databases in clinical research - mit opencourseware · pdf filedatabases in clinical research...
TRANSCRIPT
![Page 1: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/1.jpg)
Databases in Clinical Research
Harvard-MIT Division of Health Sciences and TechnologyHST.951J: Medical Decision Support
![Page 2: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/2.jpg)
Overview
• Background: History and utility of clinical datarepositories
• Strategies: Integrating the outcomes trackingdatabase into clinical workflow
• Brigham and Women’s CatheterizationLaboratory Database: Architecture,Advances, Limitations
• Examples of Data Exploration: Risk models,“drilling down”, and device safety monitoring
![Page 3: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/3.jpg)
Need for Clinical Data Repositories
•Randomized clinical trials are goldstandard for testing a hypothesis, butthere are significant limitations:
• generalizability
• timeliness
• cost $$$
![Page 4: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/4.jpg)
Cost of Randomized Clinical Trials
• Estimated cost of RCT:
® Drug Trial: $15,000/patient
• 1000 patient trial: $15MM
• Simply too expensive to answer everyrelevent clinical question with prospectiveblinded RCT.
![Page 5: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/5.jpg)
Clinical Registries
• While RCT’s test hypotheses, the real world ofclinical practice is a registry.
All patients (generalizability) Dynamic (timeliness)
• Significant Potential cost savings whenautomated clinical registry (database system)bundled with other functional requirements
clinical reporting, billing, inventory control
![Page 6: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/6.jpg)
History of Successful Clinical Registries
• Duke Database
• Washington Heart Center
• Beth Israel Hospital, Boston
• Cleveland Clinic
• Mayo Clinic
• Massachusetts General Hospital
![Page 7: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/7.jpg)
Why Clinical Cardiology?
• High volume clinical sites
• High event rates – death, MI,revascularization, rehospitalization, etc.
• High profile
• High cost to study
![Page 8: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/8.jpg)
Applications of Clinical Databases:
• Clinical Research:
® Retrospective “Hypothesis Generator”
® Data mining
® Prospective automated CRF
® Risk prediction modeling
• Quality Assurance:
® Interprovider variability
® Benchmark review – ACC NCDR
• Business and Operations Review – Turnovertimes, referral patterns
• Regulatory Requirements – State DPH
![Page 9: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/9.jpg)
Overview
• Background: History and utility of clinical datarepositories
• Strategies: Integrating the outcomes trackingdatabase into clinical workflow
• Brigham and Women’s Catheterization LaboratoryDatabase: Architecture, Advances, Issues
• Examples of Data Exploration: Risk models,“drilling down”, and device safety monitoring
![Page 10: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/10.jpg)
Strategies for Maintaining Clinical db
• Prospective/retrospective off-line chartreview
• Data extraction w/ supplemental chartreview
• Complete integration into electronicrecord system
Three Strategies:
![Page 11: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/11.jpg)
Clinical Database Strategies:
-- independent of clinical process-- focus on data quality-- maintain current workflow
ParallelChart Review
FullyIntegrated
-- requires team of coders-- COST $$$
HybridStrategy -- purely prospective
-- integral part of routine workflow-- lowest cost (??)
-- data quality issues-- data management
![Page 12: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/12.jpg)
Integration Dimensions:
Workflow
Sys
tem
s
Stand AloneRetrospective
HybridCath Lab Only
Fully IntegratedEMR
Required toachieve“synergy savings”
![Page 13: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/13.jpg)
Multi-Use Function of Clinical Cath LabDatabases:
• Clinical Outcomes Tracking Database:
® Retrospective Clinical Research
® Quality Assurance
® Administrative reporting
• Clinical report generation (structuredreporting; transcription templates)
• Technical and Professional Billing
• Inventory Management
• Increased complexity of database with eachadditional functional layer.
![Page 14: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/14.jpg)
Procedure
Post-CathEvaluation
Pre-CathEvaluation
Information Flow Integrated Into Care Process
Care Stage Output
Images
Report
Scheduling - patients - facilities - personnel
Function
Report Generation
Inventory Control
Image Distribution
Integrated Billing
ProcedureRequest
Order Placement
In-HospitalFollow-up
Post-DCFollow-up Clinical Event Monitoring
![Page 15: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/15.jpg)
Evolutionary Growth in DB Design: BWH CCL DB
0
50
100
150
200
250
Version 1 Version 2 Version 3
Clinical
Admin
DB Utility
# F
ield
s
![Page 16: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/16.jpg)
Functions Supported in Cath Lab:
• Clinical Documentation
• Clinical Outcomes Database (Research)
• Technical Billing
• Professional Billing
• Inventory Management
• Clinical/Quality Improvement Database
• Administrative Database Functionality
• State Reporting (DPH)
![Page 17: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/17.jpg)
DB: Core to Supporting Multiple Functions
• Clinical Documentation
• Technical Billing
• Professional Billing
• Inventory Management
• Clinical/Quality Improvement Database
• Procedure Scheduling
• Administrative Database Functionality
• Image archiving
![Page 18: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/18.jpg)
Tension within Medical Informatics
DatabaseRequirements
-- structure data entry-- limited vocabulary-- fixed meaning-- no free text entry-- focus on consistency
Clinical CommunicationRequirements
-- unstructured-- unlimited vocabulary-- variable meaning-- frequent revision-- focus on interpretation
Clinical Documentation
![Page 19: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/19.jpg)
Overview
• Background: History and utility of clinical datarepositories
• Strategies: Integrating the outcomes trackingdatabase into clinical workflow
• Brigham and Women’s Catheterization LaboratoryDatabase: Architecture, Advances, Issues
• Examples of Data Exploration: Risk models,“drilling down”, and device safety monitoring
![Page 20: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/20.jpg)
System Architecture: Phase I
WITT
ICIS
BICS
ReportDemographics
ReportDemographics
Parallel DBAccess 97
![Page 21: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/21.jpg)
System Architecture: Phase I
WITT
ICIS
BICS
ReportDemographics
ReportDemographics
SQL 7Transactional
DB
![Page 22: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/22.jpg)
System Architecture: Phase II
WITT
ICIS
BICS
ReportDemographicsSQL7 DBAnalytical
ReportDemographics
AdministrativeUsers
SQL 7Transactional
DB
![Page 23: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/23.jpg)
System Architecture: Phase II
WITT
ICIS
BICS
ReportDemographicsSQL7 DBAnalytical
Web Server
ReportDemographics
AdministrativeUsers
AuthorizedUsers
SQL 7Transactional
DB
![Page 24: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/24.jpg)
System Architecture: Phase III
WITT
ICIS
BICS
ReportDemographicsSQL7 DBAnalytical
Web Server
ReportDemographics
AdministrativeUsers
AuthorizedUsers
SQL 7Transactional
DB
Tracking Query-- vital status-- d/c date, labs, meds-- surgery, visits
![Page 25: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/25.jpg)
System Architecture: Phase IIIb
WITT
ICIS
BICS
ReportDemoSQL7 DBAnalytical Web Server
ReportDemographics
AdministrativeUsers
AuthorizedUsers
SQL 7Transactional
DB
Tracking Query-- vital status-- d/c date, labs, meds-- surgery, visits
ACC-NCDRExport
Application
DPHACC
Submissions
![Page 26: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/26.jpg)
Relational DB Schema: Overview
![Page 27: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/27.jpg)
Relational DB Schema: Patient
![Page 28: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/28.jpg)
Relational DB Schema: Hospitalization
![Page 29: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/29.jpg)
Relational DB Schema: Procedure
![Page 30: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/30.jpg)
Relational DB Schema: Lesion Treated
![Page 31: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/31.jpg)
Overview
• Background: History and utility of clinical datarepositories
• Strategies: Integrating the outcomes trackingdatabase into clinical workflow
• Brigham and Women’s Catheterization LaboratoryDatabase: Architecture, Advances, Issues
• Examples of Data Exploration: Risk models,“drilling down”, and device safety monitoring
![Page 32: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/32.jpg)
Applications of BWH CCL Database
Risk Prediction Model Development
Drilling Down – Novel Risk Factors
Retrospective Device Safety Analysis
Monthly QA – Cath Lab M&M
Operations Management
![Page 33: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/33.jpg)
Risk Models: Background
• Unadjusted Mortality Rates – Published 1999-2000
- NY State PTCA Registry Model: 0.9%- NNE Cooperative Model: 1.1%- Holmes et al (Mayo Clinic): 1.6%- Moscucci et al (Univ. Michigan): 3.3%
See Hannan JAMA 277(11); Holmes Circ 102:517; Moscucci JACC 34(3); O’Conner JACC 34(3)
• Risk prediction models help adjust for severity of illness_ providers assess quality of care – improve process
_ State / public compare institutions and providers
_ researchers assess effect of changes in care
Shock
0.6%
3.4%
![Page 34: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/34.jpg)
Logistic and Score Models for Death
OddsRatio p-value
2.53 0.011.93 0.081.83 0.208.14 0.006.59 0.020.50 0.088.33 0.001.69 0.172.77 0.04
Age > 74yrsB2/C LesionAcute MIClass 3/4 CHFLeft main PCIStent UseCardiogenic ShockUnstable AnginaTachycardicChronic Renal Insuf. 2.71 0.05
LogisticRegression Model
![Page 35: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/35.jpg)
Logistic and Score Models for Death
OddsRatio p-value
2.53 0.011.93 0.081.83 0.208.14 0.006.59 0.020.50 0.088.33 0.001.69 0.172.77 0.04
Age > 74yrsB2/C LesionAcute MIClass 3/4 CHFLeft main PCIStent UseCardiogenic ShockUnstable AnginaTachycardicChronic Renal Insuf. 2.71 0.05
LogisticRegression Model
![Page 36: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/36.jpg)
Logistic and Score Models for Death
OddsRatio p-value
2.53 0.011.93 0.081.83 0.208.14 0.006.59 0.020.50 0.088.33 0.001.69 0.172.77 0.04
Age > 74yrsB2/C LesionAcute MIClass 3/4 CHFLeft main PCIStent UseCardiogenic ShockUnstable AnginaTachycardicChronic Renal Insuf. 2.71 0.05
LogisticRegression Model
![Page 37: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/37.jpg)
Logistic and Score Models for Death
OddsRatio
p-value
2.53 0.011.93 0.081.83 0.208.14 0.006.59 0.02
0.50 0.088.33 0.001.69 0.172.77 0.04
Age > 74yrsB2/C LesionAcute MIClass 3/4 CHFLeft main PCIStent UseCardiogenic ShockUnstable AnginaTachycardicChronic Renal Insuf. 2.71 0.05
LogisticRegression Model
Betacoeff Risk value
0.927 2
0.659 1
0.601 1
2.097 41.886 3
-0.683 -12.120 4
0.522 1
1.020 2
0.996 2
Risk Score Model
![Page 38: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/38.jpg)
ROC Curves: Death ModelsValidation Set: 1460 Cases
See Resnic et al. Am J. Cardiol 2001 Jul 1:88(1):5-9
![Page 39: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/39.jpg)
Model Building: Artificial Neural Networks
• Artificial Neural Networks are non-linear mathematical models which incorporate a layer of hidden “nodes” connected to the input layer (covariates) and the output.
OutputLayer
InputLayer
AllAvailableCovariates
I1
I2
I3
I4
O1
![Page 40: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/40.jpg)
Model Building: Artificial Neural Networks
• Artificial Neural Networks are non-linear mathematical models which incorporate a layer of hidden “nodes” connected to the input layer (covariates) and the output.
HiddenLayer
H1
H2
H3
OutputLayer
InputLayer
AllAvailableCovariates
I1
I2
I3
I4
O1
![Page 41: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/41.jpg)
Model Building: Artificial Neural Networks
• Artificial Neural Networks are non-linear mathematical models which incorporate a layer of hidden “nodes” connected to the input layer (covariates) and the output.
HiddenLayer
H1
H2
H3
OutputLayer
InputLayer
AllAvailableCovariates
I1
I2
I3
I4
O1
![Page 42: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/42.jpg)
ROC Curves: Death ModelsValidation Set: 1460 Cases
See Resnic et al. Am J. Cardiol 2001 Jul 1:88(1):5-9
![Page 43: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/43.jpg)
Risk Score of Death: BWH ExperienceUnadjusted Overall Mortality Rate = 2.1%
See Resnic et al. Am J. Cardiol 2001 Jul 1:88(1):5-9
![Page 44: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/44.jpg)
Applications of BWH CCL Database
Risk Prediction Model Development
Drilling Down – Novel Risk Factors
Retrospective Device Safety Analysis
Monthly QA – Cath Lab M&M
Operations Management
![Page 45: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/45.jpg)
MACE Models: Impact of No-Reflow
Age > 74yrsB2/C LesionAcute MIClass 3/4 CHFLeft main PCIStent UseCardiogenic ShockUSATachycardicNo ReflowUnscheduledChronic Renal Insuff.
p-value
1.40 0.162.56 0.002.99 0.003.61 0.002.30 0.280.58 0.033.33 0.012.69 0.001.36 0.442.90 0.011.49 0.081.58 0.23
OddsRatio
Logistic RegressionModel
beta Riskcoefficient Value
0.337 00.939 21.096 21.283 30.831 2-0.539 -11.202 30.989 20.311 01.044 20.396 00.457 1
Risk ScoreModel
![Page 46: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/46.jpg)
No-Reflow: Angiographic Case Study
63yo male 4yrs s/p 4v CABG.
Presents with NQWMI W/ lateral ST depress
![Page 47: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/47.jpg)
Posis Angiojet: Rheolytic Thrombectomy
![Page 48: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/48.jpg)
Direct Stenting of Culprit Lesion
![Page 49: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/49.jpg)
No Reflow: BWH Experience 1997-2000
Risk of In-Hospital Complication
See Resnic et al. Am Heart J. In press.
![Page 50: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/50.jpg)
TIMI Flow Rates Improved Significantly
See Resnic et al. Am Heart J. In press.
![Page 51: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/51.jpg)
Lack of Effective Treatment: BWH Experience
Risk of Death or Myocardial Infarction
See Resnic et al. Am Heart J. In press.
![Page 52: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/52.jpg)
Applications of BWH CCL Database
Risk Prediction Model Development
Drilling Down – Novel Risk Factors
Retrospective Device Safety Analysis
Monthly QA – Cath Lab M&M
Operations Management
![Page 53: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/53.jpg)
Patients receiving a closure device experienced a 44%reduction in vascular complications.
See Resnic et al. Am J. Cardiol. 2001 Sep 1:88(5):493-496.
![Page 54: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/54.jpg)
This effect was preserved in those patients receiving gp 2b3ainhibitors.
See Resnic et al. Am J. Cardiol. 2001 Sep 1:88(5):493-496.
![Page 55: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/55.jpg)
Applications of BWH CCL Database
Risk Prediction Model Development
Drilling Down – Novel Risk Factors
Retrospective Device Safety Analysis
Monthly QA – Cath Lab M&M
Operations Management
![Page 56: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/56.jpg)
Coronary Procedures by Month
384 342 371 356 296 458 337 388 437 436 458 456Total Cases
375
330
364346
294
452
333
383
433 428452 447
189165 159
175159
198
148167
185 185 180202
0
50
100
150
200
250
300
350
400
450
500
Diagnostic 375 330 364 346 294 452 333 383 433 428 452 447
Intervention 189 165 159 175 159 198 148 167 185 185 180 202
Aug Sept Oct Nov Dec Jan Feb Mar April May June July
![Page 57: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/57.jpg)
Planned vs. Ad Hoc PCI
Ad Hoc PCI rate 33% 40% 36% 31% 30% 30% 35%
70 7852
67 7850
66
125 103
118115 106
121
132
0
50
100
150
200
250
Ad Hoc PCI 125 103 118 115 106 121 132
Planned PCI 70 78 52 67 78 50 66
Jan Feb Mar April May June July
![Page 58: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/58.jpg)
Internal vs. External MD Volume
Internal 77% 76% 75% 73% 69% 66% 73% 71% 71% 74% 72% 72%External 23% 24% 25% 27% 31% 34% 27% 29% 29% 26% 28% 28%
145
126119
128
110
131
108
118
131136
129
146
4439 40
47 49
67
40
4954
49 5156
0
20
40
60
80
100
120
140
160
Aug Sep Oct Nov Dec Jan Feb Mar April May June July
![Page 59: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/59.jpg)
Death 3
Stroke 1
CABG 1 (perforation of LCx)
MI* 6 (1 SAT)
TVR 2
Vascular 7 (1 transfusion reported, 1 PSA req. surg)
Renal 3
CHF 1
* MI defined as total CK-MB>3x ULN in patient w/o index AMI.
Post-Procedural Events for July, 2002
•Significant events reported during July, 2002:
![Page 60: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/60.jpg)
Clinical Event Listing by Physician 2001
![Page 61: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/61.jpg)
Applications of BWH CCL Database
Risk Prediction Model Development
Drilling Down – Novel Risk Factors
Retrospective Device Safety Analysis
Monthly QA – Cath Lab M&M
Operations Management
![Page 62: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/62.jpg)
Inpatient33%
Outpatient39%
Transfer20%
Emergency3%
Unknown5%
One-third of total case volume is noted as inpatient source in WITT.
![Page 63: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/63.jpg)
The case volume is distributed according the usual 80/20 rule. Nearly 80% of cases is referred from 20% of the MDs.
0
20
40
60
80
100
120
140
1601
14 27 40 53 66 79 92 105
118
131
144
157
170
183
196
209
222
235
248
261
Physician
Nu
mb
er o
f C
ases
![Page 64: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/64.jpg)
Rules for Designing an OutcomesDatabase
• Understand workflow in detail. Identify immutablepoints (most of these depend on perspective).
• Incremental design – identify successfulmilestones
• Open architecture – use ODBC compliantrelational databases as backbone
® Systems integration is most complexchallenge
• Goal of distributed information availability.
• Identify implementation team. Responsiblities,project plan, regular operational meeting.
![Page 65: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/65.jpg)
Acknowledgements
Decision Systems Group
Lucila Ohno-Machado, MD PhD
Robert Greenes, MD PhD
Aziz Boxwala, MD PhD
Partners Information Systems
Mark Nightingale
Cardiovascular Division
Jeff Popma, MD
Andrew Selwyn, MD
Campbell Rogers, MD
Charles Lin, MBA
Benjamin Paul
![Page 66: Databases in Clinical Research - MIT OpenCourseWare · PDF fileDatabases in Clinical Research ... •Increased complexity of database with each additional functional layer. Procedure](https://reader031.vdocuments.site/reader031/viewer/2022030401/5a75d2117f8b9a0d558cb472/html5/thumbnails/66.jpg)
Thank You!