Format

Send to

Choose Destination
J Am Med Inform Assoc. 2015 Mar;22(2):361-9. doi: 10.1136/amiajnl-2013-002538. Epub 2014 Oct 15.

Optimization of drug-drug interaction alert rules in a pediatric hospital's electronic health record system using a visual analytics dashboard.

Author information

1
Department of Anesthesiology and Critical Care Medicine, Perelman School of Medicine at the University of Pennsylvania and the Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA.
2
Department of Enterprise Analytics and Reporting, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA.
3
Department of Pediatrics, Perelman School of Medicine at the University of Pennsylvania and the Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA.
4
Department of Biostatistics in Pediatrics, Perelman School of Medicine at the University of Pennsylvania and the Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA.
5
Department of Pharmacy Services, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, USA.

Abstract

OBJECTIVE:

To develop and evaluate an electronic dashboard of hospital-wide electronic health record medication alerts for an alert fatigue reduction quality improvement project.

METHODS:

We used visual analytics software to develop the dashboard. We collaborated with the hospital-wide Clinical Decision Support committee to perform three interventions successively deactivating clinically irrelevant drug-drug interaction (DDI) alert rules. We analyzed the impact of the interventions on care providers' and pharmacists' alert and override rates using an interrupted time series framework with piecewise regression.

RESULTS:

We evaluated 2 391 880 medication alerts between January 31, 2011 and January 26, 2014. For pharmacists, the median alert rate prior to the first DDI deactivation was 58.74 alerts/100 orders (IQR 54.98-60.48) and 25.11 alerts/100 orders (IQR 23.45-26.57) following the three interventions (p<0.001). For providers, baseline median alert rate prior to the first round of DDI deactivation was 19.73 alerts/100 orders (IQR 18.66-20.24) and 15.11 alerts/100 orders (IQR 14.44-15.49) following the three interventions (p<0.001). In a subgroup analysis, we observed a decrease in pharmacists' override rates for DDI alerts that were not modified in the system from a median of 93.06 overrides/100 alerts (IQR 91.96-94.33) to 85.68 overrides/100 alerts (IQR 84.29-87.15, p<0.001). The medication serious safety event rate decreased during the study period, and there were no serious safety events reported in association with the deactivated alert rules.

CONCLUSIONS:

An alert dashboard facilitated safe rapid-cycle reductions in alert burden that were temporally associated with lower pharmacist override rates in a subgroup of DDIs not directly affected by the interventions; meanwhile, the pharmacists' frequency of selecting the 'cancel' option increased. We hypothesize that reducing the alert burden enabled pharmacists to devote more attention to clinically relevant alerts.

KEYWORDS:

Electronic health records; clinical decision support systems; drug interactions; medical order entry systems; medication alert systems; visual analytics

PMID:
25318641
DOI:
10.1136/amiajnl-2013-002538
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Silverchair Information Systems
Loading ...
Support Center