Lessons Learned in Using Hospital Discharge Data for State and National Public Health Surveillance

Abstract

The goal of the Centers for Disease Control and Prevention Environmental Public Health Tracking (EPHT) Program is to build a nationwide network of integrated health and environmental data to measure impact of environmental factors on public health. This article describes how hospital discharge data can provide essential information for public health programs, including EPHT. The state inpatient hospital discharge data systems have properties that are highly desirable for surveillance and multistate initiatives, like EPHT, yet accessing and using the data can create challenges for the end user. This article highlights the strengths and limitations of hospital discharge data and references crash outcome data and evaluation system and Healthcare Cost and Utilization Project as models for accessing, linking, and aggregating hospital discharge data. These federal-state data partnerships have overcome many of these challenges and have the potential to serve as models for the EPHT Program. The lessons learned from these "early adopters" can shortcut the implementation period for the Centers for Disease Control and Prevention EPHT Program.

Full text

Lessons Learned in Using Hospital Discharge Data
for State and National Public Health Surveillance:
Implications for Centers for Disease Control and
Prevention Tracking Program
Denise Love, Barbara Rudolph, and Gulzar H. Shah
rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr
The goal of the Centers for Disease Control and
Prevention Environmental Public Health Tracking (EPHT)
Program is to build a nationwide network of integrated
health and environmental data to measure impact of
environmental factors on public health. This article describes
how hospital discharge data can provide essential information for
public health programs, including EPHT. The state inpatient
hospital discharge data systems have properties that are highly
desirable for surveillance and multistate initiatives, like EPHT, yet
accessing and using the data can create challenges for the end
user. This article highlights the strengths and limitations of
hospital discharge data and references crash outcome data and
evaluation system and Healthcare Cost and Utilization Project as
models for accessing, linking, and aggregating hospital
discharge data. These federal-state data partnerships have
overcome many of these challenges and have the potential to
serve as models for the EPHT Program. The lessons learned
from these “early adopters” can shortcut the implementation
period for the Centers for Disease Control and Prevention EPHT
Program.
KEY WORDS: environmental public health tracking, hospital
discharge data, public health surveillance, record linkage
The primary mission of the Centers for Disease
Control and Prevention (CDC) Environmental Public
Health Tracking (EPHT) Program is to provide infor-
mation from a nationwide network of integrated health
and environmental data to measure impact of envi-
ronmental factors on public health. By systematically
J Public Health Management Practice, 2008, 14(6), 533–542
Copyright C
2008 Wolters Kluwer Health |Lippincott Williams & Wilkins
obtaining and linking environmental and health data,
the EPHT Program will inform policy and practice as
well as enhance the local analytic workforce and infor-
mation infrastructure.1The EPHT has recognized that
environmental data alone cannot document the mor-
bidity, outcomes, and financial impact of environmen-
tal exposures, committing to linkage of environmental
data with other datasets, such as health outcomes data.
When multiple data sources are linked, the result is a
source of population-based data available for state and
national research and policy decisions. Registry and
other manually abstracted data sources are expensive
to develop and maintain. National data sources, such
as the National Center for Health Statistics’ National
Hospital Discharge Survey, Agency for Healthcare
Research and Quality (AHRQ), and Nationwide Inpa-
tient Sample, are not sufficient for community-level
estimates because these datasets are samples and,
therefore, do not support substate- or community-
level analysis.2Therefore, statewide hospital discharge
databases are an important source of morbidity and
outcomes data for the EPHT Program, providing infor-
mation about health outcomes associated with environ-
mental factors.3,4
For more than two decades, state hospital dis-
charge data systems have informed researchers, policy
Corresponding Author: Denise Love, RN, MBA, National Association of Health
Data Organizations, 448 East 400 South, Suite 301, Salt Lake City, UT 84111
(dlove@nahdo.org).
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
Denise Love, RN, MBA, is Executive Director, National Association of Health Data
Organizations, Salt Lake City, Utah.
Barbara Rudolph, PhD, is Consultant, National Association of Health Data
Organizations, Salt Lake City, Utah and the LeapFrog Group, Washington, District of
Columbia.
Gulzar H. Shah, MStat, MSS, PhD, former Director of Research, National
Association of Health Data Organizations, Salt Lake City, Utah.
533

534 ❘Journal of Public Health Management and Practice
makers, purchasers, providers, and consumers to im-
prove healthcare cost, quality, and access decisions.
As the only source of hospital utilization data that in-
clude all patients and all payers from all short-term-care
providers in a state, hospital discharge data support a
growing number of community and national health in-
formation initiatives. The number of states with health
data programs is growing. In 2007, more than 47 states
and jurisdictions have implemented hospital inpatient
reporting systems and many of these states have ex-
panded reporting to include ambulatory surgery (28)
and emergency department (ED) data reporting (25)
from short-term-care providers in their states.5State
health data programs have consistently improved data
completeness and quality, enhancing their discharge
data by including important data elements such as
race and ethnicity, external cause of injury codes, and
present-on-admission (POA)∗indicator.†The addition
of a select set of electronic laboratory values is the lat-
est of the innovative additions to discharge data—it is
currently being piloted in several states.6
Despite the challenges in accessing and using the
state hospital discharge data, there are positive trends
that reflect the growing use of these data for local
and national initiatives. During the 1990s, hospital dis-
charge data reporting expanded significantly and these
systems have been the foundation of two significant
federal-state data partnerships: the AHRQ’s Healthcare
Cost and Utilization Project (HCUP) and the National
Highway Traffic Safety Administration’s (NHTSA’s)
crash outcome data and evaluation system (CODES).
Lessons learned in state hospital discharge data system
implementation in HCUP and CODES can inform an
emerging federal-state initiative, such as EPHT.
This article highlights both the benefits and the chal-
lenges associated with accessing and using hospital
discharge data for public health surveillance and in
federal-state initiatives. These observations are based
on the HCUP and CODES experiences and on more
∗Present on admission (POA) is defined in the ICD-9-CM October
1, 2007 Official Guidelines for Coding and Reporting as present
at the time the order for inpatient admission occurs—conditions
that develop during an outpatient encounter, including emer-
gency department, observation, or outpatient surgery, are con-
sidered as present on admission.
†The NAHDO estimates that 28 health data programs capture
race/ethnicity and has worked with states to adopt a national
standard. The Centers for Medicare and Medicaid Services will
require present-on-admission indicator for each diagnosis for all
Medicare inpatient hospital claims by January 1, 2008. The Na-
tional Uniform Billing Committee has approved POA and an
estimated seven health data programs are requiring POA. The
POA will add precision to ICD-9-CM (International Classification
of Diseases, Ninth Revision, Clinical Modification) coding in admin-
istrative data because it would distinguish between preexisting
conditions and complications (AHRQ, 2007).
than two decades of experience by the National As-
sociation of Health Data Organizations (NAHDO) in
providing technical assistance to states implementing
hospital discharge data-reporting initiatives. We hope
that this article will facilitate local data partnerships
among state health data organizations, EPHT grantees,
and the CDC, and encourage other public health pro-
grams to use hospital discharge data for surveillance
and community assessment.
●The EPHT Program and Hospital
Discharge Data
As the EPHT Program begins to explore the link be-
tween environmental risk factors and major causes of
morbidity and mortality associated with avoidable or
preventable diseases such as cancer, asthma, heart dis-
ease, and diabetes, multistate healthcare data will be
needed to enhance existing national data and measures,
including substate, state, regional, and national com-
parisons. Of the states and cities funded by the CDC
EPHT Program, all have mandatory or voluntary hos-
pital discharge data programs in their states.5By es-
tablishing data partnerships with the state health data
program, or data steward, the EPHT projects will have
access to a source of important information about clin-
ical care and conditions, serious enough to warrant
hospitalization. The information provides proven tools
for targeting prevention efforts7and opportunities for
community improvement for numerous health status
measures known to be conditioned by environmental
factors including adult and pediatric asthma, myocar-
dial infarction, and carbon monoxide poisoning.8–10
The linkage of environmental data with other
datasets containing health outcomes and health sta-
tus leverages existing state data, thus reducing data
collection costs. Linking across datasets helps improve
data quality across the linked datasets and fills impor-
tant information gaps.11,12 When multiple data sources
are linked, a more valuable source of population-based
data is available for state and national research and pol-
icy decisions. Asthma is an example of how a federal-
state data EPHT partnership can assess and improve
community health status. Asthma is a long-term condi-
tion with significant morbidity and cost; hospital dis-
charge data can provide essential return on investment
information about the need for prevention and the
costs associated with doing nothing. In 2004, there were
an estimated 418 789 discharges nationally from acute
care hospitals with the principal diagnosis of asthma.13
A total of 35 percent of these admissions were pedi-
atric hospital admissions (ages 0–17). It is well estab-
lished in the research literature that many of these inpa-
tient admissions could have been avoided with proper

Using Hospital Discharge Data for Surveillance ❘535
management of the condition, especially timely access
to outpatient care.14,15 Environmental exposures such as
CO2and sulfur emissions, secondhand tobacco smoke,
house dust mite, allergens, and other sources of air pol-
lution predispose people to asthma.16 Scientific studies
have established that such environmental hazards trig-
ger and worsen symptoms of asthma; without appro-
priate primary care treatment ED visits, and inpatient
hospitalizations are more likely.17,18 As we learn more
about the causes of asthma, including links to environ-
mental causes such as molds and poor air quality, the
opportunities for intervention are compounded. Yet,
because of restricted budgets, it is also critical to pro-
vide evidence of the return on investment of prevention
activities. This is just one example of uses for track-
ing; public health researchers have examined other is-
sues, including the costs (inpatient hospitalizations and
mortality) associated with cryptosporidium found in
contaminated city water. The total cost of the 1993
Milwaukee outbreak-associated illness was $96.2 mil-
lion: $31.7 million in medical costs and $64.6 million
in productivity losses.19 Hospital discharge data com-
bined with other environmental sources can motivate
change in the community, thereby reducing the envi-
ronmental risks.
Hospital discharge datasets will facilitate the EPHT
Program’s ability to link health outcomes and environ-
mental data, but accessing and using these data are not
without challenges.
●Federal-State Data Partnerships: Lessons
Learned to Inform EPHT
Data partnerships are important to improving the qual-
ity and utility of state hospital discharge data. While
the challenges of a multistate data initiative are daunt-
ing, they can be overcome. The feasibility of acquir-
ing and aggregating state health datasets, across states,
has been demonstrated by various federal-state data
partnerships. The HCUP (a federal-state-industry part-
nership sponsored by AHRQ) and the NHTSA CODES
are examples of national federally driven projects bene-
fiting from consolidating statewide hospital discharge
data and linking these data to other data sources to
produce information for research and to inform pol-
icy. The HCUP and CODES initiatives both collect and
use statewide data from multiple sources for analysis,
research, and reports. For the purposes of EPHT, the
CODES and the HCUP are most notable because of the
following common attributes:
•The reliance on existing datasets and not registry or
abstracted data,
•The aggregation of data across states and produce
state and national statistics,
•The release of aggregated, de-identified data avail-
able for research, policy, and public use,
•The provision of funding for state data, either
through data purchases (HCUP) or in the case of
CODES, cooperative agreement funding from the
NHTSA,20 and
•Incremental expansion beginning with a few states
and expanding state participation as the program
matures.
These projects have the potential to inform the EPHT
Program because of the long-term partnerships and
support they have garnered with data stewards across
states. Both HCUP and CODES have had to address
the issues of data access and rerelease as well as the
variation in data formats and quality across datasets
and across states. Understanding the structure and ap-
proaches of the two projects will be informative for the
CDC EPHT Program and other public health surveil-
lance activities.
The HCUP is funded, managed, and controlled by
the AHRQ, which manages and conducts the project.
The HCUP databases bring together the data collec-
tion efforts of state data organizations (including state
agencies, hospital associations, and private data orga-
nizations) and the federal government to create a na-
tional information resource of patient-level healthcare
data and software and analytic tools.21 The HCUP fo-
cuses almost entirely on hospital discharge datasets.
The AHRQ purchases state all payer, encounter-level
hospital inpatient, ambulatory surgery, and ED dis-
charge data from state and private health data pro-
grams and converts the state data into a uniform HCUP
format for research and public use. The HCUP has
grown from just more than 20 state inpatient databases
in 1988 to 38 states in 2007.5The HCUP approach has
been primarily a data user or customer model, purchas-
ing state health datasets from the data stewards, which
are recruited to become HCUP partners. The HCUP
provides a state dataset in uniform HCUP format back
to each data supplier and has created methods and tools
that states can use at no cost.
The CODES has taken a different approach. The
NHTSA has awarded participating states competitive
cooperative agreements and stipulates the require-
ments for participation. Believing that data linkage
fulfils expanded data needs without the additional
expense and delay of new data collection, CODES
grantees are required to link statewide population-
based crash to injury data, validate the linkages, and
analyze the linked data.22 The linkage improves data
quality and states benefit from state-specific injury
and financial outcome information about motor vehi-
cle crashes.12 The CODES funding provides resources
to build capacity in states to institutionalize data

536 ❘Journal of Public Health Management and Practice
linkage and establish local collaborations. The NHTSA
supports data aggregation and linkages across state
datasets such as death data, inpatient and ED discharge
data, emergency medical services data, crash outcomes
data, and drivers’ licenses to obtain a complete and ac-
curate assessment of the crash event.10,20 The data are
then standardized and disseminated to target interven-
tions at the local and national levels and inform policies
and facilitate research. Each CODES state has devel-
oped its own research and reporting agenda, yielding
useful information. Early efforts were merely directed
to standardizing the procedures for linking data and
less toward standardizing the reporting of CODES in-
dicators and data, which has become a focus in recent
years.23 Because of data linkage across datasets, CODES
has had to develop sound processes and tools accept-
able to multiple data stewards.24 Although the CODES
model of funding local initiatives may be more labor-
intensive than HCUP’s centralized, data acquisition ap-
proach, the CODES model may be more aligned with
the EPHT Program with its linkage across various state
datasets.
Regardless of the approach to evolve to its fullest
potential, EPHT, like HCUP and CODES, will have
to gain state data steward support and adapt to ex-
isting data standards and formats to sustain the pro-
gram. The HCUP and CODES have negotiated data use
agreements that address the complex issues associated
with data acquisition and use, including the securing
of data use agreements with health data programs and
institutional review board approval, where applicable.
Both projects have produced analytic tools and reports
that are useful to the state data stewards. The CODES
has institutionalized local linkage infrastructures and
works with data suppliers to improve data quality and
advocate for improvements. It is significant that both
HCUP and CODES have incrementally expanded state
data recruitment and participation over time. These na-
tional projects concentrated on building data partner-
ships and achieving successes and analytic utility with
a smaller group of states, adding other states as the
projects matured.
Involving the data stewards in the implementation
of the initiative is an important component to build-
ing a data partnership, and HCUP and CODES have
made data partnerships a priority. The HCUP partners
are the public and private data stewards who are con-
sulted on various aspects of the project. The NHTSA
requires local CODES grantees to form local advisory
boards comprising data stewards and other stakehold-
ers to establish a shared decision making and to insti-
tutionalize data linkage and data sharing locally.20
In summary, both the HCUP and CODES pro-
grams have acquired knowledge of certain key ele-
ments for establishing collaboration between federal
and state agencies that can inform the EPHT Program.
Although the CDC EPHT Program is still develop-
mental and in the early stages of formation, under-
standing the key elements of a federal-state data part-
nership, based on HCUP and CODES, is important.
These include a sound process for recruiting and ac-
quiring state datasets, gaining data steward support
through capacity-building, and analytic tool develop-
ment. Like HCUP and CODES, EPHT data partners
should be willing to accept the data and use the viable
data elements, while suggesting improvements to oth-
ers. Proper understanding of the various datasets and
their limitations leads to their responsible use. Mak-
ing the data partnership a win-win, through shared
analytic resources, funding for the data steward, and
acknowledging the data in all publications will facili-
tate a successful partnership, as we have learned from
early adopter initiatives, such as the AHRQ’s HCUP
and the NHTSA’s CODES programs. As we have seen
with CODES and HCUP, states are willing to participate
in multistate data initiatives and work to fill data gaps
and improve the quality of key data elements (such as
E-coding) and enhance their data with new data ele-
ments, such as race/ethnicity or present on admission.
Summarized in Table 1 are the common barriers to
access to, and use of, state data resources and solutions
proven effective to HCUP and CODES. We anticipate
that similar barriers and solutions will apply in EPHT
data acquisition and use activities.
●Understanding Hospital Discharge
Databases
For a dataset to be useful for local, state, and national
comparisons, the data must be accessible, of reasonably
high quality, automated, and linkable. Although hos-
pital discharge data meet these requirements, there is
unique information and challenges for potential users,
including the EPHT Program. Because the data cap-
ture information about a patient’s demographic char-
acteristics, diagnoses, procedures, source of payment
(including self-pay and uninsured), and utilization for
every patient discharged from an acute care hospital in
a state, they have proven to be useful for a wide range
of applications, including public health surveillance.25
All data systems have strengths and weaknesses and
hospital discharge data are no different. The major
strengths and limitations fall into general categories
of data availability/accessibility, reliability and consis-
tency, analytic utility, and enhancement potential.
Availability/accessibility
Availability and accessibility of inpatient and ED data
are critical issues for EPHT applications and present
perhaps the greatest barrier, especially when access-
ing patient-level identifiable data from the state health

Using Hospital Discharge Data for Surveillance ❘537
TABLE 1 ●Lessons learned in overcoming barriers to hospital discharge data partnershipsa
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
Barriers Solutions Examples
Gaining access to detailed,
confidential data
Prepare an overview and a summary of the project and
present it to the data steward and data stakeholders
CODES projects visit the data steward in person. CODES
may offer analytic support and funding in return for data
access
HCUP prepares and presents prospective partners with
project materials—present to stakeholders as needed
Avoid turf issues and sustain
support for the project over time
Include the data steward staff, preferably the decision
makers, in project advisory roles to develop project
policies, provide input in report development and
interpretation. No surprises
CODES projects often have local advisory panels with data
stewards playing a key role
HCUP has the HCUP partners to provide input to the project
Processing a data request in a
timely manner
Understand your “ideal” data needs as well as the
structure of the data steward’s data. Involve the data
steward in developing the data use agreement to
ensure that legal concerns are addressed
CODES negotiates the data acquisition process and
timeline in an in-person meeting with the data steward,
including institutional review board review cycles
Missing data elements or data
quality concerns
The data will not be perfect, nor will formats be
uniform. Involve the data steward closely in
interpreting the data quality. Work closely with the
data steward to improve future data and formatting.
Offer to advocate for changes
CODES accepts any format and offers suggestions for
future formats and data element improvements
HCUP also accepts any format, normalizing states’ data
into a HCUP common format. Data feedback reports to
partners are shared
Concerns about ownership of the
linked dataset
The data use agreement should address these
concerns
CODES has established a policy that databases, linked or
unlinked, are owned by their original owners and that
the data use agreement governs release for each
component dataset
Political concerns about the use of
hospital data
Invite all stakeholders to participate in interpretation of
data and provide input on reports
HCUP convenes annual partners meetings
Abbreviations: CODES, crash outcome data and evaluation system; HCUP, healthcare cost and utilization project.
aFrom AHRQ HCUP (http://www.ahrq.gov/data/hcup/) and CODES (http://www.nhtsa.dot.gov/portal/site/nhtsa/menuitem.9fef9613e59b4dd24ec86e10dba046a0/).
data program. Understanding how discharge data are
maintained is important to establishing a data partner-
ship and when conducting a multistate project, such as
EPHT. The disclosure of de-identified and individually
identifiable health information is governed by state and
federal laws. Although the data stewards are usually
public entities, they may also include private entities
such as hospital associations. In general, there are three
models of governance for state health data agencies,
which may impact release of data and allowable uses:
1. a state agency with a legislative mandate collecting
data,
2. a delegated authority, such as a hospital association
or private entity, collecting data under a state man-
date, and
3. a private agency, usually a hospital association, col-
lecting the data voluntarily from its members or com-
munity hospitals.
In 38 states that collect discharge data under a state
mandate (numbers 1 and 2 above), the release of identi-
fiable data is governed by state laws, which are usually
more restrictive than federal laws, such as the Health
Insurance Portability and Accountability Act of 1996
Privacy Regulations.5Most health data programs, both
mandated and voluntary, have mechanisms to provide
identifiable data that may be released for research pur-
poses or for public health purposes, using a limited data
set and a data use agreement. Figure 1 illustrates which
states have mandated hospital discharge data reporting
(gray) and which collect the data voluntarily (striped).
States in white have no statewide hospital discharge
data system.
In states without a legislative mandate (number 3
above), and for states with legal mandates in which
the hospital association controls the data, the data dis-
closure policies may vary the greatest. If the hospital
association or a private entity serves as a delegated
authority, the public health authority may have ac-
cess to data for public health purposes and is the log-
ical place to begin the data-acquisition process. The
state health data organization, or data steward, under-
stands the legal boundaries and is the best source of
information about working within the constraints of
these boundaries for various projects. The good news
is that most states make their data available for public
health surveillance and federal-state projects, such as
the HCUP and CODES. The key to success is establish-
ing personal relationships and building support and
trust for your project.
Another dimension of availability/accessibility is
the price of acquiring data from data stewards. Most

538 ❘Journal of Public Health Management and Practice
FIGURE 1 ●Statewide Hospital Inpatient Data Collection Practices, 2007
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
organizations charged with statewide collection of hos-
pitalization data generate some revenues through data
sales to partially offset the costs of their data collec-
tion. The price of the data established by data stewards
ranges from $75 to $30 000 per dataset per year, and
many programs price and support research and special
request datasets.5The cost of purchase is established by
the health data program and, in some states, the data
partnership terms are negotiable. In some cases, the en-
tity requesting the data provides in-kind services, such
as analytic support or tools in lieu of money. In all cases,
the purchase cost is much less than the cost to establish
a new data system.
Reliability/consistency
Once access to the data has been obtained, working with
the data poses another set of analytic challenges. Dis-
charge data are often the result of negotiated trade-offs
between the data collection burden on the providers
and the utility of data for community and quality im-
provement initiatives. What discharge data lack in clin-
ical detail, they make up in the broad coverage of the
population and relative uniformity across states. Hos-
pital discharge data are based on national standards es-
tablished by the National Uniform Billing Committee
and thus are relatively comparable across states, espe-
cially for common billing data elements. Because in-
patient data are collected by licensed/certified profes-
sional medical records coders and serve as the basis for
payment, data quality is usually higher than other in-
jury data. States vary most in their coding of fields such
as E-codes and race/ethnicity, which are not a core uni-
form billing data element. Some states collect E-codes
as a part of the diagnosis codes whereas other states
have a separate E-code field or fields. The fact that there
is variation across states in the number of diagnoses
and procedure codes captured further complicates the
matter (eg, some states collect 9 diagnosis and 6

Using Hospital Discharge Data for Surveillance ❘539
procedure codes whereas other states collect unlimited
numbers of diagnosis and procedure codes that can to-
tal more than 100). In an evaluation of 12 states’ hos-
pital ED data, based on comparisons with alternative
sources of hospital data, inpatient E-code completeness
on injury records averaged 87.2 percent and ED E-code
completeness on injury records averaged 92.5 percent.26
Scope of coverage
Because statewide discharge datasets are population-
based, representing a known population that is de-
fined by residence within geographic or political
boundaries,27 hospital discharge data are increasingly
used for community assessment, providing informa-
tion about preventable, avoidable hospitalizations, in-
cluding ambulatory care sensitive conditions (condi-
tions or diseases for which the hospitalization could
have been prevented or reduced in frequency if timely
and effective primary care had been provided).28,29
These data contain a large volume of observations on
all patients hospitalized in short-term-care facilities in
a state, although federal hospitals and specialty hos-
pitals are often excluded from state data collection
requirements. More states are adding data elements
important for public health and research to their dis-
charge data in response to national standards and user
needs. These additional data elements include patients’
race/ethnicity, “do not resuscitate order,”∗30 and a num-
ber of clinical elements.31
The information available within most state dis-
charge databases related to cost of care does not fully
determine the complete financial burden of hospital-
izations because the state hospital discharge dataset,
with an exception or two, captures hospital charges
only and not actual cost of hospitalization. Hospitals
receive varying reimbursement for their charges so that
charges do not always reflect actual cost or received rev-
enues. A growing number of states and the HCUP and
CODES projects both have developed a cost-to-charge
ratio methodology to estimate charges at the state
level.
All discharge data capture the patient’s residen-
tial ZIP code but not all states capture the patient’s
address. Discharge data will not include information
of the patient who received care in another state or
jurisdiction. Border crossing by those patients resid-
ing near state borders can skew population-based
and community-level assessments. Some states have
reciprocal agreements to exchange patient informa-
∗The “do not resuscitate” order is an order by the patient to the
healthcare provider, dictating that an individual does not desire
resuscitative measures in the case of failed breathing or cardiac
arrest.30
tion from other states to address the border-crossing
phenomenon.
Analytic utility
Discharge data support national, regional, state, and
substate statistical benchmarks. The large number of
observations or events that discharge data represent
provides statistical power to epidemiologic studies on
morbidity and hospital use at the state, community, and
hospital service-area levels. There are some limitations.
Diagnoses and procedures coding practices may vary
by hospital. When comparing hospital and community
statistics, the user must also be aware that physician
practice patterns and payer policies may contribute to
geographic variation in hospital utilization.32 Timeli-
ness is another consideration, as discharge data are typ-
ically reported to the health data program 45 days fol-
lowing the close of the previous quarter. In addition,
most health data programs build in a provider review
and validation period in which providers are required
to correct errors and verify the results. For analyses re-
quiring hospitalization statistics by number of persons,
rather than by number of discharges, analytical utility is
also hampered by the inability to convert the discharge
data into person-level data, especially in those states not
capturing a stable unique patient identifier. States vary
in their capacity to produce population estimates data
(for denominators) at the same geographic subunits (ie,
ZIP code), which may limit analytical utility for certain
analyses. Health data programs are constantly balanc-
ing the completeness and validity with timeliness. Data
partnerships improve the quality of the data, as users
provide feedback and demand for better data.
Enhancement potential
Because the data are based on billing information, dis-
charge data contain only limited clinical detail, but
when linked to other datasets, such as birth certificate
data, the linked data can provide information about
birth outcomes that discharge data or birth certificate
data alone cannot reveal.33,34 For specific purposes of
environmental tracking, additional important data ele-
ments can be added relatively easily from other sources,
such as laboratory results and medical charts.31 Link-
ages at the community level will benefit EPHT appli-
cations, as the data can be linked with environmental
factors, such as air quality. The patient’s ZIP code can
be a surrogate for community-level analyses. However,
ZIP codes are developed entirely for different purposes
and their boundaries do not always represent commu-
nity or political boundaries within a state. Including the
patient’s address among the data elements in discharge
data for states that do not currently collect it will be a

540 ❘Journal of Public Health Management and Practice
TABLE 2 ●Strengths and limitations of hospital discharge data
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
Attributes of discharge data Strengths Limitations
Availability/accessibility Data availability—inpatient data maintained in 48
states and 25 emergency department systems
maintained in 25 states
Cost-effective source of population-based information
Data access may be difficult, particularly when owned by
private entities. If access is granted by private owners, it
may be costly
Costs of public and research data vary across states
Reliability/consistency More reliable than other sources of data, such as
patient self-reporting of medical expenditures or
physician reporting of specific conditions for disease
surveillance1
Comparable across states and providers, based on
national UB-92 formats
Primarily established for billing or administrative purposes, the
hospital data may be lacking clinical detail for surveillance of
certain environmental health conditions
While many data elements are uniform across states, states
vary the most in their collection of number of E-codes,
number of diagnoses codes, race/ethnicity, laboratory
results, and patient address. Coding accuracy has also been
questioned17
Scope of coverage Inpatient discharge data allow ongoing surveillance as
these data are collected on an ongoing basis,
capturing information about every hospitalization in
acute care hospitals
State systems are population-based, including all
acute care hospitals in a state
Excludes federal or specialty hospitals and does not reflect
nonfacility outpatient care settings. Also may exclude
emergency department visits when there is not an admission
Data capture charges not actual costs
Analytic utility Available for multiple years, supporting trend analyses
over time
Benchmarks can be established for state, regional,
national rates
Large volume, number of observations, supporting
small-area and subgroup analyses
Lack of timeliness. Most states capture the data 45 d after the
close of the previous quarter. Data are aggregated in a
reporting year, edited, updated, and verified before their
public release
Limited geographic referents for patient residence (ZIP code is
not true proxy for community), coupled with limited
availability of denominator data at the community level
Coding practices vary by hospitals
Geographic and temporal variations in diagnosis, admissions,
and procedure decisions
Enhancement potential When linked with other data sources, hospital
discharge data provide important information about
health systems performance, patient outcomes, and
utilization and cost for target conditions of interest
(injuries, chronic diseases, complications of care)
Hospital data can be linked at the community level
Some states may lack a reliable, stable unique patient identifier
required for efficient record level linkage with other datasets
that would enhance the value of the discharge data
Many states do not collect the patient’s address as one of the
required elements. Addition of this variable will allow
geocoding of discharge data and in turn effective disease
surveillance at the community level
major augmentation in utility because this addition will
allow geocoding and in turn superior community-level
surveillance. Moreover, as EPHT becomes a data part-
ner, helping the data steward make the case for adding
address to core reporting requirements will be impor-
tant to improving its utility for geocoding and environ-
mental studies. Table 2 summarizes the strengths and
limitations of hospital discharge data.
●Conclusion
The EPHT Program will benefit from the use of existing
healthcare databases maintained by state health data
programs. As demonstrated with other state and na-
tional initiatives, they are an affordable source of ongo-
ing data that are relatively comparable across providers
and states.
Public health epidemiologists may prefer real-time
health records and disease registries over administra-
tive hospital discharge databases, relying on the man-
ual abstraction of clinical information from the medi-
cal record to precise specifications. However, the cost
of real-time electronic health records (EHRs) and re-
porting compliance to disease or condition-specific reg-
istries limits their feasibility and completeness.35–37 In
addition, the healthcare providers are increasingly re-
sistant to establishing new registries, citing reporting
“crowd out” due to the myriad of reporting initiatives
and the cost to produce new data systems.35

Using Hospital Discharge Data for Surveillance ❘541
Recent initiatives in data collection have focused on
the possibilities of the EHR for measuring quality, con-
ducting surveillance, and other uses for clinical infor-
mation. The vision of the EHR is that the EHR will
serve as the “core system” of health information tech-
nology for healthcare providers,38 populating public
health registries with information found in EHRs, thus
reducing the reporting burden and improving the time-
liness of data reporting as compared with the current,
manual-reporting systems. However, because of the
complex political, technical, and financial challenges
associated with the EHR, the transition from registries
to EHRs is anticipated to be lengthy. In the meantime,
there is much value in the current hospital discharge
data for local and federal partnerships and programs.
Despite limitations, hospital discharge data have been
a proven and valuable source of inpatient and nonin-
patient healthcare encounters.
As a nonprofit, membership organization represent-
ing many state and private health data organizations,
NAHDO can serve as a broker to facilitate local and
federal data partnerships and help prospective users
understand the data collection and release protocols
in play in most states. This is particularly relevant be-
cause of the crucial role NAHDO and its members have
played in ensuring the success of the HCUP. As an
EPHT partner with CDC, NAHDO is optimistic that
hospital discharge data will provide an important link
between health outcomes and environmental causes of
morbidity and healthcare costs.
REFERENCES
1. Centers for Disease Control and Prevention (CDC). CDC’s
national environmental public health tracking program: na-
tional network implementation plan (NNIP), Version 1.0.
http://www.cdc.gov/nceh/tracking/pdfs/nnip.pdf. Pub-
lished 2006. Accessed December 5, 2007.
2. National Center for Health Statistics. National hospital
discharge survey description. http://www.cdc.gov/nchs/
about/major/hdasd/nhdsdes.htm. Accessed December 5,
2007.
3. Schoenman J, Sutton J, Kintala S, Love D, Maw R.
The value of hospital discharge databases. Agency for
Healthcare Research and Quality; 2005. http://www.hcup-
us.ahrq.gov/reports/final report.pdf. Accessed June 14,
2007.
4. Centers for Disease Control and Prevention (CDC), National
Environmental Public Health Tracking Program. National
Network Implementation Plan. Atlanta, GA: Centers for Dis-
ease Control and Prevention; 2006.
5. National Association of Health Data Organizations
(NAHDO). 2006 Membership Survey Report. Salt Lake
City, UT: NAHDO; 2006.
6. Nedza S, Pine M, Merryweather P, Ho Y. Enhancing admin-
istrative data sets to support value-based purchasing. Paper
Presented at: NAHDO 22nd Annual Meeting; October 19,
2007; San Diego, CA.
7. Injury Surveillance Workgroup. Consensus Recommendations
for Using Hospital Discharge Data for Injury Surveillance. Ma-
rietta, GA: State and Territorial Injury Prevention Directors
Association; 2003.
8. Ash M, Brandt S. Disparities in asthma hospitalization in
Massachusetts. Am J Public Health. 2006;96:358–362.
9. Hammar N, Nerbrand C, Ahlmark G, et al. Identification of
cases of myocardial infarction: hospital discharge data and
mortality data compared to myocardial infarction commu-
nity registers. Int J Epidemiol. 1991;20:114–120.
10. Woolf A, Fish S, Azzara C, Dean D. Serious poisoningsamong
older adults: a study of hospitalization and mortality rates in
Massachusetts 1983–85. Am J Public Health. 1990;80:867–869.
11. Utter D. Use of probabilistic linkage for an analysis of the
effectiveness of safety belts and helmets. In: Committee on
Applied and Theoretical Statistics, National Research Coun-
cil, Federal Committee on Statistical Methodology, Office of
Management and Budget, eds. Record Linkage Techniques—
1997: Proceedings of an International Workshop and Exposition.
Washington, DC: The National Academies Press; 1999:67–70.
http://www.nap.edu/openbook.php?record id=6491&page
=R15. Accessed December 5, 2007.
12. Cook LJ, Knight S, Olson LM, et al. Motor vehicle crash
characteristics and medical outcomes among older drivers
in Utah, 1992–1995. Ann Emerg Med. 2000;35(6):585–591.
13. HCUPnet. Weighted national estimates from HCUP Nation-
wide Inpatient Sample (NIS), 2004, Agency for Healthcare
Research and Quality (AHRQ), based on data collected
by individual states and provided to AHRQ by the states.
http://hcupnet.ahrq.gov/HCUPnet.jsp?Id=2739159B754D
BC03&Form=DispTab&GoTo=MAINSEL&JS=Y. Accessed
December 5, 2007.
14. Flores G, Abreu M, Tomany-Korman S, Meurer J. Keeping
children with asthma out of hospitals: parents’ and physi-
cians’ perspectives on how pediatric asthma hospitalizations
can be prevented. Pediatrics. 2005;116(4):957–965.
15. Peat JK. Prevention of asthma. Eur Respir J. 1996;9:1545–1555.
16. Redd SC. Asthma in the United States: burden and cur-
rent theories. Environ Health Perspect. 2002;110(suppl 4):557–
560.
17. Tolbert PE, Mulholland JA, Macintosh DL, et al. Air quality
and pediatric emergency room visits for asthma in Atlanta,
Georgia. Am J Epidemiol. 2000;151(8):798–810.
18. Gruchalla RS, Pongracic J, Plaut M, et al. Inner city
asthma study: relationships among sensitivity, allergen ex-
posure, and asthma morbidity. J Allergy Clin Immunol.
2005;115(3):478–485.
19. Corso PS, Kramer MH, Blair KA, Addiss DG, Davis JP,
Haddix AC. Cost of illness in the 1993 waterborne Cryp-
tosporidium outbreak, Milwaukee, Wisconsin. Emerg In-
fect Dis. [serial online] April 2003. http://www.cdc.gov/
ncidod/EID/vol9no4/02-0417.htm. Accessed December 4,
2007.
20. NHTSA. The crash outcome data evaluation system
(CODES), technical report. http://www-nrd.nhtsa.dot.
gov/Pubs/808-338.PDF. Accessed December 4, 2007.
21. AHRQ Healthcare Cost and Utilization Project (HCUP).
Overview of the state inpatient databases (SID). http://

542 ❘Journal of Public Health Management and Practice
www.hcup-us.ahrq.gov/sidoverview.jsp#States. Accessed
December 5, 2007.
22. BenaventeM, Knodler MA, Rothenberg H. Case study assess-
ment of crash data challenges: linking databases for analysis
of injury specifics and crash compatibility issues. Transporta-
tion Res Rec. 2006;1953:180–186.
23. Maine Health Information Center for the Office of Data,
Research, and Vital Statistics Bureau of Health Maine
Department of Human Services Standardized Reporting
Using CODES (Crash Outcome Data Evaluation System).
Funded by the National Highway Traffic Safety Admin-
istration. Project No. DTNH-22-97-H-07282; November
1999. http://www-nrd.nhtsa.dot.gov/Pubs/809-048.PDF.
Accessed December 5, 2007.
24. The National Center for Statistics and Analysis, National
Highway Traffic and Safety Administration. Crash out-
comes data and evaluation system (CODES). http://www.
nhtsa.dot.gov/portal/site/nhtsa/menuitem.9fef9613e59b4d
d24ec86e10dba046a0/. Accessed September 11, 2008.
25. Rolfs R, Xu W. Uses of hospital dischargedata for population-
based health assessment. Utah’s Health Annu Rev. 1995;3:26–
38.
26. Moles E, Andrews RM. Emergency department data
evaluation. HCUP methods series, emergency depart-
ment data evaluation, report 2005–02. http://www.hcup-us.
ahrq.gov/reports/2005 02.pdf. Accessed June 18, 2007.
27. Last J, ed. A Dictionary of Epidemiology. 2nd ed. New York:
Oxford University Press; 1998.
28. Agency for Healthcare Research and Quality (AHRQ). Pre-
vention quality indicators fact sheet. http://www.ahrq.gov/
fund/ReviewofClinicalAdmDataElements 4 2 07Final.pdf.
Accessed June 15, 2007.
29. Wu J, Egger MJ, Shah GH, Haggard L, Assassnik S. Ambula-
tory care sensitive hospitalizations: more than a measure of
bed supply. Utah’s Health Annu Rev. 2000;7:1–9.
30. Lynn J, Schuster JL, Kabcenell A. Improving Care for the End of
Life: A Sourcebook for Health Care Managers and Clinicians. New
York: Oxford University Press; 2000.
31. PineM, Jordan HS, Elixhauser A, et al. Enhancement of claims
data to improve risk adjustment of hospital mortality. JAMA.
2007;297(1):71–76.
32. Payne SMC, Ash A, Restuccia JD. The role of feedback
in reducing medically unnecessary hospital use. Med Care.
1991;29(8, suppl):AS91–AS106.
33. Shah GH, Greenway JA, Yang W. Episiotomy and obstetric
trauma in Nevada: evidence from linked hospital discharge
and birth data. Nevada Public Health Assoc J. 2007;4(1):1–
10.
34. Shah GH, Freeman PAC, Cofrin K, Xu W. Cesarean deliveries
and newborn injuries: evidence from linked Utah birth cer-
tificate and inpatient discharge data. Utah’s Health Annu Rev.
2007;12:10–24.
35. Eagle K. Lessons from GRACE: the Global Registry of Acute
Coronary Events. Eur Heart J Suppl. 2002;4:E24–E31.
36. Izquierdo JN, Schoenbach VJ. The potential and limitations
of data from population-based state cancer registries. Am J
Public Health. 2000;90(5):695–698.
37. Vishwanath A, Scamurra SD. Barriers to the adoption of
electronic health records: using concept mapping to de-
velop a comprehensive empirical model. Health Informatics
J. 2007;13(2):119–134.
38. Friedman DJ. Assessing the potential of national strategies
for electronic health records for population health moni-
toring and research. DHHS Publication No. (PHS) 2007-1343.
http://www.cdc.gov/nchs/data/series/sr 02/sr02 143.pdf.
Accessed June 19, 2007.