NHLBI Workshop
Data Needs for Cardiovascular Events, Management, and Outcomes
HMO Populations and Clinical Databases as a Source for
Monitoring Trends in CVD Morbidity & Mortality -
Dr. Joseph Selby
Key Points
- Integrated HMOs are an increasingly rich source of longitudinal data
on CVD events, risk factors and quality of care
- Membership is quite stable
- Large registries of persons with CVD diagnoses are being created
- Many of these HMOs share the same electronic medical record, and data
definitions are being harmonized across HMOs
HMO Research Network (HMORN) -- 13 million members
- Fallon Health Care, Worcester, MA
- Group Health Cooperative, Seattle, WA
- Harvard Pilgrim Health Care, Boston, MA
- HealthPartners Research Foundation, Minneapolis, MN
- Henry Ford Health System, Detroit, MI
- Kaiser Permanente Colorado, Denver, CO
- Kaiser Permanente Georgia, Atlanta, GA
- Kaiser Permanente Hawaii, Honolulu, HI
- Kaiser Permanente Northwest, Portland, OR
- Kaiser Permanente Northern CA, Oakland, CA
- Kaiser Permanente Southern CA, Pasadena, CA
- Lovelace Clinic Foundation, Albuquerque, NM
- Scott and White Health System, Temple, Texas
- United Healthcare, Minnetonka, MN
Major Collaborative Projects of the HMO Research Network (HMORN)
Project |
Funder |
Coordinating Center |
CDC |
CDC |
NCI |
Group Health Cooperative |
AHRQ |
Harvard Pilgrim |
Roadmap/
NHLBI |
Group Health Cooperative |
Data are Improving
- Membership (denominators & demographics)
- Complete hospital discharge data (endpoints, procedures, comorbidities)
- Outpatient diagnoses & procedures (comorbidities and endpoints)
- Pharmacy data (comorbidities, quality of care)
- Laboratory results (risk factors)
- Outpatient measurements (BP, smoking status, BMI)
Availability of CVD Risk Factor Data among 1.2 million persons, age 45 and
above, 1/1/04; Kaiser Permanente Northern California
Men (n=580,817) |
Women (n=663,661) |
86% |
92% |
68% |
69% |
73% |
75% |
63% |
64% |
85% |
92% |
81% |
84% |
55% |
60% |
70% |
75% |
Examples of Studies Conducted in Kaiser Permanente, Northern California
- Joint Control of Three CVD Risk Factors in the KP Population, 2001
to 2003:
The 3D Study -- Support: Pfizer, Inc.
- Troponin Measurement and Trends in Acute Coronary Syndrome Hospital
Discharge Diagnosis
- To determine the role of troponin measurement in the changing
distribution of ACS hospitalization discharge diagnoses in Kaiser
Permanente Northern California (KPNC)
- All data, including cardiac biomarkers, were obtained from KPNC
electronic databases.
- All hospital discharges for ACS from KPNC hospitals between 1994-2003
were determined (primary discharge codes 410.x, 411.1, 414.x primary
with 411.1 secondary)
- Kaiser Permanente Acute Coronary Syndrome Registry (KP-ACS)
- Modeled after National Registry of Myocardial Infarction (NRMI)
- Systematic chart review of all hospitalized acute myocardial infarction
and troponin (+) unstable angina patients at 17 Kaiser facilities
since 2002 (AMI) & 2004 (UA)
- ~9,000-10,000 reviewed cases annually
- Kaiser Permanente Chronic Heart Failure Registry (KP-CHF)
- Chronic heart failure (CHF) identified from hospitalizations
and ambulatory visit (outpatient, ED) databases
- 96% of primary hospital discharge diagnoses and 85-90% of cases identified
from outpatient diagnoses are verified using Framingham criteria at chart
review
- 1996-2002: 59,772 adults > or = 20 years met registry criteria for CHF
in Kaiser No. Cal.
- Registry is updated annually and linked to treatment & clinical
outcomes
Conclusions
- Data from integrated HMOs can provide a timely look at incidence,
complications, mortality for a variety of CVD conditions
- Rich clinical data allows adjustment for population differences in
comorbidities and disease severity
- Data definitions can be standardized across settings
- Data may allow exploration of possible artifactual differences in
observed patterns
Limitations
- HMO patients do not represent the extremes of U.S. SES spectrum
- Time trends, geographic variation, subject to differences and changes
in clinical and coding practices
- Data completeness is sometimes a question
- Potentially, time trends may be affected by differing enrollment/departure
over time
Back to Workshop Agenda
|