The Future of Clinical Registries in the Cycle of Evidence (Text Version)

Slide Presentation from the AHRQ 2008 Annual Conference

On September 8, 2008, Eric D. Peterson, M.D., M.P.H., F.A.H.A., F.A.C.C., made this presentation at the 2008 Annual Conference. Select to access the PowerPoint® presentation (2 MB; Plugin Software Help).

Slide 1

The Future of Clinical Registries in the Cycle of Evidence

Eric D. Peterson, MD, MPH, FAHA, FACC
Professor of Medicine
Vice Chair for Quality, Department of Medicine (DOM)
Duke University Medical Center
Director of Cardiovascular (CV) Research and Associate Director,
Duke Clinical Research Institute (DCRI)

Slide 2

Cycle of Therapeutics

Screen shot of a diagram shows a circle with arrows rotating in a clockwise formation. The cycle starts with a Concept which leads to Clinical Evidence, to Guidelines, to Performance Indicators, to Measurement and Feedback, and finally, to Outcomes, aided by QI initiatives. In the center is a red rectangle with "Provider-led Clinical Registries". From the rectangle, arrows radiate to "Clinical Evidence," "Guidelines," "Performance Indicators," "Measurement and Feedback," and "QI Initiatives."

Adapted from Califf RM, Peterson ED et al. JACC 2002;40:1895-901.

Slide 3

Evidence Development

Therapeutic efficacy is usually best evaluated via a randomized controlled trial (RCT) ...provided such a trial is:

• Adequately sized.
• In a representative patient population.
• With typical providers and care setting.
• Well conducted and analyzed study.
• Has a meaningful outcome endpoint.

Slide 4

Unfortunately RCTs have Limitations

• Trial and community pts and setting differ:
  • Pts: older, sicker, more comorbid disease.
  • Providers: Experts vs. usual practice.
• Surrogate endpoints often used rather than patient outcome.
• Duration often short and long-term safety or efficacy not assessed.
• Certain questions not easily subject to RCT:
  • Unethical, impractical, or clinical question doesn't translate well to protocol evaluation.

Slide 5

Imagine If!

We have clinical registries that capture data:

• At "key points" in patient's life (major events, procedures).
• That can be supplemented with lab and imaging information.
• That are able to follow patients' longitudinal care and outcomes.
• That are readily assessable to researchers for scientific discovery.
• But also can be used for driving new evidence into routine practice.

Slide 6

The slide shows both a logo for the NCDR™ (National Cardiovascular Data Registry) and a header entitled "Turning Guidelines into Lifelines," with an image of a child sitting on a man"s shoulders and "Get with the Guidelines", encircled by rotating arrows.

Slide 7

CV Provider Led Clinical Registries

• Society of Thoracic Surgery: 900+ centers:
  • Coronary artery bypass surgery.
  • Valve surgery.
  • Congenital heart surgery.
  • Thoracic surgery.
• National Cardiovascular Data Registry: 1600+ Hospitals:
  • Cath/Percutaneous coronary intervention.
  • Implantable cardiac defibrillators (ICD).
  • Acute coronary syndromes (ACS).
  • Carotid stenting.
  • Ambulatory CV disease (launching).
• American Hospital Association (AHA)—Get With The Guideline (GWTG) Program: 1500+ hospitals:
  • Coronary artery disease (CAD).
  • Heart failure.
  • Stroke.
  • Ambulatory modules (also launching).

Slide 8

These Clinical Registries Currently...

• Are large and representative:
  • Patients, providers, settings.
• Contain detailed clinical data:
  • Presenting features, treatments, acute outcomes.
• Moving towards standardized data elements.
• Are of high quality:
  • Accurate, complete, audited.
• Are collaborative with one another.
• Are becoming longitudinal!

Slide 9

National Quality Forum (NQF) Evolving View of Quality Care: Importance of Longitudinal Measures

Screen shot of a diagram showing a large gray circle followed by 3 smaller, colored circles and finishing with two large rectangles atop each other. Two arrows run through the circles and split apart to go to the two rectangles. The data reads:

• Phase 1: Staying Healthy.
  • Population at Risk:
    • 1° Prevention (no known acute myocardial infarction [AMI]).
    • 2° Prevention (CAD no prior AMI.
    • 2° Prevention (Recurrent AMI events) Advanced Care Planning.
• Phase 2: Getting Better.
  • Acute Phase:
    • Episode begins—onset of symptoms
    • Assessment of Preferences.
• Phase 3: Living with Illness/Disability (T1) and Coping with End of Life (T2).
  • Post Acute/Rehabilitation Phase.
• Phase 4: Living with Illness/Disability (T1) and Coping with End of Life (T2).
  • 2° Prevention.
  • Episode ends—1 year post heart failure (HF).

• Post HF Trajectory 1 (T1): Relatively healthy adult:
  • Focus on:
    • Quality of Life.
    • Functional Status.
    • 2° Prevention Strategies.
    • Rehabilitation.
    • Advanced care planning.
• Post HF Trajectory 2 (T2): Adult with multiple co-morbidities:
  • Focus on:
    • Quality of Life.
    • Functional Status.
    • 2° Prevention Strategies.
    • Advanced Care Planning.
    • Advanced Directives.
    • Palliative Care/Symptom Control.

Slide 10

Measuring Care Across the CV Continuum: AMI Example

Screen shot of a diagram showing a timeline of AMI Care.

• IC³/GWTG Ambulatory
  • Patient Risk Factors.
• Action-GWTG:
  • NCDR.
  • STS.
  • Onset of Acute Coronary Syndrome; Admit; percutaneous coronary intervention (PCI)/coronary artery bypass graft (CABG); D/C.
• IC³/GWTG Ambulatory:
  • Post-Hospitalization: Risk factor modification, Cardiac rehabilitation.

Slide 11

National Consortium of CV Clinical Database (NC2D) Collaborative

• Funding:
  • AHRQ Centers for Education and Research on Therapeutic (CERT).
  • Individual professional societies.
• Representatives:
  • CV Registries (AHA. ACC-NCDR, STS).
  • Gov't (AHRQ, Food and Drug Administration [FDA]).
  • Payors (Centers for Medicare and Medicaid Services [CMS], United Healthcare [UHC], Aetna, Blue Cross/Blue Shield [BC/BS], Wellpoint).
  • Brookings Institute.
• Goal:
  • To facilitate collaboration among the CV registries; addressing common challenges and opportunities.

Slide 12

National Consortium of CV Clinical Databases (NC2D)

Example proposed projects:

• Promote standardized data elements.
• Address operational and methodological issues for linking:
  • Between CV clinical registries.
  • Between CV registries and claims databases.
• Facilitate post-market surveillance and comparative effectiveness studies with FDA, CMS.
• Promote the increased use of registry-based for QI and performance evaluation.

Slide 13

Linking Clinical Databases
Can We? Are We Allowed?

• What to link?
  • Clinical registry to clinical registry.
  • Clinical registry to claims data sources.
• How to link?
  • With patient identifiers: Direct match.
  • If no patient identifiers: Probabilistic matching.
• Okay to link?: Health Insurance Portability and Accountability Act (HIPAA)? Institutional Review Board (IRB?
  • Waiver of informed consent.
  • Means of protecting patient privacy.

Slide 14

ACC-NCDR PCI Probabilistic Match Example

• Steps:
  • Match hospitals in CMS and American College of Cardiology National Cardiovascular Data Registry (ACC NCDR).
  • Match records within hospital:
    • Using DOB [date of birth], gender, admit/discharge dates, etc.
• Results:
  • Site Link:
    • 697 of 720 NCDR sites (97%) were linked to CMS sites.
    • The unlinked sites contained <1% of total eligible.
  • Patient Link:
    • Among the NCDR records, 76% were linked to CMS records.
    • Non-linked= managed care, record errors, etc.

Slide 15

Uses of Longitudinal Clinical Registries:
Epidemiology and Disease Surveillance

• Track diseases, treatments, and outcomes in community-based, "real-world" settings:
  • Cross-sectional and trends.
• Post-market evaluation of drugs/devices:
  • Study rare events, late outcomes (beyond RCTs).
  • Use and outcomes of "off-label indications".

Slide 16

Tracking Diffusion and Outcome of "On Vs Off Label" DES: ACC-NCDR

The slide shows two bar graphs. The first one shows the Percentage of Use of On-label vs. Off-label for the time period 2003 Qtr2 through 2003 Qtr 4. The second one shows the Percentage of Mortality for On-label vs. Off-label for On-Label, ST MI, ISR, SVG, and CTO. The data in tabular form shows:

Source: Rao SV Am J Cardiol 2006;97:1478-1481.

Slide 17

Duke Database: 6-Month Landmark Analysis: Adjusted Cumulative Mortality Rates

Screen shot of a line graph showing the Percent Cumulative Incidence Rate for DES-C, BMS-C, BMS+C, and DES+C from 6 to 24 months. The data shows:

DES-C: 5.3
BMS-C: 4.5
BMS+C: 3.7
DES+C: 2.0

Source: Eisenstein et al. JAMA 2006.

Slide 18

AHRQ DES DEcIDE [Developing Evidence to Inform Decisions about Effectiveness] Project: ACC-NCDR CMS Merged Database

Slide 19

Preliminary Unadjusted Death or MI Rates: BMS vs DES

The line graph shows the Survival Function Estimates for Mortality or MI. The vertical axis, Survival Probability, goes from 0.5 to 1.0 and the horizontal axis, Time to Mortality or MI, goes from 0 to 1000. The line graph for BMS begins at 1.0 and ends at 0.75. The line graph for DES begins at 1.0 and ends at 0.82. There is also a small table in the line graph showing the following data:

Slide 20

Clinical Registries Linked to Pharmacy Data: Veterans Health Administration (VHA) Study of Clopidogrel Rebound

• Question:
  • Is there risk associated with cessation of antiplatelet therapy after a prolonged course?
• Study Population:
  • 127 VHA Hospital ACS Registry.
  • MI patients discharged on clopidogrel.
• Total clopidogrel duration: median (25th-75th):
  • Medical (n=1569): 281 (120-417) days.
  • PCI (n=1568): 310 (182-410) days.
• All-cause death/MI after stopping clopidogrel:
  • Determined for each 90 day period.

Source: Ho PM, et al. JAMA 2008; 299(5):532-539.

Slide 21

Event Rates after Stopping Clopidogrel

The slide is a screen shot of two separate line graphs. On the first line graph, Stent n=1568, the vertical axis, Instantaneous Incidence Rate, Per Person-Day, goes from 0 to .0004 and the horizontal line, Days Since Stopping Clopidogrel, goes from 0 to 450. A blue arrow points to a peaking in the data around 45 days before steadily declining. On the second line graph, Medical n=1569, the vertical axis, Instantaneous Incidence Rate, Per Person-Day, goes from 0 to .001 and the horizontal axis, Days Since Stopping Clopidogrel, goes from 0 to 450. A blue arrow points to a peaking in the data around 90 days before steadily declining.

Ho PM, et al. JAMA 2008; 299(5):532-539

Slide 22

Registries for Evidence Development (2): Observational Evidence

• Real World "Comparative Effectiveness":
  • Test generalizability of trial findings in broader pt populations and clinical settings.
  • Evaluate cost-effectiveness.
• Examine practices where trials either not possible or not yet conducted.

Slide 23

Comparative Effectiveness: GP IIb-IIIa Inhibitor: Results from Clinical Trials, NRMI, and CRUSADE

Screen shot of a bar graph. The graph's horizontal axis spans from 0.5 to 2.0. From 0.5 to 1.0 the graph displays: Early GP IIb-IIIa Inhibitor Better; from 1.0 to 2.0 it displays: No Early GP IIb-IIIa Inhibitor Better. The data shows the 95% CI Odds Ratio for the following studies:

• 6 RCTs ACS¹ (n=31,402): 0.91 (0.81, 1.03)
• NRMI NSTE MI² (N=60,770): 0.88 0.79, 0.97)
• CRUSADE Overall (n=56,087)³: 0.93 (0.83, 1.05)
• CRUSADE NSTEMI (n=37,433)³: 0.88 (0.77, 1.00)

Sources: ¹ Boersma et al. Lancet 2002;359:189;
² Peterson et al. J Am Coll Cardiol 2003;42:45;
³ Hoekstra et al. Acad Emerg Med 2005;12:431-438.

Slide 24

Comparative Effectiveness: low molecular weight (LMHW) VS UF Heparin

CRUSADE versus SYNERGY

Screen shot of a bar graph with the horizontal axis from 0.5 to 1.5. From 0.5 to 1.0, LMWH better; from 1.0 to 1.5, UFH better—Reference. The data shows:

• Death or MI:
  • Registry¹: 0.78 (0.70, 0.86)
  • SYNERGY²: 0.96 (0.87, 1.06)
• Any RBC Transfusions:
  • Registry¹: 0.81 (0.76, 0.87)
  • SYNERGY²: 1.04 (0.83, 1.30)

Sources: ¹ Tricoci P et al: JACC 2006 in press;
² SYNERGY Investigators JAMA 2004:45-54.

Slide 25

Excessive Antithrombotic Dosing

Screen shot of a bar graph showing the percentage of Excessive Dosing by Age for Less than 65 years, 65-75 years, and Greater than 75 years for the drugs LMW Heparin and UF Heparin. The data shows:

• LMW Heparin:
  • Less than 65 years: 12.5%.
  • 65-75 years: 12.5%.
  • Greater than 75 years: 16.5%.
• UF Heparin:
  • Less than 65 years 28.7%.
  • 65-75 years: 37%.
  • Greater than 75 years: 38.5%.

Source: Alexander KA, et al. JAMA 2005;294:3108-3116.

Slide 26

Registries Supporting Practical Randomized Trials

• Registries can help with:
  • Study design (event rates, expected pts).
  • Identify eligible sites, investigators.
  • Identify eligible patient:
    • Screen for those meeting entry criteria.
• Clarify the generalizability of trial population
  • Characterize enrolled RCT patients vs all.
• Support trial data collection and follow-up.

Slide 27

The slide first displays a flowchart, which then gets covered up by the bar graph described below. The information in the flow chart is as follows:

• Top row: 3000 patients 1st CABG leading to:
  • 2nd row: Edifoligide leading to:
    • 3rd row: 30-day Adverse Events: 1st 2400: 1-year Angiography
      • 4th row: 1, 2, 3, 4, 5-year follow-up: Death, MI, Revascularization with VG failure.
  • 2nd row: Placebo leading to:
    • 3rd row: 30-day Adverse Events: 1st 2400: 1-year Angiography
      • 4th row: 1, 2, 3, 4, 5-year follow-up: Death, MI, Revascularization with VG failure.

The bar graph shows the results for 3000 Patients: First CABG. The vertical axis goes from 0-3000 and the horizontal axis goes from August to the following October. The bar graph starts at 0 in August and ends at 3000 in October with a steady increase each month.

Slide 28

Combining Registries and Trials: FDA DES CODA Proposal

Screen shot of a timeline showing a three-year time span:

• Index PCI Randomization to 12 Months:
  • ADP Inhibitor.
  • All Patients.
• 12 Month:
  • Randomization—an arrow points back to Index PCI Randomization.
• 12 Months to 36 Months:
  • No ADP Inhibitor.
  • ADP Inhibitor.

Slide 29

New Ideas for the Large Simple Trial

The screen shot shows the home page of Aspirin: Assessment of Platelet Inhibition by Internet Response Study Web site.

Slide 30

Clinical Registries Can Facilitate Practice Change!

• Identify "opportunities" for improvement:
  • Track adoption of effective therapies.
  • Track off-label and inappropriate use.
  • Identify variability in care (e.g., disparities).
• Provider performance assessment:
  • Benchmark providers' care and outcomes.
  • Support Quality Assessment (QA) (link to Pay-for-Performance (P4P) or public reporting).
• Facilitating Provider-Led Quality Improvement:
  • Use feedback tools to stimulate practice change.

Slide 31

Clinical Registry Quality Improvement (QI) Tools

• Quarterly Feedback reports.
• Individualized Gap analysis.
• On-line Real time summaries.
• QI tool kits.
• Monthly Web-casts.
• Regional Group Meetings.

Note: Two screen shots show "Adherence Trends; Q2/06 Report" with results placed in line graphs and "Table 2: Medical History; Q2/06 Report" with results in a table and bar graphs.

Slide 32

Improving Quality and Reducing Safety Concerns

The slide shows two bar graphs. The first bar graph shows the Composite Adherence Rates from Q1-02 to Q2-04.

• Q1-02: 68.1%
• Q4-02: 72.3%
• Q3-03: 75.2%
• Q2-04: 79.3%

The second bar graph shows the Rate of Excess Dosing from Q4-05 to Q2-06.

• Q4 2005: 27.5%
• Q1 2006: 24.9%
• Q2 2006: 22.1%

Source: Mehta RH, et al. AHJ 2007.

Slide 33

Randomized Evaluation of QI: STS AHRQ "CQI in Bypass Surgery"

Screen shot showing two line graphs. The first line graph measures the Percentage of BB [beta blocker] Use from S2000 to S2002 for Control and BB. An arrow points to where the lines intersect at 61% for Intervention during F2000. The second line graph measures the Percentage of IMA [internal mammary artery] Use from S2000 to S2002 for Control and IMA. An arrow points to where the lines intersect at 81% for Intervention during F2000.

Source: Ferguson TB JAMA 2003;290:49-56

Slide 34

AHRQ CERT Individualized QI RCT

The diagram is a flowchart which shows the process of Individualized QI RCT.

• 1st row: NCDR Action Sites.
• 2nd row: Standard QI Feedback.
  or
• 2nd row: Individualized GAP Analysis:
  • Top 3 Quality or Safety Targets.
  • Targeted Data Reports.
  • Educational Modules and QI Tools.
• 3rd row: Evaluation:
  • Composite Metrics of Quality and Safety.
  • Benchmarks Achieved.
  • Surveys assessing implementation and usability.

Note: Both of the 2nd row processes lead to the Evaluation process in the 3rd row.

Slide 35

Conclusions

• Registries offer considerable promise:
  • For supplementing the discovery process.
  • Then, tracking translation to community practice.
  • And ultimately for speeding up cycle.
• Future getting brighter..:
  • Lots of new tools and resources.
  • Traditional challenges being overcome.
• Thanks to AHRQ for all the support over the year!