CDSS Notes from Lecture: Clinical Decision Support Systems

CDSS Overview

  • Speaker: Siserajithri Pulege, academic staff at the UQ Center for Online Health. Focus of the presentation: clinical decision support systems (CDSS) and their role in modern healthcare delivery.
  • Goal: examine main types of CDSS and the pros and cons of this innovation.

Clinical decision making and the role of CDSS

  • Clinician task: make treatment decisions regularly at every stage of the patient journey (patient pathway).
  • In routine practice, complex decisions are sometimes made rapidly using intuition or common sense, which can lead to suboptimal outcomes.
  • Clinical decision making is a complex process involving:
    • information processing
    • evaluation of evidence
    • application of relevant knowledge to select appropriate interventions for high-quality care and reduced risk of patient harm.
  • Challenges in decision making include:
    • increasing demand on clinicians
    • expanding clinical knowledge
    • timeliness and comprehensiveness of available data
  • Clinical decision making process may include at least three integrated phases: 33 phases
    • 1) Diagnosis
    • 2) Assessment of severity
    • 3) Management

History and types of CDSS

  • Health and medical fields have long developed decision support to aid clinicians.
  • Early CDSS were limited to clinical guidelines.
  • With advancements in information and communication technologies, CDSS design broadened.
  • Computerised clinical decision support system (CDSS) is health information technology designed to enhance clinical decisions by providing clinicians, health staff, and patients with targeted clinical knowledge and patient-specific information.
  • Purpose: improve health care delivery by assisting clinicians in the complex decision-making process.
  • CDSS focuses on knowledge management to obtain clinical advice based on multiple factors of patient-related data.
  • Core functions include analyzing patient data and using that information to formulate diagnoses and develop treatment plans.
  • After analyzing patient information, a CDSS can:
    • make suggestions on drug doses and frequencies
    • perform drug allergy checks
    • provide guidelines and reminders regarding prescription orders (e.g., glucose testing with insulin)
  • CDSS are commonly classified as:
    • knowledge-based CDSS
    • non–knowledge-based CDSS
  • Knowledge-based CDSS:
    • use a knowledge base, apply rules via an inference engine, and display results
    • typical components: data repository, inference engine, communication mechanism
  • Non–knowledge-based CDSS:
    • rely on machine learning or statistical pattern recognition to analyze clinical data
    • example: artificial neural networks, which learn from examples (usually without explicit programming)
  • Computer-based CDSS can be traced back to the 19701970s
    • early systems had poor integration, were time-intensive, and often limited to academic settings
  • In the 19901990s, hospital environments started integrating hospital information systems that generated large amounts of electronic patient information, driving the need for CDSS capabilities within health systems
  • Presently, CDSS are integrated with EHRs and CPOE systems; accessible via desktop, tablet, mobile devices, and even biometric monitoring and wearable health technology

Modern architecture, data sources, and deployment

  • CDSS can be administered through a variety of devices, including:
    • desktop computers
    • tablets
    • smartphones
    • wearable health technology and biometric monitoring devices
  • Key integration: CDSS are often integrated with Electronic Health Records (EHR) and Computerised Provider Order Entry (CPOE) systems
  • Role of CDSS with data sources:
    • leverage vast patient data in EHRs to provide targeted guidance
    • alert clinicians to potential issues and trigger predefined orders or clinical pathways

Applications and representative use cases

  • Medication management (to reduce medication errors, which are costly and harmful):
    • Drug allergy checking: compares ordered medications to patient allergies and generates alerts for potential reactions
    • Patient-specific dosing parameters: generates recommended dosing and frequency based on patient data
    • Alerts for exceeding dosing limits
    • Duplicate therapy detection: identifies when two or more drugs with the same active ingredient are prescribed concurrently; may reduce overdose and adverse effects; modeled as 22 drugs or more with the same active ingredient could trigger an alert
    • Interaction checks: considers interactions with foods/beverages, herbs, and conditions such as pregnancy or lactation
    • Many CPOE systems now include drug safety components for dose checks and duplicate therapy alerts
  • Diagnostic decision support (disease identification):
    • CDSS compare patient information with a knowledge base to generate a list of possible diagnoses
    • Often integrated within CPOE systems
    • May help decrease treatment costs by suggesting cheaper drug alternatives or identifying duplications in tests
  • Imaging and radiology decision support:
    • Knowledge-based imaging CDSS aid radiologists in selecting the most appropriate test and remind of best-practice guidelines
    • Non–knowledge-based CDS for imaging leverage AI and deep learning to enhance imaging decision making and precision radiology
    • AI examples: tumor detection, medical imaging interpretation, diabetic retinopathy diagnosis
    • Notable players: IBM Watson Health, DeepMind, Google, etc.
  • Laboratory testing and interpretation:
    • CDSS extend the utility of lab-based tests and can help avoid riskier or more invasive diagnostics
  • EHR-linked CDS and clinical pathways:
    • EHR serves as a repository for data and images that can be searched, reviewed, and compared
    • Clinical parameters in the EHR (e.g., vital signs, test results) can trigger alerts or predetermined orders, diagnostic or therapeutic bundles, or clinical pathways
    • Providing CDS at the time of diagnostic test ordering can reduce inappropriate testing
  • Overall aim of CDS across domains:
    • reduce variation in clinical practice by guiding clinicians toward best practices as determined by expert panels, professional associations, and healthcare institutions
  • Other notes:
    • Positive evidence exists for CDSs, especially those embedded in CPOE and EHR implementations
    • There is growing interest in AI-based CDS for imaging and precision radiology and broader diagnostic support

Evidence, benefits, and economic impact

  • CDSs have been shown to save hospitals substantial costs by reducing unnecessary or excessive medical testing; studies report savings in the order of exthundredsofthousandsofdollarsperyearext{hundreds of thousands of dollars per year} due to alerts signaling unnecessary or duplicative tests
  • CDSs can improve adherence to clinical guidelines and best practices
  • Imaging and diagnostic CDS can enhance decision accuracy and efficiency, particularly when paired with AI capabilities
  • EHR-linked CDS can reduce inappropriate diagnostic testing and standardize care according to expert recommendations

Benefits, workflow integration, and user considerations

  • CDSS can augment clinician decision making and support delivery of high-quality care; many applications have robust evidence, particularly when integrated with EHR/CPOE
  • When well designed, CDSs support timely, guideline-concordant decisions and can lower costs and improve patient safety

Shortcomings, risks, and challenges

  • Workflow disruption: standalone CDSS can interrupt clinician workflow; integration with existing workflows is critical
  • Initial misperceptions: early CDSS could create the impression that order verification is unnecessary, leading to safety risks
  • Over-reliance on CDS: clinicians may become over-dependent on CDS and lose skills or become complacent; requires balanced human oversight
  • Technological proficiency gaps: users may lack necessary training or comfort with the system
  • Maintenance and lifecycle management: ongoing updates to knowledge bases, inference rules, and data sources are necessary but can be neglected
  • Interoperability issues: many CDSS exist as standalone systems or cannot communicate effectively with other systems; interoperability remains a challenge
  • Standards and integration: HL7 and FHIR are central interoperability standards being used to improve system communication across vendors and health domains
  • Ongoing need for speed, ease of deployment, and affordability to maximize benefits across health systems

Interoperability standards and future directions

  • HL7 (Health Level Seven) and FHIR (Fast Healthcare Interoperability Resources) are key standards for exchanging health information and enabling CDSS interoperability
  • These standards support data sharing between EHRs, CDS tools, and other health IT systems, improving integration and reducing silos
  • Ongoing development and adoption of interoperability standards is essential to maximize the impact of CDSS across diverse health settings
  • Future directions include greater integration, faster processing, broader deployment, and more affordable CDS solutions, with continued attention to ensuring safety and preventing over-reliance

Practical considerations and governance

  • Successful CDSS implementation requires alignment with clinical workflows and processes
  • Stakeholder engagement (clinicians, IT, administration) is essential
  • Training and ongoing user support are critical to adoption and effective use
  • Data quality, privacy, and security must be prioritized
  • Continuous monitoring for safety issues and unintended consequences is necessary

Ethical, philosophical, and real-world implications

  • Trust and transparency: clinicians should understand how CDS recommendations are generated
  • Human oversight: CDS should augment, not replace, clinician judgment
  • Privacy and data security: safeguarding patient data used by CDS and AI components
  • Fairness and bias: ensuring AI-based CDS do not perpetuate or exacerbate health disparities
  • Accountability: clear delineation of responsibility when CDS recommendations impact patient outcomes

Conclusion and takeaways

  • CDSS augment healthcare providers by supporting a range of decisions and patient care tasks; they are increasingly integrated into everyday practice
  • Evidence is strongest for CDSs embedded in CPOE and EHR workflows; other areas show promise with ongoing development
  • The field continues to evolve with interoperability improvements, speed gains, easier deployment, and affordability improvements
  • Vigilance is required to mitigate downsides, including workflow disruption, over-reliance, maintenance gaps, and interoperability challenges

Key terms and concepts (glossary)

  • CDSS: Clinical Decision Support System
  • CPOE: Computerised Provider Order Entry
  • EHR: Electronic Health Record
  • HL7: Health Level Seven International (standards for health information exchange)
  • FHIR: Fast Healthcare Interoperability Resources
  • Knowledge-based CDSS: uses a knowledge base, inference engine, and data repository
  • Non–knowledge-based CDSS: relies on machine learning and statistical pattern recognition
  • AI/Deep learning in imaging: advanced algorithms for pattern recognition and interpretation in radiology and pathology
  • Diagnostic decision support: CDS focused on generating possible diagnoses based on patient data
  • Drug safety components: alerts for allergies, interactions, dosing limits, and duplicate therapy
  • Clinical guidelines: evidence-based practices used to guide CDS rules and recommendations
  • Precision radiology: imaging approaches enhanced by AI and data-driven decision support