Challenges and Incremental Benefits in Implementing Targeted Injury Detection Systems for Adverse Drug Events and Pressure Ulcers in Inpatient Settings
Shulamit L. Bernard, Ph.D., R.N.a
Asta Sorensen, M.A.a
Jonathan R. Nebeker, M.D., M.S.b
Michael I. Harrison, Ph.D.c
RTI and its health system partners developed and deployed a concurrent targeted injury detection system (TIDS) for two classes of highly prevalent inpatient iatrogenic adverse events: pressure ulcers (PUs) and adverse drug events (ADEs). TIDS was implemented in three alpha sites (Baylor Health System, Intermountain Healthcare, and UNC Health Care) and planned in three beta sites (Emory University, Providence Health System, and the Salt Lake City VA). Triggers that were established for the events included a combination of indicators taken from administrative data, medical records information, and patient assessments. UNC Health Care led the development and implementation of the PU trigger. Simple ADE triggers were developed and implemented by Intermountain Healthcare 15 years prior to the study start. The Salt Lake City VA led the development of more clinically specific triggers.
Each beta site was charged with the implementation of two trigger sets, one focusing on ADEs and the second focusing on PUs. The implementation challenges were unique to each trigger.
The ADE TIDS implementation encountered a number of technical and operational challenges. All sites had difficulty implementing the relatively complex logic of the triggers. Through two to three teleconferences with each site, project leadership and local teams of information technology (IT) and project staff resolved most coding problems. An alpha site with more ADE TIDS experience made a significant modification to its rule engine to accommodate the more complex rules. Two of the three alpha sites were not able to implement triggers based on vital signs because that information was not available electronically.
None of the beta sites implemented all of the sets of rules for the ADE trigger set. Challenges associated with the ADE trigger implementation included limited perceived utility and impact on patient management, a perceived lack of new and timely information provided by the triggers, and time and staff resources required for data collection. One of the beta sites had recently purchased a commercial system and used stock ADE triggers instead of programming the more complex, clinically specific ADE triggers implemented at the alpha sites. Another beta site lacked an electronic solution and attempted a paper-based implementation. This site was not able to collect and process laboratory results within a timeframe that allowed the TIDS to prompt interventions that would prevent or mitigate ADEs.
The variety of operational challenges to ADE TIDS implementation was met with varying success. The ADE TIDS included a data collection tool that could help guide evaluation of the triggers and aggregated analysis of the ADEs. Except at one site, the pharmacists who reviewed triggers focused on using the TIDS for clinical interventions only. All sites found the classification tool to be excessively burdensome. They tracked only whether a trigger was associated with an ADE and whether it led to a change in patient care. Another barrier was a duplication of systems. Pharmacists did not want to implement the ADE TIDS at several sites because they perceived that the system duplicated the work they did routinely when making rounds as clinical pharmacists. At two institutions, protocols for anticoagulation rendered two of the related triggers useless. One beta site was not able to implement the ADE TIDS because of a congressionally mandated reorganization of its IT processes.
One beta site committed to implement the PU trigger but experienced difficulties obtaining Institutional Review Board (IRB) approval. Implementation challenges cited by staff at the other two beta sites were related to data collection systems. Manual calculation of Braden score and age was perceived as burdensome and time consuming. Floor nurses at this site perceived the PU trigger as outside the scope of usual care and viewed it as another requirement for documentation with no obvious value. Variation in determining the Braden score was another challenge in PU trigger implementation and use. Across the sites, nurses without specialized training in wound care had difficulty accurately assessing the Braden score, suggesting that successful adoption of this trigger should be accompanied by ongoing educational efforts in wound assessment.
Many of the challenges encountered were related to the local organizational structures and investigational aspects of the study (e.g., IRB issues, the need for consistency in measurement, and difficulty programming into existing IT systems). These challenges are not likely to be generalizable to process-improvement implementation. The lack of perceived utility of the ADE trigger may be more problematic for diffusion and adoption; this may be related partly to the presence of a clinical pharmacist who monitors patients on high-risk drugs or partly to the lag in the availability of laboratory data in the manual systems, resulting in trigger information that is somewhat dated and not as relevant to clinical decisions.
A perceived duplication of some ADE triggers with selected protocols and the activity of rounding clinical pharmacists was a major barrier. Of course, the appropriate response was to inactivate triggers that did not add value. The largest barrier to TIDS implementation for system-level changes was the increased burden of data collection for both PU and ADE TIDS.
The ADE TIDS produced incremental benefits of treating new ADEs at sites that undertook electronic implementation. The triggers were designed to be more clinically specific than most previously published trigger logic. When the triggers did not overlap with existing protocols and systems, they resulted in dramatic increases in detection of some types of ADEs. Because of the perceived utility for clinical interventions, three alpha sites continued to use selected triggers after the implementation period. A beta site that implemented only part of the ADE trigger continues to develop a system through its existing commercial vendors. A beta site that implemented a paper-based ADE trigger discontinued its use; however, this site plans to implement an electronic ADE trigger system through a commercial vendor in the future.
The addition of age to the Braden score for the PU trigger was viewed as an added benefit, particularly for borderline patients who are elderly. Wound ostomy nurses emphasized that using the PU trigger resulted in better care plans, consultations for appropriate wound treatment (i.e., use of proper dressings), and face validity of the instrument itself, but that it did not improve identification of PUs. Incremental benefit was associated with alignment of the PU trigger with local and national priorities, the ability to engage clinical champions, and the low technical complexity of the trigger tool.
Incremental benefit is perceived when the trigger fills a gap in clinical knowledge and awareness. The most benefit provided by the ADE trigger was with high-risk drugs such as warfarin. The inclusion of age with the Braden score in the PU trigger highlighted the increased risk to older patients and the need to address the incremental risk through clinical care plans. In this case, the trigger served as intended: it identified patients at elevated risk for an adverse event and the subsequent need for modification of the care plan. Age adjustment to the Braden score was especially appreciated by the wound ostomy nurses, who felt that age added more face validity.
The TIDS beta sites implemented the PU and ADE triggers differently, which is not surprising given that patient safety efforts are unique and specific to local settings. Factors important to successful implementation included the similarity of the trigger tool to existing workflow, the availability of personnel, the IT resources and infrastructure, and the commitment of relevant clinical champions (e.g., pharmacy, medicine, and nursing).
At all but one site, there was a strong focus on the ability of the ADE TIDS to prompt interventions that prevented impending ADEs or mitigated ongoing ADEs. Only triggers containing logic that gave providers new information were appreciated. This barrier to adoption highlights a need for the design of more intervention-oriented trigger systems to satisfy frontline providers. The focus on ADE TIDS to prompt intervention also demonstrates the need for effective integration of frontline pharmacists into quality improvement efforts that require more diligent collection of information to guide and evaluate system-level interventions.
The perceived benefits of the ADE trigger system were few, and the amount of effort associated with data collection and trigger identification was viewed as substantial. Therefore, the perceived benefits of the ADE trigger did not exceed the effort required to use the system. However, users acknowledged that prospective ADE prevention was limited by current systems. The idea of a prospective ADE trigger system was viewed as important, and two sites are in the process of pursuing commercial vendors for prospective identification of ADEs.
Efforts to reduce pressure ulcers at all sites preceded the PU trigger implementation. Multiple PU reduction efforts already were underway through local, regional, and national initiatives. Organizations had a heightened awareness and motivation to prevent and eliminate hospital-acquired PUs in light of the Joint Commission accreditation requirements and the upcoming Centers for Medicare & Medicaid Services (CMS) reimbursement policy that, by fall 2008, would stop reimbursing costs associated with hospital-acquired PUs. These initiatives contributed to the organizations' interest and motivation to implement the PU trigger and integrate local clinical champions. However, the visibility of the PU trigger benefits was mostly limited to the specialty nurses; staff nurses at beta sites viewed it as a burden with no added value, suggesting that additional training may be needed for successful adoption of this trigger by nonspecialty nursing staff. Low technical complexity of this trigger and its alignment with existing organizational processes, as well as national priorities, contribute to the likelihood of adoption of this trigger.
a Duke University School of Nursing.
b VA Salt Lake City Geriatrics Research, Education, and Clinical Center (GRECC); Department of Medicine, University of Utah; and Intermountain Institute for Healthcare Delivery.
c Center for Delivery, Organization, and Markets, Agency for Healthcare Research and Quality.
Note: The views in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs or the Agency for Healthcare Research and Quality.
Previous Section Contents Next Section