220 likes | 659 Views
Clinical Decision Support System. 교재. Clinical Decision Support The Road Ahead Robert A. Greenes Academic Press, 2007. Chapter I. Definition, Scope, and Challenges. 1.1 Introduction. The application of computers in health care early 1960s
E N D
교재 • Clinical Decision Support • The Road Ahead • Robert A. Greenes • Academic Press, 2007
Chapter I Definition, Scope, and Challenges
1.1 Introduction • The application of computers in health care • early 1960s • Rate of adoption and degree of impact of computers and IT • Low, compared to engineering, physics, finance
1.1 Introduction • Challenges to healthcare providers • seeking to make difficult diagnoses • avoid errors • ensure highest quality • maximize efficacy • save money • Patients and the public • needs in evaluating their health • needs in making decisions • decisions are more complex • require more detailed knowledge
1.1Introduction • CDSSs are useful • simple types of CDS • laboratory test result : out of normal range • Medication: dangerous interaction with another one • patient is now due for a flu shot • usefulness evaluated over the past 40+ years • checks, warnings, and reminders - effective
1.1 Introduction • But, not widely used! • even for the more simple aids • like alerts and reminders or drug interaction checks • Using computers in decision support • much harder problem • Challenges • wider deployment beyond a single application • multi-institutional setting • Regional, national adoption
Difficulties in Knowledge Management Maintenance & updating of the knowledge Knowledge management- conflicts, overlaps, gaps Best ways to deploy various forms of DS integration with practice impact on efficiency and workflow disseminating knowledge- reuse in multiple sites making such knowledge platform-independent knowledge management economically feasible on a broad scale
1.2 Definition of Computer-Based CDS • the use of the computer to bring relevant knowledge to bear on the health care and well being of a patient. • Computer: information and communication technologies • the management of health and health care of an individual person (the patient) • Support: aiding of rather than the making of decisions • relevant knowledge : knowledge that is directly pertinent to the specific patient
1.3 Features of CDS • The general aim of CDS • To make data about a patient easier to assess • more apparentto a human • optimal problem-solving, decision-making, and actionbythe human • Users of CDS • physician, anurse, a laboratory technologist, a pharmacist, a patient • computer program • A primary task of CDS • to select knowledge that is pertinent • patient-specific data relevance of the CDS enhanced • The selection of knowledge and processing of data • inferencing process • Algorithm • Rule • association method • The result of CDS • to perform some action – usually recommendation
Why CDS Is a Hard Problem • simple forms of CDS • if.. .then rule:checking a laboratory test result • in a single computer system:is relatively easy • interact with users • not just the single use • cost effective • not too many false positives • knowledge maintain and update • interfacing with the application • how it relates to other rules • how it can be deployed in other applications or in other system platforms • how knowledge can be disseminated
Differential Diagnosis Process of A CDSS • Ascertain values for the various findings in a patient • Eliminate diseases for which finding(s) are incompatible with the presence of the disease • For pathognomonic findings, exclude all diseases that do not have the finding(s) • For those diseases remaining under consideration • perform a Bayestheorem calculation of the conditional probability of each disease
Components of CDS • Decision model • representation of the problem • e.g. combination of logical and Bayesian manipulation • Knowledge base • logical constraints • prior probabilities of disease – P(D) • conditional probabilities of findings given disease - P(F|D) • Information model • data elements needed for calculations, logic manipulations • Result specification • output or means of representing the results of the operation of the model • Application environment • interacting with a user or an information system • to obtain necessary inputs, and communicate results, • to enable a host application to interpret the results
Computer-Aided Diagnosis • deDombal system: • deDombal, 1975 • a Bayesian algorithm for diagnosis of abdominal pain • artificial intelligence models • rule-based, frame-based, and heuristic-reasoning-based • DXplain • Barnett, Cimino et al. 1987 • one of the few that remains widely available • Isabel • Ramnarayan, Tomlinson et al., 2004
Computer-aided diagnosis is rarely actually used in practice • Diagnostic challenges do not arise often in medicine • Actual clinical situation: • optimal workup strategy • choosing or modifying treatments • assessing prognosis and response to treatment. • Diagnostic problems • require analysis of many detailed data items • not readily available in the electronic health record • if present, are not sufficiently structured • must be encoded or mapped to the format required by the programs • e.g., as present/absent or high/medium/low • requires considerable manual entry of data • Differential diagnosis CDS: passive system • must be actively invoked by the physician when needed