The concept of clinical decision support system
Clinical Decision making is an essential component of healthcare service today. It involves the clinical reasoning, clinical judgment, clinical inference and diagnostic reasoning1 and this depends on accurate data, pertinent knowledge and appropriate problem-solving skills.2 and to do this healthcare professionals always expect to have patient related information and the domain knowledge within a limited time range for quality decision making during the patient encounter.
Since the last few decades, Information and Communication Technology (I&CT) has supported the healthcare professionals in managing the information related to patient care, research and education.
Decision Support System
A Decision Support System is a computer based information system that supports the personnel and organization with internal and external data to aid decision making.3 Earlier the systems were only used for the business industry but with the increase demand of Information Technology applications by the healthcare professionals have drive the adoption of clinical decision support system in healthcare practice.
Clinical Decision Support System
Clinical Decision Support System (CDSS) provides healthcare professionals with patient – specific information and domain knowledge intelligently filtered and presented at appropriate times in appropriate manner to enhance patient care.4 It forms a integral part of healthcare system typically designed to integrate knowledge base, patient related information and inference engine to generate case-specific advice.5
Musen define the clinical decision support system as any software that take the clinical situation of as an input and inferences as output that assist clinicians in quality decision making.
Wyatt j, Spiegelhalter7, describe Clinical Decision Support Systems as an active knowledge systems which use two or more items of patient data to generate case-specific advice.
Miller and Geissbuhler8 defined the Clinical Decision Support System as a computer-based algorithm that assists a clinician with one or more component steps of the diagnostic process.
Sim et al.9 describes it as software designed to be a direct aid to clinical decision-making where the characteristics of an individual patient are matched to a domain knowledge base and patient assessments data or recommendations and then presented to the clinician for the quality decision”.
2.3.2. Historical Timeline in Clinical Decision Support System
Since last few decades, the utilization of Information & Communication Technology applications has been increased among healthcare professionals for patient care, research and education. Table.3.1. shows the early timeline in design & development of clinical decision support system.
Table.2.1. Early Time in Development of Clinical Decision Support System
It uses a simple inference engine and knowledge base with 600 rules where physicians has to answer a long series of Yes/No questions and in result they will get the probability of each diagnosis.
It supports the clinician in diagnosis 100 different disease.
It is a computer based diagnostic tool for internal medicine where the clinician has to pass the sign and symptoms, lab results and patient history into the system and in turn system will provide the possible diagnosis using ranking algorithm.
It is available through World Wide Web and contained 4,900 clinical manifestations associated with over 2,200 unique diseases, yielding a total of over 230,000 unique finding-disease interconnections. It support the physicians by generating stratified diagnoses based on user input of patient signs and symptoms, laboratory results, and other clinical findings
It covers 15,000 disease manifestations compiled from hundreds of world's most respected medical resources covering internal medicine, emergency medicine, pediatrics, office OB-GYN and more. With having disease information for more than 7000 disease, It is designed to support the physician in enhancing patient care and prevent diagnostic errors.
It supports the clinicians with patient-centric visual differential diagnoses based on the patient’s signs, symptoms, medical history where the clinicians can access more than 17,000 images and expert-reviewed information for nearly 1,000 visually identifiable diseases, drug reactions, and infections represented in all age ranges and skin types.
Relative Optical Density Image Analysis perform quantitative evaluation of medical images and fracture healing monitoring and used in medical imaging, diagnostic, orthopedics and other healthcare discipline.
Leeds Abdominal Pain System 17
It supports the physicians in clinical assessment based on Bayesian probability theory where the program accept the patient related information and provide an assumption of conditional probability of various diagnosis and mutual exclusively of the seven possible diagnosis.
It is a knowledge base hospital information system with robust decision making functions by providing alerts/reminders, data interpretation, patient diagnosis, patient management suggestions and clinical protocols.
CASNET/ Glaucoma 19
This expert system used for the diagnosis and treatment of Glucoma where the knowledge is represented in a casual associational network for describing the disease process.
It is an expert system employs the casual reasoning for the management of electrolyte and acid base derangement.
It supports the physicians with a knowledge base of diseases, diagnoses, findings, disease associations & lab information & consists of medical literature of 700 diseases and more than 5,000 symptoms, signs, & labs.
2.3.3. Phase model in architecture of Clinical Decision Support System
Wright A, Sittig DF 22 (Figure. 1) has represented a four phase model in architecture of clinical decision support system since 1959. This timeline also shows the integration of clinical decision support system in clinical practice and healthcare systems.
Figure.2.1. Four Phase Model of Clinical Decision Support System
The above historical landmark represent the development of clinical decision support system in healthcare practice where the researchers have supported the healthcare team with various solutions in making quality decision during the patient encounter. Many approaches are being used during the each phase of implementation where the researchers have attempted to ease integrating clinical decision support system into clinical workflow and other information systems.
2.3.4. Clinical Decision Support System Architecture
A Clinical Decision Support System is designed and developed using various components. Figure.2. represents the architectural representation of a Clinical Decision Support System with the basic and essential components such as dialogue management, knowledge management, model management and data management.
Figure.3.2. Conceptual Model of Clinical Decision Support System22
The dialogue management builds the user interface to CDSS whereas data management builds connection for database and data warehouse to accessed and store patient related information received from other information system. The model management assists the user in analyzing the case at hand. The knowledge management support with the easy retrieval, storage and dissemination of domain knowledge receive from internal and external sources.
2.3.5. Methodology for Clinical Decision Support System
There are many methodologies that can be used for the development of a clinical decision support system to support the clinicians in quality decision making. Some of the important methodologies23 are:
220.127.116.11. Bayesian Network
Bayesian Network (BN) in Clinical Decision Support System supports the clinicians with a knowledge base graphical representation that shows the set of variables and their probabilistic relationship with the disease and symptoms. The advantage of this approach includes the domain knowledge and conclusions of expert in the form of probability assistance in decision making but it also represent the disadvantage where the system finds it difficult to get the probability knowledge for possible diagnosis. The DxPlain is one of the good example of CDSS with Bayesian methodologies where is produce a list of ranked diagnosis associate with the symptoms.
18.104.22.168. Neural Network
Artificial Neural Network (ANN) represent the non-knowledge-based adaptive Clinical Decision Support System. it uses artificial intelligence to allow the system to learn by itself from disease pattern and past experiences / examples. It form three layers; the Input (data receiver), Output (the result) and Hidden (Process data) to supports quality decision making. ANN does not require input from the experts but can process incomplete data by making educated guess due to its adaptive system learning capability. In addition to this ANN system do not require large database to store the outcome with its associated probabilities and due to this reason training the end user becomes more time consuming which lead to less acceptability of system among end users.
22.214.171.124. Genetic Algorithm
The Genetic Algorithm (GA) is based on Darvin’s evolutionary theories that dealt with the survival of the fittest. It derives the information from the patient data which go through an iterative process to produce an optimal result but the lack of transparency in reasoning and defining the fitness criteria makes undesirable for clinicians to adopt in the practice. Genetic algorithms have proved to be useful in the diagnosis of female urinary incontinence
126.96.36.199. Rule-based system
A rule-based expert system represents If-Than conditions where the user has to answer the condition to get the probable outcome. As for example, rule might read "If the patient has high blood pressure he or she is at risk for a stroke". The system evaluate the knowledge based on the rules compiled by the user during the designing phase and once it matched the system draw the conclusion of the case in hand. It is easy to store a large amount the information coming up with the rules to clarify the logic in decision making process but user may find it difficult to transfer the their knowledge into distinct rules and many rules can be required for a system to be effective. As for example, Mycin is based on 600 rules to help the end users in identifying the cause of bacterial infection.
188.8.131.52. Logical Condition
In a healthcare setup the logical conditions are used to provide alert and reminder to the healthcare team. As for example, the alert may warm the ICU nurses that the patient heart rate or respiration rate is low. The methodology in Logical condition is very simple where the system accepts the data and check whether it is within or outside the bound. Alert and Reminder assist the healthcare team in complying with guidelines but too many alert sometime overwhelm them and cause them to ignore it sometime that lead to suffering of the patient.
184.108.40.206. Casual Probability Network
Casual Probability Network work with the cause and effect methodology where the system attempt to trace a path from patients symptoms to disease classification and determine the best possible knowledge model to the end users in relation to the disease and it condition. CASNET is the first clinical decision support system uses the casual probability network to assist in the diagnosis of glaucoma. It represents a hierarchical representation of knowledge in terms of symptoms state and disease.
2.4. Selection and Implementation Guidelines for Clinical Decision Support System
The following steps are considered to be the minimum and essential in selecting and implementing decision support system in healthcare practice.24
2.4.1. Assure that end users understand the strength and limitation of the system
2.4.2. Assure that the knowledge is from a rich knowledge source, as the end users always expect to be known about the rules, evidence behind the rules, testing of the system and validation process.
2.4.3. Assuring that system is appropriate for the local site and vendors should have the answers of the end users query.
2.4.4. Assure that the end users are properly trained to understand and use the system
2.4.5. Regular monitoring regarding the proper utilization of the installed clinical decision support system should be done on a regular interval.
2.4.6. The knowledge base should be regularly monitored and maintained to support the end users with up-to-date domain knowledge.
2.5. Impact of Clinical Decision Support System
It is evident that clinical decision support system had been used as tool by many clinicians in improving the treatment and diagnostic performance on a large scale. It had been proved from the study conducted by Eta et.al where the performance of the physician and internist were assessed after the implementation of Quick Medical Reference (QMR) diagnostic decision support system. The result indicated that the physician diagnostic performance was higher in those cases where the QMR had provided the quality information.25
The impact of clinical decision support system had also been observed in the studies conducted by Rogers et.al,26,27,28 Gonzalez et.al,29 Rodman et.al,30 White R H et.al,31 Chase et.al,32 McDonald et.al,33,34,35,36 McDowell et.al,37,38 Tierney et.al,39,40 Young et.al,41 White K H et.al42 where the clinical decision support system showed a significant difference in improving the clinical performance of the clinicians as well as the patient outcome. The Implementation of a clinical reminder system by Anitha & Rajagopalan43 had shown a remarkable improvement in clinician performance in diagnosis and treatment of chronic illness and preventive care.
Dereck et.al44 on their systemic review found that the clinical decision support system is effective in drug dosing, preventive care, performance improvement of clinicians, and other aspect of healthcare practice but not for diagnosing the patient condition.
2.6. Related Work and Findings
A literature survey had been conducted to identify the available clinical decision support systems in cancer care and with specific to breast and cervix cancer. The data collected from the literature survey showed that there are several studies that have been done in the field of clinical decision support system in cancer care such as Oncocin25, Oncosifter26, OncoDoc27, Computer Aided Medical Diagnosis Tool28, CaDet29, GTDs30, OWCH31, KON332, OncoTheraper33, Isabel34, DSS using MDA35, Clinical Decision and Economic Analysis Model of Cancer Pain Management36 but there are only a few available in the domain of Breast cancer and very few in Cervical cancer care. Works of interest in this research domain were:
This decision support system is designed to provide treatment advice for cancer chemotherapy. It uses artificial intelligence to provide the recommendation to the physician on medicine, dosage and testing. The drug doses are determined on the basis of time schedule, toxicity and blood count. The system was designed by combining the Chemotherapy Protocol guidelines and knowledge provided by expert oncologist. The system consists of Interviewer i.e. a rule based expert consultant to preview the previous information and enter the present complaints of the patient and Reasoner use to get the recommendation for appropriate therapy and test in form of representing the knowledge in four main types of data structures i.e. Context, Parameter, Rules and Control Blocks. The typical users of ONCOCIN were residents and clinical assistant rather than certified physician.
Oncosifter is a search engine developed to support the clinicians in accessing related diagnosis and treatment, medical news and publication. The system is implemented using Perl-CGI and support with Keyboard Search where the query matches the metadata and the corresponding results are retrieved; Directory Interface to provide the overview of cancer by categorizing into three sections such as body location/systems, Common cancer and Childhood cancer; Hierarchical Visualization Interface to display the structural relationship of the data and the Personalization Interface where the user can create profile to include the information of his/her interest. The system is linked and mainly provides information from Medline Plus and Cancer.gov.
2.6.3. Computer Aided Medical Diagnosis Tool27
This integrated system is designed by Electronics, Informatics and Systems at the University of Calabria in Italy. This has been embedded into the Telemedicine to allow the clinicians for remote consultation. This is an automatic classifier which discriminates between benign and malignant cells from a breast cancer. Firstly the graphical computer program analyzes the cytological features of Fine Needle Aspiration samples based on digital scanning and a frame-grabber board and get it saved for the further determination of each nucleus and its boundary. The second stage consists of the creation of 30-dimensional features vector by performing the analysis for each individual on a large set of patients for which the actual diagnostic outcome is known. An automatic classifier using a Linear Programming model is used to discriminate the benign with malignant cells of breast cancer. The system works on client – server platform with common gateway interface and Java based applications.
The Oncology WorkbenCH (OWCH) is designed to support the oncologist in multi-drug chemotherapy regime where the treatment editor accepts the planned treatment regimes data from the user and send simultaneously to Simulation and Optimization Engine. Once received, both engines interact with the Information Repository for information in relation to effectiveness and toxicity of anti-cancer drug. The simulation engine evaluates the newly composed treatment and sends the result to the result viewer whereas optimization engine determines and advised the best possible treatment strategy via Treatment Editor. This result may be stored in the Information repository for future reference. The interface are delivered to the user Java applets via the WWW so that minimal set-up is required at the user end and the simulation and optimization is done on a fast server dedicated to the purpose.
Knowledge ON ONcology through Ontology (KON3) is an electronic guideline for the management of hepatocellular carcinoma. It consists of Knowledge Base to represent the knowledge at semantic level using virtual medical records, vocabulary and expression; Guideline Engine to execute the guidelines using process and rule engine whereas Guideline editor helps the user to design the guidelines associated with heptocellular carcinoma.
OncoTheraper is a Clinical Decision Support System designed and supported by a Service Oriented Architecture for the therapy planning in pediatric oncology. The technology is based on Artificial Intelligence and consists of Oncology Protocol Server to store the computerized oncology protocol in the Representation Language whereas Intelligent Monitoring Server to interpret and execute the therapy plan generated by Therapy Planning Server.
2.6.7. Clinical Decision and Economic Analysis Model of Cancer Pain Management32
This evidence based decision analytical model was designed to assist the healthcare decision makers in comparing the different strategies of cancer pain management based on guideline based care, oncology-based care and usual care. This model is constructed to facilitate the estimate of cancer pain prevalence based on demographics, epidemiology of cancer and cancer pain and evaluate the impact of cancer pain and its management in a default or user-defined healthcare population. The result evident that the guideline-based cancer pain management was lead to improve the pain control but it increases the utilization of resources.
The system is designed with an aim to support the clinician for quality decision making in care of Childhood and Lymphoblastic Leukemia. The system integrates the patient data and clinical information to provide patient specific recommendation for the treatment plan processes. This Oracle database allow the clinicians to captures, view and modify the patient information on each stage of treatment for individual children and provide decision support on dosing and scheduling of the therapy.
2.6.9. OncoDoc34, 35, 36
The system is designed to support the oncologist in providing therapeutic recommendation for breast cancer patient using decision tree and hypertext. OncoDoc allows the clinician to control the operationalization of guideline knowledge through hyper-textual reading of a knowledge base encoded as a decision tree. The system can also be used in eligibility screening system in breast cancer clinical trials.
2.6.10. Knowledge Based Approach for Diagnosis of Breast Cancer 37
This knowledge based system uses the soft computing tool such as Artificial Neural Network and Neuro Fuzzy Systems to assist the clinicians to diagnose breast cancer. The Back Propagation Algorithm, Radial Basis Function, Learning Vector Quantization and Adaptive Neuro Fuzzy Inference Engine were used to evaluate the dataset where the sources of data were Winconsin Breast cancer Diagnosis. The simulator was developed using MATLAB and the performance was compared in terms of accuracy of diagnosis, training time, number of neurons etc. The result of the comparison showed that the knowledge base approach can be effectively used by the oncologist in detection of breast cancer and to enhance the survival rate respectively.
2.6.11. Retrospect 38
Retrospect is a prototype decision support system designed to support the clinician to predict the treatment outcome and recommend optimal treatment plan for the patient with breast cancer. It supports the hybrid architecture i.e. the prediction engine is based on neural network whereas recommendation component of it is based on genetic algorithm. The evidence showed the availability of Retrospect in two forms; local that runs on 32-bits window environment and distributed consist of a thing window client and an internet server application.
2.6.12. A knowledge-based approach to assign breast cancer treatments in oncology units39
This knowledge based decision support system uses a multiple classification ripple down rules (MCRDR) for the incremental knowledge acquisition and a knowledge base for the treatment of breast cancer. The knowledge base was built using National Comprehensive Cancer Network (NCCN) clinical protocol and the knowledge was represented with the six fractions i.e. clinical stage, evaluation, findings, primary treatment, post-surgical treatment and surveillance/follow-up. The system was designed using JAVA and MySQL and connected to JBDC interface. The system consists of three important module MCRDR engine for the extraction of knowledge, Inference Engine for Inferring clinical treatment and third module to format the explanation of inferred treatment.
2.6.13. Breast Cancer Decision Support System for Rural People40
This decision algorithm system assess the values of decision variables and recommend the line of treatment based on size of tumor, size of the breast, number of masses, radiation to chest wall, collagen vascular disease, and hormonal receptor. The decision algorithm used in this system is the modification of National Cancer Control Network guidelines for the breast cancer that assist the users to classify the cancer stage in order to form the method of treatment and store the information related to the survivability rate of the patient after treatment.
2.6.14. Decision Support System for Breast Cancer Chemotherapy41
The database is designed to support the oncologist in post-operative adjuvant chemotherapy of operable breast cancer. The system is implemented in Visual Basic 4 under Windows 3.x and Windows 95. The system consists of production rules and grouped into frames invoked in a goal-driven fashion. One frame manages the rules and decides whether the patient is eligible for chemotherapy or not whereas another manages the patient pathology prior to the chemotherapy. The system provide the recommendation in tree structure where the sub-frames represent the stage of the treatment, the possible presence of pathologies, and recommend the drugs to be administered with their quantities and the date of the next drug administration.
2.6.15. Computer Aided Medical Diagnostic (CAMD) System42
CAMD assist the oncologist in identification of breast cancer in patients through the application of a well-defined set of data such as sign and symptoms and pathology result. These data are identified, collected and introduced into the database to form the classification for the detection of breast cancer. The measurements of these data were assessed based on physicians’ clinical experience, which is result in knowledge base or training set where a mathematical programming method was used. These classification and pattern recognition support the end user in making correct diagnosis and also to discriminate between benign and malignant cells for the early detection of breast cancer. The system composed of image processing tools based on cellular morphometry and an automatic classifier based on the mathematical programming tool. It uses image input from a microscope. A Web facility has been integrated into the system to allow remote diagnosis from any site by means of java-applets
2.6.16. Decision Support System for Breast Cancer patient43
This Web based clinical decision support system support the oncologist for making prognostic assessment using the characteristics of the patient through three different prognostic modeling methodologies such as Nottingham prognostic index, Cox regression modeling and a partial logistic artificial neural network with automatic relevance determination. These three models were used to obtain a more accurate prognostic assessment of the patient. The multiple imputation technique was used to overcome the issues associated with the missing data.
2.6.17. CPG Based Ontology Driven Clinical Decision Support System for Breast Cancer44
This decision support system was developed with an intention to support and guide the family physician in conducting breast cancer follow-up. The system provides the decision, recommendations and referral regarding the treatment of a breast cancer patient based on Clinical Practice Guidelines (CPG). The guideline element model had been used to convert the CPG in electronic format and the Protégé was used to develop domain ontology and finally the execution engine was used to develop IF and THEN forward rule. The system provide the recommendation by passing decision variable using IF part and result in which the THEN part execute the result using action variable.
2.6.18. Staging of Cervical Cancer with Soft Computing45
This soft computing hybrid decision support system was designed to assist the clinicians in detecting the different stages of uterine cervical cancer. The system extracts the knowledge from the knowledge base with Genetic Algorithm using Rough set theories concept and Interactive Dichotomizer Algorithm.
2.6.19. Uterine Cervical Computer Aided Diagnosis46
This Computer-Aided-Diagnosis (CAD) was designed to aid the oncologist in diagnosis of cervical cancer. The core processing system automatically analyses the data gathered from the patients and provides tissue and patient diagnosis as well as adequacy of the examination. The system captured not only the text data but also the images and videos of various forms and medium. It was built on open, modular and featured based architecture that designed for multi-data, multi-sensor, and multi-feature fusion. The embedded CAD systems make it more interactive and automate the clinical workflow to generate the patient diagnosis and recommendation automatically.
Clinical decision support system always expected to assist the healthcare profession in quality decision making during patient encounter. Cancer care is always challenging for the oncologists where the clinical decision support system helps them in getting the instant access to the domain knowledge and health information. The result of literature survey with respect to clinical decision support system in cancer care depicted that a very few clinical decision support systems are available with respect to breast and rarely in case of cervical cancer care. Each system represent a unique features but only limited to a certain class of treatment or therapy and also based on many models such as Perl-CGI, Linear programming, Artificial Neural Network, Neuro Fuzzy, Cox regression model, Artificial Engine etc. Most of the systems were found to be Window Based Applications built on Java, MsSQL, Visual Basic programming languages. As a window based application, the systems were also found to have restriction in terms of global access where it is restricting the access to the user within the Local Area networking. CDSS should be designed in such a way that leads to high end user satisfaction which resulting in maximum acceptability and sustainability of system within the practice.
Need an essay? You can buy essay help from us today!