Clinical Decision Support Systems (CDSSs): State Of The Art Review Of .

Shahsavarani A.M, et al. Clinical Decision Support Systems (CDSSs)The priority was with the review papers. The inclusioncriteria were publication date (between Jan, 1, 1995 andJuly, 1, 2015), subjective relevance to specified parts of thestudy, relevance to study aims, relevance to studykeywords, being published by academic sources, and thelevel of relevance which was determined by scientificsearch engines.2.5. AnalysisAfter primary resource collections, the Delphi method wasimplemented to increase the validity of the results anddecreased the probable latent biases. The Delphi method ismainly used to explore innovative and confident ideas inorder to collect and classify knowledge in some area ofknowledge from its experts. This method is mostly used inexploratory qualitative research with the use of variousopinions from different experts about new ideas or complexproblems by administrating several surveys and controlledfeedbacks [12, 13]. Delphi method is a dialectical processof confrontation of thesis and antithesis, and finallyconstruction of synthesis, which is the newly formedconsensus. The underlying dialectical logic of Delphimethods ensures the multidimensionality of the results andaims to construct new theoretical points of view. It helps toincrease the level of novelty and creativity in the phase ofexploration of new ideas and is mostly addressed as a novelinspiring method. [14].It has also been applied to determine and develop possiblealternatives, exploring or exposing assumptions that leadsto different judgments, generating consensus, and educatingthe respondents [15]. All typical Delphi methods comprisethree major stages. The first stage is the selection of theparticipants and is very important, because it is directlyrelated to the quality of the generated results. At the secondstage, the actual Delphi rounds are implemented. Thenumber of rounds ranges from two to ten. Each round needsto have an objective, around which the content of the surveymust be built. The final stage is the analysis of the resultsand the final written consensus. The stage also includesreflecting on the experiences gained from the Delphiprocess and applying the results and the experience inpractice [16].According to the benefits of the Delphi method and itsimplementation, authors decided to use this methodology tomaximize the optimum issues of concern in reviewing theRoundRound 1Round 2Round 3field of CDSS. This was to ensure that biases of authorswould not result in ignoring or overemphasizing somedomains of CDSS. The present study applied the form ofclassical Delphi with five features including anonymity,iteration, controlled feedback, statistical group response andstability in responses among those with expertise on aspecific issue. The participants in this type of Delphi haveexpertise and give opinions to arrive at stability in responseson specific issues [17]. In the current study, in order toadministrate the Delphi method, six experts were chosen:three PhD of industrial engineering with at least five yearsof expertise in DSS (to suggest the technical issues of CDSSin general), and three healthcare experts with at least abackground of five years dealing with CDSS (to suggest thespecific implementations and objectives of CDSS inhealthcare systems). The main question of the study (whatare the essential issues in the domain of CDSS?) was sent tothem and were asked to reply in written forms. Theiranswers were collected and unified in a checklist. Thischecklist was sent to experts and they were asked to writedown their ideas and any modification. This procedure wasrepeated three times until all experts had no modification intheir own checklist in round four. Therefore, the Delphiprocedure was terminated by participants after threedialectical stages (Table 1).Following data collection, the most related resources werechosen and devoted according to domains, which weredetermined by the Delphi method previously. The methodof data analysis were librarian and content analysis, as wellas frequent considerations of various papers to certainissues.2.6. EthicsThe ethical aspects of the study was comprised of two parts.The first ethical issue was respecting the copyrights of theauthors of resources including papers, books, book chapters,manuscripts, dissertations, etc., which was directly done inthe present study. The second part of study ethics includedanonymity and confidentiality of the participants of theDelphi method. The identity of all these experts keptanonymous. All the procedure and aims of the study werefully described to all them and they filled out a writtenconsent in which they fully understood the terms ofparticipation. The results of the Delphi methodadministration and the study were sent to theaforementioned experts as part of mutual partnership.Table 1. Results of Delphi methodsuggested topicsDefinition of decision, definition of DSS, definition of CDSS, design of CDSS, algorithms of CDSS, aims of CDSS,benefits of CDSS, classification of CDSS, types of CDSS, CDSS softwares, applications of CDSS, history of CDSS,advances in CDSS, statistical comparisons of CDSS usage, differences of CDSS in developed vs. developingcountries, CDSS in tough situations, portable CDSS, level of access in CDSS, methods of CDSS computations,degree of specifity, CDSS in different sectors of healthcare system, management of CDSS, biases in CDSS, racialand gender differences in CDSS, integration of therapeutic technologies with CDSS, limitations of CDSSDefinition of CDSS, Scopes of CDSS usage, pros and cons of CDSS, applications of CDSS, methods of analysis inCDSS, types and classes of CDSS, evidence-based (EBM) and CDSS, data mining in CDSS, domains of coverageby CDSS, medical errors in CDSS, software development of CDSS, general and specific CDSS, CDSS basic needs,CDSS benefits of healthcare system, degree of specifity in CDSS, management of healthcare system and CDSS.Definition of CDSSs, popular CDSSs, aims of application CDSS, benefits of CDSS, methods of decision support inCDSS, classification of CDSSs, Medical/clinical data mining, evidence-based medicine (EBM), reduction ofmedical error, facilitation of quality of healthcare.International Journal of Medical Reviews, Volume 2, Issue 4, Autumn 2015301

Shahsavarani A.M, et al. Clinical Decision Support Systems (CDSSs)Results3.1. Definition of CDSSThere are various definitions of CDSS. The major issues ofdefinition are emphasizing on the consideration of CDSS asan active knowledge system which uses two or moreclassifications of the data of patients to generate casespecific medical suggestions. This shows that CDSS is, infact, a DSS concentrating on the use of knowledgemanagement in medical affairs to reach a medical advice byconcluding different types of current options andinformation [7].CDSS is an analytical instrument which assists therapists todecide better about their patients by transforming rowclinical data to useful information. CDSS is indeed asoftware which provides a healthcare system withinformation for secure care. These information includesstandards and evidence-based instructions, actions andprotocols, regulations and suggestions for care, medicinereference and instruments to calculate suitable proportions,future for connection to library, digital reference books,and/or internet references [5].3. 2. Popular CDSSsNumerous clinical decision support systems are being massproduced in the market and sold day to day. In addition,many researchers have developed CDSS according to theirresearch aims [18]. Table 3 presents the popular CDSSs inclinical and healthcare systems. Indeed, there are manyCDSSs up to date, however, a few of them are being usedwidely. This list only emphasizes on the inclusive CDSSsby the time of framing the study.3.3. The aim of the implementation and application ofCDSSThe main goal in designing and developing the currentCDSSs is to assist physicians and clinical professionals inany given point of professional healthcare system. CDSS ismostly designed to help healthcare staff to decide in semistructured problems, support instead of substitution ofclinical judgments, and completion of effectiveness ofdecision-making instead of its efficacy [19].Any knowledge and information management system in theform of DSS is far beyond organizational informationarchitecture which is the main reason of development andimplementation of CDSS in healthcare systems andnetworks. It appears that in the era of informationtechnology, any healthcare system shall be equipped withCDSS to satisfy the need to best, fastest, and most confidentmedical information in all the three phases of prevention[20].According to recent review of literature in Iran, majordomains of the implementation of CDSS are disease trendmanagement (15.15%), healthcare and treatment (27.27%),prescription (27.27%), assessment and evaluation(27.27%), and prevention (12.12%) [21].3.4. Benefits of CDSSAll systems shall be cost-effective in order to be eligible tobeing incorporated to the management systems. If all prosand cons of CDSS, were simply quantifiable, then it wouldbe possible to justify their usage. Whenever pros exceedcons, the system is justified to get implemented. The302problem is that most advantages of the application of CDSSare intangible. With accurate assessments, it would bepossible to evaluate the performance and advantages ofusing DSS. Some important benefits of using decisionsupport systems in clinical settings are derived from reviewsas follows [22]:1. Saving time2. Saving expenses3. Better understanding of clinical situations4. Ability to perform unprecedented analyses5. Better use of resources6. Ability to respond fast to unexpected situations7. Extension of examined options8. Extension of available options9. Make new viewpoints10. Provide new learning11. Developmentandfacilitation of interprofessionals communication12. Provide control13. Improve decisions14. Facilitate teamwork in medical staff15. Facilitate clinical group decision-makingSome authors have divided the benefits of CDSSinto three domains [21]:1. Improvement of quality of healthcare andimprovement of patients' safety by reducingprescription errors and drug side-effects, as well asdirect following of evidence-based clinicalinstructions.2. Increase of effectiveness/cost ration by faster processof orders, reduction of medical test repetition,reduction in drug side-effects, changing drugconsumption pattern, and changing drugs with genericones in order to retrenchment the healthcare costs.3. Improvement of medical and professional knowledgeby ease of access to scientific resources, presentationof reminders, and providing useful and criticalinformation to desirable decision-making withminimum errors.3.5. Methods of decision support in CDSSsThere are various methods and techniques to decisionsupport. Nowadays, decision support mainly focuses ondecision support systems. In other words, common sense ofthe current era, identifies decision support synonymous toDSS in its general and wide meaning. Yet, most of the DSSand CDSS implement multiple-criteria decision supporttechniques. Multiple-criteria decision-making (MCDM),also known as multiple-criteria decision analysis (MCDA),is a subset of operation research which exclusively dealswith studying different qualitative and quantitative criteriain any given situation of deciding. All domains of personaland social life, whether general or specific, usually havedifferent and conflicting criteria which are needed to beinspected and evaluated before taking any decision. One ofthe mostly important and considered criteria is the cost orprice of any decision. Another controversial and conflictingdomain of decision-making is criteria for quality evaluation.Human everyday life is full of deciding points which aredone unconsciously.International Journal of Medical Reviews, Volume 2, Issue 4, Autumn 2015

Shahsavarani A.M, et al. Clinical Decision Support Systems (CDSSs)CDSSTable 2. Popular CDSSsAlgorithmsKnowledge-based, DissociativereasoningKnowledge-based, Knowledgemanagement, Clinical Rules Engine, Gstandard MFB, Andere protocollenmanufacturerPittsburg University, Pittsburg,PA, USA.Usage Period1970’s and1980’sClinical RulesDigitalis Rx, Amsterdam,NetherlandsNow activeDiagnosisPro(Free)MedTech USA, Inc.،Los Angeles,CA, USA.Now activeKnowledge-basedOnline, computersoftware, touchphone appletInfermedica Sp. Warsaw, Poland.Now activeKnowledge-basedOnlineMassachusets General Hosptial,Boston, MA, USANow activeKnowledge-based, pseudoprobabilistic algorithm, Bayesian logic,Online, computersoftwareNow activeKnowledge-based, plain INTERNISTI/QMREsagil Institute, New York, NY,USA.Isabel Healthcare Inc. AnnArbor, MI, USA.Pittsburg University, Pittsburg,PA, USA.layoutHave separatecomputer systemApplication fieldsDiagnosis of internal diseases,Educational application.OnlineMedicine prescription,consumption monitoring.1970’s and1980’sKnowledge-based, query string, HL-7,XML, APIDialog system, pattern recognition,ranking algorithm,Have separatecomputer systemNow activeOnlineLitmusdxLitmusdx Company, Kolkata,IndiaActive nowKnowledge-basedOnlineMYCINStanford University, CA, USA.1970’s and1980’sKnowledge-based, Bayesian networks,graphical models, decision treesHave separatecomputer systemDigitalis Rx Company,Amsterdam, NetherlandsNow activeWarsaw Medical University,Warsaw, Ploand.Now activeSimulConsult Inc., Chestnut Hill,MA, USA.Now activePrescriptorRODIASimulConsult(Free andPurchaseversions)Knowledge-based, Knowledgemanagement, matrix modelsNon-knowledge-based, Patternidentification, telemedicine, calibration,linear and angular measurement,phantom calibrationKnowledge-based, Bayesian inferenceengine, bioinformatics genomeannotation, statistical pattern-matchingapproachOnlineDiagnosis and differentialdiagnosis of more than 11thousand diseases and 30thousand medical conditions.Diagnosis of more than 500medical condition.Diagnosis and differentialdiagnosis of internal diseases,educational application.Diagnosis of diseases accordingto signs and symptoms, blood andurine test.Diagnosis of all medical diseases,medicine prescription.Diagnosis of internal diseases,educational application.Diagnosis and differentialdiagnosis of 11 thousand diseases,presentation of 300 therapeuticprotocols, presentation of 50thousand medicines,200 thousandmedicine usage cautions, medicaltest interpretations, medical files.Bacteria identification, bloodinfections identification, medicineprescription, blood clot diseases,educational application.Medicine prescription, Onlineaccess to medical enwww.digitalis.nlOnlineMedical imaging, diagnosis,orthopedic problems, monitoringthe bone-fracture remedialw.glinkowski@parser.com.plOnlineDiagnosis of 5300 diseasesespecially genetic ational Journal of Medical Reviews, Volume 2, Issue 4, Autumn 2015303

Shahsavarani A.M, et al. Clinical Decision Support Systems (CDSSs)People, whether ordinary or experts, usually implementmultiple-criteria probes in their routines implicitly andmight be satisfied with the decisions which are madeheuristically. On the other hand, when the capital volumeand/or value is high, or when the human life matters,accurate and correct structuring of the problem and explicitevaluation of various criteria becomes important [23].Knowing that in healthcare systems, any medical errorcould be lethal and if any of these heuristic decisions fail,patient’s health might be in harm or death threats mightoccur.CDSS uses widely MCDM methods to solve problems ofdecision-making. MCDM practically deals with structuring,deciding, and planning in multiple-criteria domains and itsgoal is supporting deciders confronting such situations.Usually, there is not just one optimum solution to suchproblems and deciders’ preferences shall be considered todiscriminate between options. Problem-solving in decisionprocess has various interpretations. This problem-solvingcould be finding and choosing the best alternatives from aset of options. In another approach, problem-solving meansto select a small set of good alternatives, or grouping ofalternatives to sets with different preferences. Some otherproblem-solving is to find all influential or non-influentialalternatives [24]. Different models and methods have beendeveloped to solve problems of MCDM (whetherevaluation, or design) that have advanced mathematicalbases and complex calculations. Today, with the use ofhigh-speed computers, all these calculations are doneautomatically. Famous methods include AHP, ANP,ELECTRE, ELECTRE-II, ELECTRE-III, DRSA, ARIM,ER, GP, GRA, MAUT, MAGIQ, NFSDSS, MAVT, IPV,WPM, EA, VA, PROMETHEE, TOPSIS, PAPRIKA,MACBETH, SIR Method, and NATA [e.g., 25, 26, 27, 28].3.6. Classification of CDSSMost of the time, CDSSs could be divided into twodistinctive groups [29]:1. Knowledge-based CDSS: Like all expert systems, mostof CDSSs have three parts of knowledge base, inferenceengine, and mechanism to communicate. Knowledgebase includes rules, regulations, and connection ofinterpreted data which often in the form of “if-then”rules. Inference engine synthesizes existing rules ofknowledge base with patient’s data. Communicationmechanism enables systems to show the results tooperators and also enables operators to present inputs tosystem2. Non-Knowledge-based CDSS: Those CDSSs which donot use knowledge-base, implement some kind ofartificial intelligence named machine learning thatallows computer learn from past experience and/ordetecting figures from clinical data. Usually nonknowledge-based CDSSs are designed and developedon the basis of artificial neural networks and/or geneticalgorithms.Another classification, divides CDSSs into sevengroups of data-access systems, data-analysis systems, futureprediction data-analysis systems, computational-models-304based systems, presentation-based systems, optimizationmodels-based systems, and suggestive-models-basedsystems [16].In a recent research, authors have reviewed CDSSpapers and classified them according to their methodologiesas follows [30]:1. Machine learning: This class represents methodologieswhich implement algorithms that enable systems tolearn from data. Such methods have an initial trainingphase to find trends in data sets of the given data base.Then, the system would be able to analyze new data withthe same parameters and suggests predictions. Thisgroup includes artificial neural networks (ANNs),support vector machines (SVM), and logistic regression.2. Knowledge representation: These methods concernthe representation of knowledge and facts which areattained from clinical expertise to generate and producea language of description which is

known as Decision Support Systems (DSS) [3]. DSS is a computer-based system of information processing which is mainly developed to support organizational and enterprise affairs. Today authors believe that DSS could be told to any system that can support decision-making processes. In other words, DDSs are information systems which support