The PRISM (Post-Licensure Rapid Immunization Safety Monitoring)

System for H1N1 Vaccine

Richard Platt, MD, MSDepartment of Population Medicine

Harvard Pilgrim Health Care and Harvard Medical School

November 18, 2009

Aims

Conduct real-time and end-of-season analyses of:

– Pre-specified outcomes, e.g., Guillain-Barré Syndrome

– New unanticipated outcomes if concerns arise

Special Issues in H1N1 vaccine

• Need for large base population for active surveillance → health plans

• Many H1N1 vaccine doses may be given by public providers and not captured in health plan data → link to state immunization registries

Role of PRISM

• Assess large populations

– Effectively monitor Guillain-Barré syndrome and other rare events

– Shorten time to identifying problems

• Evaluate H1N1 vaccine doses given by both public and private providers

PRISM – Current season surveillance

All current health plan members

State Immunization

Registries

National Health Plans

H1N1 vaccine

exposure data

Outcome data

PRISM – Historical comparison data

Health plan members

since 2005

National Health Plans

Seasonal influenza vaccine

exposure data

Outcome data

Organizations in PRISM

Public Health Informatics Institute

HHS

StateImmunization

RegistriesHealth plans

Computer Sciences Corporation

(Technical support)

America’s Health Insurance Plans(Administration)

Harvard Dept of Population Medicine

(Scientific coordination)

Health Plans

• Selection criteria– Ability to rapidly assemble data– Ability to update data every 1-2 weeks

• Participating health plans– Aetna, CIGNA, Humana,

HealthCore (WellPoint plans in CA, NY, CO; BCBS in MI)

• Total ~25 million persons

State Immunization Registries

• Selection criteria– Timely H1N1 vaccine data– Experience matching data with health plans– Size

• Participating states– MI, FL, GA, PA, AZ, WI, MN, NY, NYC

• Total ~14 million persons in the participating health plans

Collaborators – Partial List• National Vaccine Program Office

– Daniel Salmon– Kirsten Vannice

• Centers for Disease Control and Prevention– Frank deStefano

• Food and Drug Administration– Robert Ball

• America's Health Insurance Plans

– Kevin Fahey– Barbara Lardy– Victor Rhee

• Public Health Informatics Institute

– Ellen Wild– Bill Brand– Elaine Lowery– Therese Hoyle– Alan Hinman– David Ross

• Harvard Pilgrim Health Care Institute

– Richard Platt– Tracy Lieu– Grace Lee– Jeff Brown– Melisa Rett– Charlene Gay– Katherine Yih– Yury Vilk

• Health Information Systems Consulting – Robert Rosofsky

• EpiPatterns– Yinong Young-Xu

• Computer Sciences Corporation

– Daniel Foltz– Jim Van Dyke– Aurelia Ford– Jim Roddy– John Manson– Vic Tandon

• Cigna– Therese Conner– Anthony Sumner– Daniel Carmody– Ha Nguyen– Mark Regine

• Aetna– Claire Spettell– Cheryl Walraven– Joaquim Fernandes– Yihai Liu

• Healthcore (Wellpoint, BC/BS)– Peter Wahl– Alex Cannon– Lori Meyers– Greg Daniel– Chris Hetrick

• Humana– David Nau– Yihau Xu– Amy Ball

• Arizona State Immunization Information System

– Lisa Rasmussen        – Roger Volp– Richard Bradley

• Florida Shots– Susan Lincicome– Pete Garner

• Georgia Registry of Immunization Transactions and Services

– Tracy Culbreath

– Elizabeth Sullivan

– Michelle Conner

– Archie Banks

– Andre Wilson

• Michigan Care Improvement Registry

– Therese Hoyle

– Ian Hancke

• Minnesota Immunization Information Connection

– Emily Peterson

– Diana Jaeger

• New York State Immunization Information System

– Loretta Santilli

• New York City, Citywide Immunization Registry

– Vikki Papadouka

– Rezaul Kabir

– Luiz Homem de Mello

• Pennsylvania Statewide Immunization Information System

– Frank Caniglia

– Mike Jamula

• Wisconsin Immunization Registry– Dan Hopfensperger

Data Sharing and Confidentiality

• Classified as public health practice, not research

• Health plans and states encouraged to use the HIPAA public health exemption for exchange of Protected Health Information

• All person-level data remains at the health plans after transformation to standard format

• Distributed programs create counts that are shared with scientific coordinating center

Maintaining Confidentiality

Health Plananalysts

Person-level data (with merged registry immunization data)

Counts of immunizations and outcomes

Standard programs

Harvard andCSC analysts

Reports and analyses

Health plan firewall

Outcomes Being Monitored

• Guillain-Barré Syndrome (GBS)

• Central nervous system demyelinating diseases

• Neuropathies

• Seizures

• Encephalitis

• Bell’s palsy

• Myocarditis

• Ataxia

• Anaphylaxis

• Spontaneous abortion

• Pre-eclampsia

Weekly Monitoring of Counts

• Risk (exposure) windows– Periods after H1N1 vaccination– 14, 21, or 42 days, depending on

outcome

• Comparisons – Expected counts based on historical

rates after seasonal influenza vaccination

– Unexposed time periods in same persons

Statistical Analyses

• Weekly analyses of cumulative data

• Repeated testing of the same data increases the chance of false-positive results

• Adjust for this using maximized sequential probability ratio testing (maxSPRT)

Statistical Methods for Comparisons

• Expected counts based on historical rates – Poisson maxSPRT

• Unexposed time periods in same persons– Self-controlled case series analysis using

binomial maxSPRT

Signal Evaluation• Preliminary findings of increased risk will need

further evaluation

• Computerized data– Stratified analyses– Temporal scan – Logistic regression

• Chart review, if needed– Confirm or exclude cases as true– Gather data on potential confounders and other

variables

PRISM – Current Challenges

• Timeline – Processes for data sharing between

health plans and state immunization registries

– Programming and coordination

• Data latency

Implications for Vaccine SafetyPRISM is enhancing:

• Framing of vaccine safety surveillance as a public health activity

• Creation of public-private collaborations for public health

• Implementation of methods for rapid analysis of data that remains in possession of covered entities

• Development of methods for responding to findings