Benefits and Risks of CDSS To Prevent Medication Errors and Adverse Drug Events

By MedTech Intelligence Staff
February 14, 2024


• Categories: 04-Product Design & Development, Clinical Decision Support, Clinician Engagement, Topic - Product Design & Manufacturing

Computerized clinical decision support systems (CCDSS) can reduce medication errors and adverse drug events, however they do bring a number of unintended consequences that need to be addressed, according to the results of an evidence review performed by the U.S. Agency for Healthcare Research and Quality (AHRQ).

Out of 1,335 unique abstracts, 33 articles met the AHRQ’s target criteria and were included in the review (27 systematic reviews, one overview of reviews, and five primary studies). Overall, computerized provider order entry with medication-related CDSSs were associated with reduced medication errors (moderate strength of evidence) and prevention of adverse drug events (low strength of evidence). Improved or targeted medication-related CDSSs were associated with reductions of medication errors and adverse drug events (moderate strength of evidence). However, alert override rates were high and varied between studies, and the appropriateness of the overrides was largely influenced by the type of alert. Other unintended consequences included CDSS-related errors, overdependence on alerts, alert fatigue, inappropriate alert overrides, and provider burnout. An additional 48 articles focused on barriers and facilitators of CDSS implementation.

AHRQ noted that the estimates of the effects of the CDSSs on medication errors, adverse drug events, related implementation outcomes such as alert overrides, and unintended consequences of use all come from different studies with unique contexts, which makes understanding the net benefit extremely challenging.

Clinical Implications

Based in its findings, AHRQ highlighted several implications for clinical practice. Clinicians should be aware of the strengths and weaknesses of any CDSS that is employed in their practice, including the scope of the medication-related alerts, new errors potentially introduced by the system, risks of accepting inappropriate alerts, and risks of overriding high- value alerts. In addition, all alerts must be considered in the context of the specific patient, since the CDSS may not be optimized to provide tailored information based on patient-level information documented in the patient record.

Future Research Needed

Some of the challenges in applying the data from past studies to today’s usage include the increase in vendor-developed CDSSs, which are less modifiable, and not tailored to specific health systems. “As a result evidence from homegrown CDSSs is less applicable to today’s landscape,” stated the paper’s authors.

They encourage future research focusing on how CDSSs could be made more effective broadly with additional focus on defining successful collaborations between vendors, researchers, and clinicians for developing and evaluating the effects of these vendor-based CDSSs on medication errors and adverse drug events. Lessons learned from cases where CDSSs have demonstrated substantial reductions in medication errors and adverse drug events should be applied across CDSSs with different targets.

The AHRQ did highlight the promise of artificial intelligence, noting that AI algorithms and tools could substantially improve the effectiveness of medication-related CDSSs. For example, natural language processing could be used to access and extract data routinely stored in unstructured fields, such as reports or free-text notes, providing substantially more information about the patient and the context. This information coupled with structured data analyzed using complex algorithms (such as neural networks that can manage high-dimensional data) could provide more relevant and tailored alerts at the point of care.

“Future publications should provide detailed information on the characteristics of the CDSS (e.g., basic versus advanced alerts), type of decision support provided by the CDSS, populations under study, and characteristics of the healthcare organizations to clearly describe the specific context of implementation and use,” the authors wrote. “This information is critical for interpreting the results and assessing generalizability of the findings.”

Read the full study here.