Real Time Clinical Decision Support System - PowerPoint PPT Presentation

real time clinical decision support system n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Real Time Clinical Decision Support System PowerPoint Presentation
Download Presentation
Real Time Clinical Decision Support System

play fullscreen
1 / 95
Real Time Clinical Decision Support System
159 Views
Download Presentation
verena
Download Presentation

Real Time Clinical Decision Support System

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Real Time Clinical Decision Support System 林雪淳

  2. Challenge of Modern Clinical Medicine • Dilemma • Patients – increase • Clinical Time – decrease • Quality of Health Care – debase • Requirement • Patient – comfort • Clinical Time – efficiency • Quality of Health Care – improvement • Ethics • Patient Right and Privacy

  3. Strategy • RTCDSS – Real Time Clinical Decision Support System • QOL:Quality of Life • Infometrix: Information+Psychometrics Online Measurement for QOL • CDW: Clinical Data Warehouse • OLAP: Online Analytical Process • CDSS: Clinical Decision Support System • Internet: Web Services and Real-Time Technology • Foundation – • EBM: Evidence-based Medicine • HIS+ES: Hospital Information System + Expert System

  4. Scope • Operating Interface • Accessibledesign of human-computer interface • Patient- and Clinician-oriented interface • Patient-to-Clinician (P2C) communication • Flexible and Expandable modules • Web-based auto data transportation • Proposed Functionality • Traceable clinical markers for chronic diseases • Instantaneous Patient Clinical Record (PCR) • Reliable Patient Reported Outcome (PRO) • Quantitative medical informatics • Analytical online diagram

  5. INTRODUCTION

  6. Clinical Treatment • Routine treatment procedure • Physicians take much time to study patients’ clinical records (PCRs) prior to explain abstruse clinical markers to patients in clinics. • For chronic and traceable diseases, they need to refer patients’ quality of life (QOL) and their patient-reported outcomes (PROs) for prescribing the proper therapies. • Modern hospital information systems • Information technology (IT) and Web-based facility have become the major backbone of the modern hospital information systems (HIS) • Traceable clinical markers for the chronic diseases • Clinical decision support system (CDSS) with patient- and clinician-oriented interface as well as patient-to-clinician (P2C) communication.

  7. Major cancer therapy • Ability – Find and destroy tumors • Disability – Bring numbers of side effect • It implies that QOL is deeply impacted by uncertainty and after-effect due to treatment of oncology clinic. • Traps of clinical and health care • Incorrect judgments when patients embarrass on answering private questions or hiding actual conditions • Inconvincible diagnosis by computer-based CDSS (or HIS) on clinician performance and patient outcome • HIS difficulty and possible solution • Function-oriented system – uneasy to create a universal system for varied clinical requirements • Flexible platform – available to build up expandable components for specified clinical purpose by customized rules

  8. Quality of Life • Assessment of Quality of Life • EORTC: provides QOL questionnaires to highlight physicians’ awareness of patients’ status and greatly facilitate physician-patient communication • Infometrics: combines information and psychometrics technology for measurement, statistical modeling, informatics and practice, in palliative care with computerized procedure by interactive assessment system • Concept of instant QOL+PRO+PCR • Measurement and management of PROs – Infometrics technique can assist clinicians to more precisely recognize actual response of patients and improve the quality of care with instant process and real time outcomes statistics. • Graphical diagrams – clinicians thus can convince patients by presenting PRO instantly with other PCR.

  9. Needs of Clinical Practice • Problems to improve quality of clinical treatments • Clinicians may take several hours, or even a couple of days, to review PCRs but only have a few minutes to explain their opinions to patients • Patients typically find difficulty to understand their condition since clinicians may only explain the disease adequately using written descriptions • The CDSS is computerized, but it may not have online capability in many clinics • Real-time analysis is not supported by many commercial computational tools • Need: clinical data tracking  real-time decision making • The flexible Web-based CDSS with online evidence-based medicine (EBM) progress is a growing trend in advanced clinical care

  10. Clinical Decision Support • Enhanced CDSS • Analytical tools – assist clinicians in estimating the relative pretreatment parameters and for tracking the proper diagnostic guidelines on visualized interfaces • Clinician-oriented interface – improve accuracy and efficiency of decision support • RTCDSS – support interactive diagrammed interface with real-time online analysis to efficiently evaluate instant informatics and to make clinical decisions • Expandability and Feasibility – follow the model to take chronic diseases with traceable markers • Clinical Guideline • Work with medical evidences and recommend appropriate treatments by graphical interface • Allow users to traverse the algorithm by flowcharts in an interactive fashion due to Web-interfaced process

  11. Clinical Benefits • Rapid knowledge acquisitions, shareable guideline models, and robust information systems while evaluating its impacts on outcomes • The electronic guidelines improve decision quality and physician-patients interaction significantly • Encountered Obstacles • Management of workflow integration would be the most difficult tasks; ex, integrating a new RTCDSS with the PROs, PCRs, CDSS, and interactive guidelines into the legacy HIS. • The framework for interactive clinical guidelines should consider readiness of clinicians for practice, barriers to change as experienced by clinicians, and the target level of interventions

  12. DEVELOPMENT OF RTCDSS INFRASTRCTURE Model-view-controller model Object relation mapping model Clinical data warehouse model Web services model Online analytical process model Asynchronous JavaScript XML-HttpRequest model

  13. Model-view-controller model • Design patterns in software engineering • First made byintroducing 23 patterns related to creational, structural and behavioral models • For software design to progress recurrent elements (Gamma et al., 1994). • The model-view-controller (MVC) pattern • Hybrids “strategy,” “observer,” and “composite” patterns • Divides system responsibilities into the model, the view, and the controller • Software framework with MVC paradigm • Open-source framework such as “Strut,” “Spring,” “Hibernate” for development. • Use polling for its input control to solve the problems on consuming computation resources when the user is not interacting with the interface and avoid unnecessary performance loss.

  14. MVC design patterns • Model – maintains program data and logic • View – provides a visual presentation of the model • Controller – processes user input and makes modifications to the model • Modelized architecture • Expandable and reusable components of the Web-based platform • Efficient and flexible collaboration for (a) instantaneous disease evaluation, (b) risk analysis, and (c) treatment guidance • Web assessment for acquiring PROs from online infometrics system and adapting PCRswith the legacy HIS in hospital.

  15. Web-based MVC – Modelized Architecture

  16. Web-based architecture with model, view, controller components

  17. 4-Tier MVC-based RTCDSS for Clinics

  18. Components of MVC-based RTCDSS for Clinics • Four-Tier Components • Presentation – interactive guideline, real-time diagram  Browser Windows • Management – privilege administration, informatics management  Application Server • Database – data filtering  clinical data warehouse • Analysis – data analysis tools adapting  SAS, SPSS, MATLAB • Three-Level Users • Patients / Healthcare Staff  Input data • Clinicians / Decision Maker  Decide data • Engineers / Administrator  Analyze data

  19. Models – disease evaluation, risk analysis, treatment guidance, and data processing • First 3 models for clinical data computation and last one for IT modules • Disease evaluation: retrieve clinical variables, calculate pretreatment parameters, and evaluate PROs and PCRs. • Risk analysis: analyze clinical variables and parameters, identify risk indicators and criteria, and so on. • Guidance criteria: enables the generation of evidence-based diagrams, online guidance and decision support. • Data processing: supports IT-related modules such as clinical data conversion, database connection, and graphical display.

  20. Views – OLAP portal, EBM informatics, management interface, analysis view • Presentation: patient- and clinician-oriented interfaces direct OLAP portal and EBM informatics. • Management: management interface provides security administration. • Analysis and database: analysis view displays all clinical data. • Controllers – Data flow transformation, data input validation, privilege control, role identification, heterogeneous data transaction • Presentation: “data flow transformation” and “data input validation” control online inquiries • Management: “privilege control” and “role identification” secure system maintenance • Analysis and Database: “heterogeneous data transaction” coordinate clinical data

  21. Object Relation Mapping • ORM – a programming technique that converts data between incompatible type systems in relational databases and object-oriented programming languages. • session interface – conducts lightweight instances in safe as the necessary data are requested on the web tier all the time; • session factory – share many application objects and cache scripted database transaction and other mapping metadata for converting data. • configuration interface – configures the location of mapping documents and specific properties for data retrieval • transaction interface – keep applications portable between different execution environments. • query interface – performs instances to control data queries against the database • criteria interface – executes object-oriented criteria queries.

  22. ORM Data Flow

  23. Clinical Data Warehouse • Data warehouse – • an integrated, subject-oriented, time-variant and non-volatile database • provides support for decision making • builds up an integral database for historical data repository with lack of systematic arrangement • allows complex queries and analyses on the information without slowing down the operating system • unified by the extract-transform-load (ETL) procedure into database through extraction, consolidation, filtering, transformation, cleansing, conversion and aggregation • Clinic application – • integrate practical PROs and PCRs with a standard procedure from different hospital databases into the knowledge bank for advanced analysis

  24. ETL process

  25. Web Services • Web Services (WS) • an interface for describing a collection of operations that are network accessible through standardized XML messaging • W3C definition – “a software system designed to support interoperable machine to machine interaction over a network.” • Standards • WSDL (Web service definition language) - translate metadata • SOAP (simple object access protocol) - transport data • UDDI (universal description, discovery, and integration) - search information

  26. Parsing flow for web service data with XML schema

  27. Online Analytical Process • OLAP – Online Analytical Process • Keep complex query behind data mining for knowledge bank • Leave simple data transaction through dynamic views in data warehouse • OLAP in RTCDSS • Cross over the web server and database • Lead online computation within the RTCDSS • Manage and analyze infometrix and clinical data • PCR queries integrated with heterogeneous databases are primarily progressed while accessing the database server • Risk evaluations embedded within online session logs are efficiently retrieved as connecting the web server

  28. Framework withWS,MVC,ORMbeyondOLAP

  29. AJAX • Asynchronous JavaScript XML-HttpRequest – • The AJAX technique is widely applied for online interactive interface to grab instant information and to avoid lag in transportation of client-server data • Store transient data (e.g. images) at client sites to reduce redundant data query with database sites and enhance interactive patient- and clinician-oriented interface. • Process numerous data queries between database and web server • If the client site keeps sessions at online status, the browser is calling JavaScriptTM and restoring data. • Once the session needs reconnection or updating, the client-server communication is activating. • Adjust data interaction performance as adopting light-weight data like QOL questionnaires, risk evaluations or guideline indexes. • The method doesn’t need to request database all the time but load into browser’s temporary container at client site.

  30. AJAX data transportation

  31. Three stage RTCDSS with CDW Integration

  32. DESIGN OF PATIENT AND CLINICIAN ORIENTED INTERFACES Patient-oriented interface Clinician-oriented interface Integration design

  33. Interface and Infrastructure • Two types of interfaces • Patient-oriented and clinician-oriented interfaces • Process PRO with clinical infometrics and analyzing PCRs upon the EBM. • Five-layer infrastructure • “Acquisition” – acquire patient-reported outcomes • “Presentation” – present online clinical diagraph • “Management” – manage clinical information • “Analysis” – analyze patient’s clinical records • “Database” – coalesce diverse clinical databases • Practice: CIPC project in CMUH

  34. Patient-oriented Interface • Infometrics for Quality of Life • Infometrics = Information + Psychometrics • Quality of Life (QOL) • WHO – individuals’ perceptions of their position in life in the context of the culture and value systems in which they live, and in relation to their goals, expectations, standards, and concerns. • EORTC – QOL assessments in cancer clinical trials to provide a more accurate evaluation of the well-being of individuals or groups of patients and of the benefits and side-effects that may result from medical intervention. • EORTC C30 – 30 questionnaires for cancer • EORTC PR25 – 25 questionnaires for prostate cancer

  35. EORTC QOL Assessment • C30 – 30-item cancer-specific questionnaire that has often been used for patients with head and neck cancer • 5 functional scales (physical, role, cognitive, emotional, and social) with 9 multi-items • 3 symptom scales (fatigue, pain, and nausea and vomiting) • 1 global health QOL scale and 6 single items • PR25 – 25-item questionnaire for use among patients with localized and metastatic prostate cancer • Urinary symptoms (9 items) • Bowel symptoms (4 items) • Treatment-related symptoms (6 items) • Sexual functioning (6 items)

  36. CIPC – Clinical Infometrix for Prostate Cancer CIPC Scope QOL is an important healthcare index but patients probably conceal the truth because of private manners. The traditional paper-based QOL assessment usually causes reading difficulty for patients because of improper font size and print space. Most of prostate caner patients are seniors who initially might not know how to click mouse-button or scroll the browser to navigate the computer. The infometrics module of CIPC system is designed for patient orientation through sufficient accessibility and accompanies QLQ with instant PRO analysis and evaluation. Clinical Implementation for Prostate Cancer

  37. CIPC for patients • The fonts of questionnaires are enlarged for elderly patients who have poor eyesight. • the selection buttons are displayed on a touch screen for patients who are not familiar with using computer mouse. • The Web-page design is simplified by one-touch action per question before the users are well trained. • A multimedia function played with head phones is optionally provided for low education level patients who could read questions with limited literacy. • CIPC procedure • Patients are arranged privately in a consulting room to complete the questionnaires while waiting for the clinicians. • Clinicians could immediately evaluate the real time reports with online analysis according to automatic computation and statistical models.

  38. Clinic progress implemented with clinical infometrics system

  39. ~Waiting for Clinic~ ~Clinic Time~

  40. CIPC mechanism • Clinicians and researchers can immediately access infometrix data after patients completed the questionnaires. • The clinician can make cross compare overall treatment information with instant expert opinions for advanced communicate with patients. • CIPC network infrastructure • Network of CIPC needs to link hospital and campus networks, but under hospital’s information security policy, to collaborate tiers of database, analysis, management, presentation, and acquisition for clinical and research workflows. • The infrastructure bridges both networks of clinics and campus through the firewall to routinely backup clinical data and maintain the CIPC system.

  41. Network infrastructure of the CIPC system with RTCDSS components

  42. Five-tier infrastructure within CIPC network • Database tier supports clinical and infometrix data warehouse • Analysis tier assists analysts analyzing data and feeds back statistical results as resource of the knowledge bank • Management tier is the control center for administrating data flow throughout the entire system • Presentation tier presents real-time functions for online decision support and interactive guideline on a friendly interface for P2C communication • Acquisition tier becomes the data collector to execute online QOL assessment with accessibility interface.

  43. Clinician-oriented interface • Scope • Evaluate pretreatment parameters for clinical evidences • Guide clinicians to concurrently collect and analyze specific clinical markers with instant diagrams in CIPC for prostate cancer patients. • The CIPC system is proposed in urology clinic for reflecting relationship between QOL and pretreatment parameters such as PSA, clinical classification stage, and Gleason score, etc. • Function • Open source frameworks • Graphical diagrams for PROs and PCRs • Combine PCRs and biomarkers from diverse database through networks • Guidelines for decision support

  44. Prostate cancer treatment • Suspicion of prostate cancer resulting in prostatic biopsy is most often raised by abnormalities found on digital rectal examination (DRE). • PSA has evolved for the detection, staging, and monitoring of men diagnosed with prostate cancer since its discovery in 1979. • The four-stage TNM system indicates how far the cancer has spread for defining prognosis and selecting therapies: the size of the tumor (T), the number of involved lymph nodes (N), and the presence of any other metastases (M) • The Gleason grade is based on a low-magnification microscopic description of the architecture of the cancer and is the most commonly used classification scheme for the histological grading of prostate.

  45. PSA Level • PSA Density – density of prostate-specific antigen • Most prostate cancer arises as clinically nonpalpable disease with PSA between 2.5 and 10 ng/mL • PSAV – PSA Velocity • Linear regression of logarithm for PSA records • PSADT – PSA Doubling Time • The relationship of two arbitrary PSAs measured at the time T1 and T2 with respect to the doubling time TD is formulated when ln(2*P1) is estimated.

  46. TNM stage • primary tumor (T): • T1 stage presents tumor, but not detectable clinically or with imaging; • T2, the tumor can be palpated on examination, but has not spread outside the prostate; • T3, the tumor has spread through the prostatic capsule; • T4, the tumor has invaded other nearby structures. • regional lymph nodes (N): • N0, there has been no spread to the regional lymph nodes • N1, there has been spread • distance metastasis (M): • M0, there is no distant metastasis • M1, distant metastasis is found

  47. Gleason score • The predominant pattern that occupies the largest area of the specimen is given a grade between 1 and 5. • then added to the grade assigned to the second most dominant pattern • Gleason sum can be arranged between 2 and 10. • This system describes tumors as "well", "moderately“, and "poorly" differentiated based on Gleason score of 2-4, 5-6, and 7-10, respectively.

  48. Kaplan Meier survival estimation • which is known as the product limit estimator, estimates the survival function from life-time data • Let S(t) be the probability that an item from a given group of size N will have a lifetime exceeding t. • ni is the number “at risk” just prior to time ti, and di, the number of deaths at time ti, where i = 1, 2, …, N. • ti is equal or less than ti+1 • the intervals between each time typically will not be uniform. • When there is no censoring, ni is the number of survivors just prior to time ti. • With censoring, ni is the number of survivors less the number of losses. • It is only those surviving cases that are still being observed that are “at risk” of an observed death.

  49. Cox Proportional Hazard Model • h0(t) is the baseline hazard involving t but not X’s • X denotes a collection of p explanatory variables X1, X2, …, Xp • the model is nonparametric because h0(t) is unspecified. • For PSA variables correlation in prostate cancer treatment, these variables may include age, race, initial PSA, PSAV, PSAD, clinical stage, treatment, and so on.

  50. K-M Survival vs. CoxPH Model Curves for the same data set