National Health Data
Warehouse: Issues to
Lynn A. Blewett, PhD, Stephen T. Parente, PhD, Michael D. Finch, PhD, and Eileen
Peterson, RN, MPH
A B S T R A C T maintained and mined by health services
A national data warehouse that links public and private data researchers and analysts for national pol-
icy development and, more recently, to
could be used to monitor trends in healthcare costs, utilization, monitor quality of care. The patient-level
diagnosis related groups (DRGs) of
quality of care, and adherence to quality guidelines and changes Medicare’s Prospective Payment System
(PPS) were created using historical
in treatment protocols. The development of the data warehouse, Medicare claims data. Similarly,
Medicare’s physician payment system –
however, would require overcoming a number of political and the Resource-Based Relative Value Scale
(RBRVS) – was developed using physi-
technical challenges to gain access to private insurance data.
cian-level claims data. In addition, efforts
This article outlines recommendations from a national to identify national trends in costs, quali-
ty of care, and access to care by under-
conference sponsored by the Agency for Healthcare Research and served populations have also benefited
by the use of Medicare claims data.
Quality (AHRQ) on the private sector’s role in quality However, while Medicare claims data
have been critical to the development of
monitoring and provides an operational outline for the Medicare reimbursement systems, there
is no comparable data warehouse for the
development of a national private sector health data warehouse. non-Medicare, under-65 population.
Consequently, there is little national data
K E Y W O R D S W hile the debate about the useful- available to analyze trends in quality and
ness and feasibility of a national health cost for many chronic health conditions
Administrative data data warehouse for private medical care (e.g., heart disease and diabetes) that
National health data claims continues, important questions commonly develop in people under age
remain unanswered, including: What is 65. Attempts to build a national sample
the intrinsic value of a data warehouse? for this population using state Medicaid
Quality monitoring What purpose would it serve? Are the claims data have shown promise, but
Community health benefits worth the costs of development such efforts do not provide nationwide
and maintenance? coverage due to the parochial design of
Analysts and regulators have state Medicaid programs.
systems (CHMIS) addressed these questions using On the private side, only a handful of
System design Medicare claims data. For two decades, a proprietary vendors, including Medstat
national health data warehouse of and Ingenix, hold data warehouses of
Medicare fee-for-service claims has been sufficient scale to be considered national
52 Journal of Healthcare Information Management — Vol. 18, No. 1
in scope. These databases, however, are system. In the early '90s, the Hartford resenting tens of millions of patients.
built using data from a convenient sam- Foundation funded seven sites across the Incorporating private health insurance
ple of clients who are willing to consent country to construct Community Health claims data into a national health data
to sharing their patient data in a de- Management Information Systems warehouse presents a formidable chal-
identified format. (CHMIS). These systems were to serve as lenge, as there are several hundred pri-
This article explores the need for a data networks and repositories to sup- vate insurers and thousands of employ-
national health data warehouse for private port the information needs for their indi- ers who self-insure their patients.
health insurance claims and recommends Furthermore, insurers are currently
actions necessary if such a system is to be focused on compliance with the privacy
established. We begin with a discussion of and security provisions of the Health
the background of the data warehouse Insurance Portability and Accountability
concept. Next, we present the highlights
“…because a number of insurers Act (HIPAA) and are uncertain about
of a national conference sponsored by the have consolidated, only the top how to act even if they wanted to con-
Agency for Healthcare Research and tribute data to a national warehouse.
Quality (AHRQ) on the private sector’s 10 insurers would have to agree Despite these concerns, some recent
role in quality monitoring. We conclude developments make the proposition of
with recommendations for next steps in
to participate to constitute a building a private sector national health
the development of a private sector database in the tens of millions data warehouse more tractable. First,
national health data warehouse. because a number of insurers have con-
of privately insured patients.” solidated, only the top 10 insurers would
The Need for a National Health have to agree to participate to constitute
Data Warehouse a database in the tens of millions of pri-
Currently, a multitude of private insur- vidual geographic regions. The idea was vately insured patients. In addition, sev-
ance plans pay claims for millions of to link the regional data systems to cre- eral of these large insurers already pos-
Americans. This multi-payer system ate a national data repository. Due to a sess research groups that seek to publish
includes for-profit and not-for-profit confluence of political, market, technolo- peer-reviewed results from their databas-
health plans, as well as fully insured, gy and telecommunications develop- es in scientific and industry journals. The
self-insured, and managed care plans. ments, this project was largely unsuc- timing is right, but grassroots interest and
The resulting mosaic of payers and pay- cessful across all seven sites.2 motivation must now be generated to
ment schemes makes it nearly impossible The CHMIS experience, however, was bring such a system to fruition.
to get a clear picture of the healthcare instructive in several ways. First, the con-
financing structure in the United States. cept was too grand. Too many parties Data Specifications
In recent years, there has been some were involved requiring extensive coor- To build a national health data ware-
interest in combining data across pay- dination and consensus, and the upfront house, the initial specifications would
ment systems with a focus on quality investments of time and effort were too consist of three basic components. First,
measurement.1 For the most part, these high and yielded too few tangible healthcare claims data (including both
efforts have been led by the health insur- results. Second, the project failed to fee-for-service and managed care plans
ance industry and by large employers build on existing data structures and sys- data) would form the foundation of the
with programs such as: (1) the widely tems. And third, the project did not con- database and would permit the aggrega-
used Health Plan Employer Data tain a model for use and applications of tion of a single patient’s time-indexed
Information Set (HEDIS) sponsored by the system and the data. health service utilization data across all
the National Committee for Quality Despite the challenges faced by the locations of medical care where insur-
Assurance (NCQA), (2) the Foundation ambitious Hartford experiment, interest ance payment was received. Second, the
for Accountability (FACCT) quality mea- in such a concept continues to grow as data residing in the warehouse would be
sures, and (3) the IMSystem and ORYX the need to understand healthcare costs de-identified to ensure patient privacy by
programs of the Joint Commission for and utilization intensifies. Moreover, replacing the actual patient IDs with
the Accreditation of Healthcare monitoring quality indicators to support encrypted values. This process would
Organizations (JCAHO). These projects value-based purchasing will require a enable patient data to be grouped by
have focused on the creation of discrete, national database with a representation specific treatment episodes across multi-
pre-defined measures of quality and out- of millions of de-identified patients that ple encounters with the healthcare deliv-
comes that cut across health plans and combines public and private sector ery system. Third, providing for future
provider organizations. patients. To engineer this database ability to link clinical results to claims
One ambitious initiative attempted to would require addressing the missing data by medical encounter or laboratory
bring together public and private inter- piece in the mosaic of national data – test would be essential to make the
ests to develop a common information private health insurance claims data rep- database “electronic medical record”
Journal of Healthcare Information Management — Vol. 18, No. 1 53
(EMR) compatible. of design elements. To that end, we type of drug, date and dosage of med-
While we currently lack a model for begin with a discussion of the basic ication dispensed, duration of the pre-
linking clinical and administrative data, strategic design approach and issues. We scription, and amount dispensed.
initial work in linking data from inde- then move on to consider a variety of Pharmacy claims should follow the
pendent clinical labs with ambulatory specific diseases for inclusion in a proto- guidelines produced by the National
claims data has demonstrated the possi- type system. Council for Prescription Drug Programs.
bility of building a database that records Include Patient-level Demographics
not only the use of a particular lab test, through Membership Databases. It is
but also the results of that test. For universally important to include patient-
example, a diabetic patient’s linked clini- level information to enable the appropri-
cal and administrative data would not ate evaluation and comparison of out-
only describe the use of a lab test for
Hemoglobin (HbA1c), but it would also
“While the basic elements are comes. Information on age, gender, and
geographic location is the most readily
contain the clinical value of that test straightforward, the devil in is available. However, conference partici-
(HbA1c), which would indicate poor pants were adamant that data on race,
metabolic health for that patient.
the details.” income, and ethnicity must not be for-
While the basic elements are straight- gotten. Moreover, they acknowledged
forward, the devil in is the details. We System Design Elements that the routine collection and reporting
attempt to address one aspect of the Integrate Information from of these items is enmeshed with con-
detailed analysis necessary to prepare for Multiple Automated Systems. It is cerns about privacy, discrimination, and
the development a national health data important to begin by linking data sys- politics, and such issues require further
warehouse. This paper highlights the tems where feasible, but to also build in consideration and discussion.
results from a workshop on the role of flexibility for additional future linkage as Membership databases should be linked
private sector data in monitoring national technology allows. Currently, we would with claims data and unique identifiers
quality sponsored by the Agency for be able to link medical claims, pharmacy assigned members at the time of enroll-
Healthcare Research and Quality claims, and data on patient demographics. ment. These identifiers should be struc-
(AHRQ). The University of Minnesota, in Start with Medical Claims. Medical tured to allow the grouping of house-
partnership with UnitedHealth Group’s claims or encounter data are collected hold members.
Center for Health Care Policy and from all healthcare sites (inpatient, hospi- Link Patient Data with Finance
Evaluation, conducted this conference in tal outpatient, emergency room, physi- and Organization Variables. To truly
the fall of 2000. This one-day invitational cian offices, surgery centers, etc.) for vir- understand the dynamics of cost and
meeting of national experts focused on tually all covered services provided to quality interactions in healthcare, the
developing a conceptual model and cre- enrollees, including specialty, preventive, delivery and financing infrastructure of
ating an outline for a longitudinal data- and office-based treatment. A claims- the healthcare system must be examined
base using patient-level administrative based system should include claims for and understood. To do this, key indica-
data from the private sector. The objec- inpatient and outpatient services, as well tors of financial performance must be
tive of this meeting was to bring together as submissions by managed care plans. included in any data warehouse system
national researchers and policymakers to Health plans and providers should be to allow comparisons on cost, utilization,
evaluate the feasibility of establishing a able to submit claims either by mail or revenues, and expenditure trends.
private sector managed care data reposi- electronically, and all claims and coding Additional examples of performance
tory for the purposes of monitoring and must conform to industry standards, i.e., information could include: (1) the impact
evaluating changes in healthcare quality ambulatory claims, typically submitted by of health plan organization on perfor-
over time.3 These discussions identified physicians, must use the hcfa-1500 for- mance; (2) the effects of capitation and
several issues relevant to the develop- mat and claims submitted by institutions, payment approaches on provider adher-
ment of a national health data ware- typically hospitals, should be in the ence to and outcomes of practice guide-
house. These issues are presented in fur- ub-82 or ub-92 format. lines; (3) a comparison of utilization
ther detail below. Include Pharmacy Claims. Pharmacy measures by geographic area, diagnosis,
The intent of developing a national claims are critical to understanding or other sub-populations; and (4) the
private sector health data warehouse trends in costs, quality, and treatment price elasticity of demand for specialist
would be to link public and private data patterns. Claims for pharmacy services care. Again, attention must be directed
over time in an effort to monitor trends are typically submitted electronically by toward what is currently available and
in healthcare costs, utilization, quality of the pharmacy at the time a prescription what can be added over time.
care, and adherence to quality guidelines is filled and should be incorporated into Include Medical Errors Reporting.
and changes in treatment protocols. any private sector data system. Each Based on the recent Institute of Medicine
These goals should drive the discussion pharmacy claim specifies the pharmacy, (IOM) reports on medical errors and
54 Journal of Healthcare Information Management — Vol. 18, No. 1
patient safety,4, 5 a great deal of focus has from administrative records would be a reporting, and links to other data sources
been placed on medical error reporting more efficient way to monitor hyperten- for the analysis of factors contributing to
and follow-up. Again, conference partici- sion than expensive and labor-intensive higher rates of surgical error.
pants expressed a fair amount of con- retrospective medical record chart Depression. Perhaps because it is so
cern that any data systems developed to reviews. A potential incentive to encour- difficult to define, report, and monitor,
monitor medical errors not be indepen- age plans to include specific measures in another area that is often overlooked,
dent of ongoing quality initiatives. under-treated, or undiagnosed is depres-
Rather, the group agreed, that data col- sion. It may be relatively easy to identify
lection and reporting activities should be patients with depression diagnoses, and
integrated. then to evaluate whether they have
Focus Must Include Prevention.
Consideration should be given to data “HIPAA regulations received appropriate care. However, it is
extremely difficult to identify patients
elements that could be used to highlight who may be suffering from undiagnosed
prevention services based on existing
requiring the adoption of data depression. Data elements to focus on
guidelines and protocols. For an standards and administrative include prescription drug and dosing
employed, privately insured patient pop- information, monitoring changes in prac-
ulation, preventive services are a key simplification practices are tice and treatment, findings of depression
component of a quality monitoring sys- on routine health risk appraisals, and
tem. Such measures form the basis of
smoothing the way for the information on benefit sets, carve-outs,
ongoing quality monitoring systems, such creation of consolidated and employment programs such as
as HEDIS. Prevention includes screening employee assistance plans. These mea-
for PSA, breast cancer, immunizations for data systems.” sures could help address issues of
children, flu shots for those at risk, as absence of care, meaning that services
well as smoking prevention and health might have been provided through other
education programs. Prevention is critical their data files would be to exempt plans mechanisms. Efforts should focus on
to maintaining and improving the health that routinely submit these data from using existing data on diagnosis and
status of populations. HEDIS chart review. While blood pres- treatment and adding health status
Focus on a Predefined List of sure information is currently not collect- assessments or screeners over time.
Conditions. Workshop participants ed as part of an administrative data set, Heart Attacks. Workshop partici-
agreed that the list of conditions to con- there are initiatives to add some new pants agreed upon one health condition
sider for inclusion in a national private CPT codes, as blood pressure values where good data and indicators are cur-
sector health data warehouse should ini- would certainly be useful for monitoring rently available in existing data structures
tially be short and targeted. Moreover, hypertension care. – heart attack. Further, they agreed that
the conditions selected should represent Patient Safety. While avoiding med- future efforts should focus on how a pri-
high-volume, resource-intensive health- ical errors and ensuring patient safety are vate sector national health data ware-
care situations where evidence shows clear priorities, these measures are house initiative could augment the ongo-
that interventions can make a difference. among the most difficult to tackle at the ing work based on Medicare claims data.
Selection of conditions should include outset. The workshop participants rec- This issue appears to have a lot of
priority for conditions with the greatest ommended two possible areas to address momentum, and it might be a place
disease burden. In addition, the system in this area: (1) drug dosing and dispens- where quick turnaround of valuable data
should start with conditions where data ing, and (2) surgical procedures. In the may be available for reporting purposes.
is currently available and expand over area of drug dosing/dispensing, adverse Specific data elements to focus on
time to include other data linked to pri- drug reactions (ADRs) were identified as include: (1) medications after discharge,
ority conditions. an important area to address. The group (2) the use of beta-blockers, and (3) the
reached some agreement that govern- use of aspirin.
Recommendations on Priority ment leadership may be required to Basic Prevention Screening. Basic
Conditions encourage more rapid development of screening information should be routine-
The list of priority conditions could interventions in this area. It is conceiv- ly collected and reported. In addition,
vary based on the specific focus of the able that a data system could be built on practice guidelines for time frame, age,
system. In an effort to link cost and existing algorithms utilizing drug refill and risk factors should be followed.
quality monitoring the following list of schedules, potentially harmful drug inter- Screening protocols for the younger pop-
conditions is presented.6 actions, and data on prescriptions filled. ulation should include hearing and
Hypertension. While HEDIS does Surgical errors represent another area of vision screening, blood lead-level screen-
include a measure of controlled hyper- great concern to the public. There are ing, high diastolic blood pressure
tension, the automated collection of data opportunities for effective monitoring, (HDBP), antibiotic history, and obesity.
Journal of Healthcare Information Management — Vol. 18, No. 1 55
Screening protocols for an older popula- an increasing competitive managed care the development of query software to
tion should include cancer screening, environment. Information on the charac- answer key “what if” policy questions in
obesity, and depression. teristics of the health plan and/or health- a timely fashion.
care system should be linked with indi- Use an Incremental Approach. The
Recommendations for Next Steps vidual measures of quality. An effective development of a private sector quality
Despite the multitude of barriers quality monitoring system will include monitoring system that links health plan
described above, monitoring quality in system measures that currently exist and data over time is an ambitious and
the private sector is a timely and appro- develop new indicators over time. Key important task. Any such initiative
priate undertaking. Developing a proto- should start small and use an incremen-
type system to assess the practicality and tal approach, building on existing data
feasibility of a comprehensive national systems and reporting structures. The
health data warehouse infrastructure is a goal is to develop a usable, effective,
proposal that bears consideration for a data-driven system starting with existing
number of reasons. First, information
technology has advanced to where such
“Too often, legislative and administrative data at a reasonable cost.
Once a small-scale prototype is devel-
a system is not only feasible, but also policy initiatives are driven by oped and its usefulness demonstrated,
practical. Second, HIPAA regulations additional plans and data elements could
requiring the adoption of data standards
anecdotes, instead of careful be added over time. For example, a sys-
and administrative simplification prac- consideration of available data.” tem built with existing inpatient data
tices are smoothing the way for the cre- could be augmented by patient-level
ation of consolidated data systems. And and ambulatory care system data when
finally, we have seen demonstrated indicators of financial performance may it becomes available.
momentum and interest in the develop- include such measures as health plan Start with Macro-Level Indicators.
ment of effective tools to measure quali- cost, utilization, source of revenues, and A national private sector quality moni-
ty through a public-private collaborative trends in health expenditures. Some toring system should focus on major dis-
approach. Such a system can be viewed examples of performance information ease categories and start with a broader
as beneficial for the public good, as well include the effects of capitation and pay- (macro-level) set of data and informa-
as a means for improving the quality of ment approaches on quality indicators; tion. The system should focus on those
care delivered through private healthcare comparison of quality measures by geo- health conditions that represent the
systems. The following recommendations graphic area, diagnosis, or other subpop- largest disease burden, thereby enabling
address the strategic next steps required ulation of patients or health plans; the efficient, targeted data collection and
to move this collaboration forward and impact of health plan organization on monitoring. While it will be impossible
to assure success. specific quality performance measures; to obtain the universe of data on all
Establish Clear Concise Goals. and quality differences between type of conditions, the intent should be to iden-
Many quality initiatives are under way health plan structure and for-profit and tify a midpoint between having no sys-
across the country, so it is important that not-for-profit. tem and a perfect system and work
any new initiative clearly articulate its Develop System that Can Address from there.
goals and objectives, as well as build Relevant Health Policy Questions. Build the Cost of Data Collection
upon the existing research base. Too often, legislative and policy initia- into the System. If costs are to be
Additionally, while a private initiative tives are driven by anecdotes, instead of borne by private sector health plans and
can make a unique contribution, a clear careful consideration of available data. A providers, then the cost estimates for the
sense of that unique contribution must national quality monitoring health data system should include the cost of data
be articulated from the outset. Lack of warehouse would facilitate analysis in a collection. Electronic data interchange
clear goals may weaken an initiative’s timely and efficient manner, thereby (EDI) in the health arena is already
focus and contribute to the decline of enabling analysts to inform the policy being developed in the private sector.
the system’s sustainability over time. debate by answering pertinent health As systems are built, the data and infor-
Experience has shown that it is difficult policy questions in real time. Such a sys- mation necessary for quality monitoring
to justify a system’s values and costs if tem would require ongoing input from should be assessed, and a fee for access
the initiative has promised more than it the policy arena to identify: (1) key poli- to and use of the system could be
can deliver. cy questions, (2) local and national imposed. Additionally, a health plan
Link Patient and Plan Data at the trends in the health policy debate, and may choose to contribute to the cost of
Outset. The delivery system continues to (3) data and information that could be a system if the data and reports generat-
evolve in the forms of new payment used to address emerging health policy ed are found to be of value in terms of
schemes, clinical management systems, issues. One possible mechanism identi- specific quality improvements or other
vertical and horizontal integration, and fied by conference participants would be business applications.
56 Journal of Healthcare Information Management — Vol. 18, No. 1
Involve Data Collection Subjects the model presented here. Such a proto- data be analyzed and released as quickly
Early in the System Design Process. type should be designed from a clinical as possible. If the data suppliers recog-
Having buy-in at the local level will and social perspective to demonstrate nize its value and use it from the begin-
ensure that a private sector quality moni- the value of a quality monitoring system. ning, they will be more likely to contin-
toring system can be implemented and An important part of this process ue to submit data to the warehouse. In
will be used. Buy-in requires consensus involves establishing benchmarks to addition, the release of summary data
building, information exchange, and edu- measure differences in the data, identify should be obligatory, regardless of
cation. It is partly a marketing function, gaps, and develop a plan for moving whether it is favorable to the industry,
but it is also a mechanism to obtain beyond limited health plan participation providers, or the system as a whole. As
community input toward building a sys- toward broader representation. mentioned earlier, the data products
tem that will meet both public and pri- Do Not “Re-invent the Wheel.” This would include chart books, comparison
vate needs for quality monitoring. private sector quality monitoring system tables, and Web-based information.
Build in Opportunities for should be developed in the context of the
Linkages to Other Data Sources at many ongoing quality initiatives men- Conclusion
the Outset. Although it is likely that any tioned earlier in this report. Because there This article provides an operational
quality monitoring system will begin are many overlapping interests between outline for a national health data ware-
with administrative data, it is important the public and private sectors, we are house designed to detail specifications
to think about linkages to other data hopeful that this initiative would inform that, if met, would generate a resource
sources and surveys. Recently, promising the debate and make a unique contribu- of genuine value. While the challenges
strides have been made in the develop- tion to development of an effective pri- to build such a database are certain,
ment of automated medical records sys- vate sector quality monitoring system. they are not insurmountable. These
tems, physician practice management Take Advantage of “Quick-Hit.” An specifications assume the political and
systems, and linkages with population- early demonstration of the value of technical challenges to gaining access to
based surveys and data repositories. The these data will help build credibility private insurance data can be overcome.
union of detailed cost and utilization with providers and consumers. In addi- With this assumption, a practicality-
data, with patient- and member-reported tion, communication efforts can set the minded, limited, but focused scope of
perceptions, represents a powerful com- stage for future advancements. The work has been outlined to demonstrate
bination for analysis and research. In value of the health data warehouse can what could be achieved and how it
addition, there may be future opportuni- be communicated via the dissemination might be completed.
ties for follow-back surveys. Where con- of short summary reports, chart books of
tents of the database repository are data, quality measures reported by geo- About the Authors
insufficient to answer the questions iden- graphic area, etc. Reporting mechanisms Lynn A. Blewett, PhD, is an assistant
tified in the research requirements, addi- must be timely and relevant, and the professor in the Division of Health
tional data would be collected based on data included in these reports must be Services Research and Policy of the
unit samples from the existing repository. made available in an understandable School of Public Health at the University
Build in as Much Transactional and accessible format. of Minnesota in Minneapolis.
Data as Possible. An ideal system Establish Linkages Early with Stephen T. Parente, PhD, is an assis-
would be built upon detailed patient- Other Groups. To enhance the viability tant professor of healthcare management
level data at the level of patient-provider of the proposed national private sector at the Carlson School of Management in
transaction. This may not be possible at quality monitoring system, we must the University of Minnesota.
present, but, in the future, models and endeavor to move quickly through the Michael D. Finch, PhD, is a senior
data systems may allow this type of prototype phase, and through the initial researcher at the Center for Health Care
information to be utilized. In the mean- start-up activities. One goal would be to Policy and Evaluation at UnitedHealth
time, the system should take advantage get the community of health plans to Group in Minnetonka, Minn.
of technological advances to build a pro- break through organizational barriers in Eileen Peterson, RN, MPH, is a vice
totype based the finest level of transac- an effort to organize around quality. president at the Center for Health Care
tional data currently available. To effect Another opportunity involves linking to Policy and Evaluation at UnitedHealth
real quality improvement, data systems emerging private sector networks or to Group in Minnetonka, Minn.
must both collect information at the existing health plans that cover signifi-
patient-provider level and report infor- cant numbers of enrollees. These link- Note: This research was funded by a
mation at a “rolled-up” level for ongoing ages are essential for broader applica- conference grant from The Agency for
monitoring purposes. tions and will require considerable Healthcare Research and Quality (AHRQ
Develop and Test Production investments of time and communication. R13 HS10091-01). The views expressed
Prototype. A prototype database should Release Data in a Routine and are those of the authors and do not nec-
be developed to test the applications of Timely Fashion. It is important that essarily reflect the views of AHRQ.
Journal of Healthcare Information Management — Vol. 18, No. 1 57
1. The Challenge and Potential for Assuring Quality Health Care for the 21st 4. Kohn, L. T., Corrigan, J. M., and Donaldson, M. S., Eds. To Err Is Human:
Century. Washington, D.C.: U.S. Department of Health and Human Building a Safer Health System. Washington, D.C.: Institute Of Medicine,
Services for the Domestic Policy Council, 1998. 2000.
2. Starr, P. “Smart Technology, Stunted Policy: Developing Health Information 5. Crossing the Quality Chasm: A New Health System for the 21st Century.
Networks.” Health Affairs, 1997, 16: 91-105. Washington, D.C.: Institute of Medicine, 2001.
3. Blewett, L., Finch, M., and Peterson, E. Private Sector Quality Monitoring, 6. See reference 3.
Final May 2001 Report. Minneapolis: Submitted to the Agency for
Healthcare Research and Quality, 2001.
58 Journal of Healthcare Information Management — Vol. 18, No. 1