Development of an electronic trigger tool at a children’s hospital within an academic medical center

Abstract

Purpose To evaluate the validity and reliability of select recommended triggers, defined as flags found on review of the medical record that prompt further investigation to determine the presence or absence of an adverse drug event (ADE), selected from a list initially constructed based on severity, frequency, and detectability of triggers within a pediatric population. Methods This was a single-center, retrospective cohort analysis of pediatric patients admitted to University of North Carolina (UNC) Children’s Hospital who received trigger-associated medications between January 2015 and December 2016. Patient-care areas of the emergency department, operating rooms, and post-anesthesia care units were excluded. Trigger-detection encounters were evaluated by two reviewers using pre-established, consensus ADE criteria as determined by a panel of pediatric and medication safety specialists at UNC Medical Center. Events were categorized according to medication-related trigger and analyzed using descriptive statistics. Results A total of 3,836 positive triggers were included in this study. For the aggregate 12-part trigger tool package, 1,055 positive ADEs were identified, leading to a positive predictive value (PPV) of 27.5%. A 50% increase from baseline serum creatinine, resulting from co-administration of 2 or more nephrotoxic medications accounted for a total of 3,698/3,836 (96.4%). Incomplete documentation was the leading cause for event exclusion, 8/27 (30%). The triggers with the highest PPV included protamine 4/4 (100%), flumazenil 1/1 (100%), and vancomycin-related events 51/67 (76.1%), respectively. Phenytoin level >30 µg/mL or free level >2.5 µg/mL resulted in the lowest PPV, 1/12 (8.3%). Conclusion This study lays the foundation for further studies to develop a robust pediatric trigger tool that may involve developing multi-element triggers, determining sensitivity and specificity of triggers, or mobilizing the trigger tool to an automated system. Trigger tools can be individualized to meet each institutions’ needs and unique patient population.

Development and Adjustment of an Algorithm for Identifying Drug-Related Hospital Admissions in Pediatrics

“…We also achieved an improvement of the overall PPV of our algorithm, which became comparable with the values in other studies. 12,58 Only one study 13 achieved a much higher overall PPV. This can be explained by their comparatively high ADE rate in their sample (34%), which automatically raises the trigger's chance to actually indicate an ADE, even if it detects ADEs by coincidence.…”

“…Fortenberry et al 58 used a very different approach and only included and reviewed those records that were previously selected by electronic health system when the received medication was associated with the selected triggers. Therefore, their number of false-negative results remains completely neglected, and the entirety of ADEs is only assumed to be identified by their trigger tool.…”

“…Trigger tools help to identify adverse events in the healthcare system but mostly focus on adults, and only a few are dedicated to the pediatric population. 12,13,58 They further often focus on events that occur during the hospital stay but neglect those acquired in the ambulant setting, which secondarily can also lead to hospital admissions and healthcare costs. To identify these drugrelated hospital admissions in pediatrics and to fill the existing gap, we developed and adjusted an algorithm aiming to detect ambulant acquired drug-related hospital admissions.…”

“…Trigger tools help to identify adverse events in the healthcare system but mostly focus on adults, and only a few are dedicated to the pediatric population 12,13,58 . They further often focus on events that occur during the hospital stay but neglect those acquired in the ambulant setting, which secondarily can also lead to hospital admissions and healthcare costs.…”

“…Adjusted Algorithm Compared With Other Pediatric Trigger Tools12,13,58 12. Unbeck et al13 Fortenberry et al58 Number of triggers was adjusted after first validation. † Total number of records was electronically determined and only contains those with a least one activated trigger.…”