Evaluating the Impact of Drug-Disease Alerts on Health Outcomes in a Health-System

Medication errors and adverse drug events are common and can cause substantial harm. Many errors which result in injury can potentially be prevented by the use of appropriately designed computerized clinical decision support (CDS) with alerts. Electronic healthcare record (EHR) systems that embed CDS can reduce medication errors by checking for potential problems, such as drug-disease interactions. Previous studies have identified up to 44% of the general population are estimated to have drug-disease interactions; however the clinical relevance of these alerts is inconclusive.
Refining and updating clinically appropriate drug-disease alerts are important for patient safety. A previous study determined that tiered alerting in accordance to severity was associated with higher compliance rates of alerts in the inpatient setting. However, to be fully integrated, the clinical significance of these alerts need to be apparent. At Houston Methodist Health System, through the medication related clinical decision support committee (MRCDS), the decision was made to integrate drug-disease interaction alerts through a standardized model. This provided a unique opportunity to document activity taken on these alerts that could potentially increase the risk of poor outcomes for our patient population. This study integrated modern EHR technology with a commercial drug-disease interaction alert database to evaluate the clinical significance of drug-disease interaction alerts. This study specifically focused on agents that exacerbate QTc prolongation. The primary objective of this study was to assess the incidence of QT prolongation in patients identified with a QTc drug-disease interaction alert compared to similar patients in which a QTc drug-disease alert did not occur. Specific drug-disease alerts were assessed to identify clinically significant alerts vs. non-clinically significant alerts. The data from this study will help guide the health system decisions on alerts to improve patient care. Alert activity was examined on a monthly basis. A total of 41,470 patient encounters were reviewed. The baseline characteristics of patient encounters varied from each group. Overall, the majority of the sample size comprised of females (61.9%) with an average age of 55.6 years (19.5). Majority of the encounters came from the Control 3 group (- disease, - drug), followed by Control 1, Case, and Control 2 groups. The case group represents patient encounters where drug-disease interaction alerts were fired. Of the 1,341 alerts fired, 450 alerts were associated with administered doses of the drug. 421 (93.6%) alerts occurred during the order-entry stage of the medication use process, followed by order verification (21, 4.7%). 39 patient encounters (21.2%) among the case group that were administered the drug experienced QTc prolongation. In conclusion, we found a significantly high incidence of QTc prolongation in the case group. An evaluation of outcomes associated with drug-disease interaction alerts compared to those that did not elicit an alert may provide opportunities for optimizing alerts and selecting which alerts should ultimately be activated versus inactivated. Further studies are still warranted to evaluate a larger subset of drug-disease interaction alerts, to determine the impact on patient outcomes.