Pharmacogenetics

Clinical Decision
Support System
Med_Inf 406

Eric Chavez, Sean Heffernan, David W. West

March 16, 2014

Introduction
Background
The Pharmacogenetic Clinical Decision Support System aims to advance
personalized medicine by utilizing preemptive genotyping to help providers
make safe and appropriate medication orders. An underlying assumption is
that as pharmacogenetic testing becomes more and more common in the
future, providers will have a patients genotype available before they order
medications. The results of pharmacogenetic testing will be stored in a
clinical data repository of the electronic health record (EHR). This data will
either be automatically uploaded to the clinical data repository from
laboratory information systems or from genetic testing vendors, or it will be
manually entered into the EHR by staff in a standardized, coded format.
Unlike other lab data which change over time, results of genetic testing are
stable over a patients lifetime. The same data can be reused repeatedly to
help providers make decisions about future medication orders. As the field of
pharmacogenetics advances, more actionable information can be determined
from the results of these tests.
Genetic testing is becoming less expensive and more common in clinical
practice. This type of testing results in pharmacogenetic data.
Pharmacogenetics is the study of how a patients genotype and phenotype
affect his individual response to medications. Genetic testing can provide
information about an individual patients ability to metabolize medications
through genotyping the cytochrome P450 system. Patients can be classified
as rapid metabolizers, normal metabolizers, or slow metabolizers. When
multiple choices of medications exist to treat a given medical condition, this
information can be used to help providers make decisions about medication
selection and dosing. Rapid metabolizers would be expected to show a lower
clinical response to a given medication or to require higher than standard
dosing strategies. Slow metabolizers would be expected to suffer more
toxicity and adverse effects and would need lower than standard doses.
Some genetic testing can inform clinicians of the variability in efficacy of
certain medications for a patient based on his genotype. Pharmacogenetics
is believed to be on the verge of revolutionizing personalized medicine by
targeting therapies for individual patients, reducing trial-and-error
treatments, and greatly reducing toxicity and adverse effects. This will help
to reduce overall healthcare costs.
Patients respond to drugs in different ways. There are some groups of
patients that will experience severe adverse effects and there are some
groups of patients that will not respond at all to certain drugs. Factors that
may affect response to drugs include age, gender, disease state, smoking,
drug interactions, patient adherence, and genetic factors. It has been shown
that genetic factors may account for 20-95% of individual variation in drug
response (Kalow, 1998). The amount of variation depends on the class of
drug and genetic phenotype of the patient. Pharmacogenetics can provide

3
some explanation of how genetic variants change drug response
(Weinshilboum, 2003).
Examples of the benefits of personalized medicine using genetic testing
include individualized drug dosing, individualized drug selection, and
economic savings. These will be illustrated below using the examples of
warfarin, hepatitis C, and HIV.
Warfarin has a narrow therapeutic index. Inadequate or excessive
anticoagulation can lead to increased risk of adverse cardiovascular event or
bleeding complication. Warfarin dosage is complicated by individual
variability and requires regular monitoring to achieve proper anticoagulation
effects. Initial therapy is usually by fixed dosage which is then adjusted
based on the patients anticoagulation effect, as measured by international
normalized ratio (INR). Because genetic factors account for 30-35% of the
variability of warfarin response, having genetic test results could lead to a
faster time to achieve stable therapeutic dose. In one study, 46% of patients
benefitted from genetically informed drug dosing because they required
lower or higher doses of warfarin than the standard dosage (Klein, 2009).
Another study showed that patients treated with genetically informed dosing
had 28% less hospitalizations after 6 months (Epstein, 2010). The ability to
more accurately predict warfarin dosing by avoiding under-dosing (risk of
coagulation) or overdosing (risk of bleeding) can lead to improved drug
efficacy, improved patient safety, and organizational cost savings.
Genetic testing can help guide decision making when the clinical effect of a
drug is expected to vary according to patient genotype and phenotype.
Adverse clinical effects may differ according to phenotype when slow
metabolizers accumulate toxic metabolites of a drug. Genetic testing can
help guide decision making by steering toward alternative drug treatments
or dosing regimens. Patients who are slow metabolizers would benefit from
reduced dosing of the standard drug therapy or from taking a different drug
altogether. Fast metabolizers may benefit from higher than standard dosing
regimens. They would be expected to show less benefit from standard doses.
This could delay the efficacy of treatment which means longer time with the
illness. Longer time with the illness means more patient suffering and
dissatisfaction, potentially more time lost due to a decreased functional
state, and more cost to the healthcare system and the overall economy.
An example is chronic Hepatitis C which is treated with PegINF--2a or
PegINF--2b and ribavirin. Less than half of patients achieve sustained viral
repression with the medications. There is a genetic test that can predict
those who will have sustained viral repression based on a polymorphism in
the IL28B gene (Ge, 2009). Test such as these help providers know who will
respond to a given drug therapy. This can save time, save money, and save
patients from adverse drug effects when it is predicted that the benefit is
very low.
Avoiding inappropriate drug selection can save money. For example if a more
expensive drug is shown to be less effective in patients with certain

4
genotypes, the cost of trying that drug may be avoided and alternative
medications can be selected. Avoiding ineffective drug regimens can save
money by reducing patient suffering and dysfunction and by speeding time
to recovery. Avoiding serious adverse drug reactions can save money by
reducing hospitalizations or office visits and by reducing patient morbidity. It
may be more cost-effective to use generic drugs that are associated with
adverse effects in some genotypes than to use a more expensive drug
without the potential adverse effects for everyone, if genetic testing is used
to help select appropriate patients for the generic drug.
An example is in the treatment of HIV with abacavir. It is estimated that 5-8%
of patients will develop a hypersensitivity reaction to the medication in the
first 6 weeks of treatment. Re-challenging with the medication after initial
treatment may result in worsening of the reaction or even death. Genotyping
can reduce this risk by predicting who will have the hypersensitivity reaction.
In one study, screening patients for the genotype associated with the
reaction and not giving those patients abacavir lead to complete elimination
of the reaction in the study population (Gazzard, 2008). In an analysis of
cost-effectiveness, it was found that genetic testing and treatment with
abacavir when appropriate was less costly than completely avoiding abacavir
and using a more expensive alternative medication, tenofovir (Kauf, 2010).
The field of pharmacogenetics is rapidly developing. The majority of
practicing physicians lack any formal training in the field, and most do not
know how to incorporate this information into their practices to improve
safety and outcomes for patients (Baars, 2005). It would be impossible for
providers to memorize all of this information as more and more clinically
relevant information and actionable guidelines are added. The
pharmacogenetics clinical decision support system can help to solve this
problem by using a just-in-time tool that brings actionable recommendations
to providers at the point of care. This system will integrate into the
computerized provider order entry (CPOE) module of the EHR and make
recommendations to alter drug choice or alter drug dose when providers
attempt to write medication orders. In order to reduce alert fatigue the
system will issue alerts only when strong clinical guidelines exist in the
pharmacogenetics literature.
There are few studies that report on the development and implementation of
pharmacogenetic clinical decision support systems. In 2009 Deshmukh, et al.
studied two different methods of presenting genetic test results to a Cerner
clinical decision support system. They used the CYP2C9 gene and used a
data model of either single nucleotide polymorphisms (SNP) or alleles as the
discreet test reports. Both models worked for the clinical decision support
system (CDSS). The allele-based system was easier to write decision rules
for, and the SNP-based system allowed for retention of more discreet data
that could possibly be used and represented in different ways in the future.
Overby (2010) reported on the feasibility of collecting and standardizing
pharmacogenetic data from various sources, turning that data into
computable form, and writing decision rules based on the data. In 2012
Overby, et al. expanded on their previous work and described a prototype of

5
semi-active and active CDSS using translated pharmacogenetic data to drive
rules-based decisions in a Cerner EHR environment. The semi-active
component allowed providers to look up genetic testing information about a
specific drug while the active component offered recommendations for
therapy change or dose adjustments when providers attempted to order a
medication. Bell (2014) et al. describe an active CDSS in which genetic test
results are captured and encoded in the EHR problem list. Alerts were
generated when providers attempted to prescribe a medication that matched
a high risk phenotype in the problem list. In a review article, Welch and
Kawamoto (2013) report on six different studies that use genomic data as
part of CDSS to guide medication management for warfarin and antiretroviral
medications.

Stakeholders, Goals, and Objectives

For the purposes of this project we will assume that our target healthcare
organization is a large multispecialty medical center with both inpatient and
outpatient services. The stakeholders for the project are visualized in onionring format in figure 1. The innermost circle contains the product which is
the pharmacogenetics clinical decision support system. The next circle
contains the business system which are those stakeholders who do the work
in developing the product and who have operational roles. The next circle
contains the business who are the stakeholders who will benefit directly from
the product. The outermost circle contains the wider environment of
stakeholders who will be indirectly affected by the product.
The Clinical Decision Support Committee is the main driving force for the
CDS initiative. The committee may or may not include several of the other
key stakeholders listed below. The committee is responsible for developing
the goals and objectives for the CDS system and for evaluating the science
behind the system and the evidence that it will provide benefit for patients
and for the organization. The committee solicits feedback from key
stakeholders to ensure that the project is feasible. Overall, the committee
guides the development, implementation, evaluation, and monitoring of the
CDSS. It is important for the CDS committee to engage all stakeholders
through every phase of the CDS lifecycle.
Clinical leadership is necessary to evaluate the need for the CDSS and the
medical evidence to support the system. Clinical leaders need to be involved
in every stage of the project development and implementation to give
feedback on how provider workflows are effected and how patient care is
impacted. Notable among the clinical leaders is the Physician Champion who
plays a key role in the development and implementation of the CDSS. Usually
the champion is well respected among his/her peers, has excellent
communication skills, remains active clinically, and helps to grow support by
spreading enthusiasm.
Technical leadership is necessary to translate the clinical ideas into a
functional operating plan in the computer information system. It will be
necessary to involve members of the EHR and CPOE committees in the
technical leadership since this CDSS uses the EHR and CPOE system as its

6
point of interaction with providers. A vendor representative will be included
to share ideas about problems and solutions that other organizations have
faced when implementing similar CDS systems.
The administrative leadership is ultimately responsible for approving the CDS
project. They are responsible for determining whether or not the CDSS meets
the goals of the organization, for evaluating the return on investment, and
for fundraising or budgeting for the project. Administrative leadership will be
indirectly affected as it is hoped that outcomes will improve and cost savings
can be realized for the organization.
Genetic testing labs will need to ensure that results of genetic testing can be
automatically integrated into the EHR and CPOE systems
End users are the front line clinicians who will be using the CDSS on a daily
basis. They should be active in all phases of the project to ensure buy-in and
adoption. End users can give valuable feedback during various stages of the
project to make adjustments for work flow in order to help the CDSS achieve
success. Resistors and detractors among the end user group should be
included in the planning process so that potential problems can be addressed
early.
Patients may be involved during the project design and evaluation to give
feedback and to help spread support for the system. They should be made
aware of the potential benefits and risks of the new CDSS. It is hoped that
ultimately patients will be the primary beneficiaries of the new system with
improved outcomes and safety.

7
Figure 1. Stakeholders

Organizational goals and clinical goals of the project have to do with

improving patient lives, improving efficiency of physician workflows, and
saving healthcare dollars. These goals are displayed in figure 2. More
specific objectives for the project are shown in figure 3. These objectives
may be used as the metrics for a CDSS evaluation program.

 Develop cost-effective treatment by

selecting the most effective and
appropriate drug for each individual
patient

Make more effective decisions in

regards to drug selection and dosing
Predict optimal dosing for drugs with a
narrow therapeutic index
Avoid trial-and-error treatments

Minimize harmful drug effects

Optimize treatment for individual
patients
Reduce length of patient morbidity
and loss of function by selecting
effective drug regimens more rapidly

Reduce
healthcare
costs
Improve
patient
safety,
outcomes,
and
satisfaction
Improve
physician
effectivene
ss and
efficiency

Figure 2. Organizational and clinical goals.

8

Figure 3. Specific objectives.

Improve
patient
safety

reduce the number of adverse drug events

Improve
dosing
regimens

Decrease from standard dosage or select alternative

medication for patients with slow metabolizer phenotypes

Select
most
effective
therapy

Increase the number of orders written for drugs shown to

be effective for an individual patients genotype

Improve
treatmen
t
response
times

Improve the time to achieve stable therapeutic dosing for

drugs with narrow therapeutic windows (i.e. warfarin)

reduce the number of hospitalizations and office visits due

to adverse drug events

Increase from standard dosage or select alternative

medication for patients with rapid metabolizer phenotypes

Reduce the number of orders written for drugs shown not to

be effective for an individual patients genotype

Improve time to achieve reduction or resolution of patient

symptoms by more rapidly selecting the most effective
treatment regimen

Information System Inventory

The pharmacogenetics clinical decision support model shown below in figure

4 represents current and future system states, which will easily generalize to
other clinical environments, allow providers to enter metabolizer status when
ordering a drug, and support reporting and interpreting genetic test results.
The CDS will 1) integrate genetic testing results into the Clinical Data
Repository for use in the existing CPOE system and 2) store the results from
genetic labs in the CDS Knowledge Base where rules will be applied for
specifying results and preparing them for presentation to users via the
Clinical System Interface.

10
Figure 4. Clinical Decision Support Model

The EHR Application Environment supports the functional capabilities of the

decision support system. Discrete data are stored within the Clinical Data
Repository and the Clinical User can interact with the Clinical System
Interface to view data and order medication (CPOE), which may trigger
decision support implemented in the repository. Interventions and offered
medication choices display to the user through the Clinical System Interface.
The Clinical Decision Support Module implements the user interface features,
providing methods for transforming inputs (genetic test results or
medications being ordered) into patient-specific output (pharmacogenomics
links to e-resources or alert messages). Imbedding the decision support tool
in the EHR will allow the use of pharmacogenomics knowledge in making
prescribing decisions.
System Inventory Needed Acquisitions
One File System for the Result Specification XML Repository
Two Databases (one each) for the Pharmacogenomic Knowledge Base
and CDS Knowledge Base
Two application servers for deploying the new databases
One Rules Engine for retrieval and calculation of genetic tests and
related medication options
One web server for linking the CDS to the vendors e-resources

Workstations
o 10 new workstations to be deployed across Ambulatory,
Inpatient, Behavioral Health, Laboratory, and Pharmacy
Services
o Note: Although the setting for deploying the CDS is a large
urban hospital, the intended user base of the software is
limited, so the number of new workstations needed is minimal
Personnel
o Physicians
o Lab Technicians

11

o Pharmacists
o Nurse Practitioners
o Software Administrators
o IT Technicians
o Vendor Liaison
Software
o System Interface Software
o Database Deployment Software
o Rules Design/Maintenance Software
o Web Services Software

Intervention Selection and Workflow Opportunities

Medication dosing has drawn the attention of clinical decision support

systems for decades, especially after the Institute of Medicine (IOM)
published its report, To Err is Human in 1999 in which medication errors
were most prominent among all types of errors which occur in the health
care setting (Kohn, 1999). The implementation of computerized physician
order entry (CPOE), even in its simplest form, circumvents errors related to
legibility, but as a change in workflow, there has been concern in the
literature that other types of errors not previously seen are being introduced.
Since the IOM published its report, clinical decision support around dosing
has become more and more sophisticated. This support is predicated on the
enhancement in CPOE to require granular elements of the order to be
entered. These elements include dose, dose unit, frequency, and route at a
minimum. Third party expert content providers have emerged to provide
dosing support for virtually any FDA approved medications where granular
CPOE has been implemented.
Dose range checking is the most basic type of decision support, essentially
alerting the prescriber if the single or daily dosage falls outside of a standard
range. Condition-based defaults and weight-based dosing are other forms of
clinical decision support that start to take other clinical factors into account
when providing support. Lab studies such as creatinine can be incorporated
to provide further dosing support.
Genetic studies that help predict metabolism and/or effectiveness of certain
drugs is a new and growing field that shows great promise for being
integrated into clinical decision support systems. Many physicians are now
ordering these tests to assist their dosing decisions.
The proposed CDSS will integrate genetic testing results into the clinical data
repository in a manner that will effectively provide support at the point of
order entry with the goal of achieving a therapeutic, but non-toxic level of
medication. Genetic tests are an excellent example of patient-level data in
that the results do not change over time. Family history and birth history are
other examples of patient-centered data. Once the data is captured, it does
not need to be recaptured. It does need to be readily accessible at the point
of care. Therefore, avoiding duplication of genetic testing would be an
additional benefit of the proposed CDSS, though not its primary objective.

12
A number of system features will be important to integrate this information
into medication dosing decisions. The results originating in the genetic labs
need to be normalized into a value that can guide and alert dosing. The
medication repositories supplied by third party domain expert systems need
to include this relationship with genetic testing in a manner similar to age,
weight, and renal function based dosing elements. Finally, the health care
organizations EHR needs to support and integrate these various elements
into a useful presentation to the prescriber on pharmacogenetic-based
clinical guidelines.

Change Management Plan

Clinical decision support systems ultimately strive to modify and align

behaviors to best practices that will achieve optimal outcomes. Attempts at
behavior change often encounter resistance. The likelihood that behavior
change will occur tends to correlate positively to the perceived value of the
change and negatively to the pain of adoption. The change management
plan will address both of these principles.
To boost the perceived value proposition prior to implementation, data will be
gathered that helps define the opportunity to improve care. This will entail
statistics on the frequency of medication orders for which genetic data could
alter management. One can then apply knowledge about the frequency of
genetic phenotypes in the population to infer the frequency with which
medication dosing could be enhanced to avoid lack of efficacy due to under
dosing or toxicity due to overdosing.
To mitigate the pain of adoption, the effort required to incorporate the
genetic information into the dosing decision needs to be easy to find and
presented in a manner that can be quickly converted into a dosing
modification. Reports that are composed of free text, even if presented in
tabular form, need to be stored in relevant granular data elements. The
correct genetic elements need to be connected to the drugs for which a
given gene will have an impact. The connection must classify the effect of
the gene, and then offer to alter the dosing accordingly. In cases where the
genetic data suggests a pure lack of efficacy, the connection between the
genetic phenotype and the drug should lead to an offering of alternatives
that would have greater efficacy.
Post-implementation efforts can continue to try to reinforce the value
proposition. This can take the form of queries that report the frequency of
adoption of modified dosing in accordance with genetic data. This
information can be trended and shared as an organization quality
improvement objective, with more detailed drill-down data to demonstrate
inter-provider or inter-department variability. Areas of success can be
examined closely to see if practices exist that could be replicated to areas of
lower success. The opportunity to offer rationales as to why suggested
dosing was not accepted will serve to enhance clinical decision support if
other variables can be introduced that improves accuracy.

13
A continued iterative process that focuses on enhancing one or both of these
principles (increased value proposition, decreased pain of adoption) should
serve to solidify the change in the organization as a whole.

System Design
The pharmacogenetic CDSS enables the use of pharmacogenetic data each
time a relevant medication is prescribed as part of pre-emptive testing and
stores the results in the EHR before a drug is prescribed. Accordingly, the
pharmacogenetic CDSS advances personalized medicine by addressing the
inherited variation in drug responses by maximizing the drug efficacy and
minimizing toxicity for individual patients. The CDSS can also reduce costs
associated with inappropriate drug treatments or serious adverse drug
reactions.
Included in the System Design is a discussion of the:
1. Components of the System Architecture, which specifies the functional
requirements of the CDSS
2. Pharmacogenetic Use Case for Clinical Assessment and Genetic
Testing
3. Conceptual Model for the Pharmacogenetic CDS in terms of
Client/Server and Web-Based implementations

Components of System Architecture

The components of the CDSS architecture are best represented by a
specification of its functional requirements. Below in figure 5 is a depiction of
the systems functions and the interaction between processes and outputs.

Figure 5. System architecture.

14

Pharmacogenetics Use Case

The use case below shows how personal health and family health history
information is exchanged with genetic testing information between
consumers and clinicians based on two scenarios:
1. Clinical Assessment
2. Genetic Testing

15

These scenarios demonstrate how genetic test results are represented in

EHR components of a pharmacogenetic CDSS and the three phases of
performing genomic tests:
1. Pre-Analytic Phase to determine which genetic test is appropriate
2. Analytic Phase where samples are analyzed
3. Post-Analytic Phase where results are reported and interpreted

Intervention Specification
The pharmacogenetic CDSS is an alert-type intervention. Triggering occurs
when a provider uses a CPOE system to prescribe a medication for a patient.
The trigger is data-driven since a query will be performed in the clinical data
repository to determine if genetic testing results exist for the patient. If such
results are available, then a query will be performed to match the drug to a
known drug-gene pair in the pharmacogenomics knowledge base.
Decision rules in the CDSS will be based on clinical practice guidelines that
are published by the Clinical Pharmacogenetics Implementation Consortium
(CPIC). Founded in 2009, the CPIC is a group of organizations that develops
and publishes clinical practice guidelines for actionable prescribing decisions

16
based on the results of genetic testing (Caudle, 2014). CPIC guidelines are
written for drug-gene pairs where evidence clearly shows that genetic
variations affect efficacy or risk and where prescribing recommendations
such as changing dose or changing drug can be made. CPIC guidelines are
written in accordance with Institute of Medicine standards for the
development of clinical practice guidelines.
The CDSS will use the following logic (see figure 6). If genetic test results are
available for the patient and a guideline exists for a drug-gene pair, then a
modal alert box will immediately be displayed for the prescriber in the CPOE
user interface. The alert will inform the prescriber of the patients genotype
or phenotype and the expected variation in efficacy or risk of adverse effect
for the drug. Based on data from the guideline contained in the
pharmacogenomics knowledge base, the alert will offer specific advice for
dosage adjustment or alternative drug treatment as appropriate. An
infobutton will be included in the alert box which will link to more detailed
information and evidence for the recommended changes in treatment. The
prescriber will then have an option of canceling the order, modifying the
order, or overriding the alert and proceeding with the order. When overriding
an order the provider will need to justify the override by selecting a reason
from a dropdown list. If no genetic test results exist for the patient, if there
are no drug-gene matches, or if there are no important actionable clinical
practice guidelines, then no alert will be triggered and workflow will proceed
as normal. This logic step will help to prevent alert fatigue by presenting
providers with an alert only when actionable guidelines exist for a drug-gene
pair.
Because the field of pharmacogenetics is rapidly changing, the
pharmacogenomics knowledge base of the CDSS will need to be updated
with new practice guidelines from the CPIC at frequent intervals. We will
begin by updating these data sources every three months.
The CDSS will help the organization to meet the stated goals and objectives
by helping providers avoid prescribing medications for patients when a
known risk exists based on genotype. Adverse drug reactions will be avoided
by lowering doses or by choosing alternative medications. Cost savings can
be realized by avoiding adverse drug reactions, by avoiding the prescription
of drugs known not to be effective for a particular genotype, and by selecting
medications known to be more effective for a particular genotype.
Figure 6. Pharmacogenetic Clinical Decision Support System Logic

17

18

User Interface
According to a report from the HIMSS Usability Task Force (2009) user
interface design can be a key component for ensuring that clinical decision
support systems and other health information technology products meet
their intended goals of improving patient care and safety. This is because a
well-designed user interface will enhance provider adoption and lead to
routine clinical use. The design of a CDSS must be satisfying to the user,
must be effective in helping the user do his work, and must be effective at
guiding the user towards making the recommended decision. It is wise to
include end users in the planning stages of interface design so that changes
can be made in an iterative fashion in order to maximize acceptance. The
HIMSS Task Force notes several key principles for interface design. The
interface should be simple and concise and present information in a way that
is natural and familiar to the target end user. The design should aim to
minimize the cognitive load of the user who is usually a provider under time
pressure with multiple demands on his attention. Efficient interactions with a
minimum number of steps to complete tasks will enhance user satisfaction.
The design should provide feedback along the way in order to allow a user to
learn the system, especially since opportunities for formal training can often
be limited.
In a review on the topic of interface design, Horsky et al. (2012) recommend
that consistency between the CDSS and the larger EHR be maintained. This
includes visual formats, terminology, nomenclature, and processes. They
recommend presenting advice with short explanations and links to more
detailed evidence. This way, users will develop trust in the decision support
system over time since they will know where recommendations are coming
from rather than being confronted with a black box that issues advice.
Clinical decision support systems should offer guidance and
recommendations rather than commands. Physician are much more likely to
follow recommendations from decision support systems when they are
offered alternatives to carry out rather than a hard stop or a
recommendation against a particular action. Recommendations should be
written in succinct and unambiguous language that can be understood after
reading it in one pass. Alerts should be contained on one page with the
shortest amount of text possible followed by an actionable item.
The interface design of the pharmacogenetics CDSS will follow the principles
outlined above as well as the four As recommended by Kanstrup (2011):
Attention, All in one, At a glance, and At hand.
As mentioned in the intervention specification section above, the CDSS will
engage when a provider uses the CPOE system to write an order for a
medication. Input to the CDSS from the provider will come directly from the
CPOE. If an alert is triggered based on the decision support system logic, a
modal alert box will be immediately displayed. This alert box will have a
visual design that is consistent with the rest of the EHR with the exception
that it will be of a different color in order to draw attention.

19
The alert box will contain all relevant information about the drug-gene pair
and clinical guideline recommendations all in one box at a glance.
Verbiage will be minimized to present the provider with a succinct
explanation. Hyperlinks and infobuttons will be included in the alert box so
that providers can immediately have access to more detailed information
regarding the drug-gene pair and the clinical guidelines if they wish.
Immediately following the recommendations, providers will have at hand
the ability to override, modify, or cancel the order.
If providers chose to override the order, they will be asked to justify the
override using a drop down list which will include the following reasons:
patient previously tolerates this therapy, dose or therapy adjustment
already made, guideline not relevant for this patient, genetic testing is
erroneous, risks and benefits have been considered/patient will be carefully
monitored, and other. Providers will also be able to enter free text into the
override justification box. This information can be used to later fine-tune the
sensitivity of alerts.
If providers choose to modify the order, they will be taken back to the CPOE
screen with the original drug order information previously entered intact.
Providers may then change the dose or change the drug. If they choose to
cancel the order, they will be taken back to the CPOE screen with a blank
medication order entry form.
The following are two illustrations of the user interface.
Example 1: The provider attempts to order abacavir for a patient via the
CPOE. The CDSS searches the clinical data repository and finds that
genetic test results exists for the patient. The CDSS then searches the
pharmacogenetics knowledge base for abacavir and finds there is a
related drug-gene pair, abacavir and HLA-B. It then looks for HLA-B test
results for the patient. If those test results exist, it then looks for a clinical
guideline for abacavir/HLA-B. It finds that clinical guidelines do exists. If
the patient has the HLA-B*57:01 allele then the clinical guideline from
CPIC recommends alternative therapy due to high risk for hypersensitivity
reaction.

20

The clinical decision support system will then display an alert box shown in
figure 7.
Figure 7. Example clinical decision support alert box for abacavir.

The hyperlink labeled more information in the alert box will link to the
clinical guidelines describing the relationship between abacavir and HLA-B
and the evidence regarding risk of hypersensitivity reaction and
recommendation to avoid using the medication in patients with this
genotype.

The provider may then chose to override (for example if the patient has
already been taking abacavir with no problems). If he doesn't override ,
then he may cancel the order or modify it.
Example 2: The provider attempts to order codeine for a patient. The CDSS
searches for genetic test results. It then searches for codeine and finds the druggene pair codeine and CYP2D6. It then looks for CYP2D6 results for the patient
and looks for clinical guidelines for Codeine/CYP2D6. The system then reports the
clinical guideline recommendations based on the patient's phenotype as shown in
figure 8.

21
Figure 8. Example clinical decision support alert box for codeine.

Conceptual Model for the Pharmacogenetics CDSS

Based on the System Inventory, the summary below of each model

component is organized by method of implementation: Client/Server or Webbased.

Client/Server Based
The CDSS EHR Application Environment supports the functional capabilities
that were evaluated by the group and shown in the Components of System
Architecture diagram.
Discrete data elements are captured within the Clinical Data Repository. The
clinical user can interact with the Clinical System Interface to view data
elements and perform clinical tasks, e.g., ordering a medication. Such events
will then trigger decision support to fire. Triggers are implemented in the
Clinical Data Repository. Interventions and Offered Choices are displayed to
the clinical user within the CDSS Clinical System Interface.

Web-Based
The Clinical Decision Support (CDS) Module implements the IF-THEN rules
that are defined in the Rules/Execution Engine and based on each patients
genotype data combined with phenotype data stored in the CDS Knowledge
Base.
The database and rules together relate actionable genotype and phenotype
pairs to genome-informed advice messages. If defined rules are met, CDS is
delivered in real-time at the point of care through the EHR.

22
The CDS Module also implements the User Interface features (information,
warnings, recommendations) and provides the methods for transforming
input parameters (i.e., data about genetic test results and the medication
being ordered) into a specific output (i.e., Pharmacogenetic links to eResources or an Alert message).

Glossary of Terms
Clinical Data Repository
Consolidates data from multiple clinical sources and contains lab results,
patient demographics, pharmacy information, radiology reports, admission /
discharge summaries code vocabularies, progress notes, etc.
Pharmacogenetic Knowledge Base
Sponsored by NIH (pharmaKB.org), this knowledge base is the curator of
research information on genetic variations in drug responses, e.g., how
genetic factors influence a patients response to drugs
Genomics Vendor Knowledge Base
Contains genotyping and genetic testing results for patients
CDS Knowledge Base
Uses an inference engine to facilitate presentations of drug interactions,
dosing suggestions, alerts, and diagnoses by applying rules to patient data

Knowledge Engineering
Knowledge acquisition
While the interaction relationship within a drug-gene pair is likely to remain
stable once established, the quantity of relevant genotypes is growing
rapidly. It would be beyond the scope of our healthcare organization to try
and directly acquire and maintain a compendium of these pairs from primary
pharmacogenetic literature. Fortunately, as previously noted, the Clinical
Pharmacogenetics Implementation Consortium (CPIC) represents a
collaborative effort of significant academic breadth which composes clinical
guidelines for drug-gene pairs in a format that complies with Institute of
Medicine (IOM) standards. The acquisition process in this case is actually a
selection process where the organization chooses the guidelines that are
considered relevant for its needs. It may be tempting at first to think about a
wholesale adoption of guidelines, but in the long run, a process of taking
ownership by selecting and applying the guideline is an important one to
aid in the adoption of these new guidelines.
The Clinical Decision Support Committee (CDSC) for the organization is
central to have the organization take ownership of these guidelines. They
would be able to assess the needs of the organization and leverage the
resources necessary to approve and oversee a robust process of acquisition

23
and adoption. There would be merit in considering a subcommittee for this
activity to handle the anticipated volume of new guidelines emanating from
CPIC.

Knowledge representation
Taking ownership of a drug-gene pair guideline from the CPIC will require
more than a simple title/description of the relationship. These relationships
need to be parsed into a consistent set of elements which will facilitate their
incorporation in to the organizations CDSS. The proposed structure can be
seen in Appendix A using a table of drug gene pairs where three layers are
inferred. The top layer represented by the first two columns in green are the
drug-gene pair descriptions. The last three columns in white (Phenotype,
Guideline Recommendation, Recommendation Type, etc.) would have a
potential many-to-one relationship with the pair as indicated by many pairs
being seen more than once. The Genotype column in yellow has a many-toone relationship with the Phenotype recommendations since a given
phenotype could derive from multiple genotypes.
Ultimately these columns would correlate with data entry points for the
clinical decision support system to allow the elements to present in the user
interface in the method previously described. Other meta-data will also be
helpful to capture including the date approved by the CDSC and the primary
organizational researcher. A future review data could also be entered to
facilitate reporting and queuing of guidelines needing review and renewal.
These future reviews would occur at a pre-determined interval to determine
if new literature has presented any refinements or recommendations,
perhaps around lab monitoring.

Evaluation
The Pharmacogenetics Clinical Decision Support System is a relatively new
model and requires both validation and verification. The validation phase will
be a robust testing program before implementation. This testing will be
handled in three segments called unit testing, scenario testing and
integrated testing.
1. Unit Testing: During this phase of testing, each drug-gene pair result
is tested by having the particular genetic result entered on a test
patient in a test domain. The corresponding drug component would
then be entered to validate that the clinical decision support that fires
works as designed. It should offer accurate information about how to
respond. The cycle here is relatively simple, but the volume of cycles
is directly tied to the volume of drug-gene pairs being implemented.
2. Scenario Testing: During this phase of testing, a testing script is
used to elicit specific scenarios that may challenge the robustness of
the CDSS. A sample testing script is attached in Appendix B and

24
would likely be enhanced by the project team to add additional
scenarios desired by stakeholders.
3. Integrated Testing: Unit testing will occur within the confines of the
CDSS and the EHR. Integrated testing will engage all of the necessary
systems to test the entire flow of the CDSS. This will include the
addition of relevant drug-gene pairs to the relevant knowledgebase,
interfacing of gene results from the laboratory information system to
the EHR and completion of medication order including transmittal via
e-prescribing. The goal is to detect unintended consequences to
related downstream systems.
Verification will commence after implementation and will utilize a defined set
of metrics to assess impact. There will be pre-verification analysis to
determine a baseline for metrics of interest:

Medication ordering frequencies for each medication where a druggene pair exists
Dosing variability for each medication where a drug-gene pair exists
Adverse events for each medication where a drug-gene pair exists
Frequency and count of dose changes, including discontinuation and
starting, for each medication where a drug-gene pair exists

The baseline metrics above will be continued after implementation, and once
the CDSS is implemented, the following metrics will be added:

Alert frequency for absence of relevant gene testing

Percentage of corresponding medications prescribed without prior
genotype information
Alert frequency for therapeutic modification due to relevant drug-gene
pair
Response frequencies to alert (override, change medication, alter
dosing)
Percentage of medications prescribed where dosing incorporated
suggested dose modification

Specific drug-gene pairs may utilize metrics of enhanced efficacy. In the

case of warfarin based products, median time to therapeutic effect as
measured by Prothrombin Times (PT) and their related INR. Not all
medications will have this type of a metric available.

Discussion
Several previous studies have reported on the feasibility of incorporating
pharmacogenetic data into CDSS and some have reported on the
deployment of such data in CDSS in pilot studies or in a limited scope with
decision support for a particular drug or class of drugs (Bell 2014, Bielinski

25
2014, Deskhmukh 2009, Goldspell 2013, Overby 2010 and 2012). Here we
have presented ideas for a wider scope of pharmacogenetic CDS. Given the
underlying assumption that patients will already have the results of genetic
testing available in the clinical data repository (preemptive genotyping), this
system will have limited value to providers and patients in organizations
where the prevalence of such genetic testing is low. It is expected that in the
future genetic testing will be less expensive and more ubiquitous. As this
occurs, this CDSS will be at the ready to help make use of that data for the
benefits of those patients who have had genetic testing done.
The interpretation of genetic testing can be confusing, especially for
providers with limited training and limited time. Evaluating the vast amount
of research in pharmacogenetics is a daunting task. For this reason, our
system employs guidelines that have already been vetted and approved by
the CPIC.
The pharmacogenetics CDSS has been designed keeping in mind the three
pillars of the CDS Roadmap outlined by the American Medical Informatics
Association as reviewed by Lyman (2010). Best knowledge available when
needed means that this system presents only CPIC approved guidelines with
actionable decision support to providers at the point of care when they
attempt to prescribe a medication. This fits into the standard workflow and
should acceptable for providers. High adoption and effective use means
that alerts will be kept to a minimum in order to prevent alert fatigue.
Scientific evidence of each recommendation will be available to help build
trust and support for the system. Continuous improvement of knowledge
and CDS methods means that the CDSS will be evaluated regularly and
refined according to provider and organization needs. The knowledge base
will be updated at frequent intervals using information from the CPIC
guidelines. Alerts will be fine-tuned based on the number and reasons for
overrides.
Scalability is an important factor for the CDSS as pharmacogenetic
knowledge is constantly evolving. Utilizing decision rules based on the CPIC
guidelines will help to ensure scalability in the future because staff
responsible for maintaining the CDS knowledge base and decision rules will
be able to find this information in one place. New information can be taken
from the CPIC and quickly integrated into the CDSS.
Although this system was design to engage prospectively at the point of care
when providers attempt to prescribe medications, it could be expanded. The
system could potentially be used retrospectively to scan all patient data
including active medications and genetic testing results. If the system finds
patients who have been prescribed medications which have a drug-gene
match and an actionable clinical guideline, it could inform the prescribing
physician via email about the guideline. The provider could then make an
adjustment to therapy or dosing if clinically relevant. The system could also
be employed as a research tool. Providers could access the system to search
for drug-gene matches for patients prior to accessing CPOE to prescribe. This

26
may offer some benefit in that providers can discuss treatment options with
patients, considering risks and benefits.
As providers and healthcare organizations strive to offer ever more
sophisticated treatment plans based on personalized medicine, the
pharmacogenetics clinical decision support system will be an essential tool. It
will help providers by gathering and presenting information that is relevant to
an individual patient at the point of care. It will help disseminate to providers
the vast amount of pharmacogenetic data available and therefore aid in the
more rapid adoption of clinical guidelines based on pharmacogenetic data.
This system has the potential to greatly improve patient safety and reduce
healthcare costs.

27

References
Baars M. J. (2005). Deficiency of Knowledge of Genetics and Genetic Tests
among General Practitioners, Gynecologists, and Pediatricians. Genetics
in Medicine. 7(9), 605-610.
Bell G. C. (2014). Development and Use of Active Clinical Decision
Support for Preemptive Pharmacogenetics. Journal of the American
Medical Informatics Association. 21, e93-e99.
Bielinski S. J. (2014). Preemptive Genotyping for Personalized Medicine:
Design of the Right Drug, Right Dose, Right Time--Using Genomic Data to
Individualize Treatment Protocol. Mayo Clinic Proceedings. 89(1), 25-33.
Caudle K. E. (2014). Incorporation of Pharmacogenetics into Routine
Clinical Practice: the Clinical Pharmacogenetics Implementation
Consortium (CPIC) Guideline Development Process. Current Drug
Metabolism. 15(1), 1-9.
Deshmukh V. G. (2009). Efficiency of CYP2C9 Genetic Test Representation
for Automated Pharmacogenetic Decision Support. Methods of
Information in Medicine. 48(3), 282-290.
Epstein R. S. (2010). Warfarin genotyping reduces hospitalization rates
results from the MM-WES (Medco-May Warfarin Effectiveness
study). Journal of the American College of Cardiology. 55, 2804-28-12.
Gazzard B. G. (2008). British HIV Association Guidelines for the treatment
of HIV-1-infected adults with antiretroviral therapy. HIV Medicine. 9, 563608.
Ge D. (2009). Genetic variation in IL28B predicts hepatitis C treatmentinduced viral clearance. Nature. 461, 399-401.
Goldspeil B. R. (2013). Integrating Pharmacogenetic Information and
Clinical Decision Support into the Electronic Health Record. Journal of the
American Medical Informatics Association. 0, 1-7.
HIMSS Usability Task Force (2009). Defining and Testing EMR Usability:
Principles and Proposed Methods of EMR Usability Evaluation and Rating.

Horsky J. (2012). Interface Design Principles for Usable Decision Support:

A Targeted Review of Best Practices for Clinical Prescribing
Interventions. Journal of Biomedical Informatics. 45(6), 1202-1216.
Hulse M. (1997). User-Interface Design of a Web-Based Clinical Decision
Support System. Proceedings of the American Medical Informatics
Association Annual Fall Symposium. p. 951.

28

Kalow W. (1998). Hypothesis: comparisons of inter- and intra-individual

variations can substitute for twin studies in drug
research. Pharmacogenetics. 8, 283-289.
Kanstrup A. M. (2011). Four Principles for User Interface Design of
Computerised Clinical Decision Support Systems. Studies in Health
Technology and Informatics. 166, 65-73.
Kauf T. L. (2010). Economic efficiency of genetic screening to inform the
use of abacavir sulfate in the treatment of HIV. Pharmocoeconomics. 28,
1025-1039.
Klein T.E. (2009). Estimation of the warfarin dose with clinical and
pharmacogenetic data. New England Journal of Medicine. 360, 753-764.
Kohn L. T. (1999). To Err is Human: Building a Safer Health System.
Washington, DC: National Academy Press, Institute of Medicine.
Lyman J. A. (2010). Clinical Decision Support: Progress and
Opportunities. Journal of the American Medical Informatics
Association. 17(5), 487-492.
Osheroff J. A. (2012). Improving Outcomes with Clinical Decision
Support. Chicago, IL: Healthcare Information and Management Systems
Society.
Overby C. L. (2010). Feasibility of Incorporating Genomic Knowledge into
Electronic Medical Records for Pharmacogenomic Clinical Decision
Support. BMC Bioinformatics. 11(Suppl 9), s1-9.
Overby C. L. (2012). Developing a Prototype System for Integrating
Pharmacogenomics Findings into Clinical Practice. Journal of Personalized
Medicine. 2(4), 241-256.
Overby C.L. (2011) An Evaluation of Functional and User Interface
Requirements for Pharmacogenomics Clinical Decision Support. IEEE
International Conference on Healthcare Informatics.
Ross S. (2012). Promises and Challenges of Pharmacogenetics: An Overview
of Study Design, Methodological and Statistical Issues. JRSM
Cardiovascular Disease. 1(2), 1-13.
Weinshilboum R. (2003). Inheritance and drug response. New England
Journal of Medicine. 348, 529-537.
Welch B. M. (2013). Clinical Decision Support for Genetically Guided
Personalized Medicine: a Systematic Review. Journal of the American
Medical Informatics Association. 20(2), 388-400.

29

Chavez, Heffernan, West

Appendix A. Pharmacogenetics Knowledge Base Table

Type of
Drug

Gene

Genotype Phenotype

*1/*2,
*1/*3,
clopidogrel

CYP2C19 *2/*17

intermedia
te
metabolize

*2/*2,
clopidogrel

CYP2C19 *2/*3, *3/*3poor

metabolizer

phenytoin

CYP2C9

phenytoin

CYP2C9

warfarin

CYP2C9

amitriptyline

CYP2D6

amitriptyline

CYP2D6

Guideline Recommendation

consider alternative therapy such as

prasugrel or
ticagrelor due to risk of reduced platelet
inhibition and increased risk of bleeding and
adverse
events
consider cardiovascular
alternative therapy
such as
prasugrel or ticagrelor due to significant
risk of reduced platelet inhibition and
severely increased risk of bleeding

adverse
standardcardiovascular
loading dose, events
reduce maintenance
dose by 25%

*1/*2, *1/*3low
metabolizer

due to risk for adverse effects such as ataxia,

nystagmus,
dysarthria
sedation
standard
loading
dose, and
reduce
maintenance

*2/*2,
*2/*3,
*3/*3
*1/*1,
*1/*2,

very low
metaboliz
er

dose by 50%

*1/*3,

low
metabolizer

*1/*1xN,

ultrarapid

*1/*2xN
*4/*10,

metaboliz
er
intermedia
te

*5/*41

due to risk for adverse effects such as ataxia,

nystagmus,
dysarthria
and is
sedation
dose
reduction
of warfarin
recommended
due to risk of bleeding and cardiovascular
events, it is recommended to use a dosing
algorithm

*4/*4,
*4/*5,

amitriptyline

CYP2D6

*5/*5,
*4/*6
*1/*1xN,

poor
metabolizer
ultrarapid

codeine

CYP2D6

*1/*2xN
*4/*4,

metabolizer

codeine

CYP2D6

*4/*5,

capecitabine

DPYD

DPYD*2A/*
2
A

capecitabine

DPYD

fluorouracil

DPYD

2
A

fluorouracil

DPYD

DPYD*1/*2
A
DPYD*2A/*

tegafur

DPYD

tegafur

DPYD

DPYD*1/*2 DPYD
activity
A

DPYD*1/*2
A
DPYD*2A/*

therapy change

dose change

0.75 Maintenanc
e

dose change

0.5 Maintenanc
e

dose change

therapy change

metabolized
CYP2D6 in dosing due to
consider
25%by
reduction
increased risk of side effects

dose change

0.75 All

0.75 All

avoid tricyclic antidepressant due to high risk

of side effect, consider alternative drug not
metabolized by CYP2D6, if tricyclic is used
therapy change
consider
50% dose
reduction
avoid
codeine
use due
to potential for

toxicity, consider

therapy change
alternative
opiate
avoid
codeine
use due to lack of efficacy,
consider alternative opiate
poor
therapy change
metabolizer
complete
DPD avoid capecitabine due to risk of severe or
fatal drug
deficiency
therapy change
toxicity
decreased DPDstart
with 50% dose reduction then titrate dose
based on toxicity
activity
dose change
complete DPD avoid fluorouracil due to risk of severe or
fatal drug toxicity
deficiency
therapy change
decreased
start with 50% dose reduction then titrate
dose based on
DPD
dose change
activity
toxicity
complete
DPYD
avoid tegafur due to risk of severe or fatal
therapy change

deficiency
decreased

Lab
Dose Type Monitoring

therapy change

avoid tricyclic antidepressant due to lack of

efficacy, consider alternative drug not

metabolize

Modificatio

Recommendati n
on
factor

drug
toxicity
start with
50% dose reduction then titrate dose
based on toxicity
dose change

0.5 All

0.5 All

0.5 All

30

Chavez, Heffernan, West

positive

rasburicase

G6PD

abacavir

HLAB

*57:01

positive

allopurinol

HLAB

HLA
B*58:01

high risk

carbamazepine

HLAB

peginterferon alfa IFNL3

2a

*15:02
CT or TT

carrier
unfavorable

avoid rasburicase, contraindicated due to risk

of severe
therapy change
hemolysis
avoid
abacavir due to significantly
increased risk of hypersensitivity
avoid alluprinol due to significant risk
of severe cutaneous adverse reaction

therapy change
therapy change

avoid carbamazepine if patient is nave due to

significant risk of cutaneous adverse reactions,
if patient has been on carbamazepine therapy
for longer than 3 months without reaction
therapy change
may consider
continued
use due to lack of
consider
alternative
therapy
therapy change
efficacy
consider dose reduction due to increased risk

simvastatin

SLCO1B1 TC

intermediate
risk

of
myopathy

dose change

0.25 All

simvastatin

SLCO1B1 CC
*1/*2,

high risk

consider dose reduction due to increased risk

of myopathy, consider routine creatine kinase dose change
monitoring

0.5 All

azathioprine

TPMT

high risk

consider initial dose reduction of 3070% or

target dose and titrate based on tolerance
consider alternative therapy due to risk of

dose change

0.5 All

therapy change

0.1 All

*1/*3A,
*1/*3B,

*3A/*3A,
*2/*3A,
azathioprine

TPMT

*3C/*3A,
*3C/4,

extremely
high risk

*1/*2,

mercaptopurine

TPMT

*1/*3A,
*1/*3B,

myelosupression, if azathioprine is used

drastically reduce dose by 10fold and

consider three times a week dosing instead

of daily dosing

high risk

consider initial dose reduction of 3070% or

target dose and titrate based on tolerance

dose change

0.5 All

extremely
high risk

used drastically reduce dose by 10fold

and consider three times a week dosing

dose change

0.1 All

high risk

instead of daily dosing

consider initial dose reduction of 3050% or
target dose and titrate based on tolerance

dose change

0.5 All

extremely
high risk

used drastically reduce dose by 10fold and

consider three times a week dosing

dose change

0.1 All

dose change

0.7 >250
mg/m^2

*3A/*3A,

*2/*3A,
mercaptopurine

TPMT

thioguanine

TPMT

thioguanine

TPMT

*3C/*3A,
*1/*2,

*1/*3A,
*1/*3B,
*3A/*3A,
*2/*3A,

*3C/*3A,

irinotecan

UGT1A1

UGT1A1*28high risk

warfarin

VKORC1

GG, GA, AA low

metabolizer

instead initial
of daily
dosing
reduce
dose
by 30% for patients
receiving dose of
more than 250 mg/m^2 due to risk of
myelosupression
arrhythmia
dose
reduction ofand
warfarin
is recommended
due to risk
of bleeding and cardiovascular
events, it is recommended to use

dose change

0.75 All

CPK

31

Chavez, Heffernan, West

Appendix B. Scenario Testing Script

Step
1

Team

User/ Dept
prescriber

Workflow Step
order corresponding
medication for
drug/gene pair

prescriber

order corresponding
medication for
drug/gene pair

prescriber

order corresponding
medication for
drug/gene pair

prescriber

prescriber

order corresponding
medication for
drug/gene pair where
2 genes relevant to
the medication are
gene result returns
with pre-existing order
for medication

prescriber

order corresponding
medication for
drug/gene pair

prescriber

order corresponding
medication for
drug/gene pair

prescriber

order corresponding
medication for
drug/gene pair

prescriber

order corresponding
medication for
drug/gene pair where
2 genes relevant to
the medication are
gene result returns
with pre-existing order
for medication

prescriber

Actual
Testing Step/Process
Expected Outcome
Outcome
Test Patient 1 (no pre-existing prescriber alerted to absence
result) - order coumadin of any gene testing and
OUTPATIENT
advised to order as results
could affect dosing and/or
Test Patient 2 (pre-existing
Prescriber completes order for
gene result with genotype that medication without any alert
does not affect dosing and/or
efficacy) - order coumadin Test Patient 3 (pre-existing
prescriber alerted to presence
gene result with genotype that of genotype that affects
does affect dosing and/or
dosing and/or efficacy with
efficacy) - order coumadin recommended modification
Test Patient 4 (2 pre-existing prescriber sees both alerts
gene results with genotype
that affect dosing and/or
efficacy of single medication) order coumadin Test Patient 3 (pre-existing
prescriber alerted to presence
gene result with genotype that of genotype that affects
does affect dosing and/or
dosing and/or efficacy with
efficacy) - refill/reorder
recommended modification
Test Patient 1 (no pre-existing prescriber alerted to absence
result) - order coumadin of any gene testing and
INPATIENT
advised to order as results
could affect dosing and/or
Test Patient 2 (pre-existing
Prescriber completes order for
gene result with genotype that medication without any alert
does not affect dosing and/or
efficacy) - order coumadin Test Patient 3 (pre-existing
prescriber alerted to presence
gene result with genotype that of genotype that affects
does affect dosing and/or
dosing and/or efficacy with
efficacy) - order coumadin recommended modification
Test Patient 4 (2 pre-existing prescriber sees both alerts
gene results with genotype
that affect dosing and/or
efficacy of single medication) order coumadin - INPATIENT
Test Patient 3 (pre-existing
prescriber alerted to presence
gene result with genotype that of genotype that affects
does affect dosing and/or
dosing and/or efficacy with
efficacy) - refill/reorder
recommended modification

Pass/
Fail

Tester

Date
Tested

Comments
or Issues

32

Chavez, Heffernan, West

pharmacist

6
7

dispense
corresponding
medication for
drug/gene pair
order corresponding
medication for
drug/gene pair

Test Patient 1 (no pre-existing pharmacist alerted to absence

result) - dispense coumadin - of any gene testing and
INPATIENT
advised to order as results
could affect dosing and/or
pharmacist
Test Patient 2 (pre-existing
pharmacist completes
gene result with genotype that dispense for medication
does not affect dosing and/or without any alert
efficacy) - dispense coumadin pharmacist
order corresponding Test Patient 3 (pre-existing
pharmacist alerted to
medication for
gene result with genotype that presence of genotype that
drug/gene pair
does affect dosing and/or
affects dosing and/or efficacy
efficacy) - dispense coumadin - with recommended
pharmacist
order corresponding Test Patient 4 (2 pre-existing pharmacist sees both alerts
medication for
gene results with genotype
drug/gene pair where that affect dosing and/or
2 genes relevant to
efficacy of single medication) the medication are
dispense coumadin pharmacist
gene result returns
Test Patient 3 (pre-existing
pharmacist alerted to
with pre-existing order gene result with genotype that presence of genotype that
for medication
does affect dosing and/or
affects dosing and/or efficacy
efficacy) - refill/redispense
with recommended
coumadin - INPATIENT
modification
result abstractor result entry
enter genotype with incorrect result component will only
string
allow validated strings
medical
records historical
Enter patient reported
no alert fires
assistant
medication
medication (Coumadin) on