Clinical decision support for genetically guided personalized medicine: a systematic review
Transcript of Clinical decision support for genetically guided personalized medicine: a systematic review
Clinical decision support for genetically guidedpersonalized medicine: a systematic review
Brandon M Welch,1 Kensaku Kawamoto2
ABSTRACTObjective To review the literature on clinical decisionsupport (CDS) for genetically guided personalizedmedicine (GPM).Materials and Methods MEDLINE and Embase weresearched from 1990 to 2011. The manuscripts includedwere summarized, and notable themes and trends wereidentified.Results Following a screening of 3416 articles, 38primary research articles were identified. Focal areas ofresearch included family history-driven CDS, cancermanagement, and pharmacogenomics. Nine randomizedcontrolled trials of CDS interventions for GPM wereidentified, seven of which reported positive results. Themajority of manuscripts were published on or after 2007,with increased recent focus on genotype-driven CDS andthe integration of CDS within primary clinical informationsystems.Discussion Substantial research has been conducted todate on the use of CDS to enable GPM. In a previousanalysis of CDS intervention trials, the automaticprovision of CDS as a part of routine clinical workflowhad been identified as being critical for CDSeffectiveness. There was some indication that CDS forGPM could potentially be effective without the CDS beingprovided automatically, but we did not find conclusiveevidence to support this hypothesis.Conclusion To maximize the clinical benefits arisingfrom ongoing discoveries in genetics and genomics,additional research and development is recommendedfor identifying how best to leverage CDS to bridge thegap between the promise and realization of GPM.
BACKGROUNDGenetically guided personalized medicine (GPM)entails the delivery of individually tailored medicalcare that leverages information about each person'sunique genetic characteristics.1 The promise ofGPM has expanded as advances in genomics haveaccelerated over the past several decades. Thispromise of GPM is that research discoveries willone day lead to medical treatments and therapiesthat are tailored to the individual characteristics ofeach patient, including clinical data, genetic testresults, patient preference, and family healthhistory (FHx). GPM has the potential to increasethe efficacy, quality, and value of healthcareby providing individually optimized prevention,diagnosis, and treatment.2
As ongoing research continues to expand theGPM knowledge base, it has become increasinglyimportant to translate this knowledge into routinehealthcare practice in order to realize the promise ofGPM.3 However, the effective realization of GPMremains very limited.4 While this is partly due to
the need for further evidence of the clinical utilityand cost effectiveness of a genetically guidedapproach to patient care, an important additionalreason is the need for information systemsthat assist in the translation of knowledge frombench to bedside.5 Even without the complexity ofgenetics, it can often take over 15 years to translateresearch from bench to bedside.6 This translationalbottleneck is likely to be an even more significantproblem in GPM for the following reasons.
Limited genetic proficiency of cliniciansMany clinicians receive minimal training in clinicalgenetics. As a result, many physicians lack theconfidence and understanding needed for effectivelyinterpreting and using genetic information in theirclinical practices.7
Limited availability of genetics expertsCurrently, there are about 3000 board-certifiedgenetic counselors8 and approximately 1200medical geneticists practicing in the USA (S. R.DelBusso, American Board of Medical GeneticsAdministrator, October 28, 2011, personalcommunication). The growing utility of geneticinformation is putting an increasing burden onthese professionals. We cannot expect thesegenetics experts to be readily available each timegenetic information should be used to guidemedical treatment. For effective, efficient, andwidespread clinical use, the burden of geneticinterpretation and guidance must be shared by thewider clinical community.
Breadth and growth of genetic knowledge baseThere are currently over 2500 clinical genetic testsavailable to clinicians, encompassing a widebreadth of medical care.9 It is therefore unreason-able to expect a clinician to remember everyappropriate genetic test for a particular conditionin conjunction with test-specific guidelines forordering and interpretation. Compounding thisissue, the continual growth in the knowledge baseand the prospect of full genome sequencing willinevitably overwhelm clinicians' capacities tomanage and leverage this information effectivelyfor GPM unless computerized assistance is providedfor interpreting and acting on this information.Various investigators and leaders have identified
health information technology as being vitalto overcoming these barriers and realizing thepromise of GPM.2 10 In particular, clinical decisionsupport (CDS) has been identified as a criticalenabler of GPM.11 12 CDS entails providing clini-cians, patients, and other healthcare stakeholderswith pertinent knowledge and person-specific
< An additional appendix ispublished online only. To viewthis file please visit the journalonline (http://dx.doi.org/10.1136/amiajnl-2012-000892).1Department of BiomedicalInformatics and Program inPersonalized Health Care,University of Utah, Salt LakeCity, Utah, USA2Department of BiomedicalInformatics and InformationTechnology Services, Universityof Utah, Salt Lake City, Utah,USA
Correspondence toDr Kensaku Kawamoto,Department of BiomedicalInformatics, University of Utah,615 Arapeen Way, Suite 208,Salt Lake City, UT 84092, USA;[email protected]
Received 10 February 2012Accepted 27 July 2012
Review
388 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892
Published Online First22 August 2012
Editor ’s choiceScan to access more
free content
information, intelligently filtered or presented at appropriatetimes, to enhance health and healthcare.13 CDS has the capacityto process complex, disparate data and present actionable,standardized, evidence-based recommendations in a way that isusable by a clinician in everyday practice.11 As such, CDS canhelp bridge the gap between the promise and realization of GPM(figure 1). Given the criticality of CDS for realizing the promiseof GPM, and given the lack of a systematic review on this topic,we sought in this paper to assess the history and state of CDSfor GPM through a systematic review of the literature.
METHODSData sources and inclusion criteriaWe searched MEDLINE and Embase from 1990 to 2011 usinga search strategy adapted from previous systematic reviews ofCDS,14 genetic health services,15 and FHx16 (see supplementaryappendix, available online only, for full search strategy). Thefinal literature search was conducted on June 1, 2012. Theinclusion criteria for the review were as follows: English article;human focus; manuscript in peer-reviewed journal; and primaryfocus on the use of computers to deliver genetically guided,patient-specific assessments and/or recommendations tohealthcare providers and/or patients to guide clinical decision-making, as further defined in Box 1.
For all identified references, the authors reviewed titles, indexterms, and available abstracts to determine if the articlesappeared to meet all inclusion criteria. If insufficient informationwas available to make a confident decision at this stage, thearticle was included for full-text retrieval. Each full-text articlewas then reviewed to determine its final inclusion status.
Data abstractionFor each of the articles that met the inclusion criteria listedabove, we abstracted data on the clinical application area, CDStype, genetic information used, primary users, article type, studylocation, CDS purpose, and notable informatics aspects. CDStype was defined as being either stand-alone CDS or integratedCDS. A stand-alone CDS system is a CDS system that exists inisolation from a primary clinical information system containingrelevant patient data, such as an electronic health record (EHR)system. A stand-alone CDS system requires manual data inputbefore a CDS result can be produced. In contrast, an integratedCDS system is integrated with a primary clinical information
system such as an EHR system or a computerized provider orderentry system to aggregate necessary patient-specific informationautomatically and to provide guidance within routine clinicalworkflows. Clinical application area was defined as the clinicaldomain targeted by the CDS intervention. Article type consistedof system description papers and evaluation studies of varioustypes (eg, qualitative evaluation, randomized controlled trial).Genetic information used consisted of FHx, genotype, or both.Primary users were defined as the individuals who primarilyentered information and received the results. Study location wasthe country or region where the research was conducted. CDSpurpose identified the role of the CDS system within thecontext of clinical decision-making. A notable informatics aspectwas also abstracted if a manuscript utilized a methodology thatwas considered to be of potential interest to an informaticsaudience. For intervention studies, additional details regardingthe study size and study outcomes were abstracted.
Data analysis and presentationUsing the abstracted attributes, the manuscripts were groupedinto logical categories, primarily according to CDS type andclinical application area. The findings from these manuscriptswere summarized through tables and narrative discussion. Inaddition, notable themes and trends were identified anddiscussed. A quantitative analysis of CDS trials to identifyfeatures predictive of trial outcomes was considered.14 However,due to the limited sample size of CDS trials available, sucha quantitative analysis of potential success factors was notfeasible.
RESULTSThe initial MEDLINE and Embase searches identified 3416potentially relevant articles. During the title and abstract review,82 articles were rejected for not being in English, 504 articleswere rejected because they were not focused on humans, 34articles were rejected for not being a peer-reviewed manuscript,and 2494 articles were rejected because the primary focus of thework was not on the use of computers to deliver geneticallyguided, patient-specific assessments and/or recommendations.The remaining 302 articles underwent full-text review, at whichstage 37 articles were rejected for not being a peer-reviewedprimary research article and 227 articles were rejected becausethe primary focus of the work was not on the use of computers
Figure 1 Clinical decision support(CDS) as bridge overcoming barriers togenetically guided personalizedmedicine.
CDS
Where we are now
Where we want to be
The Promiseof Personalized
Medicine
The Realizationof Personalized
MedicineBarriers -Limited genetic proficiency of clinicians-Limited availability of genetics experts -Growth of genetic knowledge base
-Research Discoveries-Technology Developments
-Safer Healthcare-Improved Outcomes-Reduced Costs
"Bench" "Bedside"17+ years translating research into medical practice(Traditional Path)
Review
J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892 389
to deliver genetically guided, patient-specific care guidance(figure 2). The final set of included manuscripts consisted of 38primary research articles.17e54 The manuscripts included werepublished from 1990 to 2011, with the majority of manuscriptspublished on or after 2007. Provided below is a summary andanalysis of these earlier works, grouped primarily by CDS typeand area of clinical focus.
CDS systems for genetically guided cancer managementGenetically guided cancer management was the focus of 22primary research articles summarized in tables 1e4.17e37 54
These manuscripts include six manuscripts related to the RiskAssessment in Genetics (RAGs) system for providing FHx-drivenCDS (table 1),17e21 54 six manuscripts on other FHx-drivenCDS tools for breast cancer management (table 2),22e27 fourmanuscripts on genotype-driven CDS tools for breast cancermanagement (table 3),28e31 and six additional manuscripts
on GPM CDS tools for non-breast cancer management(table 4).32e37
RAGs system for providing FHx-driven CDSSome of the earliest and most comprehensive research on the useof CDS to support GPM was conducted by Emery55 (table 1),who identified that existing systems were not designed forprimary care and that none provided patient management advicebased on calculated risk. To address this gap, Emery developeda system known as RAGs, which helped general practitioners(GPs) in the UK collect FHx relevant to familial breast, ovarian,and colorectal cancer and provided appropriate managementguidance, primarily regarding guideline-based specialistreferrals.17e19 54 A later extension of the RAGs system wasreferred to as the GRAIDS system.20 21 This body of workincluded several favorable evaluations of these systems,18 19 21
including a cluster randomized controlled trial (RCT) across 45GP teams that found that GRAIDS significantly increased theproportion of patients referred appropriately to the regionalgenetics clinic according to evidence-based practice guidelines.21
Other FHx CDS tools for breast cancer managementBeyond the work of Emery,55 CDS research for GPM has focusedheavily on breast cancer management (table 2). Risk assessmenttools for breast cancer can enable personalized care according toan individual's level of risk.22 23 An RCT conducted in the UKfound that a stand-alone breast cancer CDS tool had limitedimpact due to lack of awareness and use by GPs.24 At the sametime, a stand-alone CDS tool that calculated risks for breastcancer, heart disease, osteoporosis, and endometrial cancer wasshown in an RCT to enhance the effectiveness of geneticcounselors using the system.25 26 Another stand-alone CDSsystem that has been found to be beneficial is HughesRiskApps,which collects relevant FHx information and provides clinicianswith various tools to support the management of patients. Anobservational implementation study of this tool in a communityhospital setting found significant adoption and impact.27
Genotype-driven CDS tools for breast cancer managementSeveral investigators have developed CDS systems that supporttreatment and decision-making once mutations have been iden-tified in the breast cancer (BRCA) genes (table 3). In the UK,Glasspool and colleagues30 31 developed a CDS tool known asREACT (Risks, Events, Actions and their Consequences overTime), which used a graphical timeline display to model real-time changes in lifetime risks as a result of risk-reduction inter-ventions for breast cancer and ovarian cancer. In addition, severalpatient-directed, stand-alone CDS systems have been developedfor improving risk communication and decision-making in breastcancer management based on BRCA genotype.28 29
CDS for other cancersBesides breast cancer, other cancers have been the focus of CDSresearch and development (table 4). Most of this CDS researchfor other cancers has involved colorectal cancer, and in particularLynch syndromeda strongly heritable type of colorectalcancer.32e34 Of note, the RAGs and GRAIDS systems describedearlier supported both breast cancer and colorectal cancermanagement.17e21 54 An additional CDS system investigated forcolorectal cancer management is CRCAPRO, similar toBRCAPRO, which used FHx to identify patients at risk ofhereditary colorectal cancer.33 In addition, a group in theNetherlands developed a CDS intervention to remind patholo-gists to order Lynch syndrome genetic testing among patients
Box 1 Manuscript inclusion criteria
< Definitions:– Healthcare provider ¼ physician, nurse practitioner,physician assistant, registered nurse, or genetic counselor
– Genetic factor ¼ genotype, gene expression profile, and/orfamily health history
< Universal inclusion criteria:– English article– Human focus– Manuscript in peer-reviewed journal
< Additional inclusion criteria (at least one):– Intervention study evaluating the impact of a CDS system inan actual patient care context
For a comparative intervention study, CDS required to bea part of the primary intervention under evaluationExcludes laboratory evaluations or simulation studies
– Methodology article whose primary focus is on how CDSsystems should be designed specifically to support clinicaldelivery of patient-specific assessments and/or recommen-dations guided by genetic factors. Includes systemdescription articles.
Figure 2 Manuscript selection process. CDS, clinical decision support;GPM, genetically guided personalized medicine.
Review
390 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892
Table1
Sum
maryof
primaryresearch
onCDSsystem
sforcancer-related
GPM
:RAGssystem
forprovidingFH
x-driven
CDS
Citationandnameof
system
(ifapplicable)
Manuscriptsummaryandtrial
details
(ifapplicable)
Users
and
studylocation
Genetic
inform
ationused
Integrated
with
primaryclinical
inform
ationsystem
CDSpurposeandclinical
focus
Manuscripttype
Notable
inform
aticsaspect
Coulson,2001
17;RAGs
System
descriptionof
RAGs,
which
was
designed
tohelpGPs
build
afamily
pedigree,calculategenetic
risk,
and
obtain
guideline-basedcare
recommendations
GPs
inUK
FHx
No
Assessm
entof
patient
riskand
provisionof
managem
ent
recommendations
forfamilial
breast,ovarian,
and
colorectalcancer
System
description
RAGsuses
PROform
a,an
argumentation-based
technology
forCDS
Emery,
1999
18;RAGs
Aqualitativeevaluationof
RAGs,
which
foundthat
thesystem
was
easy
touse
byGPs
andserved
asan
appropriate
applicationof
inform
ationtechnology
toassist
with
clinicalcare
GPs
inUK
FHx
No
Assessm
entof
patient
riskand
provisionof
managem
ent
recommendations
forfamilial
breast,ovarian,
and
colorectalcancer
Qualitativestudy
RAGsuses
PROform
a,an
argumentation-based
technology
forCDS
Glasspool,2001
54;RAGs
System
descriptionof
RAGs,
which
uses
anargumentationapproach
toassess
genetic
riskandprovidedetailed
qualitativeexplanationwith
referraladvice
GPs
inUK
FHx
No
Assessm
entof
patient
riskand
provisionof
managem
ent
recommendations
forfamilial
breast,ovarian,
and
colorectalcancer
System
description
RAGsuses
PROform
a,an
argumentation-based
technology
forCDS
Emery,
2000
19;RAGs
Com
parativeanalysisof
RAGs,
Cyrillic
(acommercially
availablepedigree
draw
ingsystem
),andpenandpaper
foruseby
GPs.Thisstudydemonstrated
that
RAGssignificantlyimproved
pedigree
accuracy
andproduced
moreappropriate
managem
entdecisionsthan
theothertwo
methods.Furtherm
ore,
92%of
GPs
preferredRAGsto
theothermethods
GPs
inUK
FHx
No
Assessm
entof
patient
riskand
provisionof
managem
ent
recommendations
forfamilial
breast,ovarian,
and
colorectalcancer
Com
parativestudy
RAGsuses
PROform
a,an
argumentation-based
technology
forCDS
Emery,
2005
20;GRAIDS
System
descriptionof
GRAIDS,a
next-generationFH
xCDStool
that
built
onboth
RAGsandCyrillicandprovided
anenhanced
user
interfaceforGPs
toassess
familialcancer
risk
GPs
inUK
FHx
No
Assessm
entof
patient
riskand
provisionof
managem
ent
recommendations
forfamilial
breast,ovarian,
and
colorectalcancer
System
description
Server-basedapplication
that
provides
both
heuristic
andstatistical
riskassessment
Emery,
2007
21;GRAIDS
AclusterRCTof
GRAIDSconducted
across
45GPteam
sintheUK.
GRAIDS
significantlyincreasedthenumberof
referralsto
theregionalgeneticsclinic
(p¼0
.001),with
thereferralsbeing
significantlymorelikelyto
beconsistent
with
referralguidelines
(p¼0
.006).
Moreover,patientsreferred
from
GRAIDSpractices
hadsignificantly
lower
cancer
worry
scores
atthe
pointof
referral(p¼0
.02)
GPs
inUK
FHx
No
Assessm
entof
patient
riskand
provisionof
managem
ent
recommendations
forfamilial
breast,ovarian,
and
colorectalcancer
RCT
Server-basedapplication
that
provides
both
heuristic
andstatistical
riskassessment
CDS,clinicaldecision
support;FH
x,family
health
history;
GP,
generalpractitioner;GPM
,geneticallyguided
personalized
medicine;
GRAIDS,GeneticRiskAssessm
entinan
Intranet
andDecisionSupport;RAGs,RiskAssessm
entinGenetics;
RCT,
random
ized
controlledtrial.
Review
J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892 391
who met certain criteria, one of which was a suspicious FHx.This intervention significantly improved pathologists' recogni-tion of patients at risk of Lynch syndrome.34 Moreover, DrHenry Lynch, for whom Lynch syndrome is named, developeda CDS system for supporting his hereditary cancer consultingservice. This CDS system expedited clinicians' decision-making
processes and resulted in a significant reduction in time spent oncases.32
Similar to the stand-alone CDS systems for breast cancermanagement described earlier,28 29 stand-alone CDS tools havebeen shown to be useful for the management of other types ofcancers, including prostate cancer37 and alcohol-related
Table 2 Summary of primary research on CDS systems for cancer-related GPM: other FHx CDS tools for breast cancer management
Citation andname ofsystem(if applicable)
Manuscript summaryand trial details(if applicable)
Usersand studylocation
Geneticinformationused
Integratedwithprimaryclinicalinformationsystem
CDS purposeand clinicalfocus
Manuscripttype
Notableinformaticsaspect
Tsouskas,199722
Evaluation of a CDS toolthat used patient-specificbreast cancer riskinformation, including FHx,to identify patients at highrisk of breast cancer. Thesystem identified nine outof 10 women with breastcancers in this study
Cliniciansin Europe
FHx No Assessment ofpatient risk forbreast cancer
Validationstudy
Expert systemdeveloped using avariant of the BASICprogramming language
Berry, 200223;BRCAPRO
Evaluation of BRCAPRO,which predicted theprobability of carryinga BRCA mutation based ona patient’s FHx. BRCAPROwas effective in predictingthe probability of carryingthe BRCA mutation
Cliniciansin USA
FHx No Assessment ofpatient risk forbreast cancer
Validationstudy
Probability calculatedusing Bayesianupdating
Wilson, 200624 RCT of a stand-alone breastcancer CDS tool to guidereferrals in everyday GPpractices. The studyconsisted of 86 GPpractices. The CDS systemdid not result in a statisticallysignificant improvement, duelargely to the limited awarenessand adoption of the tool by GPs
GPs in UK FHx No Assessment ofpatient risk andprovision ofmanagementrecommendationsfor breast cancer
RCT The deployment ofthe CDS system waspurposely pragmaticand did not involveextensive workflowintegration measures
Matloff, 200726 System description of a toolto provide patient-specificpredictions of women’sfuture risks for breastcancer, heart disease,osteoporosis, andendometrial cancer utilizingpersonal and FHx
Geneticcounselorsin USA
FHx No Assessment ofpatient risk forbreast cancer
Systemdescription
System used aMarkov model
Matloff, 200625 RCT of a CDS tool used bygenetic counselors26 to enablepersonalized risk assessmentand genetic counseling. The trialinvolved 48 cancer-free, post-menopausal women with afirst-degree relative of breast cancerwho were contemplating the useof alternative menopausal therapyoptions. This trial found that patientsin the intervention group hadincreased knowledge and a lower,more accurate perceived risk ofdeveloping breast cancercompared to the control group
Geneticcounselorsin USA
FHx No Assessment ofpatient risk forbreast cancer
RCT System used aMarkov model
Ozanne, 200927;Hughes RiskApps
System description ofHughesRiskApps and evaluationof its impact at a communityhospital. The CDS systemsignificantly increased thenumber of patients seen forrisk consultation and genetictest ordering. The implementationimproved efficiency in severalways and did not requiresignificant investment incapital or personnel
Cliniciansin USA
FHx No Assessment ofpatient risk andprovision ofmanagementrecommendationsfor breast cancer
Systemdescription;preepostcomparison
Used tablet computersto collect informationfrom patients. UsedHealth Level 7compliantinformation models.
CDS, clinical decision support; FHx, family health history; GP, general practitioner; GPM, genetically guided personalized medicine; RCT, randomized controlled trial.
Review
392 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892
cancers.36 These studies included an RCT that showed thata patient-directed, genotype-driven CDS tool for alcohol-relatedcancer risk significantly reduced alcohol consumption bypatients at increased genetic risk.36 These studies, as well as theprevious studies on breast cancer,28 29 showed that patient-directed CDS systems can be clinically useful.
CDS for pharmacogenomicsPharmacogenomics, the practice of tailoring drug therapy to thepatient's unique genetic characteristics, can be a complicatedprocess; genetically guided CDS offers a solution for simplifyingthis process. Table 5 summarizes the six primary research articlesidentified on this topic.38e43 These studies include a descriptionand validation of a CDS system for genetically guided treatment ofHIV infections,38 as well as an RCT that found that genotypingcombined with CDS-guided therapy improved outcomes overstandard of care.39 Outside of HIV therapy, other investigatorsfocused on how CDS for pharmacogenomics could be integratedwith primary clinical information systems such as computerizedprovider order entry systems.40 42 43 These studies evaluated
considerations such as developing the underlying pharmacoge-nomics knowledge base,40 representation of genetic information inthe EHR for supporting pharmacogenomics CDS,42 and theavailability of patient data required for pharmacogenomics withinthe EHR.43 The lone stand-alone system for pharmacogenomicsused genotype and clinical data to estimate and graphicallyrepresent a patient's plasma warfarin concentration over time.41
Other CDS systems for GPMTable 6 summarizes the 10 primary research articles that wereneither cancer specific nor focused on pharmacogenomics.44e53
As with CDS for cancer, there has been a substantial focus onFHx-driven CDS for other medical conditions. For example,a tool called GenInfer considered FHx and calculated inheritancerisks for genetic diseases,44 and FHx-driven CDS was included asa part of the National Russian Genetic Register.45 46 Beyondthese system descriptions, recent studies of FHx-driven CDShave focused on impact evaluation, with mixed results.48 49 51
Finally, there were four primary research studies on genotype-driven CDS systems not focused on pharmacogenomics or
Table 3 Summary of primary research on CDS systems for cancer-related GPM: genotype-driven CDS tools for breast cancer management
Citation andname of system(if applicable)
Manuscript summaryand trial details(if applicable)
Usersand studylocation
Geneticinformationused
Integratedwithprimaryclinicalinformationsystem
CDS purposeand clinicalfocus
Manuscripttype
Notableinformaticsaspect
Schwartz, 200928 RCT of patient-facing toolthat captured patient-specificinformation and providedtailored content about risks,benefits and managementoptions based on the patients’particular situations. Thisstudy found that among214 BRCA-positive womenwho were initially undecidedabout how to manage theirbreast cancer risk, patientswho used the CDS tool weremore likely to reach a managementdecision (p¼0.001), had decreaseddecision conflict (p¼0.002), andincreased satisfaction (p¼0.002)compared to women who did notuse the CDS tool
Patientsin USA
Genotype No Provision ofmanagementrecommendationsfor breast cancer
RCT CD-ROMbased,patient-directeddecision aid
Hooker, 201129 Longitudinal RCT of patient-facingBRCA decision aid.28 This studyshowed significantly higher cancer-specific distress (p¼0.01) andgenetic testing-specific distress(p¼0.01) among users of thepersonalized decision aid after onemonth. Distress levels between groupswere the same after 12 months
Patientsin USA
Genotype No Provision ofmanagementrecommendationsfor breast cancer
RCT CD-ROMbased,patient-directeddecision aid
Glasspool,200730;REACT
System description of REACT(Risks, Events, Actions and theirConsequences over Time), a breastcancer CDS tool with a graphicaltimeline display to model real-timechanges in lifetime risk as a result ofrisk-reduction interventions such astamoxifen therapy, hormone therapy,and mastectomy
Geneticcounselorsin UK
Genotype No Prediction ofresponse totreatment forbreast andovarian cancer
Systemdescription
Graphicaldisplay of riskchangesdynamicallybased onselectedinterventions
Glasspool,201031;REACT
Qualitative study of REACT by eightgenetic counselors.30 Most counselorsfound REACT effective for genetic riskmanagement, although there wereconcerns related to the tool’s potentialto alter the dynamics of theclinicianepatient interaction
Geneticcounselorsin UK
Genotype No Prediction ofresponse totreatment forbreast andovarian cancer
Qualitativestudy
Graphicaldisplay of riskchangesdynamicallybased onselectedinterventions
CDS, clinical decision support; GPM, genetically guided personalized medicine; RCT, randomized controlled trial; REACT, Risks, Events, Actions and their Consequences over Time.
Review
J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892 393
Table 4 Summary of primary research on CDS systems for cancer-related GPM: CDS for other cancers
Citation andname of system(if applicable)
Manuscript summaryand trial details(if applicable)
Users andstudylocation
Geneticinformationused
Integratedwithprimaryclinicalinformationsystem
CDS purposeand clinicalfocus
Manuscripttype
Notableinformaticsaspect
Evans, 199532 Description of a FHx CDSsystem developed fora hereditary cancer consultingservice. The system collectedFHx information, evaluatedthe FHx for familial riskpatterns, and producedpreliminary risk assessmentand managementrecommendations. Thesystem resulted ina significant reductionin time spent on cases
Hereditarycancerconsultingservice in USA
FHx No Assessment ofpatient risk andprovision ofmanagementrecommendationsfor hereditarycancer
Systemdescription;impactobservation
Expert rule-basedsystem thatmodeled thepattern recognitioncapabilities ofclinical geneticists
Bianchi, 200733;CRCAPRO
Evaluation of CRCAPRO,which used FHx ofcolorectal and endometrialcancers to identify patientswith Lynch syndrome.This study showed thatCRCAPRO has lowsensitivity and specificity
Clinicians in UK FHx No Assessment ofpatient risk forcolorectal cancer
Systemvalidation
Probabilitycalculated usingBayesianupdating
Overbeek,201034
RCT of electronic remindersto pathologists to considerLynch syndrome genetictesting among newlydiagnosed colon cancerpatients based on FHx. TheCDS reminder intervention in12 pathology laboratoriessignificantly improvedpathologists’ recognition ofpatients at risk for Lynchsyndrome (OR 2.8; 95% CI1.1 to 7.0) and increaseduse of genetic testing (OR4.1; 95% CI 1.3 to 13.2)
Pathologists inEurope
FHx Yes Provision ofmanagementrecommendationsfor colorectalcancer
RCT Electronicremindersprovidedthrough healthinformationsystem
Picone, 201135;NeoMark
System description ofNeoMark, a web-basedtool that combined medicalimages, genetic markers,and other patient databefore and after treatmentof oral cavity squamous cellcarcinoma to predictreoccurrence
Clinicians inEurope
Genotype No Assessment ofpatient risk fororal cancer
Systemdescription
Uses a service-oriented, modulararchitecture
Hendershot,201036
RCT of a web-basedgenetic feedbackintervention involving 200college students of Asiandescent. The systemprovided personalizedalcohol-related health riskinformation and feedbackbased on the patient’sgenotype. The tool resultedin significant reductions indrinking (p¼0.02) amongparticipants with thegenotype associated withhigher risk of alcohol-related cancer
Patients inUSA
Genotype No Assessment ofpatient risk;reduction ofrisky behavior(alcoholconsumption)for alcohol-relatedcancer
RCT Web-basedintervention
Wakefield, 201137 System description and pilotusability test of an onlineCDS tool that presented 22men with age and familyhistory-specific prostatecancer risk informationand management recommendations.Most participants preferred thismethod for receiving prostatecancer information
Patients inAustralia/NewZealand
FHx No Assessment ofpatient risk andprovision ofmanagementrecommendationsfor prostatecancer
Systemdescription;pilot usabilitytest
Online decisionaid using aMarkov model
CDS, clinical decision support; FHx, family health history; GPM, genetically guided personalized medicine; RCT, randomized controlled trial.
Review
394 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892
cancer.47 50 52 53 These systems included a CDS system thatretrieved genetic, radiological and clinical data from clinicalinformation systems to provide guidance on intracranial aneurismmanagement,52 as well as a portable medical device that inte-grated clinical and genetic data to provide a diagnosis for rheu-matoid arthritis and multiple sclerosis.47 In addition, GeneInsightprovides geneticists and other clinicians with patient-specificgenetic testing reports, as well as notifications regarding updatesto the presumed clinical significance of patients' previously iden-tified genotype.50 Finally, in a survey study, Scheuner andcolleagues53 found that clinicians felt their EHR systems could domuch more to meet their needs related to GPM.
Trend analysisPublication volume on CDS for GPM generally increased overtime, with a majority published since 2007 (figure 3). While all
publications before 2007 focused on stand-alone CDS, 32% ofarticles since 2007 focused on integrated CDS (figure 4). Like-wise, while 13% of manuscripts before 2007 involved the use ofgenotype for CDS, 61% of manuscripts since 2007 have involvedthe use of genotype (figure 5). As noted earlier, a major focus ofthe literature in this domain has been on FHx CDS, pharma-cogenomics, and CDS for cancer management.
DISCUSSIONSummary of findingsIn order to learn from past research efforts and to guide futureresearch into the use of CDS to enable GPM, we conducteda systematic review of the literature. Through a literature searchspanning from 1990 to 2011, we screened 3416 manuscripts andincluded 38 primary research articles. A majority of thesemanuscripts was published from 2007 to 2011, with an
Table 5 Summary of primary research on CDS systems for pharmacogenomics
Citation and nameof system(if applicable)
Manuscript summaryand trial details(if applicable)
Usersand studylocation
Geneticinformationused
Integratedwithprimaryclinicalinformationsystem
CDS purposeand clinicalfocus
Manuscripttype
Notableinformaticsaspect
Pazzani, 199738;CTSHIV
System description of the CTSHIVCDS program which manages HIVgenome data and makes virus-specific therapeuticrecommendations
Cliniciansin USA
HIVgenotype
No Provision ofmanagementrecommendationsfor HIV
Systemdescription
Uses a backwardchaining expertsystem
Tural, 200239;RetroGram
RCT of genotyping accompaniedby RetroGram, which ranked drugsuitability based on the HIVgenotype. This study showedthat genotyping combined withRetroGram use improved HIVtherapy outcomes over standardof care (p<0.05)
Cliniciansin Europe
HIVgenotype
No Provision ofmanagementrecommendationsfor HIV
RCT Contains approximately200 rules based on thescientific literature
Swen, 200840 Description of how the Royal DutchAssociation for the Advancement ofPharmacy developed guidelines forthe use of genetic information fordrug prescribing and integratedthese guidelines into automateddrug prescription and medicalsurveillance systems fornationwide use
Clinicians andpharmacistsin Europe
Genotype Yes Alert on gene-druginteractions forpharmacogenomics
Systemdescription
Recommendationsincorporated into theG-standard, an electronicdrug database usedfor CDS
Bon Homme,200841
System description of prototypeCDS tool for personalized warfarintherapy that combined genetic andclinical data to estimate the requiredwarfarin dose and the patient’splasma warfarin concentration
Cliniciansin USA
Genotype No Therapeutic doseguidance forwarfarin
Systemdescription
Provides a graphicaldisplay of estimatedplasma warfarinconcentration over time
Deshmukh,200942
This study compared the use of asingle nucleotide polymorphismdata model to the use of an alleledata model for CDS computationin an EHR system. While therewere statistically significantdifferences in computation time,this did not translate into significantdifferences in the overall clinicianordering time
Clinicians andpharmacistsin USA
Genotype Yes Alert on geneedruginteractions forpharmacogenomics
Comparativestudy ongenotype datarepresentation
CDS rules developedwithin the CernerEHR environment
Overby, 201043 This study found that thePharmacogenomics KnowledgeBase was a good source forpharmacogenomics knowledgeand that sufficient clinical dataexisted in the local EHR system tosupport 50% of the pharmacogenomicknowledge in drug labels that arecapable of being expressed as CDSrules
Cliniciansin USA
Genotype Yes Provision of therapyguidance forpharmacogenomics
Feasibilitystudy
The MINDscape EHRsystem was used inthe study
CDS, clinical decision support; CTSHIV, Customized Treatment Strategies for HIV; EHR, electronic health record; RCT, randomized controlled trial.
Review
J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892 395
Table6
Sum
maryof
primaryresearch
onGPM
CDSsystem
sforotherconditions
Citationandnameof
system
(ifapplicable)
Manuscriptsummaryandtrialdetails
(ifapplicable)
Users
and
studylocation
Genetic
inform
ationused
Integrated
with
primaryclinical
inform
ationsystem
CDSpurposeand
clinical
focus
Manuscripttype
Notable
inform
aticsaspect
FHx-driven
CDSsystem
s
Harris,1990
44;GenInfer
System
descriptionof
theGenInferprogram,
which
used
FHxinform
ationalongwith
other
inheritance
factorsto
calculategenetic
risks
and
probabilitiesof
inheritance
Clinicians
inUSA
FHx
No
Assessm
entof
patient
riskforinherited
disease
System
description
Based
onPearl’s
algorithm
forfusion
andpropagationin
aprobabilistic
beliefnetwork
Kobrinskii,
1997
45and
Kobrinsky,
1998
46;National
Russian
GeneticRegister
Descriptionof
theinform
ationsystem
used
byRussia’sfederalgeneticscenter
tomanage
patientsacross
Russiain
need
ofgeneticscare.
Thissystem
supportedpedigree
creation,
cytogenetic
analysis,riskassessment,and
inform
ationsupport
Geneticsspecialists
inEurope
FHx
No
Assessm
entof
patient
riskforinherited
disease
System
description
Utilized
both
server–client
andlocaldeploymentmodels
Orlando,
2011
48;MeTree
System
descriptionof
MeTree,
atool
that
evaluatesFH
xandprovides
managem
ent
recommendations
regardingvarious
heritable
conditionsforpatientsandclinicians.Also
provides
theprotocol
foraplannedevaluationof
thetool
inNorth
Carolinaprimarycare
clinics
Patientsand
clinicians
inUSA
FHx
No
Assessm
entof
patient
riskandprovisionof
managem
ent
recommendations
for
inherited
disease
System
description;
evaluationprotocol
description
Patient-driven
application
that
provides
CDSas
aprintout
Rubinstein,
2011
49;
CDCFamily
Healthware
RCTwith
3284
participants
oftheCDCFamily
Healthwaretool,which
provides
personalized
screeningrecommendations
formultipleheritable
conditionsbasedon
FHx.
Bothinterventionand
controlgroups
show
edimproved
adherenceto
screeningrecommendations
comparedto
the
baselinetim
eperiod,
buttherewas
nosignificant
difference
betweentheinterventionandcontrol
groups
Patientsin
USA
FHx
No
Assessm
entof
patient
riskandprovisionof
managem
ent
recommendations
for
inherited
disease
RCT
Apatient-directed,
web-based
tool
Wells,2007
51;
PRED
ICTCVD-5
Systemdescriptionofareal-timeCDSsystem
that
pulledclinicaldata
from
theEH
Rto
calculate
cardiovascular
diseaseriskandproviderisk
managem
entrecommendations.Aretrospective
analysisfoundthatincludingthepatients’ethnicity
andFH
xinto
theriskassessmentprocess
substantially
increasedthenumberof
patients
eligiblefordrug
treatm
entandlifestyle
managem
ent
Clinicianin
Australia
andNew
Zealand
FHx
Yes
Assessm
entof
patient
riskforheartdisease
System
description;
retrospectiveanalysis
Integrated
with
theMedTech
practicemanagem
entsystem
Genotype-driven
CDSsystem
s
Iavindrasana,2008
52;
@neurIST
System
descriptionof
@neurIST,
aCDSsystem
which
collectsgenetic
data,radiologicaldata,and
clinicaldata
from
clinicalinform
ationsystem
sto
provideCDSregardingintracranialaneurisms
Clinicians
inEurope
Genotype
Yes
Provisionof
managem
ent
recommendations
for
intracranialaneurism
System
description
Usesaservice-
oriented,
standards-basedapproach
Kalatzis,2009
47
System
descriptionof
apoint-of-careportable
medicaldevice
that
integrates
clinicaldata
with
genetic
data
obtained
from
aminiature
diagnostic
system
toproduceadiagnosisforrheumatoid
arthritisandmultiplesclerosis
Clinicians
inEurope
Genotype
No
Diagnostic
assistance
forarthritisandmultiple
sclerosis
System
description
Acombinationof
artificial
neuralnetworks,decision
trees,andsupportvector
machineswas
foundto
have
thebest
performance
Continued
Review
396 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892
increasing shift in focus from FHx CDS to genotype-drivenCDS, and from stand-alone CDS to integrated CDS. There havebeen nine RCTs of CDS interventions for GPM, but most CDSinterventions for GPM have not yet been rigorously assessed fortheir clinical impact.
Strengths and limitationsAs one important strength of this study, as far as we are aware,this work represents the first systematic review on CDS forGPM. As such, it contributes an important perspective ona topic that has the potential to have significant impacts in bothclinical medicine and biomedical informatics. As a secondstrength, this systematic review was based on search strategiesrefined through previous systematic reviews on relatedtopics.14e16 Third, we searched Embase in addition toMEDLINE, so as to provide greater coverage of the internationalliterature. Finally, in addition to providing a summary of rele-vant manuscripts, this review provides insights and trend anal-yses that show how this scientific field has developed over timeand where the field appears to be headed moving forward.In terms of limitations, this study does not provide a quanti-
tative meta-analysis of the impact of CDS interventions forGPM. However, such a meta-analysis was not possible due tothe limited number of outcome studies in this field and theheterogeneous nature of the various interventions and clinicaldomains. Second, we only included manuscripts written inEnglish, which may have led to some relevant manuscripts beingexcluded that were written in a different language. Third, somerelevant 2011 articles may not have been indexed by the time ofour literature search and therefore erroneously excluded.However, a literature search update in June 2012 added less than1% to the number of articles we had previously retrievedthrough March 2012, which suggests that this risk is low.Finally, there is a potential for publication bias with regard tothe clinical trials included, in which studies with successfuloutcomes were more likely to be published than studies withunsuccessful outcomes. There was a potential indication of sucha bias, in that seven of nine RCTs evaluated (77%) reportedpositive results, whereas the expected rate of positive resultswould more typically be in the range of approximately 60%.56
However, given the limited sample size, the observed discrep-ancy may simply be due to chance. Moreover, as discussed next,the high rate of successful interventions may be partly explainedby the fact that use of many of these systems was required bythe study protocol, which improved the systems' likelihood ofuse and impact.
Consistency of trial findings with expected outcomesIn a previous systematic review of CDS RCTs, we identified theautomatic provision of CDS as a part of routine clinical work-flow to be a critical predictor of the success or failure of CDSinterventions (adjusted OR of 112.1, p<0.00001).14 Whileautomatic provision of CDS was not a guarantee of success inthis systematic review, a lack of this feature was associated withnegative outcomes in all cases, generally due to the lack of use ofthe system.14 Moreover, a later RCT specifically evaluating theimportance of automatic provision of CDS directly confirmedthis finding.57
On initial examination, the results of the present systematicreview seemed to contradict this finding, as we found severalRCTs in which stand-alone CDS interventions for GPM were notprovided automatically as a part of routine clinical workflow butresulted in positive improvements in clinical practice.21 25 28 29 36 39
However, in all but one of these RCTs,21 use of the CDS systemTable6
Continued
Citationandnameof
system
(ifapplicable)
Manuscriptsummaryandtrialdetails
(ifapplicable)
Users
and
studylocation
Genetic
inform
ationused
Integrated
with
primaryclinical
inform
ationsystem
CDSpurposeand
clinical
focus
Manuscripttype
Notable
inform
aticsaspect
Scheuner,2009
53
Asurvey
ofhealth
professionals,geneticsexperts,
andEH
Rdevelopers
regardingtheability
ofEH
Rsystem
sto
document,organize,anduseFH
xand
genetic
inform
ation
Clinicians
inUSA
FHx;
genotype
Yes
Assessm
entof
patient
riskandprovisionof
managem
ent
recommendations
for
genetically-guided
personalized
medicine
Survey
EHRsweregenerally
perceived
aslackingtheability
tosupportgenomicmedicine
Aronson,2011
50;
GeneInsight
System
descriptionof
GeneInsight,aplatform
that
provides
patient-specific
genetic
testingreportsas
wellas
notifications
whenthepresum
edclinical
significanceofgenetic
variantschange
forpatients
who
have
been
previouslytested
Geneticists
andother
clinicians
inUSA
Genotype
No
Provisionof
patient-
specificgenetic
testing
reports;
notificationof
changesin
clinical
significanceof
genetic
variants
System
description
Isregistered
with
theFood
andDrugAdm
inistrationas
aclassIexem
ptmedicaldevice
CDC,Centers
forDisease
Control
andPrevention;
CDS,clinicaldecision
support;EH
R,electronic
health
record;FH
x,family
health
history;
GPM
,geneticallyguided
personalized
medicine;
RCT,
random
ized
controlledtrial.
Review
J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892 397
was mandated by the study protocol, which was an exclusioncriterion in the previous systematic review that identifiedthe critical importance of the automatic provision of CDS.14
Therefore, we believe it is premature to draw the conclusionthat automatic provision is not important when providing CDSfor GPM, as it is possible that the same CDS interventions thatled to positive results in the studies included would not haveled to positive results if use of the system was not mandatedby the study protocol, due to lack of awareness and use of thetool. With regard to other, less critical success factors identifiedin the previous systematic review of CDS interventions,14 wedid not find any trends that contradicted those findings.However, the sample size of available CDS trials was too smallin this study to allow for any meaningful analysis of these otherfactors.
Of note, in the RCT of the GRAIDS system for FHx-basedCDS, the system did have a positive impact, even though its usewas not mandated by the study protocol and the system was
not automatically provided as a part of routine clinical work-flow.14 However, the use and impact of this system may havebeen the result of exceptional circumstances specific to the studycontext and unlikely to be available in a routine clinical practicesetting. In particular, in the RCT of the GRAIDS system,designated clinicians were recruited at each practice, receivedextensive training on GRAIDS, and managed all patients in thepractice expressing concern regarding their breast or colorectalcancer FHx.21 This type of resource-intensive deploymentstrategy may not be feasible outside the context of a researchstudy, as demonstrated in another RCT of a stand-alone breastcancer CDS tool, which had limited impact due largely to thelack of awareness and adoption of the tool by clinicians.24
Therefore, while more evidence is needed before a solid conclu-sion can be drawn, we found no conclusive evidence that CDSfor GPM is unique in terms of the intervention features requiredfor successful outcomes.
Assessment of current research state and required researchIn recent years, CDS has been proposed as a promising approachto realizing the promise of GPM.10e12 58e65 However, we iden-tified only 38 primary research articles published from 1990 to2011 on the design, implementation, use, and evaluation of CDSsystems to support genetically guided patient care, whichamounts to approximately 1.7 articles per year. Even in the yearwith the most publications on this topic (2011), we identifiedonly six primary research articles. In particular, we identifiedonly nine RCTs of the impact of CDS systems for GPM, sevenarticles focused on CDS integrated with primary clinical infor-mation systems, and 16 articles involving the use of genotype todrive CDS. Furthermore, few groups have demonstrated howgenotype-driven CDS can be integrated into clinical settings andclinical information systems in a scalable, standards-based, andeffective manner.40 43 52 53
Given the tremendous volume of research being conducted inthe discovery of novel personalized medicine diagnostics andtherapeutics, we feel that much more research is required onhow CDS can and should be leveraged to take these discoveriesand to implement them in routine clinical practice. For example,even for FHx-driven CDS, which is perhaps the most well-
Figure 3 Publications included per year.
Figure 4 Publications focused on stand-alone versus integrated clinicaldecision support (CDS).
Figure 5 Publications focused on family health history (FHx)-drivenversus genotype-driven clinical decision support.
Review
398 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892
established area of research with regard to CDS for GPM, therehas been limited research on the optimal use of FHx-driven CDStools beyond hereditary cancer management. Indeed, given thelimited literature available on any one topic, we feel it would bepremature to consider any aspect of CDS for GPM to be fullymature and not in need of any further research.
In looking forward, we believe that the largest loomingresearch challenge in terms of CDS for GPM will be the devel-opment of effective approaches to manage and utilize wholegenome sequence data in the clinical setting. The pursuit of low-cost whole genome sequencing has been a priority research areafor many years, such that sequencing costs may be reduced toa level amenable to routine clinical use in the near future.66
While sequencing technologies continue to advance, the infor-matics capabilities to apply whole genome sequencing data toclinical practice is still in its infancy.67 Indeed, in our systematicreview, we did not find a single primary research articleaddressing this topic. Therefore, we recommend the prioritiza-tion and resourcing of this area of research by the scientificcommunity. In particular, to realize the full clinical potential ofwhole genome sequence data, we believe that approaches willneed to be developed for providing advanced CDS capabilitiesthat are integrated with clinical information systems andprovided automatically as a part of routine clinical workflow.
CONCLUSIONThe promise of GPM is growing with the recent advances anddiscoveries in genomics research. With this growth also comesthe growing need for translating such discoveries into everydayclinical care, so that we are able to realize the promises of GPM.CDS has the potential to bridge this gap between the promiseand realization of GPM. By systematically reviewing the litera-ture in this field and by identifying gaps in required research, wespeculate that this paper will assist with efforts to leverage CDSto enable GPM at scale.
Contributors BMW and KK both contributed to the design and conduct of the study,as well as the preparation of the manuscript.
Funding This study was funded by grant K01HG004645 from the US National HumanGenome Research Institute, the University of Utah Department of BiomedicalInformatics, and the University of Utah Program in Personalized Health Care. Thefunding sources played no role in the study design, in the collection, analysis andinterpretation of data, in the writing of the manuscript, or in the decision to submit themanuscript for publication. The views expressed in this manuscript are those of theauthors alone and do not necessarily reflect the official views of the National HumanGenome Research Institute or of the University of Utah.
Competing interests BMW is the founder and owner of SGgenomics, Inc., whichdeveloped ItRunsInMyFamily.com, a patient-centered FHx tool. KK is serving asa consultant to Inflexxion on a project funded by the National Institute on Drug Abuseto develop CDS capabilities for mental healthcare. KK receives royalties for a DukeUniversity-owned CDS technology for infectious disease management known asCustomID that he helped develop. KK was formerly a consultant for Religent, Inc. anda co-owner and consultant for Clinica Software, Inc., both of which providecommercial CDS services, including through use of a CDS technology known asSEBASTIAN that KK developed. KK no longer has a financial relationship with eitherReligent or Clinica Software.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Individuals interested in the raw data and analyses maycontact the corresponding author to obtain such data.
REFERENCES1. U.S. Department of Health and Human Services. Glossary Of Terms For Personalized
Health Care Website. http://www.hhs.gov/myhealthcare/glossary/glossary.html(accessed Aug 2012).
2. Abrahams E, Ginsburg GS, Silver M. The personalized medicine coalition: goals andstrategies. Am J Pharmacogenomics 2005;5:345e55.
3. Khoury MJ, McBride CM, Schully SD, et al. The scientific foundation for personalgenomics: recommendations from a National Institutes of HealtheCenters forDisease Control and Prevention multidisciplinary workshop. Genet Med2009;11:559e67.
4. Offit K. Personalized medicine: new genomics, old lessons. Hum Genet2011;130:3e14.
5. Fernald GH, Capriotti E, Daneshjou R, et al. Bioinformatics challenges forpersonalized medicine. Bioinformatics 2011;27:1741e8.
6. Balas EA, Boren SA. Managing clinical knowledge for health care improvement.Yearbook of Medical Informatics 2000: Patient-Centered System. Stuttgart, Germany:Schattauer, 2000:65e70.
7. Rosas-Blum E, Shirsat P, Leiner M. Communicating genetic information: a difficultchallenge for future pediatricians. BMC Med Educ 2007;7:17.
8. American Board of Genetic Counseling, Inc. About ABGC.http://www.abgc.net/About_ABGC/GeneticCounselors.asp (accessed Aug 2012).
9. GeneTests: Growth of Laboratory Directory. Seattle: University of Washington. http://www.ncbi.nlm.nih.gov/projects/GeneTests/static/whatsnew/labdirgrowth.shtml(accessed Aug 2012).
10. Glaser J, Henley DE, Downing G, et al. Personalized Health Care Workgroup of theAmerican Health Information Community. Advancing personalized health care throughhealth information technology: an update from the American Health InformationCommunity’s Personalized Health Care Workgroup. J Am Med Inform Assoc2008;15:391e6. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC130085/
11. Kawamoto K, Lobach DF, Willard HF, et al. A national clinical decision supportinfrastructure to enable the widespread and consistent practice of genomic andpersonalized medicine. BMC Med Inf Decis Mak 2009;9:17.
12. Downing GJ, Boyle SN, Brinner KM, et al. Information management to enablepersonalized medicine: stakeholder roles in building clinical decision support. BMCMed Inf Decis Mak 2009;9:44.
13. Osheroff JA, Teich JM, Middleton B, et al. A roadmap for national action on clinicaldecision support. J Am Med Inform Assoc 2007;14:141e5.
14. Kawamoto K, Houlihan CA, Balas EA, et al. Improving clinical practice using clinicaldecision support systems: a systematic review of trials to identify features critical tosuccess. BMJ 2005;330:765.
15. Scheuner MT, Sieverding P, Shekelle PG. Delivery of genomic medicine for commonchronic adult diseases: a systematic review. JAMA 2008;299:1320e34.
16. Wilson BJ, Qureshi N, Santaguida P, et al. Systematic review: family history in riskassessment for common diseases. Ann Intern Med 2009;151:878e85.
17. Coulson AS, Glasspool DW, Fox J, et al. RAGs: a novel approach to computerizedgenetic risk assessment and decision support from pedigrees. Methods Inf Med2001;40:315e22.
18. Emery J, Walton R, Coulson A, et al. Computer support for recording andinterpreting family histories of breast and ovarian cancer in primary care (RAGs):qualitative evaluation with simulated patients. BMJ 1999;319:32e6.
19. Emery J, Walton R, Murphy M, et al. Computer support for interpreting familyhistories of breast and ovarian cancer in primary care: comparative study withsimulated cases. BMJ 2000;321:28e32.
20. Emery J. The GRAIDS Trial: the development and evaluation of computer decisionsupport for cancer genetic risk assessment in primary care. Ann Hum Biol2005;32:218e27.
21. Emery J, Morris H, Goodchild R, et al. The GRAIDS Trial: a cluster randomisedcontrolled trial of computer decision support for the management of familial cancerrisk in primary care. Br J Cancer 2007;97:486e93.
22. Tsouskas LI, Liaros A, Tzitzikas J, et al. The use of an expert system ofcomposite risk factors in breast cancer screening. Stud Health Technol Inform1997;43:859e63.
23. Berry DA, Iversen ES Jr, Gudbjartsson DF, et al. BRCAPRO validation, sensitivity ofgenetic testing of BRCA1/BRCA2, and prevalence of other breast cancersusceptibility genes. J Clin Oncol 2002;20:2701e12.
24. Wilson BJ, Torrance N, Mollison J, et al. Cluster randomized trial of a multifacetedprimary care decision-support intervention for inherited breast cancer risk. Fam Pract2006;23:537e44.
25. Matloff ET, Moyer A, Shannon KM, et al. Healthy women with a family history ofbreast cancer: impact of a tailored genetic counseling intervention on risk perception,knowledge, and menopausal therapy decision making. J Womens Health2006;15:843e56.
26. Matloff ET, Shannon KM, Moyer A, et al. Should menopausal women at increasedrisk for breast cancer use tamoxifen, raloxifene, or hormone therapy? A frameworkfor personalized risk assessment and counseling. J Cancer Educ 2007;22:10e14.
27. Ozanne EM, Loberg A, Hughes S, et al. Identification and management of women athigh risk for hereditary breast/ovarian cancer syndrome. Breast J 2009;15:155e62.
28. Schwartz MD, Valdimarsdottir HB, DeMarco TA, et al. Randomized trial ofa decision aid for BRCA1/BRCA2 mutation carriers: impact on measures of decisionmaking and satisfaction. Health Psychol 2009;28:11e19.
29. Hooker GW, Leventhal KG, DeMarco T, et al. Longitudinal changes in patientdistress following interactive decision aid use among BRCA1/2 carriers: a randomizedtrial. Med Decis Making 2011;31:412e21.
30. Glasspool DW, Oettinger A, Smith-Spark JH, et al. Supporting medical planning bymitigating cognitive load. Methods Inf Med 2007;46:636e40.
31. Glasspool DW, Oettinger A, Braithwaite D, et al. Interactive decision support for riskmanagement: a qualitative evaluation in cancer genetic counselling sessions.J Cancer Educ 2010;25:312e16.
Review
J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892 399
32. Evans S, Lynch HT, Fusaro RM. Clinical results using informatics toevaluate hereditary cancer risk. Proc Annu Symp Comput Appl Med Care1995:834e8.
33. Bianchi F, Galizia E, Bracci R, et al. Effectiveness of the CRCAPRO program inidentifying patients suspected for HNPCC. Clin Genet 2007;71:158e64.
34. Overbeek LI, Hermens RP, van Krieken JH, et al. Electronic reminders forpathologists promote recognition of patients at risk for Lynch syndrome: cluster-randomised controlled trial. Virchows Arch 2010;456:653e9.
35. Picone M, Steger S, Exarchos K, et al. Enabling heterogeneous data integration andbiomedical event prediction through ICT: the test case of cancer reoccurrence. AdvExp Med Biol 2011;696:367e75.
36. Hendershot CS, Otto JM, Collins SE, et al. Evaluation of a brief web-based geneticfeedback intervention for reducing alcohol-related health risks associated withALDH2. Ann Behav Med 2010;40:77e88.
37. Wakefield CE, Watts KJ, Meiser B, et al. Development and pilot testing of an onlinescreening decision aid for men with a family history of prostate cancer. Patient EducCouns 2011;83:64e72.
38. Pazzani MJ, See D, Schroeder E, et al. Application of an expert system in themanagement of HIV-infected patients. J Acquir Immune Defic Syndr Hum Retrovirol1997;15:356e62.
39. Tural C, Ruiz L, Holtzer C, et al. Clinical utility of HIV-1 genotyping and expert advice:the Havana trial. AIDS 2002;16:209e18.
40. Swen JJ, Wilting I, de Goede AL, et al. Pharmacogenetics: from bench to byte. ClinPharmacol Ther 2008;83:781e7.
41. Bon Homme M, Reynolds KK, Valdes R Jr, et al. Dynamic pharmacogenetic modelsin anticoagulation therapy. Clin Lab Med 2008;28:539e52.
42. Deshmukh VG, Hoffman MA, Arnoldi C, et al. Efficiency of CYP2C9 genetic testrepresentation for automated pharmacogenetic decision support. Methods Inf Med2009;48:282e90.
43. Overby CL, Tarczy-Hornoch P, Hoath JI, et al. Feasibility of incorporating genomicknowledge into electronic medical records for pharmacogenomic clinical decisionsupport. BMC Bioinformatics. 2010;11(Suppl. 9):S10.
44. Harris NL. Probabilistic belief networks for genetic counseling. Comput MethodsPrograms Biomed 1990;32:37e44.
45. Kobrinskii BA, Tester IB, Demikova NS, et al. A concept of the federal computersystem for families with genetic diseases and its practical implementation. BiomedEng 1997;31:172e4.
46. Kobrinsky B, Tester I, Demikova N, et al. A multifunctional system of the nationalgenetic register. Stud Health Technol Inform 1998;52:121e5.
47. Kalatzis FG, Giannakeas N, Exarchos TP, et al. Developing a genomic-based point-of-care diagnostic system for rheumatoid arthritis and multiple sclerosis. Conf ProcAnnu Int Conf IEEE Eng Med Biol Soc; 2e6 September 2009, Minneapolis,Minnesota, 2009:827e30.
48. Orlando LA, Hauser ER, Christianson C, et al. Protocol for implementation of familyhealth history collection and decision support into primary care using a computerizedfamily health history system. BMC Health Serv Res 2011;11:264.
49. Rubinstein WS, Acheson LS, O’Neill SM, et al. Clinical utility of family history forcancer screening and referral in primary care: a report from the Family HealthwareImpact Trial. Genet Med 2011;13:956e65.
50. Aronson SJ, Clark EH, Babb LJ, et al. The GeneInsight Suite: a platform to supportlaboratory and provider use of DNA-based genetic testing. Hum Mutat2011;32:532e6.
51. Wells S, Kerr A, Broad J, et al. The impact of New Zealand CVD risk chartadjustments for family history and ethnicity on eligibility for treatment (PREDICT CVD-5). NZ Med J 2007;120:U2712.
52. Iavindrasana J, Lo Iacono L, Muller H, et al. The @neurIST project. Stud HealthTechnol Inform 2008;138:161e4.
53. Scheuner MT, De Vries H, Kim B, et al. Are electronic health records ready forgenomic medicine? Genet Med 2009;11:510e17.
54. Glasspool DW, Fox J, Coulson AS, et al. Risk assessment in genetics: a semi-quantitative approach. Stud Health Technol Inform 2001;84:459e63.
55. Emery J. Computer support for genetic advice in primary care. Br J Gen Pract1999;49:572e5.
56. Jaspers MW, Smeulers M, Vermeulen H, et al. Effects of clinical decision-supportsystems on practitioner performance and patient outcomes: a synthesis ofhigh-quality systematic review findings. J Am Med Inform Assoc 2011;18:327e34.Epub 2011 Mar 21.
57. van Wyk JT, van Wijk MA, Sturkenboom MC, et al. Electronic alerts versus on-demand decision support to improve dyslipidemia treatment: a cluster randomizedcontrolled trial. Circulation 2008;117:371e8.
58. Ullman-Cullere MH, Mathew JP. Emerging landscape of genomics in theelectronic health record for personalized medicine. Hum Mutat 2011;32:512e16.
59. Snyderman R, Yoediono Z. Prospective care: a personalized, preventative approachto medicine. Pharmacogenomics 2006;7:5e9.
60. Samwald M, Stenzhorn H, Dumontier M, et al. Towards an interoperableinformation infrastructure providing decision support for genomic medicine. StudHealth Technol Inform 2011;169:165e9.
61. Hoffman MA. The genome-enabled electronic medical record. J Biomed Inform2007;40:44e6.
62. Hoffman MA, Williams MS. Electronic medical records and personalized medicine.Hum Genet 2011;130:33e9.
63. Ginsburg GS, Willard HF. Genomic and personalized medicine: foundations andapplications. Transl Res 2009;154:277e87.
64. Downing GJ. Policy perspectives on the emerging pathways of personalizedmedicine. Dialogues Clin Neurosci 2009;11:377e87.
65. Chan IS, Ginsburg GS. Personalized medicine: progress and promise. Annu RevGenomics Hum Genet 2011;12:217e44.
66. Cordero P, Ashley EA. Whole-genome sequencing in personalized therapeutics. ClinPharmacol Ther 2012;2:51.
67. Ali-Khan SE, Daar AS, Shuman C, et al. Whole genome scanning: resolvingclinical diagnosis and management amidst complex data. Pediatr Res2009;66:357e63.
PAGE fraction trail=12.75
Review
400 J Am Med Inform Assoc 2013;20:388–400. doi:10.1136/amiajnl-2012-000892