Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 41
Review of the NCHS Plan for the
National Health Care Survey
OVERVIEW
As discussed in Chapter 2, major changes in the organization, financ-
ing, and delivery of health care have outpaced the capabilities of existing
national health care surveys to provide relevant and timely data to monitor
the health care of the nation and evaluate the impact of the changes in the
system. Although extensive national data activities exist, they are not ide-
ally designed to yield the kinds of data needed for monitoring and evaluat-
ing the changes occurring in the nation's health care system or for health
policy analysis and development. (Brief descriptions of some of the nation-
al health data systems maintained by the U.S. Department of Health and
Human Services are presented in Appendix C).
In an effort to be responsive to various emerging national health care
issues and to the associated data needs, the National Center for Health
Statistics (NCHS) in 1986 undertook a review of its existing surveys of
health care providers. As a result of this review, NCHS has developed a
plan for restructuring its existing health care provider surveys into a new
National Health Care Survey with the aim of providing a more complete
and useful picture than now exists of the medical care provided in the
United States. The center's plan for the survey is described in the docu-
ment entitled, "The National Health Care Survey" (reproduced in Appendix
A). This document was supplemented by additional material provided by
4
1
OCR for page 42
42
TOWARD A NATIONAL HEALTH CARE SURVEY
NCHS as it became available as well as presentations to the panel by NCHS
officials.
According to the plan, the proposed National Health Care Survey will
build on the four existing health care provider surveys, which will be merged,
expanded over time, and linked geographically with the National Health
Interview Survey (NHIS). The provider surveys are the National Hospital
Discharge Survey, the National Ambulatory Medical Care Survey, the Na-
tional Nursing Home Survey, and the National Master Facility Inventory.
Table 1 is a profile of the current survey design for the NCHS health care
provider surveys and the National Health Interview Survey. The primary
objective of this new survey will be "to produce annual data on the use of
health care and the outcomes of care for the major sectors of the health care
delivery system. These data will describe the patient population, medical
care provided, financing, and provider characteristics" (see Appendix A).
The NCHS proposes to implement the National Health Care Survey over a
period of years as resources petit. The phased implementation of the new
survey design began in 1988.
The proposed survey has four main features:
(1) The survey will be conducted on an annual basis to eliminate gaps
in data and fluctuations in resource requirements;
(2) The types of providers and settings sampled would be expanded to
include an array of additional sites of health care, such as hospital
emergency and outpatient departments, ambulatory surgery centers,
home health agencies, and hospices;
(3) Sampling designs will be revised. An integrated three-stage cluster
design will be used in which providers would be sampled from a
subsample of the primary sampling units (PSU)~ of the National
Health Interview Survey, and
(4) A capability would be developed to conduct routine and specialized
patient follow-up studies to examine issues related to the outcome
and subsequent use of medical care.
By expanding the types of providers surveyed and by adding a follow-
up component, NCHS hopes to enhance the kind and amount of information
currently available about health care events.
iFor the first stage of sample design for the National Health Interview Survey in which the
civilian noninstitutionalized population residing in the United States is sampled, the United
States is considered to be a universe composed of approximately 1.900 geographically defined
primary sampling units. A primary sampling unit consists of a county, small group of contig-
uous counties, or a metropolitan statistical area. The PSUs collectively cover the 50 states and
the District of Columbia.
OCR for page 43
REVIEW OF THE NCHS PLAN
43
The panel reviewed the center's plan for an integrated National Health
Care Survey from two perspectives: (1) technical features of the plan and
~2) the extent to which the survey would enable NCHS to meet the changing
data needs associated with the health care policy issues discussed in the
previous chapter. The following sections present the panel's review.
PERIODICITY
One feature of the center's plan is the annual data collection schedule
planned for each provider survey. In recent years, primarily because of
resource limitations, the average time between cycles of these surveys has
increased si~,r~ificantly. The center correctly argues that funding and staff-
ing these surveys on an on-again, off-a~ain basis is not sound from a man-
agement, fiscal, or statistical point of view.
in fielding annual surveys, including, maintaining a small group o~ wc~-
trained staff, reducing the budgeting and scheduling problems associated
with periodic surveys, minimizing recurrent start-up costs for survey com-
ponents, addressing the seasonal nature of illness, and providing current
statistics.
In order to offset the increased costs of conducting the provider surveys
annually, however, the center clans to reduce their sample sizes. Reducing
It points to several advantages
-berm- By ,
ample sizes obviously works against solving the general problem of pro-
ducing data for subdomains of the population. The sample sizes of these
surveys are too small now for subnational estimates, for estimates of diag-
nosis and treatment of rare conditions, and for statistics on minorities. With
reduced sample size the problems of developing estimates for small areas
and various subpopulations of interest appear to be aggravated.
The center plans to compensate for smaller sample sizes by aggregating
data for more than one year. Although multiyear aggregation will be a
necessity with the small samples, it is not clear how many years of aggrega-
tion will be necessary to produce some of the statistics needed. Further-
more, such aggregation may be less appropriate for characteristics that are
not stable across years. It also will affect the timeliness of data production
and analysis.
The users of data who were surveyed expressed interest in estimates
below the national level, at the state level, and even at the county or com-
munity levels.
The list of subpopulations of interest identified includes
blacks, Hispanics, American Indians, Asians, disabled people, rural popula-
tions, specific industrial groups, elderly people, children, especially dis-
abled children, groups defined by socioeconomic level, and people in insti-
tutions. The Disadvantaged and Minority Health Improvement Act of 1990,
which contains the reauthorization of NCHS, also requires the center to
develop better systems to collect data on minority subpopulations. Al-
OCR for page 44
44
-
e~
~C
-
e~
o
._
-
Ct
4
C~
C~
a'
;-
._
o
~:
-
Ct
~:
V)
:r:
V
z
C'
C)
; -
t
E~
o
C~
~:
._
C~
V)
V
o
-
s ~ _
CĒ ~ _
o,
·_ C) ~
~ ~ Z
~ _
._ ;,- ~.
~ C)
- , ~
Z ~ _
_ V) _
o ~ :C
·- ~ Z
C~ o Z
z :e _
_ ~
~ C~
._ ,
_ ° ~ V
_ ~ V
.° ~ Q) ~
CC ~ ~ z
z ~ v _
-
C~ ~ _
._ , .
U. ~s oo
o V' C~
_ 04
Ct
t: ~
o ~,
C;: · _
Z ~
-
U'
::
z
-
o ~
_ ;
V)
3
~ ._
~ _
-
;^ o
~ o
U' V
-
._
-
._
C~ ~
O cn ~ C)
00 ·_
· _ eC L e~ ~
O _ _ ~ ~ ~ O
~5
_
~ c~
O -`
_ O .c> 0 .0 ,~
V~ ~ ~ ts _ ~) _
, e 0 0 , ~
D ~ ~ e 0 ~ ~ U;.! ° e c,
_ ~
. ~ _
.O _
~ 0
Ģ =~o
o ~ =, C,' ~ ~
~ v' ~ ~ ~ e c ° ' e"
C~ ~ 8)
~ CC
o 0 3 v' ~
_ ,4_ _ ._
4) ;>' V, ° ~ ~ o °
~ _ ~ ~ ~ o ~ _
D ~ _ O ~ ~ ~ ~ ~,~ ~~ D
C)
0C _
e ~ _
_ ~ ~~; 5
c' _ c~ `_
_ ~ ~ o: O
c~ ~ O
V, ~ C)
OCR for page 45
45
c
o
- ~
. - -
~ :>
o
_ 5 i>, ~ o ;^
O Ci.0 ~ ~ -'' O ~ O
~ U: O ~ C.) ,_ c~ .
=o 0 a=~=x~
an
O ~ ~ ~ ~ O
~ ~ .' ~ ~ ~ _
c)
Ģ
Cat
~ i_
_ o ~
o ~ ~ ~ so
~ ,,, ~ o ~ ._ ~
7~ sit At ~ ~ U)
~ ~ ~ '= V' I'm
Cal
s
-
o
us C,)
ID
Be;
C)
-
o
C)
-
e~
Cal
U) C~)
U'
'Cal °
o s
~ on ~
.= ~ to
-
Cal
o
o ._
~ _
~ en
_
c5
~ A:,
o
-
C~
Ģ
lo
-
-
C~
-
C~
-
C)
on
Ct
Cal
Cal
._
OCR for page 46
46
TOWARD A NATIONAL HEALTH CARE SURVEY
though not all of these expectations are realistic for national survey-based
data systems, the reduced annual sample sizes planned for the National
Health Care Survey will undoubtedly further aggravate the problems of
developing estimates for small areas and various subpopulations at a time
when NCHS is looking for ways to improve responsiveness in these areas in
a cost-effective manner. The panel seriously doubts the possibility of pro-
ducing small area estimates even with the proposed aggregation of data
over multiple years.
The panel is informed that the NHIS sample is being redesigned to
increase the sampling of PSUs in which blacks and Hispanics are concen-
trated. Since the National Health Care Survey will be conducted in a sub-
sample of the NHIS PSUs, the NHIS redesign should somewhat strengthen
the samples of blacks and Hispanics in the National Health Care Survey as
well. However, the panel does not believe that this increase in the sampling
of NHIS PSUs in which blacks and Hispanics are concentrated will improve
significantly the ability of the National Health Care Survey to provide health
care statistics for minority populations.
Despite the sample size deficiency, the panel believes that there are clear
advantages to fielding the provider surveys on an annual basis, to reduce
fluctuations in resource requirements and to ensure currency of the statistics.
Recommendation 3-1: The panel endorses the NCHS plan to
conduct the components of the National Health Care Survey on
an annual basis.
SCOPE AND COVERAGE
.
Current NCHS plans call for extending the coverage of the National
Health Care Survey beyond the current four major health care provider
surveys to include collection of national data for selected alternative sites of
the three major types of health care and health care providers hospital
care, ambulatory care, and long-term care:
· The Hospital and Surgical Care Component, based on the Nation-
al Hospital Discharge Survey (NHDS), is being modified to include
hospital-based and free-standing ambulatory surgery centers.
The Ambulatory Care Component, based on the National Ambu-
latory Medical Care Survey (NAMCS), is being modified to in-
clude medical care provided in emergency rooms and outpatient
departments and clinics in nonfederal, short-term (i.e. average length
of stay less than 30 days) hospitals. Long-range plans call for
inclusion of additional settings, such as community and neighbor-
hood health clinics.
OCR for page 47
REVIEW OF THE NCHS PLAN
47
The Long-Term Care Component, based on the National Nursing
Home Survey (NNHS), is being expanded to include board and care
homes that provide assistance with activities of daily living, but not
nursing or medical care. This addition will extend the survey to
group homes, rest homes, and other supported living arrangements,
including sites that provide service to the mentally and physically
limited population. The long-term care component will be modi-
fied further by the inclusion of a home health care and hospices
survey.
.
The National Master Facility Inventory is being renamed the Na-
tional Health Provider Inventory (NHPI) and expanded to in-
clude hospices, home health care agencies, and licensed residential
care facilities. Long-range plans call for the addition of providers
of acute ambulatory care and community-based long-term care to
the inventory.
The panel endorses the center's efforts to extend the coverage of the
provider surveys to include a range of the alternative health care settings
that have emerged in recent years. The proposed extensions constitute an
important step in extending the coverage of these surveys. However, they
are only a first step if NCHS is to meet its responsibilities to provide a
comprehensive coverage of national health care statistics in the United States.
An important further extension needed is the inclusion of services pro-
vided by nonphysicians who deliver health care. The panel's survey of
users identified at least 37 types of providers and health care settings not
currently being surveyed by NCHS, but for which health care statistics are
needed (see Appendix B for the list of providers identified by users). Some
of these health care settings are already under consideration for inclusion by
NCHS. Others may not be appropriate for national surveys. Some obvious
examples of additional providers for consideration are psychologists, den-
tists, physical and occupational therapists, pharmacists, podiatrists, chiro-
practors, nurses, nurse practitioners, physician assistants, nurse midwives,
and optometrists, all of whom deliver some form of health care.
Another specific concern is the distinction made historically by NCHS
between collection of data on the physical health services that it plans to
cover, and collection of data on mental health and dental services, which
are not included in the center's plan. Such a distinction seems to reflect
bureaucratic issues rather than theoretical coherence. Obviously, it would
not be cost-effective for NCHS to attempt to duplicate the data collection
efforts of the National Institute of Mental Health, the National Institute of
Dental Research, or the Health Resources and Services Administration. The
panel believes, however, that NCHS should consider the collection and/or
OCR for page 48
48
TOWARD A NATIONAL HEALTH CARE SURVEY
dissemination of adequate statistics about mental health and dental services
to be part of its responsibility; it should furthermore ensure, perhaps through
collaborative efforts, that the needed data in these areas are available from
appropriate agencies and that they are collected in a form that is consistent
with that used in the other provider surveys.
Finally, though ambulatory care coverage is the area most in need of
expansion beyond the current plan, the hospital care component also needs
to be expanded. The omission of federal hospitals and long-term care hos-
pitals from the hospital care component is a significant deficiency. NCHS
has made efforts in the past to obtain data on federal hospital discharges,
and the task has proven difficult. While recognizing the difficulties, the
panel nevertheless considers that it would be desirable to have integrated
data on all hospitalizations, including those in federal hospitals.
Recommendation 3-2: The panel recommends that NCHS ex-
tend its coverage of providers of health care to include a fuller
range of health providers than currently planned both physi-
cians and nonphysicians. The panel further recommends that
the universe for the hospital care component be extended to
include long-term care hospitals and federal hospitals.
The panel does not mean to suggest that all possible health care provid-
ers should be surveyed each year. A strategy needs to be developed for
sampling types of providers at varying periodicities, with priorities based
on their relative importance. For instance, periodicities for specific provid-
er groups could be based on such considerations as expenditures for care,
distribution and volume of users, political or policy issues of importance,
and availability of resources. Certain categories of providers may have to
be approached through periodic special studies.
At the present time, however, no objective process exists for deciding
which providers should be given priority for inclusion in the surveys, ex-
cept availability of resources from the parties interested in the data. A
mechanism is needed t-or developing standard criteria for setting priorities
for incorporating nets ad emerging settings and the full range of health
care providers, as we .~ for periodic review and revision of these criteria
as needed.
Recommendation 3-3: The panel recommends that NCHS put
in place a mechanism for developing criteria and for setting
data collection priorities across the full range of health care
providers and service settings, and that this mechanism and
process be dynamic and include periodic review and revisions
of both the criteria and coverage as necessary.
OCR for page 49
REVIEW OF THE NCHS PLAN
49
DATA COLLECTED
The panel was asked to review the content of the questionnaires and
definitions used in the provider surveys in terms of appropriateness and
adequacy. The panel is in general agreement with the content areas includ-
ed in the center's plans; however, it finds the coverage of the data content
deficient in terms of what is needed to address current and future health
care policy issues, as described in Chapter 2. The panel recognizes that any
significant expansion or change in the data collected in the National Health
Care Survey will require a major increase in the level of effort and associat-
ed resources. Nonetheless, the panel has identified several key issues that
are not reflected either in the content of existing surveys or in the plan for
the proposed National Health Care Survey in its current stage of develop-
ment; they are discussed below.
The provider surveys that serve as the basis for the National Health
Care Survey proposed by NCHS currently collect data only on events, such
as individual physician visits or individual hospital discharges. Historical-
ly, NCHS has provided basic information for a broad range of issues related
to the supply and volume of health care, but has lacked person-based infor-
mation on health care received by individuals over time or over the entire
progression of an illness episode, from onset to completion. However,
policy analysts and other data users also need data on persons and about the
medical care received by them in order to answer questions about episodes
of illness, outcomes, costs and expenditures for care, insurance coverage,
the use of multiple providers, etc. The person-based longitudinal approach
would also allow for linkage to the death record through the National Death
Index (NDI)2 to obtain the mortality status of those persons being followed
longitudinally. Linkage with birth records will be needed in pregnancy
related to outcome studies.
The cost of health care is one of the most important health policy issues
confronting the nation today. Although the NCHS statutory mandate ~n-
cludes monitoring health care costs and financing, including trends in health
care prices and costs, the sources of payment for health services, and the
federal, state, and local government expenditures for health services, it col-
lects almost no information on payers or costs and expenditures for care
(including the component paid by insurance) in a survey of health care use
2The National Death Index is a central computerized file of death record information for all
deaths in the United States. It is compiled from magnetic tapes submitted to the NCHS by the
state vital statistics offices. These tapes, beginning with deaths occurring in 1979, contain a
standard set of identifying information for each decedent for use in searches of the NDI file to
identify and locate state death records. The NDI is used by researchers conducting prospective
and retrospective studies to determine if persons in their studies may have died.
OCR for page 50
so
TOWARD A NATIONAL HEALTH CARE SURVEY
(see Appendix D). The panel recognizes the difficulties in obtaining this
information, but it firmly believes that a National Health Care Survey is a
logical vehicle to collect data on costs and expenditures of care.
The NHDS collects information on discharges, not individuals, so data
on readmissions are not available. In addition, information on the function-
al status of individuals would make the information on hospitalization more
valuable; at the present time there is nothing on functional status in hospital
discharge abstracts. The panel recognizes that the feasibility issues with
respect to measuring functional status of hospitalized patients are numer-
ous, but, if they could be overcome, research on outcomes of hospital care
would be greatly strengthened. This need for more information on health
and functional status of the patient for the purpose of evaluating outcomes
extends, of course, to all provider sites and should be recognized in all the
provider surveys.
Information on the longitudinal dimensions of care is of critical impor-
tance in order to assess the effect of treatment. Current data sets do not
permit construction of meaningful patterns of care over time. This has been
particularly problemat ^n long-term care, although there have been notable
efforts to begin to correct this problem in the most recent National Nursing
Home Survey. The panel is encouraged to note that some of the needed
information on the residents will be available from the follow-up study to
the 1985 National Nursing Home Survey. However, there is no indication
in the center's plan that collection of such information will continue as part
of the long-term care component of the proposed new survey.
There is need to link multiple admissic*~;` of individuals rather than treat-
ing each admission as a new patient, as is cure^-itly the case in NHOS, NAMCS,
and NNHS. Such linkage would especially enhance the value of the National
Nursing, Home Survey. Almost one-third of all nursing home discharges are
to hospitals (NCHS, 1981~. Estimates suggest that one in five nursing home
residents has a hospital admission each year. Research suggests that inappro-
priate use of the hospital by nursing home residents is a significant hidden cost
of long-term care and amounts to cost shifting in a substantial proportion of
cases from Medicaid to Medicare, with possible adverse health outcomes for
patients (Aiken, et al., 1985; Shaughnessy,l990~. Yet it has not been possi-
ble to study the patterns of care and interchange of patients between hospi-
tals and nursing homes from the current NCHS data sets.
To be maximally useful, existing provider surveys do not include suffi-
cient data on characteristics of providers and their practice patterns. A
valuable addition would be to increase the amount of information about the
various facilities, settings, and providers in which sampled events occur.
For the hospitalization component, data should be collected about corporate
structure, and other related variables of short-term hospitals should be add-
ed from the American Hospitalization Association's files. For ambulatory
OCR for page 51
REVIEW OF THE NCHS PLAN
51
care in private offices, the kind of corporate arrangement in which physi-
cians are practicing, the mix of specialties, the number of physicians in-
volved in a group (if it is a group), and relationships to managed insurers
constitute information that would be useful in tracking what is happening to
ambulatory care.
In recent years there has been major growth in specialized units in
general hospitals such as specialized psychiatric units, alcohol treatment
units, and drug detoxification and treatment units. Patients treated in these
units are different from those treated in "scatter" beds, and patterns of
service, length of stay, and costs differ for patients in these units compared
with patients with comparable diagnoses treated in general medical units.
NCHS needs to develop plans to obtain information on patients seen in such
special units of general hospitals. This deficiency is easily corrected by a
simple coding change to indicate whether the patient was treated in a spe-
cialized unit.
An important new data collection activity that NCHS should explore is
the addition of data from medical records or charts to the content of the
National Health Care Survey. As mentioned previously, a major limitation,
not only of NCHS surveys, but also of virtually all data sets about health
care services, is the lack of detailed information on the tests and services
performed during the various treatment events. The addition of such data
from medical records or charts to the content of the surveys is critical to at
least three important areas of research: (1) studies related to the appropri-
ateness of care, (2) studies of outcomes and efficacy of treatments, and (3)
studies of costs of treatment.
As a first step, a sample of charts from hospital discharges and from
ambulatory care visits should be drawn. The information to be abstracted
from these charts should be able to address issues such as:
.
What tests were performed.
Details on how patients were treated. This information is critical to
outcome studies to evaluate the efficacy of alternative ways of treating
patients. Details about treatments are also essential for evaluating
information about the cost of care and for producing comprehen-
sive statistics on health care use. In particular, insurance claims do
not have information about noncovered services. For example, the
administrative system of Medicare does not provide information
about the use of prescription drugs. Moreover, providers outside
the traditional fee-for-service system, such as health maintenance
organizations (HMOs), typically do not file insurance claims in
detail. As a result, in order to get any information about the value
of delivered care, it is necessary to collect the detail that appears
only on hospital and medical charts.
OCR for page 52
52
TOWARD A NATIONAL HEALTH CARE SURVEY
NCHS will play a major role in the years to come in monitoring the
progress of the department's Health Objectives for the Year 2000 (IJ.S.
DHHS, 1991a). NCHS has responsibility for tracking about 160 of more
than 300 objectives. Although most of these objectives involve promotion
of healthful behaviors in the population, rather than health systems perfor-
mance per se, 28 specific objectives do relate to health care. Although
NCHS is not currently collecting the needed information, the panel is in-
formed that 20 of these objectives could be addressed by adding questions
about physician practices to the physicians induction questionnaire for the
National Ambulatory Medical Care Survey in order to assess patient treat-
ment and health promotion activities.
NCNS should start collection of the additional information beginning
with the next cycle of the National Ambulatory Medical Care Survey so
that baseline data will be available to measure progress toward the depart-
merzt's Health Objectives for the Year 2000.
The content areas identified above are needed now and will continue to
be needed in the foreseeable future, for policy analysis of the critical health
care issues both for the 1990s and into the next century. It is essential
therefore that NCHS move boldly into new areas of data collection and new
methods that are required to obtain the needed data to meet the challenges
of the next century.
Recommendation 3-4: The panel recommends that `~e National
Health Care Survey include collection of person-based longitu-
dinal information, expanding the data collected to include, but
not be limited to, information on the health care received, costs
and gross expenditures for health care, and outcomes.
DESIGN FEATURES
Cluster Survey Design
A central feature of the center's plan for the National Health Care
Survey is to move from the existing four independently designed provider
surveys to a three-stage cluster sample designs and to link the sample selec-
tion of these surveys geographically with the sampling design of the Na-
tional Health Interview Survey (NHIS). Under this design the health care
providers are to be sampled at the second stage from a first-sta~e sample of
geographic areas, rather than selecting, the providers at the first stage. The
geographic areas currently being used for fielding the new survey are a
3Cluster sampling, frequently used by statisticians in designing surveys, is a method of
sampling in which a population is first defined as groups or clusters of primary units and then
from these groups a sample is selected.
OCR for page 53
REVIEW OF THE NCHS PLAN
53
subsample of the primary sampling units for the NHIS. This linkage with
NHIS is consistent with the center's decision to base the NCHS Integrated
Design Program on the NHIS sample and to establish survey linkages to the
other NCHS population surveys (see Appendix A). Other features being
considered by NCHS include the possible use of the same samples of health
care providers over time and for several components of the National Health
Care Survey in order to increase data quality and to reduce sample induc-
tior~ costs.
The center has identified in its plan several potential advantages: (1) in-
creased analytic utility, since health care use is examined in relation to health
status indicators from the NHIS; (2) reduced interviewing costs, since sample
providers are concentrated in specific geographic areas; (3) increased potential
for record linkage across settings, which aids in tracking patients and in
differentiating multiple episodes of the same condition; and (4) the possibil-
ity of producing local area statistics, for sampled areas or communities.
The stated advantages of moving to a cluster sample design are being
examined by NCHS. Research is currently under way for the next cycle of
the NHIS redesign based on the 1990 census, including not only issues
related to the NCHS population surveys, but also particular requirements of
the NCHS provider surveys. Factors related to subdomain statistics, alter-
native design options, including the redefinition of PSUs in terms of health
service areas, the effect of conducting the National Health Care Survey
components in a subsample of the NHIS PSUs, and the analytical utility of
such a design are being explored as part of the redesign effort. Results of
this research are beginning to be available to NCHS.
In reviewing the center's proposed survey design, it becomes clear that
there are two related but distinct aspects to the issue of the use of geograph-
ic areas as the primary sampling units in the cluster survey design: (1) the
use of NHIS PSUs and (2) the definition of PSUs, i.e., current NHIS PSUs
that are population based or PSUs defined in terms of health service areas.
Use of NHIS Primary Sampling Units
The panel believes that there may be some practical advantages to hav-
ing, the National Health Care Survey and the NHIS conducted in the same
PSUs. It is possible that there will be efficiencies in sharing data collection
staff (trained Census Bureau field staff), although the real benefits cannot
be assessed until the actual strategies for data collection are fully in place.
Furthermore, the assumed cost savings of having the providers geographi-
cally clustered needs to be demonstrated. The trade-off between reduced
costs through use of a clustered provider sample and the impact of this
design on the efficiency of the sample needs to be examined further. NCHS
staff estimates suggest that, for national estimates, the loss of efficiency
OCR for page 54
54
TOWARD A NATIONAL HEALTH CARE SURVEY
will be modest, less than 6 percent for the NHDS (Shimizu and Cole, 1990~.
The panel is not aware of similar research for the NAMCS but would be
surprised if the results were worse, given the apparent correlation between
physician and population geographic distributions. The panel seriously doubts
the possibility of producing local area estimates by conducting the survey in
a subsample of the NfITS PSUs, even with the planned multiyear aggrega-
tion of data. As stated earlier in the chapter, given the current sample sizes,
the panel questions if NCHS will be able to produce reliable estimates for
states, let alone for specific PSUs.
NCHS obviously has given considerable thought to the concept of an
integrated National Health Care Survey and has presented arguments that its
plan implies integration. The panel supports the use of the NHIS PSUs as
an important first step in that direction. It should be emphasized, however,
that simply conducting the provider surveys in the NHIS PSUs does not by
itself result in a meaningfully integrated survey. There are a number of
approaches that, taken together, would achieve greater integration with the
NHIS than envisioned in the center's plan.
Within the context of this
survey, the panel believes that a National Health Care Survey would be
considered integrated if it had the capability of obtaining relevant health
care data at the person, provider, visit, or episode level that can be inter-
linked directly at the individual rather than at the aggregate level. Thus for
example, behavioral and health status data for a sample person can be tied
to such things as the use of health care facilities, the care received, the
costs, and the outcome of health care received for the illness episodes that
the person is experiencing.
Definition of Primary Sampling Units
The panel is skeptical about the assumption of increased analytic utility
resulting from geographic linkage and by defining the PSUs in teas of
health service areas. NCHS defines health service areas as one or more
counties that are relatively self-contained with respect to the provision of
routine hospital care. NCHS plans to be able to examine health care use in
a geographic area in relation to the availability of health resources and the
health status indicators from the NHIS. This would permit the computation
of ecological correlations using the health service area as the unit of analy-
sis. Recent research by Makuc and others on defining health service areas
showed that, in a group of health service areas representing roughly 94
percent of the U.S. population, less than 25 percent of the hospital stays of
people residing in those areas occulted outside the service area (Makuc et
al., 1990~. When examining patterns for ambulatory care, Kleinman and
Makuc found that nearly 20 percent of physician visits occur outside the
county of residence, with substantial variation according to metropolitan
OCR for page 55
REVIEW OF THE NCHS PLAN
55
status., proximity to the metropolitan area, and population density (Klein-
man and Makuc, 1983~. No parallel research has been completed for nurs-
ing homes.
Makuc presents some examples of the kinds of analyses that could be
undertaken based on the first-stage area design. Each area can be character-
ized by census variables and variables from other sources, such as the Area
Resource File.4 Health care data can be produced for the same areas so that
such questions as the following could be answered: to what extent is one
subcomponent of care being substituted for another, e.g., ambulatory surgi-
cal centers as a substitute for inpatient hospital care, hospice and home care
for nursing home care? To what extent can variation in rates of utilization
of health care be explained by variation in supply of providers? Small area
variation in the use of health services might be used to monitor the effec-
tiveness of the medical outcome initiatives. There are several issues that
need further research before the utility of this kind of analysis can be deter-
mined. One example is the use of self-representing and non-self-represent-
ing PSUsi in the current National Health Interview Survey design.
Even if all the issues can be resolved, a major problem would still
remain: the use of ecological correlations to draw inferences about rela-
tionships that refer to individuals. The PSUs defined in terms of health
service areas may contain too diverse a population to permit valid charac-
terization by some of the kinds of variables listed above. The internal
variation may simply be too large. However, individual-level analysis can
also be done by linking information about the individual health service area
to the individual's health information.
In an attempt to resolve some of the issues identified above, research
was undertaken by both the NCHS and the Bureau of the Census, as part of
the overall NHIS redesign research, to attempt to define health service areas
that would maximize the proportion of individuals receiving, their health
care within the defined areas, as well as to measure the effects of using
these areas as PSUs on the precision of the survey estimates. The panel has
learned that, on the basis of the findings of this research, NCHS has made
the design decision to retain for the 1995 NHIS redesign the existing defini-
4The Area Resource File is a county-level health resources database containing over 7,000
health professions and related data elements for each county in the United States. The purpose
of the file is to summarize statistics from many disparate sources into a single file in order to
facilitate analysis and planning of the geographical distribution of health professionals and
health resources. The Area Resource File contains data on geographic descriptor codes and
classifications, health professions, health characteristics and economic data, expenditures, and
environment. The data are drawn from major national surveys completed by numerous federal
agencies and professional associations.
HA self-representing primary sampling unit is one that is included in the sample with
certainty.
OCR for page 56
56
TOWARD A NATIONAL HEALTH CARE SURVEY
talon of PSUs. There is also some concern that defining PSUs in terms of
health service areas increases the NHIS sampling errors, the design effects
being, somewhat larger for the Hispanic population. Another factor influ-
encing the decision is the fact that the health service areas cross state boundaries,
thus making the NHIS redesign less appropriate for producing state statis-
tics. Analytic justifications for changing the PSU definition are lacking at
this time. This decision does not preclude consideration of redefining the
PSUs for the next NHIS redesign; in the interim, however, more research
needs to be undertaken before the next NHIS redesign to determine the
analytic gains, if any, and the cost consequences of redefining the NHIS
PSUs.
In summary, the concept of defining a geographic area in which most of
the providers serving a population are located and most of the population
receiving services from the providers in that area are also located is appeal-
ing. However, there are major difficulties in defining a health service area
across a range of services and different types of providers. For example,
the seNice areas for treatment of rare or serious conditions may not coin-
cide with the service areas for primary care. Moreover, there appears to be
serious loss in efficiency in using these areas as PSUs in the sample design.
The panel has not seen any definitive evidence at this time of the feasibility
or analytic justification of redefining the PSUs in terms of health service
areas on a national level, particularly since the health service areas can be
used in analyses of NCHS surveys without using them to define PSUs. The
panel, however, supports the center's intention to continue the needed re-
search in this area.
Recommendation 3-5: The panel endorses the NCHS decision
to use the primary sampling units from the National Health
Interview Survey for the National Health Care Survey, to re-
tain their existing definition at this time, and to continue the
needed research in this area.
Patient Follow-up Component
The center's plan for the National Health Care Survey includes devel-
opment of the capability to conduct routine and specialized patient follow-
up studies of the sample event (visit, discharge, or admission) to obtain
information beyond what is available in provider records. Data will be
collected from the patient or the patient's family about the patient outcome,
including subsequent use of medical care and morbidity, changes in health
status, and quality of care. The development of the follow-up component
will build on the experience gained from the 1985 National Nursing, Home
Survey Follow-up and the various follow-up studies of their population-based
OCR for page 57
REVIEW OF THE NCHS PLAN
.
.
.
.
57
surveys. NCHS also anticipates linking these data with other data sources,
just as the National Nursing Home Survey is being linked to the National
Death Index to obtain mortality status information on former patients.
Discussed below are some of the potential values of the design involv-
ing follow-up of patients whose visits and discharges are sampled for the
provider surveys; some of the methodological, legal, and feasibility issues
that will need to be resolved before such studies can be undertaken; and the
limitations of such an event-based approach to obtain the kinds of health
care data needed.
There are a number of potential benefits of the planned follow-up com-
ponent to the National Health Care Survey:
Data could be collected in the follow-up component to enable pa-
tients to be characterized in terms of their health status, demographic
characteristics, health behaviors, and other use of medical care.
Data could be collected from the patient or the patient's family
over a period of time about the results of treatment for the sampled
event, including use of medical care and morbidity subsequent to
the sampled event, changes in health status, hospital readmissions,
and quality of care.
A patient follow-up study could be conducted to focus on specific
issues such as a financing mechanism; particular demographic or
other groups of interest; a particular disposition at discharge; a
diagnosis, condition, or procedure; or any emerging issue.
Follow-up interviews with patients open the possibility of collect-
ing meaningful data on gross expenditures from providers of the
medical care and from insurers who paid for the service, at least for
the specific sampled event.
Patterns of care could be described for at least a subset of health
care events, which could enhance the analytic value of the data
about those services provided.
An important current concern in health policy research is the evaluation
of the effect of various patterns of treatment and medical care on patient
well-being. To address such issues, it is necessary to collect data directly
from patients about their health status, symptoms, and quality of life at one
or more points after they have received health care services and treatments.
Once a mechanism is in place for contacting patients who are sampled in
ambulatory or hospital settings; it would become possible to initiate special
studies of the results of those treatments and the medical outcomes associat-
ed with those sampled events.
OCR for page 58
58
Unresolved Issues
TOWARD A NATIONAL HEALTH CARE SURVEY
One of the problems inherent in the event-based follow-up design is the
need to get consent from patients for their health provider to release to
NCHS the name and address information needed in order to approach the
individual patients selected for follow-up. For patients whose visits were
sampled in an ambulatory setting, two alternative designs present them-
selves: (1) ambulatory care providers could be asked ex post facto to con-
tact the sample patients to gain permission for subsequent follow-up or to
allow for their patients to be contacted directly for follow-up and (2) inter-
viewers could interact with the patient directly in the physician's office.
The latter approach is similar to the design used in the Medical Outcome
Study (Tarlov et al., 1989~. In that project, an interviewer was continually
present in the physician's office during the sampling period. The interview-
er can explain the study, obtain needed release permissions, obtain informa-
tion needed for subsequent contact, and obtain data from patients. Al-
though this procedure would obviously require feasibility testing, and certainly
would constitute a major change in current survey procedures, the fact that
such a design has been used by others makes it worth serious consideration.
The feasibility, costs, and effectiveness of each of these procedures would
need to be investigated. A field study of the cost and effectiveness of these
approaches might also be considered in conjunction with any study to test
feasibility of alternative procedures to obtain names and addresses of sam-
pled patients.
Similar types of issues must be addressed for a sample of discharged
patients who were in the hospitalization survey sample. Permission will be
needed also from the hospital administrator to release to NCHS information
about the hospitalized patient, either at the time of admission or after the
discharge. One approach that can be considered is to have sampled hospi-
tals routinely ask for a release from all patients upon admission during a
sample period. An alternative approach is a passive consent procedure
whereby hospitals would mail to the discharged patients notification that
their names (nothing about their conditions) would be transmitted to NCHS
unless patients request that they not be. Matching the patient rec' -; s to
interview responses would occur only after patients had been informed and
had signed a consent form for the interviewer. Both approaches would
require considerable effort on the part of hospital employees. The feasibil-
ity and acceptability to hospitals and patients of alternative ways to gain
patient consent for follow-up need to be explored and tested. Moreover,
state laws vary widely on release of information on hospitalized patients.
In addition to the issues identified above, there are several other meth-
odological issues that need to be addressed to assess the feasibility of the
event-based follow-llp design, such as the efficacy of alternative ways to
OCR for page 59
REVIEW OF THE NCHS PLAN
59
deal with the problem of linkage between providers, visits, and persons; the
extent of attrition occurring during the process of contacting patients for
follow-up through the provider setting in which they were identified (i.e.,
hospital or ambulatory care provider); and the problem of constructing sam-
pling frames for providers other than those covered by existing surveys in
order to conduct the follow-up studies as planned by NCHS.
Research needs to be undertaken to evaluate the feasibility and effec-
tiveness of each of these issues and procedures, including an examination of
the literature to assess how effective these procedures have been when used
in previous studies.
Limitation of Event-based Follow-up Design
Although the panel supports the center's desire to develop capability
for collecting data directly from patients through follow-up, a key issue in
the design suggested in the plan is the appropriateness of an event-based
sample for follow-up studies. A major limitation of the National Health
Care Survey as presently designed is that all the provider surveys begin
with a sample of events and not persons. That means that the probability
of selection of any individual patient depends on the number and type of
eligible services that he or she receives. In the ambulatory care component
of the survey, for instance, people are sampled proportionate to their use of
ambulatory services in physicians' offices. Although Hat may be a s~ghdorward
way to sample events, it is not an efficient design for producing person-
based statistics. Many health conditions produce multiple visits to an am-
bulatory care setting; NCHS needs to determine the extent of the confound-
ing problem resulting from the use of an event-based survey. The conceptual
development of the issues involved in samples of events versus persons as a
unit of analysis has lagged; there is a clear need to think about these issues
in a more concrete and systematic manner than is possible within the con-
straints of this study. Laumann and Knoke (1987), for example, devote an
extended theoretical discussion in their book The Organizational State to
the actor-event interface and the analytic and methodological problems en-
gendered by taking the focal analytic unit to be the actor or the event, and
the need to see that these analytic units are, in fact, intimately interrelated
in myriad ways. Their perspective is especially congenial to researchers
who have adopted network-related analytic techniques.
Analyses of costs and outcomes depend on identifying meaningful clin-
ical events. Some service events are meaningful units to sample. The best
examples are those surgical procedures that almost always occur in hospital
settings and that usually occur only once to an individual. For example,
sampling hospital discharge records is a meaningful way to sample people
who experience gall bladder surgery or appendectomy. Starting with such a
OCR for page 60
60
TOWARD A NATIONAL HEALTH CARE SURVEY
sample, one could collect data about related services and interview patients
at a later date about the complications and benefits of surgery. All hospital-
izations, however, do not always produce meaningful data. For example,
hospital admissions for cancer could serve as a sampling frame to describe
the characteristics of those hospitalized for cancer and the services that they
utilize, their costs, and how they were feeling at some later point. Howev-
er, the same case of cancer could produce multiple admissions, and for
some analytic purposes the omission of those who had cancer but were not
hospitalized at the time (and hence were not represented in the statistics)
could constitute a serious problem.
In summary, the follow-up design described by NCHS potentially adds
to the value of the data collected from the NHDS and the NAMCS for
studying certain conditions. NClIS has made a case for sampling clinically
meaningful events from provider-based samples. Treatments that usually
occur only once constitute one class of health care that can be studied from
such samples. Health conditions that almost invariably lead an individual
to seek care from a provider, such as a broken hip, a broken leg, or a heart
attack, also probably can be meaningfully sampled from event-based sample
frames. There are real practical problems, however, associated with obtain-
ing permission and contacting a high percentage of sampled patients. There
are multiplicity problems associated with sampling to be resolved, arising
from the fact that an individual's chance of selection depends on the num-
ber of health care events the individual had for the condition (Sirken, 1972a,
1972b; Nathan, 1976~.
CONCLUSION
The panel supports the center's primary objective for the National Health
Care Survey. The panel is concerned, however, that, as currently designed,
the survey appears to be limited mostly to modest modifications and cover-
age extensions of the existing provider-based surveys, with minimal, if any,
true integration of design and data. Yet the concept of a National Health
Care Survey is a broad and challenging concept. If the center's objective,
as stated in its plan, is to move in the direction of an integrated survey
design to provide comprehensive health care data that are urgently needed,
especially on access, expenditures, disease episodes, and outcomes, then
such an objective should have a major impact on the design and content of
their surveys. NCHS cannot get that kind of statistical information from a
sample of events, which is its current approach.
Although NCHS provider surveys have considerable strengths in mea-
suring the health services provided in traditional medical care settings and
in some alternative care settings in the near future, they reflect only part of
the care provided in the United States today. Each of the NCHS provider
OCR for page 61
REVIEW OF THE NCHS PLAN
61
surveys on health care is oriented toward providing data on services re-
ceived for the sampled event in the respective facility. When taken togeth-
er, they do not represent a comprehensive approach to the collection of the
needed data on the nation's health care system.
NCHS has proposed strategies to address some of the gaps in the data
from its existing provider-based surveys. The panel supports these initia-
tives; however, the panel believes that the center's vision for a National
Health Care Survey falls short of a plan for meeting anticipated health care
information needs today and into the next century, even with the sugges-
tions contained in this chapter for changes and improvements. The panel
strongly urges the center to become more comprehensive in its approach to
meeting the data needs identified in this report. A strategy that is flexible
enough to adapt to changes and new fast-breaking directions requires a
more integrated and a more visionary course of action than currently planned
by NCHS. The next chapter provides the panel's recommendations for a
design framework that would lead to a more integrated National Health
Care Data System that will provide the basis for such a flexible, long-term
data collection and implementation strategy.
Representative terms from entire chapter:
provider surveys
