Clinical Decision Support

The Road to Broad Adoption

Robert Greenes, Robert A. Greenes

• Publisher: Academic Press
• Year: 2014
• ISBN:
  9780128005422
• Edition: 2
• Topic: Informatica
• Subtopic: Bioinformatica

With at least 40% new or updated content since the last edition, Clinical Decision Support, 2nd Edition explores the crucial new motivating factors poised to accelerate Clinical Decision Support (CDS) adoption. This book is mostly focused on the US perspective because of initiatives driving EHR adoption, the articulation of 'meaningful use', and new policy attention in process including the Office of the National Coordinator for Health Information Technology (ONC) and the Center for Medicare and Medicaid Services (CMS). A few chapters focus on the broader international perspective. Clinical Decision Support, 2nd Edition explores the technology, sources of knowledge, evolution of successful forms of CDS, and organizational and policy perspectives surrounding CDS.

Exploring a roadmap for CDS, with all its efficacy benefits including reduced errors, improved quality, and cost savings, as well as the still substantial roadblocks needed to be overcome by policy-makers, clinicians, and clinical informatics experts, the field is poised anew on the brink of broad adoption. Clinical Decision Support, 2nd Edition provides an updated and pragmatic view of the methodological processes and implementation considerations. This book also considers advanced technologies and architectures, standards, and cooperative activities needed on a societal basis for truly large-scale adoption.

• At least 40% updated, and seven new chapters since the previous edition, with the new and revised content focused on new opportunities and challenges for clinical decision support at point of care, given changes in science, technology, regulatory policy, and healthcare finance
• Informs healthcare leaders and planners, health IT system developers, healthcare IT organization leaders and staff, clinical informatics professionals and researchers, and clinicians with an interest in the role of technology in shaping healthcare of the future

Section IV

The Technology of Clinical Decision Support

Outline

Section IV. The Technology of Clinical Decision Support

Chapter 15 Decision Rules and Expressions

Chapter 16 Guidelines and Workflow Models

Chapter 17 Ontologies, Vocabularies and Data Models

Chapter 18 Grouped Knowledge Elements

Chapter 19 Infobuttons and Point of Care Access to Knowledge

Chapter 20 Formal Representations and Semantic Web Technologies

Chapter 21 The Role of Standards: What We Can Expect and When

Section IV. The Technology of Clinical Decision Support

Prevalent forms of clinical decision support are decision rules, guidelines, and groupings of knowledge elements such as in order sets and documentation templates. These approaches are discussed in Chapters 15, 16, and 18, respectively, in terms of both the methodologies and the evolution of the approaches over the years, and the challenges that they face.

Chapter 17 explores a common need for all of the above models, the ability to reference specific data elements in a decision support artifact, by an unambiguous data reference–requiring not only a formal data model for accessing the data element, but also a method of coding or standardizing the format and content of the data itself. This also requires schemas for organizing and referencing particular data elements and their coding through ontologies, taxonomies, and vocabularies.

Another form of clinical decision support, infobuttons, provides context-specific methods of informing a retrieval engine of the kind of information needed in a particular context, e.g., when reviewing of a laboratory test result. This methodology does not tailor advice but rather selects it based on context. Chapter 19 describes this useful method.

In Chapter 20, we provide an overview of a large undertaking of creating formal methods for knowledge modeling using the Semantic Web being actively pursued in many fields. Health care informatics experts are heavily engaged in this work, but it has not yet had impact on deliverable forms of clinical decision support. Nonetheless, because of the scope and momentum of this set of activities, it is introduced here to give the reader an understanding of its potential and possible future role.

As the final chapter in this Section, Chapter 21 provides an overview of the role of standards and interoperability initiatives in furthering the adoption of the above technologies.

Chapter 15

Decision Rules and Expressions

Robert A. Jenders

A decision rule is a representation of knowledge in a particular domain that encapsulates the flow of logic employed in deterministic reasoning to make a decision. Decision rules, then, represent a form of algorithm, typically represented as discriminating questions or logical IF-THEN statements that may be followed to reach some conclusion. They map the circumstances of a particular situation, such as the case of an ill patient for whom a diagnosis must be chosen, to a particular choice, whether that be a diagnosis, a treatment plan or an inferred observation that, in turn, may lead to another decision. In a computer-based clinical decision support (CDS) system, decision rules often are represented in one of two formats: procedures and production rules. This chapter examines the use of decision rules as a knowledge representation formalism for CDS. The details of such a formalism are explored, including inference mechanisms that are employed in order to make decisions using knowledge encoded in this fashion. Facilitating the use of this approach through standardization, with an emphasis on the Arden Syntax and a common expression language, is explored. Additional efforts to facilitate transfer of knowledge so encoded through the use of standard data models are reviewed. Advantages and disadvantages of these approaches are explored.

Keywords

Arden Syntax; clinical decision support; decision rule; Healthcare Quality Measure Format; HL7; HQMF

15.1 Introduction

Deterministic reasoning is a key type of decision-making process in which a decision maker applies branching logic and deduction against the information of a particular situation in order to arrive at a plan of action. A decision rule is a representation of knowledge in a particular domain that encapsulates the flow of logic employed in deterministic reasoning to make a decision. Decision rules, then, represent a form of algorithm, typically represented as discriminating questions or logical IF-THEN statements that may be followed to reach some conclusion. They map the circumstances of a particular situation, such as the case of an ill patient for whom a diagnosis must be chosen, to a particular choice, whether that be a diagnosis, a treatment plan or an inferred observation that, in turn, may lead to another decision.

In a computer-based clinical decision support (CDS) system, decision rules are often represented in one of two formats: procedures and production rules. Like a subroutine in a programming language, a procedure is a collection of references to data together with logical statements that manipulate them and execute, largely serially, using control structures to direct the flow of decision making through the procedure. In a system based on production rules, each unit of knowledge is a single IF-THEN logical statement, and an inference engine, evaluating the available data and statements, chooses which statement to execute next.

Although these formalisms have been applied to address a wide range of problems, lack of specificity for the medical domain and lack of standardization have impaired both use and sharing of knowledge bases encoded using them. Recognizing these impediments led in the 1990s to the development of a standard approach that combines these formalisms, represented by the Arden Syntax. Perceived limitations with this standard and the need to encode a growing body of computable clinical practice guidelines have led to the examination of other approaches, including the use of a standard expression language in the context of a guideline formalism.

This chapter examines the use of decision rules as a knowledge representation formalism for CDS. The details of such a formalism are explored, including inference mechanisms that are employed in order to make decisions using knowledge encoded in this fashion. Facilitating the use of this approach through standardization, with an emphasis on the Arden Syntax and a common expression language, is explored. Additional efforts to facilitate transfer of knowledge so encoded through the use of standard data models are reviewed. Advantages and disadvantages of these approaches are explored.

15.2 Procedural knowledge

As noted by Miller (1994), some of the earliest work in implementing decision rules for CDS used procedures written in conventional programming languages. Two key features characterize this representation. First, clinical knowledge and inferencing or control knowledge are mixed in the same representation. This means that instructions to the computer about how to use the clinical knowledge, such as which statement to execute next, is mixed with logical statements about the clinical domain, such as a laboratory test threshold that must be exceeded in order for the diagnosis of a particular disease state to be made.

Second, the flow of control is made explicit. A procedure typically is a series of statements that are executed serially – in the order that they appear in the unit of knowledge. Control statements, such as GO TO and iterations, interrupt the serial execution but still specify explicitly the next statement to be executed, although that may be dependent on data available only at the time of execution. Control knowledge includes not only specification of the flow of execution but also how communication with users occurs (e.g. synchronously via a computer terminal), conditions under which the procedure will execute (e.g. when called from an electronic medical record) and methods for displaying output (e.g. sending a message to a clinician).

Decision rules characterized by an explicit flow of control in accord with a series of branching questions or logical statements are sometimes represented graphically as decision trees (Figure 15.1) or flow charts. In a typical decision tree, each node in the tree may ask a different yes/no question, and the appropriate branch of the tree is followed depending on the response to the question. Ultimately, a conclusion of the decision rule is reached when the traversal encounters a terminal or leaf node of the tree that offers no further refining questions.

Figure 15.1 Decision rule represented as a decision tree.
The decision rule helps determine the diagnosis in a case of a patient with a sore throat based on physical examination findings.

This approach offers many advantages. Nearly any programming language that supports subroutines, functions or procedures can be used to encode the clinical knowledge in executable format. In turn, this means that commonly available programming tools for these languages, such as compilers or debuggers, can be used. If a programming language used is one that is supported on many different types of computers, development and maintenance of the knowledge can be done on multiple platforms without the need to acquire specialized software. Further, because flow control is explicit, the knowledge engineer can tightly contro...