Hong Kong Population.pptx

1、Hong Kong Population7,298,600-Hong Kong Healthcare System- Dual System -nPrivateSelf-financed by patientsPublicHighly subsidized by govt2.6% GDP2.5% GDP12% inpatients88% inpatients69% outpatients31% outpatientsPublic Healthong ong Public HealthcareHospitals, Specialist Outpatient Clinics,Public Heal

2、th and Screening Servicesand General Outpatient ClinicsHospital Authority of Hong KongDoctor10%Non-PatientCare29%Nurse36%Supporting16%AH10%70,000 staffs7 Clusters42 public Hospitals28,000 beds47 Specialist Out-Patient Clinics73 General Out-Patient Clinics102.2M A&E attendances p.a.119M Specialist ou

3、tpatient clinic attendances p.a.13Development in the HA System“Green Fields”1991 Patient Administration1992 Pharmacy System1993 Lab results online1994 Radiology Information System1995 Clinical Management System (CMS) Direct Clinical documentation and order entry2000 CMS Phase II Electronic Patient R

4、ecord (ePR)2003 eSARS2004 ePR Image Distribution2006 ePR sharing with private sector2008 Move towards CMS III2009 Filmless HA2013 CMS III2016 Electronic Health Record (EHR)35,000 clinical users17,000 clinical workstations184.4M CMS transaction p.d.700,000 ePR access p.d.10M patient recordsUnique Pat

5、ient Identifier Using Hong Kong Identity Number (HKID #) HKPMI, Admissions/Discharges and AppointmentsBooking implemented across all HA hospitals andclinics HA HKPMI contains 8 million peoples recordsUniquely identify all patientsand facilitate linkingtogether episodes of careHA-wide Standards & Con

6、trolled MedicalTerminologyDiagnosis/ProceduresGCDCDFHACVTICD9CMICD 10ICPCSNOMEDLOINCICFCDARSePRCDR ePRExtensive Use of BarcodingAdultPaediatricClinical Systems - Service Coverage34 Radiology Centers700+ Old40 hospitalsAge Homes130 Pharmacies14 Psychiatric UnitClinical SystemService Coverage31 Mortua

7、ry16 A&E Units120 PathologyCenters74 GOPCs230OperationTheatre9 ObstetricsUnits48 SOPCs7/6/20162525Click to edit Master title styleClick to edit Master subtitle styleCurrent Status: ePR / CDRAcquisitionTransformation & StandardizationInformation SupportSource System200 Sybase servers51 SQL servers187

8、 Unix servers10 NT serversStaging6 Sybase servers2 Informix serversUnix file systemClinical Data RepositoryePR : 2 Informix servers (prim + sec)CDR : 2 Informix servers (prim + sec)Front-end ApplicationData Service (ePR)Data Interface (CDR)HKPMIIPASOPASAEISLISJavarpcePRCDRStaging(Database)RISreplica

9、tionMOECDDHMDLOTMSOTRSERSPowerCenterServerEDIEAIESBStaging(File System)EISGCD.SFTPPL/SQLInterfacefilesPTNTPBRCClinical Systems essential in HATo Date.Each Day.10m patient records1,000m Lab records20,000 clinician users100,000 patients300 Terra Bytes clinical datavolume11m online transactions- Peak t

10、ime: 600 Tran./ sec1.8m ePR transactions800 Terra Bytes Rad Images7/6/201628Business Drivers Quality & Safety Risk management Evidence-based medicine Clinical audit Disease surveillance Performance Performance trend analysis Benchmarking (internal & external) Strategy & Planning Fees & charges revie

11、w Clinical costing Financial modeling Data Sharing with Other Sectors Government departments Universities Private organizationsBusiness Drivers Information dimensions and complexities from hospitals to corporate from single to multiple data subjects from retrospective analysis to real timeOur Belief

12、s (1) Quality Patient Care through Improvement of ClinicalEffectiveness and Efficiency Provision of more updated relevant information fordifferent users according to their requirements forachievement of their works Alignment of expectation on performance measurementamong different key stakeholders M

13、anagementFrontlineOur Beliefs (2) Improvement of Data Quality Completeness, timeliness and quality of data would beimproved through better utilization of data Better key stakeholders engagement andempowerment Customerization and personalization of information anddisplay Transparency of information a

14、nd better communicationClinical Information Management CDARS Adhoc Query and Standard Reports ofdata in Clinical data Repository (CDR) MIPo (Management Information Portal)Overview of Source SystemsCMSI PASDx, Px, Specialist I/C, MOI/C, OTRS, ROPatient DemographicsEpisode MovementsAppointments & Atte

15、ndancesInformationO PASAEISAttendances InformationPHS/CDDHOBCISDrug DispensedObstetric InformationClinicalDataRepositoryLISRISLab test resultsRadiology ExaminationOT RecordOTRSOTMSPSYCISGRRCDARSCMS /IE6+OT Management SystemPsycisGRR FormFM ModuleFMCDARS Ad-hoc enquiry ModulesStep 1. Define basic dat

16、a scopeStep 2. Specify selection criteriaStep 3. Group criteriaStep 4. Design report layoutIn-Patient Episodes Analysis ExampleReport TypeSummary ReportDischarge DateInstitutionsCriteria01/04/2003 30/06/2003PWH, PYN, QEH, QMH, TMHCA breast patients who were treated withTamoxifenAny DiagnosisLayoutRo

17、w DimensionMeasurementInstitution (IPAS)No. of EpisodeStep 1 - Define basic data scopeSpecify the report nameSpecify the studydate type andrangeChoose theanalysis hospitaland specialtiesStep 2 - Specify selection criteriaSelect thedesiredreportingcriteria fromthis tree andadd to theSelectedThis pane

18、l showsall the addedcriteriaCriteria onthe rightStep 3 - Group criteriaHighlight any two criteria(those in pink background) at atime and define theirrelationship by choosing one ofthe “And”, “Or ”, “Not”operatorsStep 4 - Design report layoutSelect and drag thedesired dimension fromthis tree to the b

19、lue areaon the report layoutSelect and drag thedesired measurementfrom this tree to thepink area on the reportlayoutLaboratory Information - ExampleAnalysis TypeIn-Patient Episodes AnalysisDetailReport typeReport TitleInstitutionLIS HbA1CXXHDischarge Date01/06/2003 30/06/2003CriteriaLayoutLab test H

20、bA1C performed with result “levelequal to HIGH”HN NumberLIS ResultsCDARS Standard ReportsStandard reports are customized reports for target end-users. Simple steps to specify the search criteria like study period,institutions, etc Report layout is fixed and business logic is incorporated inthe repor

21、t.Standard ReportsStandard Report ResultOperating Theatre utilization for Elective theatreCMS III - 4 PerspectivesManagementOperationalClinicianPatientGoals : To provides at-a-glance managementinformation To supports management who need to beaware of directions and trends in order tomake informed de

22、cisionsService AccessibilityWinter Surge ManagementDiabetes Mellitus ManagementService MonitoringAdult Acute Stroke ManagementKey Performance IndicatorCT / MRI Brain donewithin 12 hour of admissionSupport communication with end-usersAuto-Alert Notification by EmailPlease be informed that the percent

23、age on chart A is less than 30%.Managemen DashboardBeta version1 October 2015http:/mipo.home/Last Updated Date: 12 Feb 20161. A&E Performance Reports1. A&E Performance ReportsProfessional developmentClinical Research & AuditHA Convention PresentationsAuditNo. of CDARS Request by different groups1600

24、00140000120000100000800006000040000200000ClinicalHead OfficeMRO & FinanceOther2003200420052006200720082009201020112012YearNumber of CDARS Requests for Calendar Year 2003-2015350000300000250000200000150000100000500000308019300193281212243197211990178721143392967877754963638482764274930095200320042005

25、2006200720082009201020112012201320142015Health Informatics andClinical Informatics Governance44 groups160 doctors100 others98Data Competence Centre Goal(established in June 2015)Integrated“With HA data & advanced analyticalmethods to be further implemented in ourIT infrastructure to help HA make mor

26、einformed decisions for improvement of HAhealthcare service delivery.”AutomaticProactiveInteractiveInstant, timelyActionablePredicitvieClinical / Business cases to be solved withBig data analyticsInitiatives1a.Detection from free-text and structured clinical data (e.g. laboratory,radiological report

27、s b.Track & trace the follow-up those results withoutproper care with structured & unstructured clinical activities records2Detection from free-text and structured data allergy & ADR checking duringmedication prescription34Identification of determinants of frequent HA service usersWorkflow analysis

28、to identify pressure points for workflow improvement (e.g.medication prescription and dispensing)56Disease prognostic / progression modelling (e.g. Cancer) Visualization andbenchmarking of HA IT system performanceVisualization and benchmarking of HA IT system performanceBig Data Enablement ProjectJu

29、l 2015Background1. Acquisition of Professional Services to gain the hands onexperience on the Big Data technology2. Demonstration of a few simple infrastructure analytic queries3. Technical skills transfer and training4. Use of temporary server hardware and storage5. Temporary Big Data software lice

30、nse to be provided by theservice providerA big data use case- flu activity monitoring from automatic identification ofpositive lab results with free text lab result dataPOC Requirements& resultsPositive flu result identification by manual processITHICICOIT- Draw newresult patterns inCDR periodically

31、to HI- Indicate eachpositive fluresult in excelfile with a “Y”- Verify andconfirm the newpositive fluresults- Add newpositive fluresults to dataretrieval criteriaExploration of automatic positive flu resultidentification by Big Data Analytics A simple business case to start exploring the conceptand

32、application of Big Data Analytics - Automaticidentification of positive lab results from free-textlab result data in HA database Objectives: To explore the precision of automatic result identificationby Proof of Concept (POC) of Big Data analytics technology To examine the worth of adopting the new

33、technology inthe Hospital Authority for existing business caseBig Data Technology Fast (hardware) by parallel computing (cluster of computers withsupercomputer power) Reliable / Fault-tolerance (software) HADOOP creates and stores copies of data Machine learning (software) build models by data scien

34、ce transform data into insights (e.g. data with predictionrate) PredictionExpected output of the POC Reports same as the previous flu activitiesmonitoring report by automatic identificationof positive results from structured flu lab resultdata & unstructured free text flu lab dataInformation provide

35、d for the POC1. Sample dataset from HA CDR M&V test results (excluding C&ST), with testname and test result Jan 2011 Aug 2015 (2015 data for prediction)2. Rules of positive flu results for “supervised”machine learning The excel file with manually marked positive flu result With some known flu relate

36、d test and known positiveresult pattern removed (for testing the ability of test andresult prediction)POC prediction results Flu Positive results identification (Prediction rate 0.9) Sensitivity = 0.999970791 68469 positive records identified in 2015 data versus 68471 positive records in the dataset

37、 retrieved by golden rule of CDARSstandard report #46 All tests are flu related tests All results are positive flu results Able to predict tests removed in the rules Able to predict some positive results removed in the rules Failed to predict 2 positive records, which positive resultpattern was remo

38、ved in the rules.POC result demonstration ofBig data use casePrescription SheetMOE Prescription 10,000,000/Year27,397/DayDrug Allergy CheckingNo. of Alert Prompted846,630No. of Alert Accepted389,768 (46.04%)No. of Alert Overridden456,862 (53.96%)G6PD Deficiency Contraindication CheckingNo. of Alert

39、Prompted27,374No. of Alert Accepted16,450 (60.09%)No. of Alert Overridden10,924 (39.91%)Pregnancy Contraindication CheckingNo. of Alert Prompted43,765No. of Alert Accepted31,660 (72.34%)No. of Alert Overridden12,105 (27.66%)Drug Drug Interaction CheckingNo. of Alert Prompted545,671No. of Alert Accep

40、ted75,339 (13.81%)No. of Alert Overridden470,332 (86.19%)Adverse Drug Reaction CheckingNo. of Alert Prompted61,845No. of Alert Accepted18,355 (29.68%)No. of Alert Overridden43,490 (70.32%)Background Current situationSystem checking againststructured text allergy /ADR data onlyBackground After eHR li

41、ve run(Tentative 13 Mar 2016)No system checking againstfree text allergy / ADR data +eHR dataChecking between free text drug allergydata and HA Corp Alert Allergen listHA Corp Alert Allergen ListFree text drug allergy data(golden rule)CheckingUser Interface1. Input the drug namewith “|” behind2. Cli

42、ck check3. The top e.g. three resultswith highest match rate willbe shown first.Clinical / Business cases to be solved withBig data analysticsInitiatives1a.Detection from free-text and structured clinical data (e.g. laboratory,radiological reports b.Track & trace the follow-up those results withoutp

43、roper care with structured & unstructured clinical activities records2Detection from free-text and structured data allergy & ADR checking duringmedication prescription34Identification of determinants of frequent HA service usersWorkflow analysis to identify pressure points for workflow improvement (

44、e.g.medication prescription and dispensing)56Disease prognostic / progression modelling (e.g. Cancer) Visualization andbenchmarking of HA IT system performanceVisualization and benchmarking of HA IT system performance“Enabler ”From Individual Patient Care toCorporate-wide Healthcare Service DeliveryEnabling systemic improvement ofHA Clinical Service Deliverywith real evidence ofrich pool of clinical dataBenefits most if not all people inthe HA family, and moreimportantly,our Patients.Thank you