Public Health: A Health Status Indicator for Targeting Federal Aid to
States (Letter Report, 11/13/96, GAO/HEHS-97-13).
Pursuant to a congressional request, GAO identified measures of the
health status of states' populations that could be used to target
federal funds, focusing on whether: (1) premature mortality rates and
the number of people in poverty would be appropriate measures of and
proxies for health status; and (2) other measures might also be used.
GAO found that: (1) premature mortality is the best single proxy for
reflecting differences in the health status of states' populations as
measured by both the Centers for Disease Control and Prevention's
Healthy People 2000 indicators and the ReliaStar Financial Corporation's
composite health status index; (2) using premature mortality to
distribute federal funding for core public health functions would
systematically target federal assistance to states on the basis of their
populations' rates of mortality, disease incidence, and risk for
mortality and morbidity; (3) a number of other variables, including the
proportion of states' populations that are poor or minorities, were also
found to be correlated with health status differences, but including
these variables along with premature mortality did not significantly
enhance GAO's ability to differentiate the health status of state
populations; and (4) improving the targeting of funds beyond that
obtained using premature mortality alone would require using several
additional variables, which would add to the complexity of the
allocation formula.
--------------------------- Indexing Terms -----------------------------
REPORTNUM: HEHS-97-13
TITLE: Public Health: A Health Status Indicator for Targeting
Federal Aid to States
DATE: 11/13/96
SUBJECT: Health statistics
Health care programs
Block grants
Public health legislation
State-administered programs
Statistical methods
Disadvantaged persons
Federal aid to states
Population statistics
IDENTIFIER: HHS Healthy People 2000 Program
ReliaStar Composite Health Status Index
******************************************************************
** This file contains an ASCII representation of the text of a **
** GAO report. Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved. Major **
** divisions and subdivisions of the text, such as Chapters, **
** Sections, and Appendixes, are identified by double and **
** single lines. The numbers on the right end of these lines **
** indicate the position of each of the subsections in the **
** document outline. These numbers do NOT correspond with the **
** page numbers of the printed product. **
** **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced. Tables are included, but **
** may not resemble those in the printed version. **
** **
** Please see the PDF (Portable Document Format) file, when **
** available, for a complete electronic file of the printed **
** document's contents. **
** **
** A printed copy of this report may be obtained from the GAO **
** Document Distribution Center. For further details, please **
** send an e-mail message to: **
** **
** **
** **
** with the message 'info' in the body. **
******************************************************************
Cover
================================================================ COVER
Report to the Chairman, Committee on Labor and Human Resources, U.S.
Senate
November 1996
PUBLIC HEALTH - A HEALTH STATUS
INDICATOR FOR TARGETING FEDERAL
AID TO STATES
GAO/HEHS-97-13
Health Status Indicator
(118130)
Abbreviations
=============================================================== ABBREV
AIDS - acquired immunodeficiency syndrome
CDC - Centers for Disease Control and Prevention
HIV - human immunodeficiency virus
YPLL - years of potential life lost
Letter
=============================================================== LETTER
B-274670
November 13, 1996
The Honorable Nancy Kassebaum
Chairman, Committee on Labor
and Human Resources
United States Senate
Dear Madam Chairman:
The public health system plays an essential role in improving the
health of the American population. It promotes the prevention of
communicable diseases and of exposure to toxic environmental
pollutants and helps guard against harmful products and poor-quality
health care. During fiscal year 1993, federal, state, and local
governments spent an estimated $14 billion on activities such as
preventive services, health surveillance, outreach, training, and
planning.
The federal government fulfills its role in improving public health
in a variety of ways. For example, it coordinates with the states to
set and implement national health policy. It also sponsors and
administers programs for developing health resources and preventing
and controlling diseases and alcohol and drug abuse. In fiscal year
1993, the federal government financed an estimated $3 billion, or
about 20 percent, of the $14 billion spent on core public health
functions by all levels of government.
As part of the federal effort, the Centers for Disease Control and
Prevention (CDC) funds a variety of grant programs to state and local
governments for such projects as immunization, human immunodeficiency
virus (HIV) prevention, preventive health activities, and lead
poisoning prevention. In fiscal year 1995, CDC distributed nearly $2
billion to the states for public health services.
On January 4, 1995, you introduced the Public Health Enhancement Act
of 1995 (S. 142).\1 Under your proposal, 12 of the federal public
health grants that are administered by CDC would have been
consolidated into one integrated health system block grant (see app.
I for a list of these programs). The goal of your proposal was to
increase the efficiency and flexibility with which the public health
system attends to state and regional health problems.
To help ensure that this block grant reflected this goal, you asked
us to identify measures of the health status of states' populations
that could be used to target federal funds. More specifically, you
asked whether premature mortality rates\2 and number of people in
poverty would be appropriate measures of--and proxies for--health
status and whether other measures might also be used.
Using state-level data, we examined the relationship of premature
mortality, people in poverty, and other measures to information on
states' health status from two sources and assessed the
appropriateness of including these proxies in an allocation formula
for a federal public health block grant. The two sources we used
were the Healthy People 2000 indicators, which consist of 18
indicators compiled by CDC, and a composite health status index
developed by the ReliaStar Corporation.
To determine if cross-state differences in health status could be
largely explained--and, therefore, represented in a funding
formula--by one or just a few proxies, we undertook a statistical
analysis to identify those variables that best accounted for
cross-state differences in the Healthy People 2000 indicators and the
ReliaStar index. First, we used principal component analysis to
reduce the total variation associated with the Healthy People 2000
indicators to a smaller number of more general components. Next, we
conducted correlation and regression analyses to determine the extent
to which individual Healthy People 2000 indicators, premature
mortality, and other selected variables could serve as proxies for
the entire set of Healthy People 2000 indicators. Similarly, we
examined the correlations of these indicators with the ReliaStar
index to determine if they corroborated the results from our analysis
of the Healthy People 2000 indicators (see apps. II and III for more
detail). We conducted our work from December 1995 through August
1996 in accordance with generally accepted government auditing
standards.
--------------------
\1 S. 142 did not become law.
\2 We used years of potential life lost (YPLL) as our measure of
premature mortality. YPLL is defined as the number of years between
the age at death, for those who die before age 65, and age 65. While
YPLL is typically calculated over an age range from birth to 65 years
of age, it is also calculated using other age ranges, such as birth
to age 75.
RESULTS IN BRIEF
------------------------------------------------------------ Letter :1
We found that premature mortality is the best single proxy for
reflecting differences in the health status of states' populations as
measured by both the Healthy People 2000 indicators and the ReliaStar
index. Premature mortality accounted for 36 percent of the variation
in the Healthy People 2000 indicators and 75 percent of the variation
in the ReliaStar index. Our analysis showed that using premature
mortality to distribute federal funding for core public health
functions would systematically target federal assistance to states on
the basis of their populations' rates of mortality, disease
incidence, and risk for mortality and morbidity.
A number of other variables, including the proportion of states'
populations that are poor or minorities, were also found to be
correlated with health status differences as measured by the Healthy
People 2000 indicators and the ReliaStar index. However, including
these variables along with premature mortality did not significantly
enhance our ability to differentiate the health status of state
populations. Moreover, improving the targeting of funds beyond that
obtained using premature mortality alone would require using several
additional variables, which would add to the complexity of the
allocation formula.
BACKGROUND
------------------------------------------------------------ Letter :2
Health status is a multidimensional concept encompassing such
elements as the presence of disease, quality of life, and risk of
death. The Healthy People 2000 program, administered by the
Department of Health and Human Services, is a national strategy for
improving the health of the American people in the decade preceding
the year 2000. As part of this strategy, CDC established 18
indicators with which to assess different dimensions of states'
health status. The indicators address health status outcomes, such
as mortality and disease incidence, and potential health risk
factors, such as low-weight births and childhood poverty. (See app.
II for more detail.)
In addition to Healthy People 2000, the ReliaStar Financial
Corporation annually publishes a composite health status index. In
contrast to the Healthy People 2000 indicators, the ReliaStar index
presents a single summary indicator of health status. When the index
was first developed, a panel of public health experts reached
consensus on 17 indicators to use to reflect a variety of health
status outcomes and risks, including lifestyle, disease, and
mortality. The panel then agreed on how to weight each of the
individual indicators to create an overall summary indicator. (See
app. III for more detail.)
The Healthy People 2000 indicators and ReliaStar index include a
number of the same health indicators, but there are several that are
unique to each. For example, both use infant mortality and motor
vehicle deaths, but only the Healthy People 2000 indicators include
low-weight births, births to adolescents, childhood poverty, and air
quality. In contrast, the ReliaStar index includes the prevalence of
smoking, access to primary care, and state support for public health
care, but the Healthy People 2000 indicators do not.
PREMATURE MORTALITY BEST
REFLECTED CROSS-STATE
DIFFERENCES IN HEALTH STATUS
------------------------------------------------------------ Letter :3
Our analysis indicated that premature mortality better explained the
cross-state variation in the Healthy People 2000 indicators than any
other measure that we examined: Independently, premature mortality
accounted for 36 percent of the cross-state variation. Once
differences in premature mortality were accounted for, the efficacy
of using other indicators became questionable. In some cases, adding
a particular indicator could modestly increase the amount of
variation in the Healthy People 2000 indicators that was accounted
for, but doing so might not be appropriate. Adding states' suicide
rates as an indicator, for example, would increase the explained
variation by about 9 percentage points, but the subjectivity involved
in identifying suicide as a cause of death would limit its value as a
variable in the funding formula. That is, the potential influence on
funding levels of using the suicide rate indicator could affect how
reliably states reported it as a cause of death.
In other cases, including additional indicators contributed little to
explaining the variation in health status presented by the Healthy
People 2000 indicators. For example, including the percentage of a
state's population that lived in a rural area increased the amount of
variation that was accounted for by only 5 percentage points.
Further, indicators such as deaths due to work-related injuries,
cardiovascular deaths, the percentage of a state's population living
in an area with poor air quality, teen births, and motor vehicle
deaths each increased the percentage of explained variation by 3
percentage points or less.
We also examined the relationship between the ReliaStar index and
premature mortality and other selected indicators. We found that
premature mortality was the best single indicator for explaining
cross-state variation associated with this index: It accounted for
75 percent of the cross-state variation in the ReliaStar index.
Adding motor vehicle deaths increased the proportion of explained
variation by only 7 percentage points, to 82 percent. Adding the
proportion of people in poverty as a third indicator increased the
share of explained variation to 85 percent.
CONCLUSIONS
------------------------------------------------------------ Letter :4
Premature mortality is an appropriate health status indicator for
allocating federal funding for the core public health functions
administered by the states. If using a single indicator in an
allocation formula is considered desirable to reduce its complexity
and the burden of administering it, premature mortality is the best
indicator to choose.
Using premature mortality along with states' suicide rates would
better reflect cross-state differences in states. However,
determining suicide as a cause of death is somewhat subjective, and
including it as a variable in a funding formula could affect the
reliability of data reported on suicide rates.
---------------------------------------------------------- Letter :4.1
Because this report does not directly affect agency operations, we
did not obtain comments from CDC or any other agency of the
Department of Health and Human Services. We did, however, submit our
report for review by outside experts and included their technical
suggestions, where appropriate.
We are sending copies of this report to appropriate congressional
committees and subcommittees, the Secretary of Health and Human
Services, and the Director of the Centers for Disease Control and
Prevention.
This report was prepared under the direction of Jerry Fastrup,
Assistant Director, Health Financing and Systems Issues. If you have
any questions about this report, please contact me on (202) 512-7119
or Jerry Fastrup on (202) 512-7211. Other individuals who made
contributions to this report include Mark Vinkenes and Michael
O'Dell, both Senior Social Science Analysts.
Sincerely yours,
William J. Scanlon
Director, Health Financing
and Systems Issues
CDC PROGRAMS THAT WOULD HAVE BEEN
CONSOLIDATED UNDER S. 142
=========================================================== Appendix I
Total fiscal
year 1995
award amounts
Programs (in millions)
------------------------------------------ --------------
Immunization grant program $355
Preventive health services programs for 227
human immunodeficiency virus\a
Preventive health and health services 154
block grant
Preventive health services programs for 108
tuberculosis
Breast and cervical cancer 84
Prevention and control of sexually 82
transmitted disease
Lead poisoning prevention 26
Cancer registries 14
Preventive health services programs for 11
diabetes
Preventive health services programs for 8
disabilities
Preventive health services programs for 5
tobacco use
Infertility and sexually transmitted \b
disease prevention
----------------------------------------------------------
\a Human immunodeficiency virus = HIV.
\b Included in the funding amount for prevention and control of
sexually transmitted disease.
HEALTHY PEOPLE 2000 INDICATORS
========================================================== Appendix II
As part of Healthy People 2000, the Centers for Disease Control and
Prevention (CDC) monitors the health status of state populations
using 18 health status indicators. On the basis of a statistical
analysis of these indicators and selected variables, we concluded
that premature mortality is the best single proxy to reflect
differences in health status as measured by the Healthy People 2000
indicators. Moreover, we concluded that once premature mortality is
accounted for, no other single measure appears to appreciably improve
the differentiation of states' health status. Instead, several
measures would have to be used to differentiate health status among
states, and this would make a grant allocation formula more complex
and difficult to use. This appendix presents the statistical
analyses that support these conclusions.
BACKGROUND
-------------------------------------------------------- Appendix II:1
Healthy People 2000 is a national strategy for improving the health
of the American people in the decade preceding the year 2000. The
strategy, which was unveiled in September 1990 by the Secretary of
the Department of Health and Human Services, has three broad goals:
-- to increase the span of healthy life for Americans,
-- to reduce health disparities among Americans, and
-- to provide access to preventive services to all Americans.
These goals are supported by 300 objectives that address 22 priority
areas. For example, one objective is to reduce the prevalence of
cigarette smoking to no more than 15 percent of the population aged
20 and older. Another objective is to increase basic immunization
levels to at least 90 percent among children under age 2. For each
priority area, a U.S. Public Health Service agency was designated
both to develop an implementation plan and to coordinate activities
directed toward attaining the objectives under that area.
CDC was delegated responsibility for the priority area concerning
health surveillance and the development of supporting data systems.
As part of this responsibility, CDC developed a set of 18 health
status indicators that are used to track the general health status of
state populations.
The indicators were chosen to facilitate national, state, and local
efforts in tracking the Healthy People 2000 objectives and to help
communities assess the general health status of their populations. A
committee consisting of federal, state, and local health officials
and representatives from academic institutions selected the 18
indicators by consensus.
STATES DIFFERED IN TERMS OF
MORTALITY, DISEASE INCIDENCE,
AND RISK
-------------------------------------------------------- Appendix II:2
Each of the Healthy People 2000 indicators is based on data that are
produced by the federal government. We have classified the 18
indicators into three groups: mortality, disease incidence, and
indicators of health risk (see table II.1).
Table II.1
Healthy People 2000 Indicators
Minimum/
National maximum
average rate state rate
------------------------------ ------------ ------------
Mortality indicators (per 100,000 population)\a
----------------------------------------------------------
Total mortality 504.5 392.0-608.0
Cardiovascular deaths 180.4 137.2-237.1
Lung cancer deaths 39.3 17.1-53.6
Breast cancer deaths 21.9 16.4-27.5
Motor vehicle deaths 15.8 8.7-31.6
Suicides 11.1 5.9-22.9
Homicides 10.5 0-19.6
Infant mortality 8.5 5.6-11.9
Work injury-related deaths 3.2 0-15.5
Disease incidence indicators (per 100,000 population)\b
----------------------------------------------------------
Acquired immunodeficiency 31.2 1.4-79.0
syndrome\c
Syphilis 10.4 0-67.5
Tuberculosis 9.8 0-21.7
Measles 0.1 0-5.2
Risk factors\d
----------------------------------------------------------
Poor air quality, as measured 23.5% 0-96.9
by the proportion of people
living in counties exceeding
U.S. Environmental Protection
Agency standards for air
quality during the previous
year
Prenatal care, as measured by 22.3% 11.5-38.3
the percentage of mothers
delivering live infants who
did not receive prenatal care
during the first trimester
Childhood poverty, as measured 20.8% 9.5-39.4
by the proportion of children
less than 15 years of age
living in families at or
below the poverty level
Low birth weight, as measured 7.1% 4.9-9.9
by the percentage of live-
born infants weighing less
than 2500 grams at birth
Births to adolescents (females 4.9% 1.9-9.4
aged 10 to 17 years) as a
percentage of total live
births
----------------------------------------------------------
\a All mortality indicators are based on 1992 data except for
work-related death and infant mortality rates. Work-related death
rates are based on 1993 data. Infant mortality rates are based on
1992 data except for people of Asian, Native American, and Hispanic
origin. Infant mortality rates for these groups are based on an
average of 1989 to 1991 data. Also, the infant mortality rate is
determined by the rate (per 1,000 live births) of death among infants
less than 1 year of age.
\b All disease incidence indicators are based on 1993 data.
\c Acquired immunodeficiency syndrome = AIDS.
\d All risk indicators are based on 1993 data.
As shown in table II.1, individual states differed from the national
average on each of the indicators. In some cases, these differences
were substantial. While the national average total mortality rate
was 505 per 100,000 people, this rate ranged from 392 in Hawaii to
608 in Mississippi. States differed most with respect to deaths due
to work-related injuries. For this indicator, the highest rate was
nearly 400 percent higher than the national average. States differed
the least in terms of total mortality and deaths resulting from
breast cancer. The highest rates differed from the national averages
on these indicators by about 20 percent and 26 percent, respectively.
When considering the variation across all of the 18 indicators,
states differed most with respect to the disease incidence indicators
and least with respect to the indicators of mortality and risk.
Among the former indicators, the maximum rates differed from the
national averages by a range of about 120 to 5,100 percent. In
contrast, for the mortality and risk factors, the maximum rates
differed from the national averages by a range of about 20 to nearly
400 percent.
STATISTICAL ANALYSIS
-------------------------------------------------------- Appendix II:3
Using all of the Healthy People 2000 indicators to reflect states'
health status would result in a complex formula with 18 discrete need
indicators. A less complex formula is possible, however, by
reflecting the cross-state variation in the 18 Healthy People 2000
indicators with a smaller subset of proxies.
To identify such proxies, we followed a two-step process. First, we
performed a principal component analysis to reduce the Healthy People
2000 indicators to a smaller number of components.\3 We then used
correlation and regression analyses to determine how well premature
mortality, poverty, the individual indicators, and other selected
measures or combinations of these variables accounted for the
cross-state variation contained in the Healthy People 2000
indicators, and their suitability as proxies for health status in an
allocation formula.
--------------------
\3 For the measles indicator, the mean was quite small and the
standard deviation quite large. Consequently, we omitted this
indicator from our analysis. Hence, our principal component analysis
involved 17 of the original 18 Healthy People 2000 indicators.
PRINCIPAL COMPONENT ANALYSIS
------------------------------------------------------ Appendix II:3.1
Principal component analysis is a statistical technique that creates
new "synthetic" variables, called principal components, to reflect as
much of the total variation within a group of variables as possible
but with a smaller set of components. Principal components are
created sequentially, so that the first component reflects as much of
the variation in the original group of variables as possible.
Succeeding components are each created to reflect the largest
possible shares of the remaining variation. The total number of
principal components equals the total number of original variables
and, taken together, all of the principal components explain all of
the variation in these variables. By using this technique, it is
often possible to reflect a substantial proportion of the variation
of a large number of variables with only two or three principal
components. In a principal component analysis, all variables are
treated as equally important. For example, a high rate of mortality
is treated as being as important as a high incidence of measles.
FOUR PRINCIPAL COMPONENTS
ACCOUNT FOR
THREE-QUARTERS OF THE
VARIATION IN THE HEALTHY
PEOPLE 2000 INDICATORS
---------------------------------------------------- Appendix II:3.1.1
The principal component analysis of the Healthy People 2000
indicators produced four components that accounted for over 75
percent of the variation in the original set of indicators (see table
II.2). The first component accounted for 42 percent of the total
variation, with each remaining component adding successively less to
the explained variation. Components produced beyond the fourth one
contributed about 5 percent or less to the explained variation;
therefore, we eliminated them from the remaining analysis.
Table II.2
Principal Components
Percentage
Percentage of
of Cumulative explained
Component variation percentage variation
---------------------- ---------- ---------- ----------
1 42.1 42.1 55
2 17.4 59.5 23
3 10.7 70.2 14
4 6.3 76.5 8
----------------------------------------------------------
Each principal component is calculated as a weighted sum of the
original 17 indicators. A better understanding of the components can
be obtained by examining the weights.\4 Indicators with large weights
have a greater influence on a principal component and indicators with
small weights have less influence. The weight on each of the
indicators is shown in table II.3. Weights above 0.60 appear in bold
type.
Table II.3
Weight on Each Healthy People 2000
Indicator Associated With Each Principal
Component
1 2 3
Healthy People 2000 Indicator (42%) (17%) (11%) 4 (6%)
-------------------------------------- ------ ------ ------ ------
Mortality indicators
----------------------------------------------------------------------
Total mortality .9264 - - .2557
.0553 .1459
Cardiovascular deaths .8802 - - .0626
.1506 .3086
Lung cancer deaths .6476 - - .3749
.2370 .3921
Breast cancer deaths .0447 - - .2801
.5010 .4876
Motor vehicle deaths .5644 .7379 - .1008
.0742
Suicides - .7414 .3660 .4271
.0116
Homicides .8646 - .3097 -
.0970 .0020
Infant mortality .7819 .0662 - -
.3049 .1345
Work injury-related deaths .0182 .6458 - -
.0688 .0482
Disease incidence indicators
----------------------------------------------------------------------
AIDS .2441 - .4661 -
.7371 .0459
Syphilis .8241 .0194 - -
.1338 .1413
Tuberculosis .5072 - .5359 -
.4229 .2064
Risk factors
----------------------------------------------------------------------
Poor air quality - - .4027 .6837
.0779 .3571
Prenatal care .5331 .3660 .5341 .0038
Childhood poverty .7737 - .0288 -
.0606 .1107
Low birth weight .8538 - .0940 -
.1136 .1468
Births to adolescents .8969 .3067 .0773 .0034
----------------------------------------------------------------------
--------------------
\4 These weights are generally referred to as loading factors in a
principal component analysis.
INTERPRETATION OF THE
PRINCIPAL COMPONENTS
---------------------------------------------------- Appendix II:3.1.2
The first component appears to be most strongly related to both
mortality and risk factors. Total mortality has a weight of 0.93,
and deaths as a result of cardiovascular disease, homicide, and
infant mortality also have high weights. Three of the five risk
factors (childhood poverty, low birth weight, and teen births) also
contribute heavily to the first component.
The second component is most strongly related to suicide, motor
vehicle deaths, and work injury-related deaths. The component is
negatively related to the incidence of AIDS. No single indicator or
set of indicators is highly related to the third component. The
fourth component is most strongly associated with areas with poor air
quality (weight = 0.68).
STEPWISE REGRESSION
------------------------------------------------------ Appendix II:3.2
We used correlation and regression analyses to see if premature
mortality, poverty, and other selected variables could serve as
proxies for the principal components. Through correlation analysis,
we identified as potential proxies any of the variables that had a
correlation coefficient of at least 0.60 with a component.\5 We then
used stepwise regression to determine the extent to which variation
in each of the principal components could be explained by these
variables.
The stepwise regression procedure allowed us to determine which
statistical models most simply explained each of the components.
This technique selects the variable that is most strongly related to
the dependent variable and then includes only those additional
variables that increase the explanatory power of the model.
Moreover, a variable with a relationship to the dependent variable
that is likely to have occurred by chance is omitted from the
equation.\6
--------------------
\5 The measures included premature mortality, minority population,
poverty population, population under age 18, population over age 60,
population between ages 16 and 24, and rural population.
\6 Variables were excluded from a regression equation when their
relationship to the dependent variable had more than 5 chances in 100
to have occurred randomly.
COMPONENT 1
---------------------------------------------------- Appendix II:3.2.1
As mentioned before, the first component accounted for the largest
proportion of variation in the Healthy People 2000 indicators--42
percent. The component was correlated at 0.60 or more with nine of
the Healthy People 2000 indicators and two of the other selected
measures--premature mortality and poverty. The correlations exceeded
0.80 for eight of these variables.
Since premature mortality had the highest correlation with the first
principal component, the stepwise regression technique selected it as
the first variable to be included in the model. Premature mortality
alone explained 86 percent of the variation in this component (see
adjusted R\2 in table II.4). Adding cardiovascular-related deaths
contributed an additional 7 percentage points to the explained
variation, increasing the R\2 to 93 percent. Teen births contributed
an additional 4 percentage points to the explained variation, while
homicide and low birth weight added 1 percentage point each.
Table II.4
Component 1 Stepwise Regressions
Model 1 Model 2 Model 3 Model 4 Model 5
------------------- ------------ ------------ ------------ ------------ ------------
Adjusted R\2 0.86 0.93 0.97 0.98 0.99
Intercept -5.40 -6.66 -5.90 -5.35 -5.53
(t-value) (-17.04) (-22.85) ( -30.35) (-26.21) (-30.26)
Premature mortality 5.70 3.77 2.31 1.40 1.23
(t-value) (17.29) (10.29) (8.46) (4.57) (4.60)
Cardiovascular 3.16 2.67 2.76 2.57
deaths
(t-value) (6.85) (9.43) (11.57) (12.12)
Births to 1.14 1.06 0.98
adolescents
(t-value) (9.12) (9.97) (10.46)
Homicides 0.41 0.33
(t-value) (4.52) (4.09)
Low birth weight 0.68
(t-value) (4.00)
-----------------------------------------------------------------------------------------
COMPONENT 2
---------------------------------------------------- Appendix II:3.2.2
The second principal component accounted for 17 percent of the total
variation in the Healthy People 2000 indicators. Only three Healthy
People 2000 indicators had correlations at 0.60 or more (see table
II.3); only one of the other selected measures--population less than
18 years of age--had a correlation of 0.60 or more. The second
component is most strongly related to specific kinds of mortality,
for example, accidental death and suicide.
Suicide alone explained 54 percent of the variation in the second
component (see table II.5). Work injury-related deaths increased the
explained variation by 17 percentage points, and motor vehicle deaths
added another 7 percentage points, raising the explained variation to
78 percent. Finally, population less than 18 years of age added 2
percentage points.
Table II.5
Component 2 Stepwise Regressions
Model 1 Model 2 Model 3 Model 4
--------------------------------- ------------ ------------ ------------ ------------
Adjusted R\2 0.54 0.71 0.78 0.80
Intercept -2.89 -2.97 -3.38 -5.35
(t-value) (-7.42) (-9.60) (-11.75) (-5.88)
Suicides 2.64 2.10 1.62 1.38
(t-value) (7.65) (7.21) (5.80) (4.80)
Work injury-related deaths 0.57 0.41 0.28
(t-value) (5.43) (4.07) (2.56)
Motor vehicle deaths 1.03 1.06
(t-value) (4.02) (4.33)
Population less than 18 years of 2.31
age
(t-value) (2.27)
-----------------------------------------------------------------------------------------
COMPONENT 3
---------------------------------------------------- Appendix II:3.2.3
The third component accounted for about 11 percent of the total
variation in the Healthy People 2000 indicators. This component was
moderately correlated with only one measure--rural population--which
explained about 43 percent of its variation.\7
--------------------
\7 Because of the negative sign associated with rural population, by
implication, the factor is positively correlated with urban
population.
COMPONENT 4
---------------------------------------------------- Appendix II:3.2.4
Finally, the fourth component accounted for the smallest percentage
of the total variation in the Healthy People 2000 indicators--about 6
percent. This component was moderately correlated with just one of
the Healthy People 2000 indicators, the proportion of people living
in counties exceeding U.S. Environmental Protection Agency air
quality standards, which explained about 46 percent of the variation
in the fourth component.
PREMATURE MORTALITY IS THE
BEST SINGLE PROXY OF HEALTH
STATUS THAT WE TESTED
------------------------------------------------------ Appendix II:3.3
We combined the results of the principal component and regression
analysis to determine how much total variation each potential proxy
accounted for in the 17 Healthy People 2000 indicators. Premature
mortality accounted for 36 percent of the total variation since, by
itself, it accounted for 86 percent of the variation in the first
component, which itself represented 42 percent of the total variation
in the set of indicators. The results for the other proxies are
summarized in table II.6.
Table II.6
Proportion of the Cross-State Variation
in the Healthy People 2000 Indicators
Associated With Various Proxies
Explained
percentage
Variable of variation
-------------------------------------------- ------------
Premature mortality 36
Suicides 9
Rural population 5
Work injury-related deaths 3
Cardiovascular deaths 3
Poor air quality 3
Births to adolescents 2
Motor vehicle deaths 1
Homicides \a
Low birth weight \a
Population less than 18 years of age \a
----------------------------------------------------------
\a Less than 1 percent.
RELIASTAR STATE HEALTH STATUS
INDEX
========================================================= Appendix III
The Northwestern National Life Insurance Company, a subsidiary of the
ReliaStar Financial Corporation, has published state rankings of
health status since 1989 that are based on a methodology they
developed for this purpose. We found that premature mortality alone
can explain 75 percent of the cross-state variation in the ReliaStar
index. As with Healthy People 2000, once premature mortality is
accounted for, none of the other variables that we used adds
appreciably to the explanation of the variation in the ReliaStar
index. Further, no other measure that we tested was shown to be a
reasonable single proxy for both the Healthy People 2000 indicators
and the ReliaStar index. This appendix presents our analysis of the
ReliaStar index.
DEVELOPMENT AND COMPOSITION OF
THE RELIASTAR INDEX
------------------------------------------------------- Appendix III:1
The ReliaStar state health status index is an overall measure by
state of the general health of the population in the United States.
The index was first published in October 1989 and has been published
annually ever since. Except for one major refinement in 1990, the
methodology used to produce the index has generally remained the
same. Therefore, versions of the index since 1990 are essentially
comparable and can be used to note shifts in measured health status
from year to year.
The ReliaStar index is currently based on an overall score produced
from 17 health status measures that are grouped into five categories:
lifestyle, access to care, occupational safety and disability,
disease, and mortality (see table III.1). Each measure is assigned a
weight on the basis of a consensus judgment of health experts that
determines the measure's percentage of the overall score. An overall
score is constructed by summing the measures after multiplying each
one by its weight. We have converted the score to an index number so
that a score of 1.0 indicates the national average.
Table III.1
Measures and Weights of ReliaStar State
Health Rankings
Minimum/
Weight maximum
Measure (in percent) Mean state values
----------------------------------------------- ------------ ------------ ------------
Lifestyle
-----------------------------------------------------------------------------------------
Prevalence of smoking--the percentage of the 10.0 22.2% 14.3-29.3%
population over age 18 that smokes tobacco
products regularly
Motor vehicle deaths--the annual number of 5.0 1.8 0.9-3.0
deaths per 100 million miles driven
Violent crime--the annual number of murders, 5.0 746.0 82.0-
rapes, robberies, and aggravated assaults per 1,206.0
100,000 people
Risk for heart disease--a measure of three 5.0 \a \a
criteria: obesity, hypertension, and sedentary
lifestyle
High school graduation--the percentage of ninth 5.0 71.1% 56.3-89.1%
graders who graduate within 4 years
Access to care
-----------------------------------------------------------------------------------------
Unemployment--the average percentage of the 5.0 6.1% 2.9-8.6%
civilian, noninstitutional labor force that is
unemployed during the year
Adequacy of prenatal care--the percentage of 5.0 70.2% 49.1-84.1%
pregnant women who had at least one prenatal
visit with a health professional within the
first trimester of pregnancy and at least 9
visits within the 36 weeks of gestation
Lack of access to primary care--the percentage 5.0 10.6% 2.5-27.4%
of population areas that are underserved by
primary medicine practitioners residing in
designated Health Manpower Shortage Areas
Support for public health care--the percentage 5.0 Index = 1.44 0.69-2.69
of a state's expenditures for public welfare,
health care, and related services divided by
the percentage of the state's population with
an annual income below $15,000
Occupational safety and disability
-----------------------------------------------------------------------------------------
Occupational fatalities--the incidence per 2.5 7.1 1.3-43.4
100,000 workers of fatalities over a 5-year
period
Work disability status--the percentage of a 2.5 4.2% 2.3-8.4%
state's population that says a disability
prevents them from working
Disease
-----------------------------------------------------------------------------------------
Heart disease--a 3-year average death rate per 7.5 151.0 116.0-188.0
100,000 due to heart disease that is age-and
race-adjusted
Cancer cases--the number of projected cases per 7.5 485.0 200.0-648.0
100,000 for the current year
Infectious disease--a 3-year average per 5.0 53.4 10.8-103.1
100,000 that includes the occurrence of AIDS,
tuberculosis, and hepatitis (all types)
Mortality
-----------------------------------------------------------------------------------------
Total mortality--a 3-year average rate per 10.0 521.0 430.0-590.0
100,000 that is age-and race-adjusted
Infant mortality--a 3-year average rate per 7.5 8.5 5.4-9.7
1,000 births that is race-adjusted
Premature death--the loss of years of 7.5 5,348.0 3,552.0-
productive life per 100,000 due to death 7,388.0
before age 65
-----------------------------------------------------------------------------------------
\a Data not readily available.
STATE VARIABILITY ACROSS
MEASURES
----------------------------------------------------- Appendix III:1.1
Table III.1 shows that, as with the Healthy People 2000 indicators,
the individual states can differ significantly from the national
average on the ReliaStar measures. For example, the incidence of
occupational fatalities in Alaska was about five times the national
average. Also like the Healthy People 2000 indicators, however, the
states tended to be more similar in terms of the mortality measures.
For example, the highest rate of total mortality was only about 13
percent higher than the national average.
DISTRIBUTION OF THE
RELIASTAR INDEX
----------------------------------------------------- Appendix III:1.2
Each state's index score, as determined by the ReliaStar methodology,
is reported in table III.2. The national average is represented by
an index score of 1.00. Poorer health status is represented by
higher index scores and better health status by lower ones. For
1995, the health status index was lowest in the states of Minnesota,
New Hampshire, and Utah. Conversely, the health status index was
highest in the states of Louisiana, Mississippi, and South Carolina.
Table III.2
ReliaStar Health Status Index Scores,
1995
ReliaStar
State index
-------------------------------------------- ------------
Alabama 1.11
Alaska 1.12
Arizona 0.99
Arkansas 1.14
California 0.99
Colorado 0.91
Connecticut 0.85
Delaware 1.01
Florida 1.10
Georgia 1.04
Hawaii 0.85
Idaho 0.98
Illinois 1.00
Indiana 0.97
Iowa 0.89
Kansas 0.91
Kentucky 1.10
Louisiana 1.23
Maine 0.90
Maryland 0.92
Massachusetts 0.88
Michigan 1.00
Minnesota 0.83
Mississippi 1.24
Missouri 1.04
Montana 0.96
Nebraska 0.90
Nevada 1.14
New Hampshire 0.81
New Jersey 0.92
New Mexico 1.12
New York 1.10
North Carolina 1.02
North Dakota 0.92
Ohio 0.96
Oklahoma 1.06
Oregon 0.97
Pennsylvania 0.95
Rhode Island 0.93
South Carolina 1.19
South Dakota 0.94
Tennessee 1.10
Texas 1.03
Utah 0.83
Vermont 0.86
Virginia 0.91
Washington 0.92
West Virginia 1.16
Wisconsin 0.89
Wyoming 1.03
----------------------------------------------------------
Note: States in bold type have the three highest and three lowest
ReliaStar rankings.
Figure III.1 shows that the health status index is typically lower
(better health status) in the New England states and in the
Midwestern states like Iowa, Minnesota, and Wisconsin. Conversely,
the health status index tends to be higher (poorer health status) in
the southern states, for example, Alabama, Louisiana, Mississippi,
and South Carolina.
Figure III.1: ReliaStar Health
Status Index Scores
(See figure in printed
edition.)
COMPARISON OF THE RELIASTAR
INDEX AND HEALTHY PEOPLE 2000
INDICATORS
------------------------------------------------------- Appendix III:2
To facilitate a comparison of the Healthy People 2000 indicators with
the ReliaStar index, table III.3 displays the health status measures
found in each set according to whether they reflect mortality,
disease incidence, health risk, or some other factor.
Table III.3
Measures Included in the ReliaStar Index
and Healthy People 2000 Indicators
Healthy
Measure ReliaStar People 2000
------------------------------ ------------ ------------
Mortality (deaths per 100,000 population)
----------------------------------------------------------
Total mortality X X
Cardiovascular deaths X X
Lung cancer deaths X
Breast cancer deaths X
Motor vehicle deaths X X
Suicides X
Homicides X
Infant mortality X X
Work injury-related deaths X X
Premature deaths (premature X
mortality)
Disease incidence
----------------------------------------------------------
AIDS X
Syphilis X
Tuberculosis X
Measles X
Cancer cases X
Other infectious diseases X
Risk factors
----------------------------------------------------------
Poor air quality X
Prenatal care X X
Childhood poverty X
Low birth weight X
Births to adolescents X
Smoking X
Violent crime X
Risk for heart disease X
Other
----------------------------------------------------------
Work disability status X
High school graduation X
Unemployment X
Access to primary care X
State support of public health X
care
----------------------------------------------------------
STEPWISE REGRESSION
------------------------------------------------------- Appendix III:3
To determine how well premature mortality could serve as a proxy for
the ReliaStar health status index, we fitted a regression model with
these two variables. We also used stepwise regression to test if any
additional measures would improve our ability to account for
variation in the ReliaStar index. For the stepwise regression, we
used those Healthy People 2000 indicators and selected measures that
were correlated at 0.60 or more with the ReliaStar index.
The results of the regression analysis are summarized in table III.4.
Model 1 demonstrates that premature mortality alone can account for
three-quarters of the cross-state variation in health status as
measured by the ReliaStar index. Including motor vehicle deaths
increases the proportion of explained variation to 82 percent, and
adding poverty to the equation increases this proportion to 85
percent.
Table III.4
Predictors of the ReliaStar Health
Status Index
Model Model Model
1 2 3
---------------------------------------------- ------ ------ ------
Adjusted R\2 0.75 0.82 0.85
Intercept 0.44 0.44 0.46
(t-value) (9.66) (11.37 (12.62
) )
Premature mortality 0.58 0.46 0.37
(t-value) (12.12 (9.62) (6.91)
)
Motor vehicle deaths 0.11 0.08
(t-value) (4.57) (3.33)
Poverty 0.10
(t-value) (2.98)
----------------------------------------------------------------------
*** End of document. ***