Decision Support Systems (DSS): W13 Lecture II PDF

Decision Support Systems IB RAH IM AD H A M B IN TA IB , P H D IN TE R N AT ION A L IS L A M IC U N I VE R S IT Y M AL AY S I A (I IU M ) Outline of lecture  Introduction to AI  Brief history of AI  Foundational concepts of AI  Value propositions of AI  Definition of DSS  Applic...

Decision Support Systems IB RAH IM AD H A M B IN TA IB , P H D IN TE R N AT ION A L IS L A M IC U N I VE R S IT Y M AL AY S I A (I IU M ) Outline of lecture  Introduction to AI  Brief history of AI  Foundational concepts of AI  Value propositions of AI  Definition of DSS  Applications of DSS  Benefits of DSS  Barriers to DSS  DSS in Malaysia  Legal and regulatory considerations Introduction to AI  A computer program or system that can make decision on its own either because of training or because its ability to learn on its own  AI has changed from being just faster than humans, to making decision using rules (expert driven) and later learning (data driven) Brief History of AI  Advances in AI due to engineering, biology, psychology, mathematics, philosophy, linguistics, etc  In 1950, Alan Turing raised the question of whether a computer can learn  1956, Dartmouth AI Workshop was held Brief History of AI  Computers continue to advance, but do not demonstrate strategic thinking  In 1997, IBM’s Deep Blue defeated Gary Kasparov at chess using rules and algorithms developers provided  In 2017, Google’s DeepMind created Alpha Go Zero which defeated grandmasters of Go using its own strategies it developed from learning https://www.youtube.com/watch?v=PeMlggyqz0Y Foundational Concepts of AI  Machine Learning (ML): An automated system able to process large volumes of data and extract meaningful information from it (data mining) as well as to use this information to address practical problems (decision support). Foundational Concepts of AI  Knowledge-Based Systems: Use expert knowledge (e.g., guidelines, rules) to offer predefined solutions. Best for established scenarios but prone to false alarms and less effective in novel situations.  Data-Driven Systems: Leverage AI and ML to analyze large datasets, identify trends, and support diagnoses. Highly adaptive but struggle with unknown scenarios and lack flexibility.  Model-Based Systems: Utilize simulation and mathematical models to predict outcomes and optimize decisions. Ideal for planning and scenario analysis but limited to predefined models. Foundational Concepts of AI  Classifiers: Processes that map input data into categories or classes, also referred to as predictions. Classifiers are trained on a data set for which the proper classification is known, i.e., labelled, so that new and unlabelled data can be correctly categorized Decision Tree Foundational Concepts of AI  Artificial Neural Networks: machine learning algorithm modelled on biological nervous systems  Comprises of interconnected neurons that receives input, process input, and produce output in the form of a prediction https://www.youtube.com/watch?v=-8se4mWn058 https://www.youtube.com/watch?v=6M5VXKLf4D4 Foundational Concepts of AI  Natural Language Processing (NLP): Automated language analysis intended to parse unstructured text to respond to queries or otherwise extracting data in analyzable form  Image Analysis: The process of extracting meaningful information specifically from images as opposed to numeric, categorical, or text data.  Speech Analysis: Similar to image analysis, however, focused on extracting meaningful diagnostic and prognostic insights from patterns discernible in recorded speech Value: Information Synthesis  Too much information to handle  amount of data on patients  data complexity  amount of medical literature Value: Augmenting Human Performance  Rare diseases or unusual presentations may be difficult to determine  Time pressure for decision making  Possibility of interaction effects Value: Surveillance  Public health surveillance  Diseases spread quickly due to travel  Oversight such as post-market surveillance  How medicine is practiced, fraud, etc WHAT IS CLINICAL DECISION SUPPORT SYSTEM (CDSS)? 25 “…ANY COMPUTER PROGRAM DESIGNED TO HELP HEALTHCARE PROFESSIONALS TO MAKE CLINICAL DECISIONS.” 26 27 Basis for CDSS – From the earliest days of computing, people have recognised that computers can help sift through vast collections of possible decisions and symptoms. – Healthcare practice involves decision-making; hence, computers can have direct or tangential effect on quality of Applications Patient Safety: Drug-Drug Interaction (DDI) Alerts: Prevent harmful medication combinations. Barcode Medication Administration: Ensures accurate drug dispensing and administration. 28 Applications 29 Diagnostics: Diagnostic Decision Support Systems (DDSS): E.g., Peripheral neuropathy diagnosis with 93% accuracy. Imaging Assistance: Guides appropriate imaging requests (e.g., lumbar MRI for back pain). Laboratory Support: Combines test results for non-invasive liver fibrosis diagnostics.  Applications Cost Optimization: Suggests cost-effective medication alternatives. Reduces duplicate tests and inpatient stays. Patient-Facing Systems: Personal Health Records (PHR): Tools for self-management Wearable Technology Integration: Monitors glucose for diabetes management. 30  APPLICATIONS AI and Machine Learning: Deep Learning in Imaging: Detects diabetic retinopathy and tumors with expert-level accuracy. Automated Tumor Grading: Assists in cancer diagnosis and treatment planning. 31 Benefits of CDSS Diagnosis CDSS supports accurate and timely diagnosis by analyzing clinical evidence and providing suggestions to healthcare providers. Benefits: Reduces human errors in diagnostic decision-making. Prevents late diagnoses by automating test result analysis. Assists practitioners in interpreting medical images and pathology results. Offers tailored diagnostic suggestions based on patient data. 37 BENEFITS OF CDSS Prevention CDSS helps in the early detection and prevention of diseases, aiming to reduce the severity and progression of conditions. Benefits:  Increases screening rates by automating and guiding prevention protocols.  Facilitates early-stage diagnosis, reducing long-term healthcare costs and complications.  Mitigates disease severity through predictive algorithms and adherence to clinical practice guidelines. 38 Benefits of CDSS 39 Management and Planning CDSS improves operational efficiency and resource allocation in healthcare settings.  Benefits:  Streamlines clinical workflows, saving time and reducing costs.  Enhances the management of hospital resources, such as beds, surgical rooms, and equipment.  Improves documentation through automation, such as auto- selecting diagnostic codes and retrieving electronic medical records.  Assists in human resource planning (e.g., staff allocation) and logistics (e.g., drug and blood bank management).  Guides patients in making informed decisions based on real-time data. BENEFITS OF CDSS Prescriptions CDSS ensures safer and more effective medication practices by assisting with drug selection, dosing, and monitoring. Benefits:  Reduces medication errors and adverse drug interactions.  Improves patient safety by preventing inappropriate prescriptions.  Optimizes dosage management and monitors side effects.  Decreases costs by recommending alternative, cost- 40 41 Barriers to CDSS Interoperability Issues. Data Quality and Completeness Usability Concerns Ethical and Regulatory Barriers 42 Barriers to CDSS Resistance to Adoption Focus on Early Decision-making Phases Insufficient Stakeholder Involvement Evaluation and Feedback Mechanisms CDSS IN MALAYSIA  Slower adoption than EMR  Study reveals that physician in Malaysia feel threaten by CDSS because  CDSS affect natural flow of work  Their knowledge may be codified at shared with others  Uncomfortable about receiving instructions from CDSS  Misconception that CDSS will replace physicians 43 44 LEGAL AND REGULATORY CONSIDERATIONS Liability borne by those who use and those who choose not to use? Who determines whether such systems are safe for use in healthcare? End of lecture. Thank you for listening. 45