Overview
What is Clinical Decision Support?
Clinical Decision Support (CDS) is a service that enables healthcare providers to make well-informed decisions by supplying guidance, knowledge, and patient-specific information at relevant points in the patient journey, such as diagnosis, treatment, and follow-up. CDS uses a range of mechanisms to assist users in this process. Examples of these mechanisms include automated alerts or reminders, clinical guidelines, contextually relevant reference information, conditional order sets, diagnostic support, and patient-focused reports, forms, or templates. The beneficiaries of the information derived from CDS may include patients, clinicians, and others involved in the delivery of health care.
It is important to distinguish the general practice of clinical decision support from the application of tools designed to enhance decision support practices. One is performed by humans who make decisions based on knowledge they possess and information they consume. The other is computed by systems and engines using rules and predefined conditions. Although both are important, the technical components of CDS are designed to assist rather than replace the subtle judgment and guidance provided by the clinician.
Applications and tools that provide clinical decision support are known as as (CDSS). A clinical decision support system is defined as a computer system or software application designed to assist clinicians, caregivers, or patients in healthcare and/or treatment decisions.
Notes
Typically a clinical decision support system responds to triggers, such as specific signs or symptoms, diagnoses, laboratory test results, medication selections, or complex combinations of such triggers. The system then provides information or recommendations relevant to the specific patient.
History
It has been suggested that the origins of clinical decision support (CDS) can be traced back to the 1950s and 1960s.. Since then, there have been countless developments and advancements in the area of decision support. Many theories have been proposed as to how CDS should be approached and applied in clinical practice.
The Five Rights
When implemented properly, CDS has the potential to enhance patient care, reduce errors and duplication of effort, and introduce efficiencies to the clinical workflow. Conversely, CDS tools can also be distracting and disruptive, even producing unwanted consequences. It is therefore important to consider the lessons learned from previous implementations of CDS and conduct thorough requirements analysis prior to designing or procuring a CDSS. One of the best practice frameworks that has been developed to guide those considering a CDS implementation is the. "The Five Rights" suggests that to realize the full potential of CDS, solutions should:
Supply the right information (evidence-based guidance, address the clinical need)
To the right people (entire care team, including the patient)
Using the right channels (e.g., EHR, mobile devices, patient portals)
In the right intervention formats (e.g., order sets, flow-sheets, dashboards, patient lists)
At the right points in the workflow (for decision making or action)
Example
A typical application of CDS is shown in the diagram below:
The clinical setting in which this hypothetical tool has been applied is the prescribing of a medication. In this example, the patient has previously had an 91936005 | Allergy to penicillin| recorded. When prescribing a new drug, such as 27658006 | Amoxicillin|, an alert is displayed to remind the clinician of the previously diagnosed allergy. The application may also provide a mechanism to search for alternative medications. Note that the mechanics of this workflow uses a predefined rule which specifies a condition to be evaluated and an action to be taken if the condition evaluates to true.
Functional Areas
This section addresses the functional scope of clinical decision support. CDS may be represented in a variety of formats or tools which depend on the clinical situation or environment. These tools are often referred to as CDS formats, types, or interventions and can be deployed to a wide variety of systems and platforms, such as mobile devices.
. Some of the more common CDS functions are described briefly in the table below. Use of these functions may be appropriate in a variety of clinical domains or use cases, some of which are discussed in the Clinical Areas section.
Alerts or Reminders
One of the more common types of decision support is computerized alerts (or reminders). These are triggered by rules and designed to interrupt clinicians or patients at the appropriate time. These alerts are also referred to as “best practice advisories” and can be implemented as pop-ups on a users screen or in monitoring tools such as a dashboard. Alerts can also be used to trigger other communication mechanisms such as paging or faxing. Examples of alerts include drug to drug interactions, or drug allergy warnings triggered when medications are prescribed.
Clinical Guidelines and Reference Information
These CDS functions are often implemented as links to external references which are published by third party, knowledge experts. Guidelines may be represented in a standardized format to facilitate interoperability - for example, the. References can be based on relevant, context-dependent data captured in a patient health record or another electronic artifact such as an order or clinical document.
Diagnostic Support Tools
These tools use a combination of patient data, context-based suggestions and clinical knowledge links to aid the clinician in making a diagnosis. An example would be a tool that prompts a physician for additional findings and suggests additional tests or procedures to help differentiate the diagnosis.
Automatically Triggered Smart Forms
These documentation tools, which include reports and summaries, are aimed at high quality records, the reduction of errors, and more complete information. These tools can be triggered when a specific patient condition is detected or when a finding is deemed reportable to a jurisdictional health body. These can be represented as focused patient data reports or summaries and are often utilized at the point of care (POC) in real time.
Conditional Order Sets and Pathway Support
These are typically designed for complex ordering scenarios. They may be comprised of a proposed set of orders or a treatment regimen which is based on an explicit situation or medical condition. These interventions can ensure compliance with established protocols. They can also be utilized to guide clinicians though complex care pathways.
Clinical Areas
The focus of this section is the clinical application of CDS tools or how the functional components described earlier can be used in practice. Stakeholders from various clinical domains interact with clinical systems, such as EHRs with CDSS and CPOE (computerized physician order entry). The table below lists some of the clinical areas in which SNOMED CT enabled CDSSs can assist clinicians in making well informed decisions.
A clinician uses an EHR with CDS to prescribe 375374009 | Warfarin sodium 4mg tablet|. The CDSS queries the EHR and discovers that the patient is 77386006 | Pregnant|. The CDSS determines that the proposed drug has 372756006 | Warfarin| as an ingredient. As warfarin in contraindicated during pregnancy, the system triggers an alert to be displayed to the clinician. Relevant clinical guidelines are also displayed to the user. These guidelines suggest a safe alternate, such as 714788005 | Dabigatran|, which the clinician then safely prescribes to the patient.
Diagnosis
A clinician uses an EHR with CDS in a case analysis scenario to aid in diagnosis. The clinician records the patient’s age and gender, then prepares to enter specific clinical findings, history, symptoms, etc. As the physician records symptoms of 55350005 | Hunger| , 84229001 | Fatigue|, and 87715008 | Dry mouth|, a ranked list of common diseases, associated with these clinical findings, is dynamically presented to the clinician. At the top of this list is 73211009 | Diabetes mellitus|. A scale is used to indicate the level of support for each disease. The CDSS then prompts the clinician for additional findings to help differentiate between diseases. Once a confirmed diagnosis is made, the differential diagnoses can be marked as 2667000 | Absent|, 52101004 | Present|, or 261665006 | Unknown|. An additional finding of 17173007 | Always thirsty| is recorded and the level of support for each disease in the list is adjusted accordingly. Support for 73211009 | Diabetes mellitus| has now increased from minimal evidence to sufficient evidence. The clinician then selects 44054006 | Type 2 diabetes mellitus| which opens an evidence screen displaying the recorded findings which either strongly support, support, or do not support the chosen disease. The clinician is then presented with a link that displays all the PubMed articles associated with 44054006 | Type 2 diabetes mellitus|.
Laboratory
A patient presented at Emergency complaining of 29857009 | Chest pain| and was subsequently admitted to the hospital. The attending physician ordered a series of lab tests including a 271236005 | Serum potassium measurement|. Laboratory tests are completed and published to the laboratory information system (LIS). The CDSS then queries the LIS and learns that the 365760004 | Potassium level| is 166690008 | Low serum potassium level| and considered critical. The CDSS then queries the EHR to confirm the patient has been prescribed 350608001 | Oral form digoxin|, which has 387461009 | Digoxin| as an active ingredient. A knowledge base rule has been defined which stipulates, if the drug prescribed contains 387461009 | Digoxin| and the laboratory test indicates a 166690008 | Low serum potassium level|, then inform the user. An alert, in the form of an urgent pager message, is generated and sent to the attending physician.
Radiology
(e.g. Contraindication)
An ordering physician has requested a 1343710007 |Plain X-ray series of upper gastrointestinal tract with barium contrast|, which uses 25419009 | Barium sulfate| materials. The patient presents at the imaging clinic on the day of their exam. During study protocoling, the imaging department uses the CDSS to query the patient record and determine the patient has a 161524000 | History of hay fever|. An alert is triggered to advise the imaging technician about the risk of an allergic reaction. The imaging department, in consultation with the GI radiologist, calls the ordering doctor to discuss the associated risks. Additional guidelines related to preparing for reactions and symptom management ( 247472004 | Hives|, 418290006 | Itching|, 65124004 | Swelling|, etc.) are provided via the CDSS. An additional medication is administered prior to the contrast material to reduce the risk of an allergic reaction. The imaging department proceeds with the planned procedure.
Radiology
(e.g. Appropriate Imaging)
A clinician records notes into the appropriate fields of an EHR. For example, Clinical notes: “Pt is 75 yo. LBP (lower back pain) for the past 2 weeks. On exam normal SLR (straight leg raise)…” Using NLP, these notes are encoded as part of the record storage process. (For example, as 279039007 | Low back pain| and 298686006 |Straight leg raising normal|.) The clinician orders a series of imaging tests. The CDSS, based on specific quality metrics (e.g., appropriate use criteria or AUC), evaluates whether or not imaging guidelines are being followed by analyzing the patient's health record together with the proposed tests. If the guidelines were not followed, the CDSS will display an alert informing the clinician that they may want to consider alternative imaging or additional tests. For example, an alert may indicate: “The patient has 279039007 | Low back pain| and 309537005 | Numbness of lower limb|. A 394451000119106 | MRI of lumbar spine without contrast| for this case has an appropriateness rating of 8 (scale of 10) and is recommended.”
Emergency Department
(e.g. Order sets)
A patient has presented at the Emergency Room (ER) complaining of 267036007 | Shortness of breath| . The attending physician records the appropriate clinical finding codes in the EHR. She then prepares a condition-specific order set in a Computerized Physician Order Entry (CPOE) system. The selection of the order set triggers the presentation of new clinical guidelines based on an analysis of the patient record with the proposed treatment. The physician then choses alternative treatment. Suggested dosage guidance is provided by relevant contextual links within the order set.
Infectious Disease Reporting
A primary care physician logs on to their EHR with CDS and opens a patient chart to record a condition deemed communicable, such as 36989005 | Mumps| or 14189004 | Measles|. The CDSS then triggers an alert to advise the provider that this condition is considered reportable to the jurisdictional public health office. The CDSS then provides a pre-populated smart form which facilitates quick, consistent, and accurate reporting of the condition to the local officer of medical health. The smart form is completed and submitted to the jurisdictional health office. The clinical findings in the report are terminology-encoded which promotes interoperability and facilitates population based health reporting.
Clinical Treatment Audit
A department head uses an EHR with CDS to conduct a treatment analysis. She uses the system to generate a list of all inpatients with a confirmed diagnosis of 128053003 | Deep venous thrombosis|. She then uses the system to determine which of these patients have received 103746007 | Heparin therapy| for at least 72 hours. The patients which have not met this criteria are flagged for appropriate treatment.
Acute Asthma Management
t of 281239006 |Acute asthma| in adults. The guidelines help document indications and contraindications to determine eligibility. A triage nurse queries the EHR and learns that the patient is over 16 years of age, has an 281239006 |Acute asthma|, and one or more episodes of 56018004 | Wheezing| which necessitated 1366004 | Breathing treatment|. The CDSS then triggers an alert to follow the pathway’s medical directives, which are carried out by a Respiratory Therapist (RT). The directives, in this case of 370218001 | Mild asthma|, include 47101004 | Heart rate monitoring|, establishing various baseline 251880004 | Respiratory measurements|, and administration of a 372580007 | Bronchodilator| and 374072009 | Prednisone 50mg tablet| . The RT then notifies the attending physician who fills out and signs discharge instructions which a nurse then reviews with the patient. The desired clinical outcomes of this pathway include improved adherence to evidence-based management and improved patient outcomes such as reduced number of hospitalizations and lower ED return rates.
Nursing Interventions
Research has provided evidence to show that patients receiving 40617009 | Mechanical ventilation| are at high risk for |Pneumonia| : |due to| = |Aspiration|. recommend 423171007 | Elevation of head of bed| from 30° to 45°, if not contraindicated, to reduce risk of 233604007 | Pneumonia|. A nursing supervisor uses a dashboard-like tool in an ICU to monitor patients in her ward. Patients who meet the criteria for risk of 422588002 | Aspiration pneumonia| are automatically flagged in the system using CDS logic so that the appropriate action may be initiated by nursing staff in the ward. Once the angle of the patient's bed is adjusted, the system is dynamically updated and the flag is removed.
SNOMED CT Features
This section contains a brief summary of key SNOMED CT features and explains how they may be useful in CDSSs.
Concepts
SNOMED CT concepts are used to represent clinical meanings. Every concept in SNOMED CT is uniquely identified by a distinct SNOMED CT Concept Identifier. For example, 195967001 is the concept identifier for the concept 195967001 | Asthma| .
SNOMED CT concepts play an important role in CDS by enabling actions to be triggered based on the meaning of data recorded in the patient records.
Descriptions
SNOMED CT descriptions provide the human-readable terms associated with SNOMED CT concepts. A concept may have one or more descriptions, which act as synonyms for the same clinical meaning. This is also how SNOMED CT supports different dialects and languages.
SNOMED CT descriptions allow common CDS rules to be consistently applied across patient records recorded using different synonyms, dialects and languages.
Relationships
SNOMED CT relationships link concepts together to formally define the meaning of each concept. For example, one type of relationship is the 116680003 | is a| relationship which relates a concept to a parent or supertype. These 116680003 | is a| relationships define the subtype hierarchy of SNOMED CT concepts.
For example, the concepts 53084003 | Bacterial pneumonia| and 75570004 | Viral pneumonia| both have an 116680003 | is a| relationship to 233604007 |Pneumonia| which has an 116680003 | is a| relationship to the more general concept 128601007 |Infectious disease of lung|. Subtype relationships can be used by CDS rules to refer to codes in an EHR that are any specific type of a relevant clinical concept.
Additional attribute relationships help to define the meaning of a concept. For example, the concept 75570004 | Viral pneumonia| has a 246075003 | Causative agent| relationship to the concept 49872002 | Virus| and a 363698007 | Finding site| relationship to the concept 113255004 |Structure of parenchyma of lung|.
Attribute relationships can be used by CDS rules to refer to codes recorded in an EHR that have a specific meaningful relationship with a concept of interest.
Concept Model
The SNOMED CT concept model is a set of rules that govern the ways in which SNOMED CT concepts are permitted to be modeled using relationships to other concepts. It defines the types of relationships that may be used on each type of concepts, and the permitted values for each relationship type. The Machine Readable Concept Model (MRCM) represents the rules in the SNOMED CT concept model in a form that can be read by a computer and applied to test that concept definitions and expressions comply with these rules.
The SNOMED CT concept model plays an important role in CDS by providing the rules by which the clinical meaning of SNOMED CT encoded health records can be queried. The MRCM makes it possible to process these rules in a machine-processable way.
Expressions
SNOMED CT provides a mechanism which enables clinical phrases to be represented by a computable expression, when a single concept does not capture the necessary level of detail. For example, the following expression represents a right hip:
182201002 |Hip joint|:
272741003 |Laterality| = 4028007 |Right|
SNOMED CT expressions enable additional clinical meanings to be captured in a health record, without requiring the terminology to include countless combinations and permutations of precoordinated concepts.
SNOMED CT expressions facilitate CDS over an expanded set of clinical meanings that extends beyond individual concepts. For more information about expressions, please refer to the SNOMED CT Compositional Grammar - Specification and Guide.
Reference Sets
SNOMED CT reference sets are a flexible and standardized approach used to support a variety of requirements for the customization and enhancement of SNOMED CT. These include the representation of subsets, language preferences for use of particular terms, mapping from or to other code systems, and ordered lists.
Reference sets may be used in the following aspects of CDS:
Representing subsets of SNOMED CT concepts that may trigger a CDS action
Representing non-standard aggregations of concepts for specific CDS use cases
Defining language or dialect specific sets of descriptions over which term searches can be performed
For more information about reference sets, please refer to the SNOMED CT Reference Set Guide.
Description Logic Features
Description Logic (DL) is a family of formal knowledge representation languages and used as the formal foundation of meaning in SNOMED CT. The way that concepts have been modeled in SNOMED CT permits them to be represented using Description Logic. DL helps computers to make useful inferences about concepts, and to classify SNOMED CT using a DL reasoner. Description Logic also helps by testing expressions for subsumption and equivalence.
The logical inferences supported by DL can be useful when executing CDS rules. For example, when a CDS rule requires an action to be performed when the patient has any type of 195967001 | Asthma| , a DL reasoner may be used to determine that 281239006 |Acute asthma| and 427603009 | Intermittent asthma| are both types of 195967001 | Asthma| and should therefore both trigger the action to be performed.
Abbreviations
The following table contains the definition of abbreviations used in this document. Please refer to the SNOMED Glossary for additional definitions.
CDS
Clinical decision support, which is defined as a service that assists clinicians, caregivers, or patients in healthcare and/or treatment decisions.
Notes
A clinical decision support system is a computer system or software application designed to assist clinicians, caregivers, or patients in healthcare and/or treatment decisions.
CDSS
Clinical decision support system, which is defined as a computer system or software application designed to assist clinicians, caregivers, or patients in healthcare and/or treatment decisions.
Notes
Typically a clinical decision support system responds to triggers, such as specific signs or symptoms, diagnoses, laboratory test results, medication selections, or complex combinations of such triggers. The system then provides information or recommendations relevant to the specific patient.
EHR
Electronic health record, which is defined as a systematic collection of health information about individual patients or populations that is stored in digital form.
KB
Knowledge base, which is defined as the underlying set of facts, assumptions, and rules which a computer system has available to answer a question or solve a problem.
UI
User interface, which is defined as the way in which a software application presents itself to a user.
NLP
Natural language processing, which is defined as a service in which a computer system converts human-readable text and/or spoken language to formal representations of information.
POC
Point of care, which is defined as the time and location at which healthcare professionals deliver healthcare products and services to patients.
Last updated