Potential Drug-Drug Interaction (PDDI) CDS IG : STU Ballot 1

This page is part of the Potential Drug-Drug Interaction (PDDI) Clinical Decision Support (CDS) (FHIR IG) (v0.1.0: STU 1 Ballot 1) based on FHIR v3.5.0. . For a full list of available versions, see the Directory of published versions

6.0.0 Documentation

6.1.0 Preliminaries

This section contains documentation on how to implement PDDI CDS artifacts from a clinical and technical perspective. Implementation details are described using two specific knowledge artifacts as examples.

6.2.0 CPOE Workflow Hooks

Figure 1 depicts hook triggers for Level 1 and 2 Implementations. The primary difference in the Level 2 Implementation is the additional hook and defining the initial trigger at the top of the CPOE workflow. The Level 1 Implementation follows the CDS Hooks medication-prescribe specification, which does not necessarily define the triggering event.

Figure 1: Level 1 versus Level 2 Implementation Hooks
CPOE_workflow_2.svg

6.3.0 Prefetch Role

For technical implementers, the intended role of prefetch is to improve the CDS service performance. This is achieved by minimizing CDS service network calls to external resources such as a FHIR server. When a client program subscribes to the PDDI CDS service, the service MUST return a prefetch specification in the response. This specification identifies resources that the PDDI CDS service SHOULD receive upon request. As described below, the prefetch requirements are different for medication-select and medication-prescribe services. The ideal scenario for both implementations and services is to send prefetch data with the CDS Hooks request. The implementor has flexibility on when and how to fulfill the prefetch templates (e.g., data in EHR memory or server call), which will likely result in a solution that reduces the burden of the server and improves the CDS service efficiency. If the CDS service does not receive prefetch data in the request it MUST query the server via network call.

7.0.0 Warfarin + NSAIDs

7.1.0 The Warfarin + NSAIDs Knowledge Artifact

The Warfarin + NSAIDs knowledge artifact represents a relatively complex contextualized PDDI CDS algorithm. The knowledge artifact contains logic that uses both drug and patient contextual factors. The original rule was developed by clinical experts as part of the W3C Community Group effort to develop a PDDI minimum information model. Table 1 is the Warfarin + NSAIDs knowledge artifact at the narrative level using the minimum information model.

Table 1. Warfarin + NSAIDs Minimum Information Model

Drugs involved: Warfarin and non-steroidal anti-inflammatory drugs (NSAIDs)
Comment: The drugs involved in a PDDI MUST be explicitly stated. To support a computable representation of the PDDI, the drugs involved SHOULD be listed as sets of terms from a terminology such as RxNorm or the Anatomical Therapeutic Chemical Classification System (ATC). Such so called value sets MAY be referenced by a URI to a public repository such as the Value Set Authority Center that is maintained by the United States National Library of Medicine.
Clinical Consequences: Increased risk of bleeding including gastrointestinal bleeding, intracranial hemorrhage, and cerebral hemorrhage
Comment: The clinical consequences associated with a PDDI MUST be reported if known. Clinical consequences SHOULD refer health outcomes as specifically as possible. To support a computable representation of the PDDI, clinical consequences SHOULD be represented as one or more sets of terms from a terminology such as ICD-10 or SNOMED-CT. Such so called value sets MAY be referenced by a URI to a public repository such as the Value Set Authority Center that is maintained by the United States National Library of Medicine.
Seriousness: Bleeding is a serious potential clinical consequence because it can result in death, life-threatening hospitalization, and disability.
Comment: A PDDI clinical consequence MUST be noted as serious if it can result in death, life-threatening hospitalization, congenital anomaly, disability, or if it requires intervention to prevent permanent impairment or damage.
Severity: While bleeding is a serious potential clinical consequence, severity can vary from easily tolerated to incapacitating
Comment: The severity of a PDDI clinical consequence MUST be reported if known. The severity of a PDDI clinical consequence MUST be noted using non-ambiguous terms or phrases. Any of the existing terminologies for adverse event severity, such as Common Terminology Criteria for Adverse Event (CTCAE), MAY be used for describing a PDDI clinical consequence.
Mechanism of Interaction: Non-steroidal anti-inflammatory drugs (NSAIDs) have antiplatelet effects which increase the bleeding risk when combined with oral anticoagulants such as warfarin. The antiplatelet effect of NSAIDs lasts only as long as the NSAID is present in the circulation, unlike aspirin’s antiplatelet effect, which lasts for up to 2 weeks after aspirin is discontinued. NSAIDs also can cause peptic ulcers and most of the evidence for increased bleeding risk with NSAIDs plus warfarin is due to upper gastrointestinal bleeding (UGIB).
Comment: The mechanism of a PDDI MUST be reported if known. The description SHOULD be written for a clinician audience and include details that help the clinician decide what course of management action to take. To reduce ambiguity, the description MAY refer to specific drugs or health conditions using codes from terminologies.
Recommended Action: If the NSAID is being used as an analgesic or antipyretic, it would be prudent to use an alternative such as acetaminophen. In some people, acetaminophen can increase the anticoagulant effect of warfarin, so monitor the INR if acetaminophen is used in doses over 2 g/day for a few days. For more severe pain consider short-term opioids in place of the NSAID.
Comment: Any recommended actions that apply to all patient exposures SHOULD be stated using clear and concise language. The recommended action statement SHOULD also provide citations to evidence for a suspected drug-drug interaction (not provided in this example). Recommendations that depend on contextual information/modifying factors SHOULD be mentioned separately to support context-specific presentation of such information.
Contextual information/modifying factors:
  1. The NSAIDs is topical diclofenac
    • Recommended Action: No special precautions
    • Evidence for a Suspected Drug-Drug Interaction: Topical diclofenac has relatively low systemic absorption; in one study a topical gel (16 g/day) produced about 6% of the absorption seen with systemic administration of 150 mg/day. A higher than recommended dose of topical gel (48 g/day) produced 20% of a systemic dose of diclofenac.
  2. The NSAID is NOT topical diclofenac but the patient is concomitantly taking a proton pump inhibitor or misoprostol
    • Recommended Action: Assess risk and take action if necessary
    • Evidence for a Suspected Drug-Drug Interaction: Proton pump inhibitors and misoprostol may reduce the risk of UGIB in patients receiving NSAIDs and warfarin.
  3. The NSAID is NOT topical diclofenac, the patient is NOT concomitantly taking a proton pump inhibitor or misoprostol, and the patient has one or more of the following risk factors:
    • History of upper gastrointestinal bleeding (UGIB) or peptic ulcer or age > 65 years old
      • Recommended Action: Use only if benefit outweighs risk
      • Evidence for a Suspected Drug-Drug Interaction: Patients with a history of UGIB or peptic ulcer may have an increased risk of UGIB from this interaction. The extent to which older age is an independent risk factor for UGIB due to these interactions is not firmly established, but UGIB in general is known to increase with age.
    • Concomitantly taking systemic corticosteroids, aldosterone antagonists, or high dose or multiple NSAIDs
      • Recommended Action: Use only if benefit outweighs risk
      • Evidence for a Suspected Drug-Drug Interaction: Both corticosteroids and aldosterone antagonists have been shown to substantially increase the risk of UGIB in patients on NSAIDs, with relative risks of 12.8 and 11 respectively compared to a risk of 4.3 with NSAIDs alone (Masclee *et al.* 2014)
Comment: Contextual information/modifying factors are necessary for alerts that are both sensitive and specific. Like clinical consequences, each known factor SHOULD be stated as specifically as possible. The factors SHOULD be amenable to implementation as executable logic using value sets from clinical terminologies such as RxNorm, the Anatomical Therapeutic Chemical Classification System (ATC), ICD-10, and SNOMED-CT. As is used in this example, each factor SHOULD be related to a specific recommended action that is supported by the evidence for a suspected drug-drug interaction
Frequency of Exposure to the PDDI: Unknown
Comment: Frequency of exposure and frequency of harm information is rarely available but can help a clinician assess the risk/benefit trade-off of exposure to PDDI. Such information SHOULD be provided if available.
Frequency of Harm for persons who have been exposed to the PDDI: Unknown
Comment: Frequency of exposure and frequency of harm information is rarely available but can help a clinician assess the risk/benefit trade-off of exposure to PDDI. Such information SHOULD be provided if available.

7.2.0 Level 1 Implementation - Warfarin + NSAIDs Knowledge Artifact (semi-structured)

Figure 2 depicts how a PDDI CDS implementer would translate a minimum information model narrative to a semi-structured knowledge artifact. The Level 1 Implementation uses a single CDS service call and response with the medication-prescribe hook. The decision tree results in three warning indicators (i.e., green, orange, red) and contextual factors that MAY be passed to the clinician. After processing the CDS Hooks medication-prescribe request, the CDS service MUST return CDS Hooks Cards that MAY include actions with associated FHIR resources. Figure 3 builds on Figure 2 by depicting a Card display example within the order entry workflow. The decision points, medication-prescribe request, and Card(s) response are discussed further in the sections below.

Figure 2: Level 1 – Warfarin + NSAID Semi-Structured Knowledge
Basic_Warfarin_NSAID.svg

7.2.1 Decision Points Summary

Many PDDI CDS scenarios have similar drug and patient related decision points. Implementers SHOULD develop CQL artifacts that work in conjunction with the medication-prescribe context and prefetch to support the base decision points. The Warfarin + NSAIDs PDDI exemplar as three main decision blocks that include:

  1. whether the prescribed NSAID is topical diclofenac,

  2. whether the patient is taking a proton pump inhibitor, and

  3. the presence or absence of risk factors involving age, exacerbating medications, and history of upper gastrointestinal bleed

7.2.2 CDS Hooks Request

The medication-prescribe request includes context and prefetch elements with FHIR resource attributes or entire resources. The contents of these elements for the Warfarin + NSAIDs CDS artifact are shown below.

context

prefetch

Note: The use of multiple medication resources is to ensure a comprehensive capture of medications the patient may be taking. In some cases the implementing institution may only have access to MedicationRequest (prescription order), and in other cases they may have access to several resources for a specific medication intervention (e.g., prescription order from medical office, prescription product picked up from pharmacy). The 100-day look-back period is a general starting point. Implementors SHOULD refine this based on the available data. For example, MedicationAdministration is typically documented in the inpatient setting when a nurse administers a medication. This data source may be a more reliable proxy for blood concentrations and could be used to refine CDS logic.

  • Age of patient on current date.
  • Rolling 5-year look-back period for history of upper gastrointestinal bleed.

7.3.0 CDS Hooks Cards

Figure 3: Level 1 – Warfarin + NSAID Response Card Example
Basic_W_N_Cards.svg

7.3.1 Card Actions

The CDS Hooks service response supports providing actionable information to clinicians at the time of medication order entry. A response Card has an action element within the suggestion attribute. The action element is defined by three types including create, update, and delete. Depending on the type of action, resources may be provided that facilitate the suggestion. For example, if a suggestion recommends substituting naproxen for acetaminophen, a create action may be used to apply a MedicationRequest for acetaminophen to the current order entry task. The actions, types and associated resources are listed below.

  • create
    • Substituting NSAID for acetaminophen – MedicationRequest for acetaminophen
    • Adding proton pump inhibitor – MedicationRequest for proton pump inhibitor
  • update
  • delete
    • Remove current NSAID order

Note: These actions are options that SHOULD be customized to an institutions needs and capabilities.

7.4.0 Level 2 Implementation - Warfarin + NSAIDs Knowledge Artifact (semi-structured)

The Level 2 Implementation for the Warfarin + NSAID artifact is split into two separate hooks and services. Figures 4 and 5 depict the decision tree for warning indicators (i.e., green, orange, red) and contextual factors for both Hooks (i.e., medication-select and medication-prescribe). Figure 6 provides a Card display example for each CDS Hooks instance within the order entry workflow. In the provided Card display example, the clinician decided to order the NSAID medication but adds a proton pump inhibitor, in response to the card suggestion. This action results in a downgrade of the medication-presecribe response card (i.e., “hard-stop” – red to “warning” – orange). The blue task boxes highlight the DetectedIssue status indicator, which informs the EHR of additional needed resources (whether or not to fulfill the medication-prescribe service prefetch template), and medication-prescribe service if it needs to perform a FHIR server request in the event prefetch data are not provided in the request.

Figure 4: Warfarin + NSAID medication-select logic
Warfarin_NSAID_select.svg
Figure 5: Warfarin + NSAID medication-prescribe logic
Warfarin_NSAID_prescribe.svg

7.4.1 CDS Hooks request

context

medication-select 1.0
Field Optionality Prefetch Token Type Description
patientId REQUIRED Yes string Describe the context value
encounterId OPTIONAL Yes string Describe the context value
medication REQUIRED No object STU3 - FHIR MedicationRequest resource
medication-prescribe 1.1
Field Optionality Prefetch Token Type Description
patientId REQUIRED Yes string The FHIR Patient.id of the current patient in context
encounterId OPTIONAL Yes string The FHIR Encounter.id of the current encounter in context
detectedissue REQUIRED Yes object STU3 - FHIR Bundle of DetectedIssue resource for current order entry task
medication REQUIRED No object STU3 - FHIR Bundle of draft MedicationRequest resource for the current order entry task
medication-select prefetch
medication-prescribe prefetch
  • Age of patient on current date.
  • Rolling 5-year look-back period for history of upper gastrointestinal bleed.

7.4.4 CDS Hooks Cards

Figure 6: Warfarin + NSAID Cards
Advanced_W_N_Cards.svg

8.0.0 Digoxin + Cyclosporine

8.1.0 The Digoxin + Cyclosporine Knowledge Artifact

Table 2. Digoxin + Cyclosporine Minimum Information Model

Drugs involved: Digoxin and Cyclosporine
Comment: The drugs involved in a PDDI MUST be explicitly stated. To support a computable representation of the PDDI, the drugs involved SHOULD be listed as sets of terms from a terminology such as RxNorm or the Anatomical Therapeutic Chemical Classification System (ATC). Such so called value sets MAY be referenced by a URI to a public repository such as the Value Set Authority Center that is maintained by the United States National Library of Medicine.
Clinical Consequences: Increased risk of digitalis toxicity that may lead to cardiac arrhythmias
Comment: The clinical consequences associated with a PDDI MUST be reported if known. Clinical consequences SHOULD refer health outcomes as specifically as possible. To support a computable representation of the PDDI, clinical consequences SHOULD be represented as one or more sets of terms from a terminology such as ICD-10 or SNOMED-CT. Such so called value sets MAY be referenced by a URI to a public repository such as the Value Set Authority Center that is maintained by the United States National Library of Medicine.
Seriousness: Digitalis toxicity is a serious potential clinical consequence because it can result in death, life-threatening hospitalization, and disability.
Comment: A PDDI clinical consequence MUST be noted as serious if it can result in death, life-threatening hospitalization, congenital anomaly, disability, or if it requires intervention to prevent permanent impairment or damage.
Severity: While digitalis toxicity is a serious potential clinical consequence, it can produce a range of cardiac arrhythmias and rhythm disturbances that vary in severity, from manageable bradycardia to life-threatening ventricular fibrillation.
Comment: The severity of a PDDI clinical consequence MUST be reported if known. The severity of a PDDI clinical consequence MUST be noted using non-ambiguous terms or phrases. Any of the existing terminologies for adverse event severity, such as Common Terminology Criteria for Adverse Event (CTCAE), MAY be used for describing a PDDI clinical consequence.
Mechanism of Interaction: The mechanism of this interaction appears to be mediated through P-glycoprotein inhibition by cyclosporine. P-glycoprotein is a major transporter for digoxin efflux.
Comment: The mechanism of a PDDI MUST be reported if known. The description SHOULD be written for a clinician audience and include details that help the clinician decide what course of management action to take. To reduce ambiguity, the description MAY refer to specific drugs or health conditions using codes from terminologies.
Recommended Action: For patients with a reliable plasma digoxin concentration in normal range, it is reasonable to anticipate an increase in plasma concentrations after the initiation of cyclosporine. Following initiation, close monitoring and adjusting the digoxin dose as needed is recommended.
Comment: Any recommended actions that apply to all patient exposures SHOULD be stated using clear and concise language. The recommended action statement SHOULD also provide citations to evidence for a suspected drug-drug interaction (not provided in this example). Recommendations that depend on contextual information/modifying factors SHOULD be mentioned separately to support context-specific presentation of such information.
Contextual information/modifying factors:
  1. Relevant digoxin concentration in normal range
      Recommended Action: Assess risk and take action if necessary
      Evidence for a Suspected Drug-Drug Interaction: For patients with a reliable plasma digoxin concentration in normal range, it is reasonable to anticipate an increase in plasma concentrations after the initiation of cyclosporine. Following initiation, close monitoring and adjusting the digoxin dose as needed is recommended. (Dorian et al. Clin Invest Med 1988; 11(2):108-112) (Dorian et al. Transplant Proc. 1987; 19(1):1825-1827). Taken 4-6 hours after the most recent dose, the approximate therapeutic range is 0.8 to 2.0 ng/mL (1 to 2.6 nmol/L); however, toxicity has been observed with concentrations in this range.
  2. Risk factors that may increase digoxin concentration or toxicity independent of cyclosporine include: reduced renal function and electrolyte abnormalities
      Recommended Action: Assess risk and take action if necessary
      Evidence for a Suspected Drug-Drug Interaction:
      50-70% of digoxin is excreted unchanged in the urine. Changing renal function may increase serum concentrations and risk of toxicity, (Digoxin-FDA [prescribing information] NDA 20405/S-004)
      Hypokalemia, hypomagnesemia, and hypercalcemia may potentiate digoxin toxicity. (Lip et al. Postgrad Med J. 1993; 69(811):337)
Comment: Contextual information/modifying factors are necessary for alerts that are both sensitive and specific. Like clinical consequences, each known factor SHOULD be stated as specifically as possible. The factors SHOULD be amenable to implementation as executable logic using value sets from clinical terminologies such as RxNorm, the Anatomical Therapeutic Chemical Classification System (ATC), ICD-10, and SNOMED-CT. As is used in this example, each factor SHOULD be related to a specific recommended action that is supported by the evidence for a suspected drug-drug interaction
Frequency of Exposure to the PDDI: Unknown
Comment: Frequency of exposure and frequency of harm information is rarely available but can help a clinician assess the risk/benefit trade-off of exposure to PDDI. Such information SHOULD be provided if available.
Frequency of Harm for persons who have been exposed to the PDDI: Unknown
Comment: Frequency of exposure and frequency of harm information is rarely available but can help a clinician assess the risk/benefit trade-off of exposure to PDDI. Such information SHOULD be provided if available.

8.2.0 Definitions

The Digoxin + Cyclosporine artifact logic depends on whether the patient is stable on digoxin or cyclosporine before the current medication order event. This section defines terms used in the subsequent flow diagrams. Certain terms are defined by assumptions that may be taken based on the presence of resources in context versus prefetch elements of the request.

  • Incident Order – context medication is not in prefetch medications and, thus, is presumably the first occurrence.

  • Prevalent Order – context medication is in prefetch medications and, thus, is presumably a medication order that is continued or repeated.

  • Normal – observation that is within a specified time period, and the measure is within a therapeutic window or below/above a certain acceptable threshold.

  • Abnormal – observation that is not within a specified time period, or the measure is not within a therapeutic window or below/above a certain threshold.

Note: Parameters for “normal” and “abnormal” observations SHOULD be modified by the implementor. The provided artifacts use a simplistic approach of querying for the most recent measure in a specific time frame. This approach SHOULD be modified to capture and present the most clinically relevant information. For example, clinicians may want a look-back period that captures several measures for serum creatinine to determine the status and prognosis for renal insufficiency.

8.3.0 Level 1 Implementation - Digoxin + Cyclosporine Knowledge Artifact (semi-structured)

Figure 7 shows how a PDDI CDS implementer would implement the Digoxin + Cyclosporine PDDI knowledge artifact using the CDS Hooks medication-prescribe hook. The figure shows the CDS Service processes the PDDI CDS logic after receiving a medication-prescribe request. Figure 7 progresses through the decision tree and includes warning indicators (i.e., green, orange, red) and contextual factors that may be presented to the clinician. Figure 8 builds on this artifact and provides a display of Cards example.

Figure 7: Level 1 Digoxin + Cyclosporine logic
Basic_Digoxin_Cyclosporine.svg

8.3.1 Decision Points Summary

The Digoxin + Cyclosporine exemplar artifact has two main decision blocks:

  1. whether the patient is taking digoxin and/or cyclosporine at the time of the current order for digoxin or cyclosporine, and

  2. whether the patient has risk factors that may potentiate the risk of digitalis toxicity.

8.3.2 CDS Hooks Request

context

prefetch

8.3.3 CDS Hooks Cards

Figure 8: Level 1 Digoxin + Cyclosporine Cards
Level_1_D_C_Cards.svg

8.3.4 Card Actions

The actions, types and associated resources for the Digoxin + Cyclosporine Level 1 Implementation are listed below:

  • create
    • Adding order for digoxin level measure – ProcedureRequest for serum digoxin trough within 24 hours
    • Add consultation for either prescriber of digoxin or cyclosporine depending on which the patient is currently taking – ProcedureRequest consultation
    • Add order for electrolyte panel(s) (i.e., calcium, magnesium, potassium) and renal function (e.g., eGRF and serum creatinine) – ProcedureRequest for labs to be drawn
  • update
    • Reduce digoxin order currenlty on the patient’s profile – MedicationRequest for low-dose digoxin
  • delete
    • Remove current order for digoxin or cyclosporine

Note: These actions are options that SHOULD be customized to an institutions needs and capabilities.

8.4.0 Level 2 Implementation - Digoxin + Cyclosporine Knowledge Artifact (semi-structured)

As described under the Getting Started tab, the Level 2 Implementation proposal requires several changes to the current standard specifications. Changes to the CDS Hooks context are specified below. The Level 2 Implementation proposal for the Digoxin + Cyclosporine artifact is split into two separate services. Figures 9 and 10 depict the decision tree for warning indicators (i.e., green, orange, red) and contextual factors for both services (i.e., Medication Select and Medication Prescribe). The blue task boxes highlight the DetectedIssue status indicator, which informs the EHR of additional needed resources (whether or not to fulfill the Medication Prescribe Service prefetch template), and Medication Prescribe Service if it needs to perform a FHIR server request in the event prefetch data are not provided in the request. Figure 11 depicts a Card display example. In this scenario, the medication-prescribe Cards are filtered since the clinician’s action indicated that the patient was no longer taking digoxin.

Figure 9: Digoxin + Cyclosporine medication-select logic
Digoxin_Cyclosporine_select.svg
Figure 10: Digoxin + Cyclosporine medication-prescribe logic
Digoxin_Cyclosporine_prescribe.svg

8.4.1 CDS Hooks Request

medication-select 1.0 context

Field Optionality Prefetch Token Type Description
patientId REQUIRED Yes string Describe the context value
encounterId OPTIONAL Yes string Describe the context value
medication REQUIRED No object STU3 - FHIR MedicationRequest resource

medication-prescribe 1.1 context

Field Optionality Prefetch Token Type Description
patientId REQUIRED Yes string The FHIR Patient.id of the current patient in context
encounterId OPTIONAL Yes string The FHIR Encounter.id of the current encounter in context
detectedissue REQUIRED Yes object STU3 - FHIR Bundle of DetectedIssue resource for current order entry task
medication REQUIRED No object STU3 - FHIR Bundle of draft MedicationRequest resource for the current order entry task

medication-select prefetch

medication-prescribe prefetch

  • Rolling 100-day look-back period for digoxin concentration
  • Rolling 100-day look-back period for serum creatinine
  • Rolling 100-day look-back period for electrolytes including: potassium, magnesium, and calcium

8.3.3 CDS Hooks Cards

Figure 11: Level 2 Digoxin + Cyclosporine Cards
Level_2_D_C_Cards.svg

9.0.0 Warfarin + NSAIDs and Digoxin + Cyclosporine CDS Artifacts (structured)

The structured artifacts following the aforementioned logic and behavior are available for the Level 1 Implementation under Artifacts.