Clinical Decision Support Systems

Clinical decision support systems are the informatics layer that delivers pharmacogenomic guidance to clinicians at the moment of prescribing. They link a stored genotype result to drug-ordering workflows in the electronic health record, so that when a clinician prescribes a relevant medication, the system can surface the applicable guidance automatically. Effective decision support is widely regarded as essential for translating pharmacogenomic results into action, but it must be designed to inform without overwhelming.

Definition

A pharmacogenomic clinical decision support system is software integrated with the electronic health record that links a patient's stored genotype to medication orders and presents relevant pharmacogenomic guidance to the clinician at the point of prescribing.

Scope

This entry describes how pharmacogenomic clinical decision support is built into the electronic health record, the role of pre- and post-test alerts, the challenge of alert fatigue, and why decision support is considered a prerequisite for usable pre-emptive testing. It draws on institutional implementation experience. It is a reference description of decision-support design and is not a recommendation about any drug, alert, or prescribing action.

Core questions

• How does decision support connect a stored genotype to a drug order?
• What are pre-test and post-test alerts, and what do they do?
• Why is alert fatigue a central design problem?
• Why is decision support considered necessary for pre-emptive testing to be useful?

Key concepts

• Electronic health record integration
• Point-of-prescribing alerts
• Pre-test and post-test decision support
• Alert fatigue
• Stored genotype reuse
• Automated genotype-to-guidance linkage

Mechanisms

A decision-support system stores discrete, computable genotype results in the electronic health record and associates them with the drugs they affect. When a clinician orders a relevant medication, the system fires an interruptive or passive alert presenting the applicable guidance; post-test alerts use an existing result, while pre-test alerts can prompt ordering a test when one is missing (Hicks et al., 2016; Hicks et al., 2016). Because a germline result is durable, the same stored genotype can drive alerts for many drugs over time, which is why decision support is treated as the mechanism that makes pre-emptive testing actionable (Pulley et al., 2012; Roden, 2019). A persistent design tension is alert fatigue: too many or poorly targeted alerts lead clinicians to dismiss them, so systems must balance completeness against signal quality.

Clinical relevance

Decision support determines whether a pharmacogenomic result actually reaches the clinician when it is relevant; understanding its design helps explain why some results influence care and others go unused. This entry describes the informatics infrastructure at a system level and is not a source of prescribing guidance; the content of any alert and any resulting decision rests with qualified clinicians applying current guidelines.

Evidence & guidelines

Institutional reports describe how pharmacogenomic interpretations were automated and integrated into the electronic health record and how decision support was paired with consultation services (Hicks et al., 2016; Hicks et al., 2016). Prospective programs demonstrated decision support as part of pre-emptive testing infrastructure (Pulley et al., 2012), and reviews describe its role across the field (Roden, 2019). These are implementation accounts rather than individualized clinical advice.

History

As institutions began storing pharmacogenomic results, it became clear that results sitting in a record had little effect unless they were surfaced when relevant. Programs such as Vanderbilt's PREDICT built decision support into prospective genotyping from the outset (Pulley et al., 2012), and subsequent health-system implementations developed automated pipelines to load pharmacogenetic interpretations into the electronic record and trigger alerts at prescribing (Hicks et al., 2016; Hicks et al., 2016). Managing alert burden emerged early as a defining challenge of these systems (Roden, 2019).

Debates

Balancing completeness against alert fatigue

Comprehensive alerting maximizes the chance that relevant guidance is seen, but excessive or low-value alerts cause clinicians to ignore them; tuning the number, timing, and interruptiveness of pharmacogenomic alerts is an ongoing design problem.

Key figures

• J. Kevin Hicks
• Kristine R. Crews
• James M. Hoffman
• Dan M. Roden

Seminal works

• pulley-2012
• hicks-2016-ehr
• hicks-2016

Frequently asked questions

Why is decision support needed if results are already in the record?

A stored result has no effect unless it reaches the clinician when a relevant drug is prescribed; decision support automatically links the genotype to drug orders and surfaces the guidance at that moment.

What is alert fatigue?

It is the tendency of clinicians to dismiss or ignore alerts when they are too frequent or low in value, which can cause important pharmacogenomic guidance to be overlooked and is a key reason alerts must be carefully targeted.