four records open at the time of ordering (unrestricted mode). Assignments will be made prior to the start of the study, and will remain constant throughout the study. A computer programmer working in IT, who is not an investigator of this study, will use Microsoft Excel to generate random numbers and assign one number to each provider. Providers assigned odd numbers will be in the restricted cohort, and those assigned even numbers will be in the unrestricted cohort. Providers who join Montefiore after the start of the study will be assigned a random number from Excel when assigned a new user log in for the EHR from a computer programmer not affiliated with the study, and will be added to the appropriate group based on their assigned random number. At the start of the randomized controlled trial, investigators will explain the purpose of the study to clinical staff via email and directly from within their IT systems, using a message crafted by the study team. The message will assure clinicians that data will be kept confidential and cooperation will carry no risk to them. Montefiore uses the Epic Systems Corporation, or Epic, EHR system. Epic will implement the Retract-and-Reorder (RAR) tool, an automated method for identifying wrong-patient electronic orders, as well as capture the number of records open at the time of placing an order. This study will examine the effect of having the EHR system in restricted mode vs. unrestricted mode on RAR events. The goal is to obtain an estimate of the effect size and the intra-class correlations to provide preliminary data for a larger-scale health IT implementation research project. The unit of analysis will be the order. First, the RAR event rate for orders placed in the restricted vs. unrestricted mode will be calculated, testing the difference in rates using rank sum tests. Next, the relationship between the RAR event rate in restricted vs. unrestricted mode in subsets of providers and settings will be examined to determine whether specific types of providers or settings carry increased risk. Finally, a mixed-effects logistic regression model will be fitted with RAR event as the outcome and mode of the EHR system (restricted vs. unrestricted) as the independent variable of interest. The model will include random effects at the provider and order-session level because previous work has suggested substantial within-provider and within-session correlation. Orders will be nested in sessions and sessions will be nested in providers. To address the threat of confounding, the model will include fixed-effects variables including provider, patient, order-session, and order level covariates. To safeguard against the possibility that the intervention actually worsens (increases) the RAR event rate, and to prevent unnecessary continuation of a study that is already conclusive, a data safety monitoring committee will conduct one interim review of the data in the randomized controlled trial.