Two of the major reasons put forth to justify the need for health
care reform are increasingly higher costs of medical care, and the lack of access
to health insurance for a significant portion of the population. These two issues
are not independent. High medical costs make insurance more expensive, and, as a
result, universal coverage is virtually unattainable under the current system.
A number of analysts argue that recent low health care cost trends
result from the market's reaction to pending health care reform. In addition, the
Administration's Health Security Act imposes future limits on annual cost increases
that are significantly below historical levels. As a result of these issues, we
feel it is essential to analyze the driving economic forces relating to the historical
health care cost increases. An understanding of such relationships will provide
a background for more informed decisions about health care reform.
Our analysis utilizes over 30 years of annual National Health Care
Expenditure (NHE) data produced by the Health Care Financing Administration (CMS).
We also use a number of economic variables in modeling the historical relationships
to determine the factors that drive or correspond to the increases in NHE.
The variables studied in relation to the NHE increases produce the
following results:
1. Growth in real personal income (or GDP) three to five years prior.
This means that as the economy grows, the health care sector grows three to five
years later. This relationship may also reflect increasing investment in health
care research as the national wealth increases.
2. Physicians per capita and the composition thereof. This is the
strongest of the factors with NHE rising at a multiple of the increase in physicians
per capita. More limited data also indicates that the increasing proportion of specialists
also contributes to increasing NHE growth. These factors are clearly related to
demand/supply relationships. Our model, however, cannot indicate whether physician
supply creates its own demand, or is a reaction to meet the demand that already
exists in the marketplace. These variables may also be proxies for technological
change.
3. Out-of-pocket payments by individuals for medical services. As
the individual's share of direct medical costs (excluding insurance premiums) decreases,
the growth of medical costs accelerates. Out-of-pocket costs as a percentage of
NHE declined from about 50% in 1960 to about 20% today. This shrinkage is encouraged
by favorable tax treatment for both employers and employees and by legislative and
judicial expansion of coverage.
4. Underlying inflation. Medical care costs increase with underlying
inflation, although the relationship is not as strong as might be assumed. Some
of the other variables representing marketplace supply/demand relationships and
technology change may be reducing the impact of this relationship.
5. Managed Care. Under a number of model configurations, growth in
HMO penetration, as a proxy for managed care, corresponds to a slowing of health
care cost growth. However, the relatively low current HMO penetration in the U.S.
may mask the potential for major changes in cost relationships once a critical volume
of enrollment is achieved. Furthermore, to the extent managed care reduces the physician
per capita growth due to more efficient utilization of resources, then the managed
care impact may be implicit from that variable. Our final model does not include
a factor for the managed care variable.
6. Demographic changes. As the population ages, demand for health
care increases. This results in a small average increase in annual health care costs.
Since this is reflected in increased demand which is satisfied by increased physician
supply, the actual demographic variable is not significant in our final model.
These relationships raise the issue of what is politically possible,
or desirable, in restraining health care costs. Obviously, personal income and the
aging of the population cannot be slowed solely for the purposes of controlling
health care growth. Furthermore, the causes of overall inflation reach well beyond
the health care sector. This leaves the controllable factors as physician supply,
level of benefits and managed care.
Reducing the growth of physician supply could have undesirable consequences
-- such as: rationing, additional shortages in under-served areas, and reduced technological
growth. Future reductions in physician supply is a possible consequence of premium
caps and budget limits.
By reducing the level of benefits covered, and requiring more individual
responsibility for health care, the growth in costs can be slowed. However, expanding
the covered population and increasing or establishing a generous level of minimum
benefits would increase health care costs unless current levels of personal responsibility
are maintained. Alternatively, if health benefits are made taxable, this would inevitably
lead to a natural preference for reduced benefits and would slow costs.
Managed care has grown in recent years due to competition in the market
place. HMO penetration has only reached about 18% of the population at a national
level as of 1993. On the other hand, some regions have reached HMO penetration of
over 30%. Our own proprietary health care cost trend database, the Health Insurance
Trend Model® (HITM), shows that some regions have been experiencing significantly
lower health care cost growth than the national average. Further study of regional
data is necessary to determine whether some critical mass of managed care penetration
may result in a paradigm shift, resulting in lower future health care cost growth.
Moreover, to the extent managed care growth results in slower physician growth due
to more efficient resource allocation, this should also result in lower health care
spending.
The consumption of health care is a complex process, driven by such
diverse factors as: the health status and demographics of the population, inflation,
the prevalence and level of medical insurance, the financial incentives of providers,
and more recently, intervention by third party payers. While some of these factors
impact the microeconomic decisions affecting the individual consumption of health
care, other factors affect the macroeconomic aspects of health care. Such factors
include: increases in wealth, price inflation, physician supply, benefit richness
and the specialization of physicians. In fact, increases in wealth (income) appear
to be a leading indicator of health care consumption. Econometric models that forecast
health care trends further into the future and with greater accuracy can be designed
once this lagged relationship is properly understood. The existence of this lagged
"wealth effect" is noted in other research, notably a paper Macro Forecasting
of National Health Expenditures by Thomas E. Getzen, in Advances In Health Economics
and Health Services Research, Volume 11, 1990.
This report presents the results of our research into modeling health
care costs, extending the work done by Getzen. The models are based on government
statistics of health care consumption. Our proprietary HITM database of health care
costs trends supports the results of our models with government data, and is more
suited to applying these techniques on a timely basis in private health insurance.
Government statistics provide the basis for our analysis in this research
report. This allows us to publish results on information that is not proprietary.
These models provide a framework for an in-depth discussion into the nature and
cause of the observed relationships. Finally, we will discuss the implications for
health care reform and other policy issues.
Many of the factors normally considered to drive individual consumption
are unimportant in pursuing a macroeconomic approach to model building. This is
because the inherent volatility and unpredictability of individual health care consumption
obscures the more stable nature of aggregate demand decisions. Microeconomic determinants
of health care consumption that we did not consider are: health status of the individual,
availability and scope of insurance, access to care, and actions of primary care
physicians. When modeling consumption in aggregate, however, these factors apparently
have little or no impact on changes in the rate of consumption of health care (trend).
In his paper, Getzen calls this the "composition fallacy."
The set of variables we use to model consumption from a macroeconomic
perspective will be proxies for supply and demand effects. A variable to measure
the "wealth effect" is required. Other variables are chosen to test hypotheses
that are presented regarding other determinants of health care spending. Among these
are physician supply and composition, managed care impact and richness of benefits.
And finally, the effect of demographics must be considered. The discussion below
presents the variables that are considered and what hypotheses are tested by their
inclusion in the models.
As the wealth of the population increases through economic growth,
each marginal dollar of income is allocated to some form of savings or consumption.
As wealth increases health care consumption should increase. The variable we chose
to use is personal income less transfer payments and taxes. Personal income is then
stated on a constant dollar basis to eliminate inflation. This variable may also
reflect increases in investment in health care research, in both the public and
private sector, that occur as the nation's wealth increases.
Chart 1 illustrates the nature of the income/consumption relationship.
The chart compares annual increases in per capita National Health Care Expenditures
(NHE) on a constant dollar basis versus weighted time lagged Personal Income (PI)
on a constant dollar basis. PI is a weighted composite of three, four, and five
year lags using the relative weights estimated in our model.
Regular market factors driving prices higher are captured in a variable
measuring economy-wide inflation. One would expect the same market forces affecting
economy-wide price inflation to impact the health care sector. We use the CPI-W
All Items as this variable.
Chart 2 compares the CPI-W All Items growth rate vs. NHE per capita
growth rate.
As physician supply grows, does health care consumption grow or decline,
and at what marginal rate? The supply of physicians should impact price, quantity
and quality of health care consumption. Increasing physician supply should work
to hold down prices but increase quantity and quality by improved access to and
usage of medical technology. Although we believe the net impact is positive, we
want to quantify the relationship between physician supply and medical expenditures.
We do so by including a variable representing physicians per 1,000 population.
Chart 3 compares the growth rate of physicians per capita vs NHE per
capita.
It is argued that the increasing proportion of physicians in specialty
fields has led to, or at least is coincident with, the greater use of technology
and more medically intense therapies, thereby raising costs significantly. We were
hoping to model this hypothesis by including a variable measuring the percent of
all physicians that are specialists. However, sufficient historical specialist information
is not available for a complete analysis.
For a number of years managed care has been considered the greatest
hope for controlling health care costs. Some data from our HITM database indicates
that trends are lower in regions with higher managed care penetration. We were hoping
to test this premise by using the market penetration of HMOs as a proxy variable
in our model. One might anticipate that managed care's effectiveness in holding
down costs would depend on the level of market penetration. Evidence of managed
care effectiveness could either be viewed as a distortion of normal economic decision-making
(i.e. rationing) or as a spur on reform that encourages the growth of competition
as a cost saving mechanism.
Chart 4 compares the growth in HMO penetration vs. the growth rate
of NHE per capita.
As the population ages, more health care is consumed due to natural
physical deterioration of the aging process. We tested a measure of this demographic
impact which reflects the distribution of the relative use of health care by age
and sex of the population.
Theoretically, as individuals are required to pay a smaller share
of their direct health care bill (excluding insurance premiums) their sensitivity
to price is reduced. As a greater percentage of dollars are covered by third-party
payers (e.g., insurance, Medicare, Medicaid...) health care consumption should rise
at the margin. This has significant health policy implications. The variable is
the out-of-pocket payments as a percent of NHE.
Chart 5 illustrates the rate of change in out-of-pocket payments as
a percent of NHE vs the rate of change in NHE per capita.
We have built econometric models to test the viability of the hypothesized
relationships of the variables described above. We chose the Box-Jenkins Transfer
function methodology as a framework. This method can test for correlation in the
error terms (which should be random). If necessary an Autoregressive Integrated
Moving Average (ARIMA) structure can be used to adjust for correlation of the errors.
In addition, this framework enables us to identify the need for the significance
of different year's lags in the independent variables. Finally, we also applied
intervention detection techniques to search for and, if necessary, to adjust for
outliers or level shifts in the data. Interventions, as well as error term problems,
are usually evidence of model mis-specification or omitted variables. It is essential
to correct for these problems in order to interpret the relationships of the independent
variables.
The modeling process is intended to identify the combinations of variables
that best achieve the following goals:
The NHE (dependent) variable comes from CMS's medical expenditure
database. We have used the national health care expenditures (NHE) data set with
annual observations from 1960-1991. The models were built using logarithmic transformations
on all variables.
The results presented below represent what we feel is the most appropriate
model -- given the selection criteria. The inclusion or exclusion of variables,
as in all statistical modeling, is based on arbitrary significance criteria. Therefore,
it is possible other model forms could have been considered appropriate.
We are able to model the annual government data on health care consumption
with a high degree of accuracy. The personal income, CPI, physician supply, and
out-of-pocket percentage show statistically significant relationships to health
care consumption. The following table summarizes the results from our selected model:
| Variable |
Coefficient |
T-Statistic |
| Constant |
.033 |
2.75 |
| Autoregressive Order 1 Term |
.64 |
4.97 |
| CPI-W All Items |
.48 |
8.58 |
| Out of pocket payment percentage |
-.19 |
-2.81 |
| Physicians per capita |
2.33 |
13.08 |
| Real personal income - lagged three years |
.17 |
2.07 |
| Real personal income - lagged four years |
.39 |
5.30 |
| Real personal income - lagged five years |
.33 |
4.40 |
Number of Residuals = 30
Degrees of Freedom = 22
R2 = .999921
Sum of squares of residuals = .0018796
In order to test the stability and forecasting ability of the model,
we performed an ex-post forecast analysis. This entailed withholding five years
of data, re-estimating the model, forecasting the withheld observations using actual
data for the explanatory variables and measuring the resultant forecast accuracy.
This process tests whether the relationships between the variables are stable or
if they are dependent on the chosen observation period. The model remained consistent
and forecast the withheld values remarkably well. Our forecast of NHE five years
out erred by only .9%. The largest forecast error was 1.6%.
In general, the coefficients present in the above table can be interpreted
in a uniform manner. Since the model is in a log-linear form, the coefficients represent
the sensitivity (elasticity) of health care consumption to that factor. In other
words, if CPI-W has a coefficient of .48, then for every 1% increase in CPI, NHE
increases by .48%.
The important model statistics above are the t-statistics of the estimated
variable coefficients and the sum of squares of the model error terms. The t-statistics
are a test of statistical significance. The greater the absolute value of the t-statistic,
the more likely that the variable actually correlates to health care consumption.
A t-statistic greater than 2.0 (absolute value) generally indicates a high probability
that the variable is significant. The sum of squares of the model errors is used
in calculating the R2 which measures the goodness of fit of the model.
The PI per capita variable is a significant explanatory variable when
lagged three, four and five years. This implies that economic growth results in
increased health care consumption three to five years later depending on current
market conditions. In order to have a lag of this length, supply and demand relationships
must clear very slowly. Among the possible factors contributing to the lags are:
The length of the lag relationship is important because it allows
for increased accuracy in forecasting up to five years into the future. However,
since other variables included in our models exhibit contemporaneous relationships
to health care consumption, future projections will still depend on forecasts or
scenarios of those variables, which are sometimes more stable (growth in physicians
due to known medical school enrollment) or readily available (inflation forecasts).
The PI variable has a cumulative impact of .88. Thus a 1% growth in
real PI will result in a .88% increase in health care consumption. This is referred
to as the income elasticity of health care consumption. This is significantly less
than the impact that Getzen reported in his models. Getzen reported an impact of
1.45, which is likely due to the omission of other important explanatory variables
from his model. Even though the proxy for this variable is real personal cash income
after taxes, this does not recognize the impact of fringe benefits. In fact, a large
portion of the actual health care spending growth occurs in employer provided health
insurance. Thus, much of the health care cost growth corresponding to increasing
personal income does not necessarily directly reduce that income.
An important implication from these results is that the decline in
trends (as generally reported in the press) in recent years, and particularly the
past two, is partly driven by the weakness in the economy around 1990 and low concurrent
inflation. Some health care analysts speculate that recent low rates of increase
in health care consumption are the result of behavioral changes of health care providers
in the face of health care reform. Based on the models presented here this appears
unlikely. Rather the calls for health care reform are more a manifestation of the
growing costs of health care to which the normal market forces have already begun
to respond.
General price inflation appears to be a significant explanatory variable
for health care consumption. It does, however, have a smaller impact than was expected.
The estimated coefficient was .48, whereas the a priori estimate would be 1.00.
The deviation from the expected impact has numerous potential explanations. In a
multivariate model two or more variables may explain similar factors. This is referred
to as multi-colinearity. In essence, the CPI is a proxy for the market forces causing
health care prices to rise. The overall CPI as a proxy is generally a poor one due
to the special characteristics of the health care market. Much of this inflation
component is explained by other variables.
Other potential explanations for the smaller than expected impact
of the CPI include the following: health care may react differently to general price
pressures, such as labor market conditions and productivity changes; it is also
possible that errors in measuring the variables used in the modeling process have
distorted the results; and finally, problems in model specification could result
in biased estimates of the coefficients.
The impact of physician growth on health care consumption is a potentially
controversial subject. If the health care market worked like a classic free market,
the increase in physician supply would work to hold down the price of health services,
all other factors remaining equal, by increasing competition. In contrast to the
impact on price, increased physician supply may help to increase the quantity and
quality of health care delivered. Thus the overall impact of physician supply on
NHE is difficult to predict. Evidence from our model indicates that there is a net
positive relationship between physician supply and NHE. In fact, as physician supply
becomes more plentiful, health care costs appear to rise by an even greater amount.
This could be explained by the distortions of the normal economic decision process
due to the insurance/third party payment mechanisms found in the health care market
and the virtually unlimited demand for medical care. In addition, physicians play
a key role in the decision to consume health care; they direct the setting and services
provided. One further explanation is that physician supply is also a proxy for the
induced demand for medical services caused by advances in medical technology.
Our estimates of the impact of physician supply range from over 100%
to over 300% impact depending on the model formulation. In our final model the estimated
coefficient was 2.3. Thus, a 1% increase in physician supply corresponds to a 2.3%
increase in health care costs. This occurs despite the fact that payments for physician
services account for only 40% of privately insured health costs and only about 20%
of total national health expenditures.
On the surface this relationship implies that physician supply is
an important policy control variable. Rationing the physician supply might lend
some degree of control over health care costs. However, it is likely that the physician
supply variable is a proxy for the impact of other factors, such as technology growth
and managed care. Thus, controlling physician supply might not have the full impact
implied by these models.
This variable is closely related to physician supply. It attempts
to isolate the impact of the increasingly higher specialist concentrations over
time vs. general increases in physician supply. Specialists are more likely to utilize
high-tech and higher cost treatments than primary care physicians. It is also possible
that the growth in specialists has been caused by advances in medical technology
and this variable would serve as a proxy.
Unfortunately we were unable to obtain data covering our full analysis
period. Furthermore, most of the growth in physicians per capita was occurring in
specialists. Therefore, most of the impact of physician growth is also attributable
to the specialists. Like the physician supply variable, this might imply that discouraging
specialization would help hold down costs. In fact, government reimbursement policies
and some of the health reform proposals are already aimed at this goal.
It is possible that physician specialists have become more prevalent
in conjunction with advances in medical technology. If this were the case, then
limiting specialization might only serve to force GPs to acquire some of the professional
skills of specialists. This could also result in increased demand for the fewer
specialists (and full-time-equivalent GPs) and could cause prices to rise. This
ultimately could be less efficient than the current system of specialization and
serve to raise costs. Alternatively, such action could discourage the development
and dispersion of technology. This could restrain costs, but at the price of future
improvements in medical treatments.
We use the percent market penetration of HMOs as a proxy for the impact
of managed care on health care costs. There should be a significant inverse relationship
if managed care helps hold down costs. Managed care theoretically helps hold down
costs in a number of ways. Care is directed to the most cost effective setting.
Negotiated reimbursement arrangements with providers result in increasing discounts
from charges, since market power should increase the effectiveness of the reimbursement
negotiations. Because of these factors, the impacts may be non-linear in nature.
That is, managed care's effectiveness may increase more than linearly with increasing
market penetration. At lower market penetrations providers are able to effectively
shift costs to other payers and little net impact results. Finally, utilization
controls and provider financial incentives can result in the elimination of unnecessary
care, resulting in a one-time reduction in costs.
We have found that, under a number of configurations, HMO enrollment
has a marginally significant impact on health care costs. According to our final
model, the HMO variable was not significant. Our belief is that as HMO penetration
increases to higher levels this variable may become significant. Currently, about
18% of the population is enrolled in an HMO. This figure varies greatly by area.
In our proprietary regional HITM models, the trends in the Pacific region have been
generally lower than the national average for some time. At the same time, the HMO
penetration in the Pacific region is and has been much higher than the national
average. HMOs, by their nature, attempt to provide more efficient use of health
care resources and, as a result, may slow physician supply growth. Thus, some of
the impact of the HMO variable may be showing up in the physician supply variable.
In fact, physician per capita growth in the Pacific region, with the highest HMO
penetration, has been slower than the US as a whole. Chart 6 compares the HITM Pacific
regional trends versus the US as a whole over the past twenty years. Beginning in
the early 1980's and continuing to this day, the Pacific region has experienced
consistently lower trends than the remainder of the US. As further evidence, in
the study "HMO Market Penetration and Hospital Cost Inflation," James
C. Robinson, Ph.D. found that California markets with relatively high HMO penetration
in hospital admissions experienced significantly lower hospital cost inflation than
markets with lower HMO penetration.
As the U.S. population ages, demand for health care increases and
health care consumption per capita should increase. For example, the average Medicare
eligible consumes about three times the average insured health care benefits than
the younger insureds not yet eligible for Medicare. Furthermore, within the non-Medicare
population, the average adult consumes about 30% more and the average child about
50% less than average per capita costs. This results from the natural aging process.
We constructed a variable to measure this demographic shifting. The variable attempts
to isolate the change in cost related to the growing percentage of elderly in our
population. We have used estimates of insured relative claim costs by age and population
distribution by age over time to construct this demographic variable.
Due to the construction of this demographic variable, we would expect
a coefficient of 1. However, we did not observe a statistically significant relationship
for this variable. Since increasing demand is met by increasing physician supply,
that variable is likely to have absorbed the impact of the demographics.
The increasing amount of health care payments that come from third
parties affects individual economic-decision making. An increasing percentage of
payments by third parties (insurers, governments) would increase marginal health
care consumption relative to other goods. This would result in the demand for health
care rising faster than it would have, given all other factors. This growth in third-party
coverage is encouraged not only by favorable tax treatment, but also through the
expansion of coverage by legislative actions and court decisions extending coverage
well beyond original insurance limits. These expansions result in a distortion of
the rational economic decision-making process.
We include a variable which measures the percent of payments for health
care that come from direct consumer payments (excluding premiums). This variable
shows that increasing consumption results. The coefficient is estimated to be -.19.
That is, as the consumer's direct share of payments for health care declined, overall
consumption rose at nearly 20% of the corresponding rate of decrease.
This variable has significant health care reform implications. The
desire to include more persons in the insured system or increased benefits (in total
or in certain classes of benefits) would lead to further reductions in personal
responsibility for the health care consumed and increase overall demand. However,
a reform plan that carefully balances coverage with personal responsibility could
actually help contain costs.
Medicare has twice, since 1983, made major changes in its provider
reimbursement mechanisms. The first was the Prospective Payment System (PPS) which,
beginning in late 1983, reimbursed inpatient hospital admissions on a Diagnosis
Related Group (DRG) basis. The second was the move to reimburse physician services
on the Resource Based Relative Value Scale (RBRVS) beginning in 1992. Based on our
models, we found no evidence that the DRG change produced any significant net savings
in the country's aggregate health bill. The Medicare program saved significant amounts
under the PPS program but it seems providers have been able to shift costs or increase
revenues from other payers. Data is not yet available to test the RBRVS impact.
We use this model to forecast National Health Expenditures for three
years, from 1992-1994. Since only the Personal Income variable has a leading indicator
relationship we must specify a scenario for the other model variables for calendar
year 1994. The following table summarizes the growth rates used.
| Year |
CPI-W |
Physicians/Capita |
Out-of-Pocket Payment Percent |
| 1992 |
2.9% |
1.9% |
-4.8% |
| 1993 |
2.8% |
1.9% |
-4.8% |
| 1994 |
3.5% |
1.9% |
-5.1% |
Our models forecast that NHE will grow by 27.65% over the three years
for an annual rate of increase of 8.48%. NHE will grow from $751.7 billion in 1991
to $959.6 billion in 1994. NHE as a percent of GDP was 13.14% in 1991. We forecast
that this will rise to over 14% in 1994. The following table summarizes these results.
Please note that these results are not stated on a per capita basis.
| Year |
NHE (billions) |
Growth Rate |
NHE as a Percent of GDP |
| 1991 |
$751.771 |
11.4% |
13.14% |
| 1992 |
$818.142 |
8.8% |
13.55% |
| 1993 |
$889.529 |
8.7% |
13.95% |
| 1994 |
$959.644 |
7.9% |
14.04% |
Chart 7 illustrates our forecast of the rate of increase in NHE (per
capita) along with the fitted values from the model.
The results from this model have significant policy implications.
In the health reform debate it is often taken as a given that spending more on health
care is bad. From an economic perspective, more is not bad unless it is caused by
distortions in the economic decision-making process. These models have shown that,
to a large extent, we consume more health care as a nation because we earn more.
In conjunction with other factors, this has led to a situation where our spending
on health care is outstripping the growth in our income. This is commonly measured
by our health care spending as a percentage of GDP. The three main controllable
causes appear to be related to physician supply, growth of managed care, and the
reduction of personal financial responsibility for health care.
The relationship of physician supply to health care consumption appears
obvious but may cause controversy. It is not clear whether this correlation with
consumption is due to a casual relationship. Do we spend more on health because
there are more physicians or are there more physicians because we demand more health
care? Do specialists cause higher costs by using higher technology or do we need
more specialists because of higher technology? The answers to these questions are
difficult and conjectural. However, these questions do raise concerns about the
nature and structure of proposed health care reforms.
Evidence from our analysis indicates that managed care in the marketplace
is currently having only a marginal impact on health costs. This may be due to the
marketplace's belated response to the distortions caused by the employer-provided
insurance mechanism. And HMOs often provide lower out-of-pocket costs, thus otherwise
increasing demand. Also, managed care may ultimately control costs by reducing physician
supply through more efficient resource deployment, which is already a direct variable.
Finally, we demonstrated that the reduction in an individual's relative
burden of health care costs resulted in higher expenditures. Will health care reform
lower or raise this aggregate responsibility? The inclusion of more people in the
insurance system will tend to continue the trend to lower out-of-pocket responsibility.
Some reform proposals call for increased personal responsibility; others do not.
Alternate forms of cost-sharing, premiums vs. copayments, may have differing impacts
on consumption. Therefore, the mix of these factors will determine the effects under
reform.
Milliman & Robertson produces a proprietary database of health
care trends called the Health Insurance Trend Model (HITM). The HITM is analogous
in many ways to the government data analyzed in this report. The HITM measures the
net payments to providers of health care, excluding MEDICARE. Thus, it is aimed
at the private sector. We have developed similar models to those presented in this
report for our database and have arrived at similar results and conclusions. These
models have led to an increase in our ability to model and forecast health care
costs for our clients (HMOs, Blue Cross/Blue Shield Plans, commercial insurers).
The HITM has these distinct advantages over available government (NHE)
data:
Milliman USA uses the HITM, in conjunction with other variables, as a basis
for analyzing our clients historical claim cost experience and developing future
projections.
Recent experience from the HITM has shown sharp declines in health
care cost increases. This is consistent and wholly explained by the economic phenomena
discussed here. The continued weakness in inflationary pressures and the lagged
impact of the recession of 1990-1991 and slow recovery have caused trends to decline.
Thus, the more robust economic recovery experienced beginning in late 1993 will
lead to higher trends in the future. If general inflation were to increase concurrently,
significant increases in trends could be experienced.
Chart 8 compares the NHE vs. the HITM.
This report was intended to dispel some myths regarding health care
costs and point toward new macro economic approaches to modeling trends. In building
these models we can test hypotheses regarding the impact of variables with important
policy implications. The impact of managed care, physician supply and personal responsibility
are all themes that have echoed around the health care debate. Their contribution
to health care consumption should be carefully considered in evaluating reform proposals.
Box G.E.P., Jenkins G.M.. "Time Series Analysis: Forecasting
and Control." Holden-Day, 1976.
Burner, S.T., D.R. Waldo and D.R. McKusick. "National Health
Expenditures Projections Through 2030." Health Care Financing Review 14:1,
Fall 1992.
U.S. Department of Labor. Bureau of Labor Statistics. (Various Issues
and Years). CPI Detailed Report.
Getzen, T.E. "Macro Forecasting of National Health Expenditures."
Advances in Health Economics and Health Services Research 11:27-48.
Historical Statistics of the United States, Colonial Times to 1970.
Consumer Prices, Population, Physician Series.
McCleary R., Hay R. "Applied Time Series Analysis for the Social
Sciences." Sage Publications 1980.
National Health Expenditures, 1960-1991, CMS Tapes.
Robinson, J. "HMO Market Penetration and Hospital Cost Inflation
in California." JAMA November 20, 1991 266:19.
"Statistical Abstract of the United States." 1993 and earlier editions.