NHLBI Workshop
Data Needs for CVD Events, Management, and Outcomes

May 26-27, 2005

TABLE OF CONTENTS

Executive Summary

Rationale for Surveillance

The 20th century experienced a rise of chronic diseases, specifically cardiovascular disease and cancer, to epidemic proportions in the industrialized world. These chronic diseases are now leading causes of morbidity and mortality. In the 21st century, these illnesses will become the main causes of death in the developing world as well.

While there is an effective national cancer surveillance system (Surveillance Epidemiology and End Results (SEER) and CDC’s National Program of Cancer Registries (NPCR)), there is no similar system for the more common cardiovascular diseases that affect the nation’s health. The lack of such a system inhibits our understanding of these leading causes of death and the ability to plan and allocate prevention, treatment and research resources. This lack of reliable information is exemplified by the peak and decline of acute age-adjusted coronary heart disease in the mid-1960s, which was not recognized until the late 1970s and was not confirmed or even partially explained until the late 1980s.

It is crucial to have timely and comprehensive information on the ongoing cardiovascular disease burden to better understand disease patterns, presentations, short- and long-term outcomes, treatment, and community level intervention/health promotion programs in an evolving picture. Costs for cardiovascular disease threaten to overwhelm the Medicare insurance system and quality of care is an ongoing issue. Finally, there are large regional differences and inequalities in disease patterns, care and outcomes.

A national surveillance system for cardiovascular disease will provide data for the control of this leading cause of morbidity and mortality. The implementation of a surveillance system is past due.

Recommendations

The workshop participants called for the establishment of a national surveillance system for cardiovascular disease (CVD). To that end, they emphasized the importance of:

Providing unique individual health identifiers to allow linkage across databases and within databases for multiple entries per patient
Establishing and implementing uniform data standards, electronic medical record with data transfer capability, and data interoperability

The group recommended a national CVD surveillance system with state- and local-based sampling and data-reporting process that allows these entities to leverage resources to collect more data at the state and local levels for planning purposes. They also called for the establishment of sentinel surveillance research centers that would serve to validate the data collected, provide additional in-depth individual data collection, and generate novel ideas on methodological research in this area. In short, a framework under which this national CVD surveillance system could operate involves a three-tier approach:

Tier 1: Systematic nation-wide reporting of CVD

Tier 2: Validation and collection of a limited number of standardized core measures and outcomes at selected time points

Tier 3: Detailed hypothesis-driven studies of these patients in specific surveillance centers

It is recognized that the establishment of a comprehensive national CVD surveillance system will likely need to be implemented in a stepwise fashion. For example, similar to strategies initially employed in cancer surveillance, state and local efforts to collect more comprehensive data than at the national level are appropriate next steps. The ultimate goal would be the standardized nationwide collection of a core set of data with additional data collected at the state and local levels tailored to the specific needs of that jurisdiction.

The workshop participants also recommended specific short- and long-term steps that can be taken to improve existing data collection and analysis efforts as well as to move the field toward a coordinated nationwide surveillance program:

Short-term:

Solicit proposals to

Enhance/expand existing electronic medical record linkage

Conduct studies that propose/test creative efficient CVD data collection

Better track out-of-hospital CVD events and outpatient care

Investigate how HIPAA can facilitate efforts at national surveillance
Begin efforts to establish a nationwide individual health ID
Seek more uniformity across electronic medical record system platforms
Enhance and promote the use of existing data collection systems. Examples include

Adding National Committee for Quality Assurance (NCQA) performance indicators to National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS) to better understand delivery of care

Adding health-related quality of life (HRQL) markers to CMS’ Minimum Data Set (MDS) and CDC’s National Nursing Home Survey (NNHS)

Long-term:

Expand existing community-based surveillance to increase geographic and race/ethnic diversity
Implement nationwide individual health identification numbers to allow health record linkage
Seek a mandate for CVD reporting (as currently exists for some infectious diseases)
Implement a nation-wide CVD event surveillance system

Back to Table of Contents

Workshop Summary

Working Group Objectives

Healthy People 2010 goals include primary and secondary prevention through risk factor reduction; improved early recognition and response to out-of-hospital cardiac arrest, acute coronary syndromes (ACS), and stroke symptoms; increases in timely reperfusion treatment; and reduced mortality from coronary heart disease and stroke. Although cardiovascular disease (CVD) remains the country’s leading cause of death, health care dollars spent, and lost productivity, a nationally representative surveillance system to track progress toward these goals does not exist. Current data collection efforts to describe ACS case presentation, treatment practices, and outcomes are limited in their scope, data collection procedures, and generalizability. Ongoing community-based surveillance studies generally have small numbers of predominantly white communities, emphasize disease incidence, and include little longitudinal follow-up. In addition, they are often restricted to in-hospital events and do not capture the evolving detection and treatment of CVD in the outpatient setting. Industry-sponsored registries collect primarily inpatient data as well, from participant hospitals that are mostly urban and academic centers. Despite these limitations, the potential exists for using and linking existing data sources and to use existing systems as models for new and more comprehensive study designs. This workshop was convened to explore these opportunities.

Obtaining high quality nationally representative CVD surveillance and outcomes data is of interest to and ideally will entail cooperation among a broad range of governmental and non-governmental entities. Recognizing this, the planning group that met to explore possible designs for a workshop to address this topic included representatives from the NHLBI, the Centers for Disease Control’s National Center for Chronic Disease Prevention and Health Promotion, and the American Heart Association, in addition to other experts in the field of CVD surveillance. The workshop itself was co-sponsored by these three entities, and the workshop presenters represented the breadth of expertise and stakeholder interest in this topic.

The objectives of the workshop were:

To describe the current knowledge base provided by ongoing registries, surveys, and research studies on the incidence/prevalence of acute cardiovascular events and the clinical management and patient outcomes following these events.
To describe the current data needs concerning the incidence and prevalence of acute CV events and the short- and long-term medical practice patterns, quality of care, and health outcomes following these events.
To consider the appropriate study designs (including potential for linking existing data-gathering infrastructures vs. starting de novo) to address the data needs.

Back to Table of Contents

Workshop Design

The morning and afternoon sessions of the first day of the workshop were structured to address various aspects of the above objectives. Four overview presentations described current knowledge and data needs concerning the incidence and prevalence, medical practice patterns, quality of care, and patient outcomes for sudden cardiac arrest, acute coronary syndromes, stroke, and congestive heart failure. The presentations that followed addressed opportunities and limitations of relevant existing data collection systems including federally-funded registries, various health care delivery systems, quality improvement measurement systems, and other large surveillance systems or studies. Particular operational challenges in collecting such data were also considered.

Workshop participants met in one of three small working groups in the evening session to consider appropriate study designs to address the identified data needs and to develop recommendations from one of three perspectives: hospital surveillance, out-of-hospital surveillance, or assessment of quality of care and outcomes. The working group leaders presented their groups’ conclusions and recommendations the following morning for consideration and discussion among all workshop participants. The last segment of the workshop focused on developing final conclusion and recommendations.

Workshop Agenda and Summary of Presentations
List of speakers and planning group members

Back to Table of Contents

Recommendations of Small Working Groups

A. Out of Hospital Surveillance

This working group concluded that the U.S. lacks national data on CVD incidence and selected outcomes. They described the need to track data on sudden cardiac death, stroke, acute and chronic coronary syndromes (including myocardial infarction, unstable angina, and stable angina) and heart failure. The out-of-hospital settings where data-tracking would be important include emergency medical services (EMS), offices/clinics, emergency departments, and deaths that occur outside of the healthcare system.

The group highlighted several potential barriers that prevent the collection of national CVD incidence and outcomes data. First, the U.S. does not mandate the use of unique individual health identifiers, which is important in allowing individual health information to be tracked longitudinally across healthcare systems. Second, there are inadequate standards and no interoperable electronic systems that permit linking individual data for the entire U.S. population. Having such a system would allow seamless linkage from in-hospital to out-of-hospital settings, across federal agencies, and also among other institutions. Third, no public mandate exists to report CVD data. Finally, there is a need for a national agency to be designated to coordinate and collect such data.

In order to achieve the goal of obtaining national CVD incidence and outcomes data, the group indicated that new data strategies and strategies to optimize existing systems need to be implemented. Standards-based interoperable electronic medical record (EMR) infrastructures should be accelerated to become readily available for all healthcare providers. The group recognized that as medical records become increasingly electronic, it may also be more feasible for CVD to become a reportable disease. Ongoing efforts to standardize definitions for CV diseases and outcomes should be encouraged. Once data are gathered, they should be tested and validated to assure quality and accuracy. A unique health identifier should be provided for all individuals; as technology advances, individuals may also be given health information data cards to carry with them. The group encouraged learning from other models such as those from Canada and Scandinavian countries that have successfully implemented national health identification number systems.

The group discussed 10 year expectations. By then, a coordinated, cost-effective system of CVD surveillance should be established that has core data on the national level, becoming increasingly specific at the state levels, and collecting even more detailed information at the local levels. They also envision widespread use of uniform data definitions, EMR systems, and secure data exchange standards.

The group recommended some immediate steps that can be taken now:

Develop an inventory and map of all existing data collection systems, including those that collect outpatient data.
Hold a meeting with stakeholders to develop a vision document, such as that akin to the “EMS Agenda of the Future.” Additional outcomes from the meeting should include establishing strategies to achieve the vision and determining who should lead the effort.
Continue to standardize definitions for classification of disease and associated outcomes, incorporating standard vocabulary amenable to electronic capture.
Link current existing CVD databases from federal, state, local, and private institutions.
Require institutional compliance with electronic data standards (e.g., Health Level 7, National Health Information Network, and Public Health Information Network) before providing federal funding to grantees/contractors.

The long-term steps recommended by the group include the following:

Provide a unique healthcare identifier to individuals in the U.S.
Have national uniform definitions for reporting clinical data on CVD.
Make CVD encounters reportable as electronic medical record systems become operational.

The group recognized that implementing these changes will require various resources, some of which are fairly expensive. They emphasized that it is crucial to convey the positive return on such an investment in documenting and improving the nation’s health.

Back to Table of Contents

B. Hospital Surveillance

With respect to surveillance of CVD in the hospital setting, this working group unanimously agreed that there is a gap in present knowledge on nation-wide incidence data on CVD, including MI, unstable angina, stroke, heart failure (HF) and atrial fibrillation. The collected CVD incidence data should reflect national demographics (e.g. in age, sex, ethnic/racial distributions). In prioritizing data needs in light of feasibility, the group decided to focus chiefly on MI and stroke (including stroke sub-types).

Possible approaches to capture CVD incidence and outcomes data were discussed. Capitalizing on the expanding use of the EMR is attractive but presently problematic given the incomplete penetration of the use of EMR, which may introduce an unknown degree of bias to the collected data. The following databases are potential data sources: Joint Commission on Accreditation of Healthcare Organizations (JCAHO); National Hospital Discharge Survey (NHDS); and National Health and Nutrition Examination Survey (NHANES). However, it is important to underscore that unique individual identifiers are needed to derive true incidence from these various data sources.

In addition to the initial case finding steps, validation procedures relying on uniform standardized criteria are needed, which could use a sampling approach. Part of the validation procedures could include a limited number of standardized core measures including, for example, atrial fibrillation (for stroke) and other risk factors.

Case identification and outcome measurements for MI and stroke:

For MI, the group recommends using laboratory results of biomarkers, with the understanding that this may be problematic given false positive rates, especially with troponin. However, biomarkers are preferred over reliance either on the electrocardiogram or on physician diagnosis, which is more subjective and more prone to bias (e.g. due to lack of adherence with guidelines or under-ascertainment of post-procedure MI).

For stroke, the group recommends using imaging data as part of the procedures for case finding.

As for events after MI or stroke, the group recommended measuring a) case-fatality rates at selected time intervals and b) recurrent MI or stroke. They felt the other non-fatal outcomes were presently not feasible as part of a nation-wide system. For mortality, there is a need to rely on mortality at a fixed point in time (not in-hospital mortality given temporal declines in duration of hospital stay and likely inter-site variations).

Other disease targets include unstable angina, heart failure, and atrial fibrillation. For unstable angina, the group saw a need to track the invasive procedures percutaneous coronary interventions (PCI) and coronary artery by-pass grafting (CABG). Although it is important to also track unstable angina, they pointed out the difficulty in case definition and that validation procedures are needed before accurate data can be obtained.

The group felt that surveillance of heart failure and atrial fibrillation present unresolved challenges with regards to standardized definitions and event ascertainment. For heart failure, a major challenge remains how to define it in a standardized fashion. One option is to track heart failure with low ejection fraction, though doing so will exclude diastolic heart failure. The group suggested that perhaps heart failure could be the focus of center-specific surveillance research efforts (e.g. Veterans Administration or Kaiser). For atrial fibrillation, the group suggested collecting these data while validating stroke; otherwise, the data may be better collected in the outpatient setting. For unstable angina, a major challenge remains the need to validate the events, as discharge codes are often unreliable.

The group felt that optimizing existing systems is an important intermediate step while progressing towards nationwide mandatory reporting. The current system can be optimized by adding sites to existing surveillance programs that would increase ethnic diversity, leading to the establishment of appropriately diverse surveillance networks. The Veterans Administration’s data system appears attractive for enhancing diversity of surveillance. The group also underscored that having uniform standardized criteria was essential.

The group recommended a 3-tier approach for gathering hospital CVD data nation-wide.

Tier 1: Systematic nation-wide reporting of MI and stroke

Tier 2: Validation and collection of a limited number of standardized core measures and outcomes at selected time points

Tier 3: Detailed hypothesis-driven studies of these patients in specific surveillance centers

The group also called for the establishment of heart disease centers that are modeled after the U.S. national cancer registries (SEER and NPCR).

Back to Table of Contents

C. Assessment of Quality of Care and Outcomes

The collection of data regarding quality of care and patient outcomes should be included in a CVD surveillance program. This is because these data can inform policy makers regarding the need for, and success of, quality improvement initiatives. The translation of science into practice faces many barriers, one of which is the lack of data regarding the success of that translation process. Measurement of quality of care and patient outcomes would provide impetus to quality improvement efforts and a basis for evaluation. Better healthcare delivery and improved patient outcomes would represent a substantial positive return on the investment in a CVD surveillance system.

The working group made specific recommendations to assess quality of care and outcomes in the U.S. by optimizing existing systems applicable to the out-patient and in-patient settings. As an underlying assumption, the working group affirmed that acute coronary syndromes, stroke and heart failure were of primary interest. Time constraints prevented the group from generating specific recommendations concerning sudden cardiac arrest; however, they recognized that surveillance of the quality of care provided to, and the outcomes of, patients served by the pre-hospital care system, including the emergency medical care system and disease management programs, is also of great interest (though substantial work is required in these areas to develop, test and implement systems capable of capturing these important data, the group recommended that such developmental work be conducted and supported).

To improve data collection on outpatient quality of care, the group suggests a) expanding and optimizing CDC’s National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS), and b) making use of CMS data when medication data become available. With NAMCS and NHAMCS, the sample sizes of these surveys could be increased to provide more reliable data regarding cardiovascular conditions. A validation method is needed to ensure accuracy of the CVD data. Performance indicators, such as those from the National Committee for Quality Assurance (NCQA), could be added to the surveys. The group felt that these expansions to NAMCS and NHAMCS, if endorsed by CDC and assuming adequate funding presently available, could be accomplished within a short-term timeline. Regarding CMS data, the group pointed out the need for capturing managed care claims as well as fee-for service claims. Similar to using NAMCS and NHAMCS, data validation and inclusion of performance indicators should be considered. A limitation to using CMS data is that information is predominantly limited to those 65 years or older. The group felt that enhancing CMS data in such a manner may require a long-term timeline.

To better measure outpatient health-related quality of life (HRQL), the group suggested a) expanding and optimizing the use of the Medicare Current Beneficiary Survey (MCBS) or b), alternatively, by expanding the scope of work of CMS Quality Improvement Organizations (QIOs). Either option limits data to mostly those 65 years or older. The group suggested that MCBS be expanded to cover more HRQL measures and be linked to claims data as well as medication use. The sample size may be increased to capture cardiovascular conditions. With acceptance from CMS and assuming availability of adequate funding, the group felt that these changes could be implemented in a short-term timeline. As for expanding the QIO’s role to include collection of outpatient HRQL data from persons recently hospitalized for cardiovascular disease, the group sensed that this process would first require pilot testing to assess long-term feasibility.

As for data on inpatient quality of care and HRQL, the group cautioned against relying on voluntary databases (e.g., National Cardiovascular Data Registry, National Registry of Myocardial Infarction, Get With the Guidelines, CRUSADE, ADHERE) in the long-term. The group recommended the expansion and optimization of a) the Nationwide Inpatient Sample (NIS) database from AHRQ and/or b) CDC’s NHDS. Such data would also need validation, as with other aforementioned databases. Incident versus recurrent events should be distinguished. Performance indicators such as those from JCAHO or ACC/AHA should also be measured. Data to assess the appropriateness of procedures ordered by clinicians should also be obtained if possible. The group felt that these changes, if agreed by AHRQ and CDC, could be accomplished over a medium to long-term timeline. They also felt that inpatient HRQL should be placed as a lower priority, except possibly as a baseline. The committee recognized that inpatient HRQL is generally expected to be poor at the time of discharge, and that the assessment of transitional or outpatient HRQL (e.g., 15-day, 30-day or 90-day post discharge) may be more relevant.

To improve measurement of quality of care and HRQL during the transitional period between hospital discharge and outpatient care, the group recommended a) expanding and optimizing MCBS and, alternatively, b) expanding the scope of work of the QIOs. Once again, the age limit of 65 years or greater applies here. For MCBS, the group noted that sample size could be optimized to provide more reliable estimates of measures for beneficiaries who recently experienced CVD events, a process for data validation is needed, incident versus recurrent events could be distinguished, performance indicators such as those from NCQA could be added, and HRQL could also be included. Unlike expanding the work of QIOs, which would require a longer time line to implement, the group felt that changes to MCBS, if acceptable to CMS and assuming availability of adequate funding could be implemented in a short-term timeline.

The group also discussed long-term care settings. To measure quality of care and HRQL, the group recommends the expansion of CMS’ Minimum Data Set (MDS) and CDC’s National Nursing Home Survey (NNHS). At least MDS, if not also NNHS, already captures some HRQL measurements. They suggested that these databases be optimized to include larger sample size to provide more reliable estimates among persons with CVD events, establish a process for data validation, and include performance indicators such as those from NCQA. If approved by such agencies and assuming adequate funding is available, the group felt that these systems could be optimized within a short time frame.

The group discussed other issues concerning data on CVD quality of care and HRQL. They concluded that high quality, representative data from a national perspective is currently lacking. They recommended that state-based sampling and data collection processes may enable states and local entities to leverage resources to collect more data at the state and local levels for planning purposes. They also suggested the establishment of sentinel surveillance research sites that would serve to a) validate the data collected from these national surveys or claims data, b) provide additional in-depth individual data collection, and c) generate novel ideas on methodological research in this area. They emphasized the importance of establishing and implementing uniform data standards, electronic medical record data transfer capability, and data interoperability. As for carrying out the process to obtain data on CVD quality of care or HRQL, the group also suggested that Request for Applications be created to establish various sites to a) analyze existing surveillance data, b) enhance/expand current systems, and c) develop new methods and systems. They highlighted the need for unique patient identifiers, which would allow linkage across databases. They also pointed out the need to link within the databases, in order to account for multiple entries per patient.

Additional comments provided later from members of the group who were unable to participate in the discussion include the following:
a) Utilization of the National Healthcare Quality Report and the National Healthcare Disparities Report, both of which are prepared annually by AHRQ, would be an excellent method for assuring that quality of care data on CVD are tracked annually, both overall and by race/ethnicity/socioeconomic status.
b) Further investigations are needed to assess how HIPAA can be used, and modified if necessary, to facilitate efforts at national surveillance of CVD.

Workshop Agenda and Summary of Presentations
List of speakers and planning group members

Back to Table of Contents