IIR 03-298
Surveillance of Medical Device-Related Problems in Intensive Care Units
Matthew H. Samore, MD VA Salt Lake City Health Care System, Salt Lake City, UT Salt Lake City, UT Funding Period: November 2006 - April 2010 Portfolio Assignment: Healthcare Informatics |
BACKGROUND/RATIONALE:
This study addresses problems of patient safety in intensive care units (ICUs) in VA hospitals. The modern ICU embodies the convergence of technology and medical science that has transformed care of the critically ill. Progress and innovation are manifested by the presence of new, increasingly complicated devices. While these medical devices are intended to augment the quality of care, their use engenders new threats to patient safety. Questions about the nature of these hazards and how to prevent them form the basis for this proposed investigation. Errors and adverse events associated with medical devices are an understudied area in patient safety. In addition, the organizational context of acute nursing units wields important influences on adverse events. The broad goals of this study are to define the incidence of problems associated with medical devices, to evaluate their causes, and to identify potential solutions. This study was designed to examine two major hypotheses. The first hypothesis is that adverse medical device events (AMDEs) and hazards occur at a clinically significant frequency across a spectrum of ICUs and types of therapeutic and monitoring devices. The second hypothesis is that error-producing conditions are important causes of these problems. Error-producing conditions are factors that are recognized to affect reliability and human performance in non-healthcare industries. Five types of error-producing conditions will be specifically addressed: 1) task interruption; 2) task familiarity; 3) time pressure; and 4) task ambiguity. This hypothesis will be evaluated drawing from methods from the disciplines of cognitive psychology, human factors, and epidemiology. OBJECTIVE(S): The immediate objectives of this study are to 1) Systematically characterize nursing tasks related to medical device use in six VA intensive care units (ICUs) and one non-VA ICU, 2) Estimate the incidence of device-related hazards and adverse medical device events (AMDEs) in the participating ICUs, 3) Ascertain the occurrence of error-producing conditions within the organizational context of the participating ICUs, 4) Assess the causal effects of error-producing conditions on device-related hazards and AMDEs, and 5) Feedback results of the analyses to ICU and patient safety personnel at each institution. Achieving the immediate objectives listed above will advance progress toward the following long-term objectives: a) Devise a strategy to ameliorate factors that are barriers to improved care related to medical device use, and b) Implement and evaluate a strategy to improve the safety of medical device use in ICUs. The broad goal of this study is to acquire knowledge that enhances patient outcomes within the VA healthcare system. The more specific focus is to improve detection and control of patient safety problems associated with medical device use in the ICU. METHODS: This study relies on the technique of direct observation to ascertain AMDEs, potential adverse events (hazards), and error-producing conditions. The observational data is supplemented with survey and patient record abstraction. Analyses include descriptive summaries, estimation of events rates, as well as the use of regression models. Various mixed effects regression models were fitted to examine the association between error-producing events (e.g., task interruptions) and hazards. In these models, the multiple tasks performed, or required to be performed, were nested within the specific observation period, to account for the lack of independence introduced by this clustering. Study site and ICUs were included in the models as fixed effects to examine their influence, using indicator variables. Mixed effects logistic regression, mixed effects Poisson regression, and mixed effects Cox regression with multiple events were tried to check the robustness of the analysis approach. The logistic regression approach treated each task as its own "time window", but required the assumption that the time required for each task was essentially equal, or equal time at risk for a hazard event. The Cox regression approach took time into account, but assumed that the time sequence across the observation period had a meaningful interpretation. The Poisson regression approach had the most intuitive interpretation, modeling hazards per unit of time, while not assuming any underlying meaning to the time sequence. The Poisson approach was ultimately adopted as the best approach. In these models, a running sum of the number of previous tasks, and running sum of the number of previous interruptions, were included as candidate predictors, possibly reflecting time pressure. FINDINGS/RESULTS: During the sample period from April 2008 through December 2009 we completed 1148 hours of observations in six VA and one non-VA intensive care unit in three geographically diverse locations. All observers were trained and inter-rater reliability was computed using a frequency analysis at the subtask level for simultaneous, independent field sessions conducted by observer pairs. Overall agreement was 84%, a prevalence and bias-adjusted Kappa was 0.67 (95% confidence interval 0.57 - 0.76). During the same period, a total of 73,452 nursing tasks were recorded during 1148 hours of observation. Task interruptions comprised 9.6% (7105) of the total nursing tasks. Subgroup analysis for the interrupted tasks indicated that 28% were direct care tasks, 60% were indirect care tasks, and 12% were medical device-related tasks. The three most frequent sources of task interruptions were nurses (18.8%), alarms (10.8%), and technicians (8.1%), and 50.4% of interrupting tasks were unrelated to the tasks in progress. Nurses responded to task interruptions by multi-tasking (44.6%), switching to the interrupting task (37.3%), delaying the interrupting task (11.3%), or ignoring the interruption (5.6%). The two most common interrupting tasks were for the purpose of work clarification (45.1%) and alarm-related monitoring (5.8%). A total of 558 hazards were documented as coded data associated with qualitative information. Overall, 0.76% of nursing tasks were accompanied by the presence of a hazard. The likelihood of a device-related hazard occurrence was significantly higher for interrupted tasks than for non-interrupted tasks (odds ratio = 2.1, 95% confidence interval 1.0 - 4.0). IMPACT: The results of this study make a novel contribution to the critical care medicine and quality improvement fields and significantly inform understanding of the links between intensive care unit work system problems and their impact on patient safety. External Links for this ProjectDimensions for VADimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.Learn more about Dimensions for VA. VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address. Search Dimensions for this project PUBLICATIONS:Book Chapters
DRA:
Health Systems Science
DRE: none Keywords: Adverse events, Informatics, Safety, Surgery MeSH Terms: none |