HMOs' consumer-friendliness and preventive health care utilization: exploratory findings from the 2002 Medical Expenditure Panel Survey

HMOS’ CONSUMER-FRIENDLINESS AND

PREVENTIVE HEALTH CARE UTILIZATION:

EXPLORATORY FINDINGS

FROM THE 2002 MEDICAL EXPENDITURE

PANEL SURVEY

QIAN XIAO

West Texas A&M University

GRANT T. SAVAGE

University of Missouri

ABSTRACT

Research should move beyond the simple dichotomy between

HMO and non-HMO care provision, and embrace the multidimensional

aspects of HMOs. Doing so, we argue, helps address the issue of HMO

performance. We used a consumer-centered approach to distinguish

multiform HMOs and asked the questions, “Do HMOs differ in their

consumer-friendly characteristics?” and if so, “Are these characteristics

associated with different preventive health care utilization outcomes?”

In this exploratory study, the consumer-friendly characteristics of both

Medicaid HMOs and private HMOs were examined in relationship to

consumers’ utilization of preventive care services. HMOs did differ in

their consumer-friendly characteristics, and some of these

characteristics were significantly associated with the utilization of

preventive care services.

The health care system in the United States has

undergone rapid change over the past decade in response to

years of escalating costs. The most visible evidence of this

response has been the growth of health maintenance

organizations (HMOs). HMOs differ from indemnity health

insurance in the way they manage both the cost of health

care and the range of health care services they offer. These

differences often include selecting a network of providers;

relying on primary care physicians to be care managers

260 JHHSA FALL 2008

who must approve referrals to specialists; using capitation

or other financial incentives to encourage cost-effective

care; and employing a variety of utilization management

tools, such as profiling service use, prior authorization, and

case management of high-cost cases. At the same time,

HMOs typically lower consumer cost sharing by requiring

only modest co-payments rather than the high deductibles

and co-insurance more common in other types of insurance.

Because of differences in care management and consumer

cost sharing, as well as other differences, it is reasonable to

expect that access to care, and use of heath care services

should improve with HMOs.

However, the research evidence thus far to justify

good HMO performance is mixed and inconclusive. Results

differ as to whether HMOs increase or decrease health care

utilization, and whether this relationship has evolved over

time (Rizzo, 2005). Reviews of the literature by Robinson

and Steiner (1998) and Miller and Luft (1997, 2002) found

no conclusive evidence in one direction or the other on the

relationship between managed care and quality of care.

Studies that focused specifically on the relationship

between HMOs and health care utilization found little

difference in health outcomes (Rizzo, 2005). For example,

using two data sets from Massachusetts, Cutler, McClellan

and Newhouse (2000) found very similar results for cardiac

medication treatment and outcomes for heart disease

patients in HMOs and traditional health insurance plans. In

a study comparing cancer outcomes, Merrill et al. (1999)

reported that colorectal cancer-specific mortality did not

differ significantly between the patients in an HMO setting

and those in a fee-for-service setting.

Reasons for the inconclusive evidence may in part

reflect the difficulty of measuring quality of care, a

complex, multidimensional concept. A further complication

may be that HMOs are heterogeneous. Robinson (2000) has

noted that a limitation of the literature on managed care is

JHHSA FALL 2008 261

that very few studies compare performance in terms of

different models of managed care. Previous studies that

have evaluated HMO performance generally compare

HMOs with all other forms of insurance plans,

dichotomously coding HMOs versus other insurances.

However, such treatment may be misleading if HMOs are

not a unitary construct. If different forms of HMOs exist

and if they are associated with different outcomes,

aggregating finding across the various forms will yield

misleading results. We further argue that different forms of

HMOs can be distinguished or represented by the structural

characteristics of HMOs. That is, distinct characteristics of

different HMOs may be the underlying explanatory factor

for the diverse performance of HMOs. For instance, Miller

and Luft (1994) cautioned researchers about drawing

conclusions and generalizing from the literature evaluating

HMOs. They pointed out that many factors can affect

managed care organizations’ health care use, expenditure,

and quality performance. They divided the most likely

factors into three groups: characteristics of the managed

care organizations, characteristics of the managed care

benefit plans, and characteristics of the markets in which

managed care organizations operate. Based on Miller and

Luft’s distinctions, we examine whether the structural

characteristics of different HMOs may be associated with

diverse HMO performance in terms of health care

utilization.

HMOs can be structurally characterized by either

‘provider-driven’ or ‘consumer-centered’ approaches.

‘Provider-driven’ approaches distinguish HMOs based on

the relationship between HMOs and physicians such as the

payment method to providers, the type and amount of risk

and reward sharing, provider selection, etc. The ‘provider-

driven’ approach results in four essential HMO types: staff

model HMO, group model HMO, IPA (Independent

Practice Association) HMO, and network HMO. However,

262 JHHSA FALL 2008

in this analysis, we chose the ‘consumer-centered’

approach and used consumers’ experience with HMOs to

distinguish among different HMOs. Our choice of a

consumer-focus can be warranted for two reasons. First,

consumers often have major roles in choosing health care

and health plan coverage, which ultimately may have

implications for the future viability of different forms of

health care delivery and financing. These public

perceptions and attitudes also affect the formulation of

public policies regarding the regulation and provision of

health insurance. Second, the consumer-centered approach,

as a market-based solution, highlights consumers’ positive

roles in the health care by taking into account consumers’

values, expectations, and medical needs; and therefore

corresponds to the emerging advocacy of consumer-driven

health care (Herzlinger, 2004).

Therefore, this paper sets out to resolve the

inconclusive arguments about the performance of HMOs

by asking the questions, “Do HMOs differ in their

structural (consumer-friendly) characteristics?” and if so,

“Are these characteristics associated with different

preventive health care utilization outcomes?” Our logic is

that if multiform HMOs can be differentiated by their

structural characteristics, and these characteristics do relate

to different health care outcomes, we expect to resolve the

debates about the HMO performance by turning to the in-

depth study of HMO structural characteristics which may

serve as the potential explanatory variables for the diverse

performance of HMOs. In an attempt to answer these

questions, we used consumer-based measures to distinguish

different levels of HMOs’ consumer-friendliness; both

Medicaid HMOs, and private HMOs, consumer-friendly

characteristics were examined in relationship to consumers’

utilization of preventive care services.

The study’s focus on preventive care utilization is

warranted because preventive medicine is the cornerstone

JHHSA FALL 2008 263

of good medical care. It has important effects on disease

progression, morbidity, and mortality. Despite the

importance of good preventive care, research suggests that

such care has been underutilized in the United States

(Schauffler & Rodriguez, 1993). In a report identifying key

problems with quality of care in the United States, the

President’s Advisory Commission on Consumer Protection

and Quality in the Health Care Sector (“The State of Health

Care Quality”, 1998) specifically cited problems with

underutilization of preventive care, including flu shots,

mammography and screenings for colorectal cancer. While

HMOs should create incentives that influence preventive

care access and utilization, studies of preventive care yield

conflicting evidence. Therefore, this study seeks to bridge

these gaps in the literature.

Miller and Luft (1994) addressed several other issues

in their literature review. They pointed out that the

performance of managed care organizations differ

considerably depending on which local market areas are

used for analysis. Consequently, evaluation findings based

on data from a small number of plans, providers, or local

market areas cannot necessarily be generalized to the

nation. They also cautioned that much of the literature

relied on relatively old data—data that become less relevant

in today’s changing health care marketplace. They

recommended that future research focus on multiple

dimensions of performance and investigate effects on

important subgroups. This paper represents a significant

step toward addressing Miller and Luft’s recommendations

and avoiding the limitations of previous studies. Unlike

many other studies that used convenience samples and

narrowly focused on patients in a particular hospital (e.g.,

Pearson et al., 1994) or members of a small number of

health plans (e.g., Manning et al., 1984), the present study

has a national focus. In addition, these data were collected

in 2002 and, thus, are among the most recent available. The

264 JHHSA FALL 2008

remainder of the paper is designed to examine the impact of

an HMO’s consumer-friendliness on preventive care

utilization.

METHODS

Data and Sampling Procedures

Data for the analysis were obtained from the 2002

Medical Expenditure Panel Survey (MEPS). This database,

co-sponsored by the Agency for Healthcare Research and

Quality (AHRQ) and the National Center for Health

Statistics (NCHS), provides nationally representative

estimates of medical treatments and health care

expenditures, health status, health insurance coverage, and

sociodemographic and economic characteristics for the

civilian, non-institutionalized population in the United

States. The MEPS sample was derived from a subset of

respondents to the National Health Interview Survey

(NHIS) in 2000 and 2001, allowing linkage to information

collected from those surveys as well.

The NHIS utilized a multi-stage sample design. The

first stage of sample selection was an area sample of

primary sampling units (PSUs), where PSUs generally

consisted of one or more counties. Within PSUs, density

strata were formed, generally reflecting the density of

minority populations for single or groups of blocks or block

equivalents that were assigned to the strata. Within each

such density stratum “supersegments” were formed,

consisting of clusters of housing units. Samples of

supersegments were selected for use over a 10-year data

collection period for the NHIS. Households within

supersegments were selected for each calendar year the

NHIS was carried out. A household may contain one or

more family units, each consisting of one or more

individuals. Analysis can be undertaken using either the

individual or the family as the unit of analysis.

JHHSA FALL 2008 265

Generally, about three-eighths of the NHIS

responding households were made available for use in

MEPS (“MEPS HC-070”, 2004). A subsample of these

households was then drawn for MEPS interviewing. The

MEPS further validated information on medical care

utilization by contacting health care providers and

pharmacies identified by survey respondents. In order to

produce annual health care estimates for calendar year 2002

based on the full MEPS sample, full calendar year 2002

data collected in rounds 3 through 5 for the MEPS panel 6

sample were pooled with data from the first three rounds of

data collection for the MEPS panel 7 sample. Overall there

are 37,418 person-level survey respondents, and the

combined response rate is 64.7% (“MEPS HC-070”, 2004).

Several steps were taken to make the data fit the

purpose of analysis. First, we included in the analysis only

adults (older than 18 years of age) who were insured by

either Medicaid HMOs or private HMOs. Second, because

consumer-friendly characteristics variables for Medicaid

HMOs and private HMOs were collected respectively in

two different data files (2002 full year consolidated data

file and 2002 person round plan file) we combined relevant

variables into one data file for the convenience of analysis

by using the common person identifier variable. Third, as

noted before, the MEPS has a complex survey design,

involving sample stratification into primary sampling units,

clustering, and oversampling of certain subgroups. As a

result, we performed all statistical analyses using weights

provided in MEPS to correct mean values, coefficient

estimates, and standard errors to be reflective of national

averages. Fourth, outliers were removed when casewise

diagnostics showed that some cases were outside three

standard deviations; and listwise deletion on key constructs

in this study was adopted when respondents refused to

provide the answer to focus constructs, the questions were

inapplicable to the respondents, or the answer was not

266 JHHSA FALL 2008

ascertained. Thus, the final sample included 521 cases, with

126 insured by Medicaid HMOs and 395 insured by private

HMOs. To assess the presence of non-response bias in our

data, we compared usable responses after the data-cleaning

process against non-usable responses on five

sociodemographic characteristics: age, gender, education,

ethnicity and poverty level. The non-response bias test

showed that the original sample was well represented by

our data in terms of percent distribution of selected

characteristics; however, two demographic variables were

only represented by less than 5 cases for Medicaid HMO

enrollees. Under race, only 3 American Indians, 3 Asians,

and 1 Pacific Islander were sampled; while only 1 person

with a master’s degree was sampled for education.

Therefore, the results of our analysis should be interpreted

with some care on these variables.

Measures

HMO consumer-friendly characteristics. Consumers’

experience with HMOs was used as measures to distinguish

different levels of consumer-friendliness of multiform

HMOs (in this case Medicaid HMOs and private HMOs

respectively). Respondents who were insured either by

Medicaid HMOs or by private HMOs were interviewed as

to their experience with the insurance plans. Question

wording was based on an AHRQ-sponsored family of

survey instruments designed to measure quality from the

consumer’s perspective. Ten question items addressed the

following topics which represent the major consumer-

friendliness characteristics: difficulty getting a personal

doctor or nurse, delays waiting for plan approval for care,

problems finding or understanding plan information,

problems getting help from customer service, problems

with paperwork, and rating of experience with plan.

Subjects provided answers as to whether it was a problem

JHHSA FALL 2008 267

(yes or no), or ranked the problem as a big one, a small one,

or not a problem.

Examination of the correlation matrix of 10 variables

revealed that variables LKINFOM (lack information on

how plan works) and PRBINFOM (problem finding

information), both concerning the problem of information

communication on how plan works, demonstrate significant

correlation as high as .899, a potential signal of

multicollinearity. The same problem relates to variables

PPRWRKM (fill out paperwork for plan) and PRBPWKM

(problem with plan paperwork), which are used to examine

the problem with plan paperwork. The correlation between

these two variables is .910. Although the high correlations

may be the artifact of value range restriction of variables,

we still chose to retain one variable of the highly correlated

two while deleting another in order to avoid the

multicollinearity issue. A high level of multicollinearity can

result in unstable regression coefficients in linear

regression models (Pedhazur, 1982; Barringer & Bluedorn,

1999). Collinearity statistics (VIF) of these four variables

are respectively 7.827, 7.749, 8.432, and 8.204, and

therefore warranting this choice. Taking into account the

sample size requirement, we chose to give up variables

with too many invalid values such as -9 through -1; and

thus we deleted variables PRBINFOM and PRBPWKM

and kept variables LKINFOM and PPRWRKM. For other

variables, although we also detected that some variables are

significantly correlated with each other, the collinearity

statistics (VIF) of these variables are under 1.5, which

demonstrates that the correlations between these variables

are not high enough to cause serious multicollinearity

problem in a regression analysis.

An exploratory factor analysis (EFA) with oblique

rotation was further to assess whether the remaining 8

variables represented different factors. The results showed

that variables with similar value ranges tended to combine

268 JHHSA FALL 2008

into one composite, which did not reflect reasonable

conceptual meaning. Therefore we retained 8

characteristics items as individual discriminants of

multiform HMOs in our model. This treatment follows

MacCallum and Browne’s (1993) mathematically

equivalent model to avoid possible issues related to the

formative indicators. Thus, the eight indicators were treated

as individual explanatory variables that directly influence

preventive care utilization variables rather than being

turned into a composite when examining the relationship

between HMOs’ consumer-friendly characteristics and

enrollees’ preventive care utilization. The same method

was adopted in dealing with the preventive care utilization

measures.

The correlation matrix, as well as the

multicollinearity test, is included in Table 1. Appendix I

presents the description of consumer-friendly

characteristics variables and their coding.

Table 1

Correlation, Collinearity Statistics, Mean and Standard

Deviation of Characteristics Variables

JHHSA FALL 2008 269

Preventive care utilization. Studies of preventive

care yield conflicting evidence. Kenkel (1994) found that

HMO members used less preventive care as measured by

breast examinations and PAP smears. In contrast, Miller

and Luft’s (2002) review of the literature reported that most

studies on preventive care pointed to greater use of

preventive care in HMOs. However, the researchers also

noted that most of these studies considered cancer

screening rather than broad-based comparisons of

preventive medicine. In the empirical analysis to follow, we

seek to gain insight into the effects of HMOs on a variety

of preventive care treatments.

In the MEPS database, survey respondents were

asked questions pertaining to whether they had received

specific types of preventive medicine. Using this

information, we created the following variables indicating

whether or how often subjects had received blood pressure

checks (BPCHEK), cholesterol screenings (CHOLCK), flu

shots (FLUSHT), and physical examinations (CHECK).

For females, we also included variables indicating whether

or how often subjects had received a breast exam

(BRSTEX), a PAP smear (PAPSMR), or mammography

(MAMOGR). Subjects were asked to provide information

of about how long since their last check-up of various

preventive care services: within the past year, within the

past 2 years, within the past 3 years, within the past 5 years,

or more than 5 years. Our general assumption is that more

preventive care is better care (Rizzo, 2005). It is difficult to

argue that receiving more frequent blood pressure

checkups, cholesterol screenings, physical examination, or

breast examinations, for example, is on average bad. More

controversial are mammography screenings, which expose

the patient to small amounts of radiation per exam, and

which have fairly high false-positive rates (Rizzo, 2005).

For item information on the preventive care utilization

variables refer to Appendix I.

270 JHHSA FALL 2008

Control variables. Sociodemographic factors and

health status were included in the empirical model to

reduce confounding effects due to the difference of the

populations enrolled in various types of insurance plans.

Demographic indicators included age, gender, race,

education degree, and poverty status. Kenkel (1994) has

demonstrated the importance of age and education on

preventive medicine, at least with respect to PAP smears

and breast examinations. The impact of age involves a

trade-off. On the one hand, older age raises the probability

of discovering a problem, making preventive screening

more beneficial. At the same time, advanced age lessens

the potential benefits in terms of increased longevity in the

event that screening helps to prevent a medical problem

from occurring or worsening. The age variable was recoded

into four categories: 18-24 years old, 25-44 years old, 45-

64 years old, and 65-85 years old. Better-educated

individuals possess greater knowledge of the benefits

associated with preventive care and hence are more

proactive in obtaining such care. Kenkel (1991) found

evidence that better health knowledge explained part of the

relationship between schooling and healthy behaviors. At

the same time, he noted that most of the educational effects

on healthy behaviors remained even after differences in

health knowledge were controlled. The education variable

was recoded into four categories: high school and less,

bachelor’s degree, master’s degree, and other degree. Race

may relate to preventive medicine because disadvantaged

minorities may have less information about the benefits of

preventive care, or less generous health insurance plans and

/ or fewer financial resources generally. Race-specific

differences in certain types of disease may also prompt

differential use of preventive care. The race variable

included 6 categories: white, black, American Indian,

Asian, Pacific Islander, and multiple races. We also

JHHSA FALL 2008 271

controlled for gender to see whether males and females

behave differently in general preventive care utilization.

A poverty status variable was constructed by dividing

family income by the applicable poverty line based on

family size and composition, with the resulting percentages

grouped into 5 categories: poor (less than 100%), near poor

(100% to less than 125%), low income (125% to less than

200%), middle income (200% to less than 400%), and high

income (greater than or equal to 400%). Poverty status

relates to disposable financial resources; poor people may

tend to underuse preventive care services, resulting in even

poorer health status. Evidence suggested that HMO

members tend to be younger and healthier (Glied, 2000;

Scitovsky, et al., 1978; Jackson-Beek, et al., 1983; Ellis,

1989; Langwell, et al., 1989), perhaps because sicker

individuals are discouraged from joining HMOs, or because

such individuals tend to avoid HMOs (Rizzo, 2005).

Moreover, health status may affect preventive care to the

extent that the provider and / or the patient recognize that

the need for screening increases as the patient’s health

status declines. In order to avoid the likely understatement

of the impact of HMOs on preventive medicine, our

multivariate models controlled for global measures of

health status, which were indicated by the subject’s

responses about their overall health status as excellent, very

good, good, fair, or poor.

ANALYSIS AND RESULTS

In this section, we provide a demographic and health

status profile of the sample enrollees. ANOVA and

ordinary least squares regression methods were then used to

investigate the two research questions; and regression

results assessed the effects of both control variables and

consumer-friendly characteristics on preventive care

utilization.

272 JHHSA FALL 2008

Sample Profile

Table 2 provides a demographic and health status

profile of the analyzed sample enrolled in Medicaid HMOs

and private HMOs. As expected, there were statistically

significant differences between Medicaid HMO and private

HMO enrollees across gender, education, poverty status,

and perceived health status, which support their inclusion

as control variables. Since Medicaid is aimed at poor and

low-income Americans, it is understandable that a

predominant proportion of Medicaid HMO enrollees were

near or under the poverty line, while the private HMO

enrollees were characterized by middle and high income

levels. Correspondingly, Medicaid and private HMO

enrollees tended to demonstrate relevant features in terms

of their education and health status since, as noted above,

education level can influence poverty status; and poverty

status may be the reason for health status. The larger

proportion of female enrollees in Medicaid HMOs versus

male enrollees reflects the categorical policies of Medicaid,

with its emphasis on children and women with infants.

However, the distribution data for Medicaid HMO and

private HMO enrollees suggested that the two samples

were quite similar across age and race variables.

JHHSA FALL 2008 273

Table 2

Demographic and Health Characteristics of Sample Data in

Medicaid HMO and Private HMO (United States, 2002)

Results

To answer our research questions, we used two steps

in our analysis. First, we investigated whether Medicaid

HMOs and private HMOs differ in their 8 consumer-

friendly characteristics by analyzing the variance

274 JHHSA FALL 2008

(ANOVA) between the two types of HMOs. Second, we

treated preventive care variables as continuous, since

frequency of preventive care utilization was measured by

year duration; thus, we used ordinary least squares

regression (OLS) to examine the research question of

whether HMO consumer-friendly characteristics

differentially impact preventive care utilization.

The ANOVA results are summarized in Table 3.

Seven out of the 8 consumer-friendly variables were

significantly different between Medicaid HMOs and private

HMOs at the .05 level. The only non-significant variable,

PPRWRKM, examines whether enrollees need to fill out

paperwork for the health plans; both Medicaid HMOs and

private HMOs were similar in this aspect. We also

conducted a one-way multiple analysis of variance

(MANOVA), controlling for the differences in the sample

enrolled in each type of insurance by entering six control

variables (age, sex, race, education, poverty level, and

health status) in the model and obtained similar results.

Because of space considerations, we only reported the

ANOVA results. In general, we found evidence that

different types of HMOs (in this case Medicaid HMOs and

private HMOs) can be distinguished by their consumer-

friendly characteristics. Further efforts were made to

examine the impacts of these characteristics on preventive

care utilization.

JHHSA FALL 2008 275

Table 3

ANOVA of Consumer-Friendly Characteristics between

Medicaid HMOs and Private HMOs

The seven preventive care variables were regressed

against the eight consumer-friendly variables, controlling

for differences in sociodemographic characteristics and

health status. The OLS regression results, which allow us to

address our second research question, are reported in Table

4.

276 JHHSA FALL 2008

Table 4

OLS Regression Beta Results, with Preventive Care

Utilization as Dependent Variable

As shown in Table 4, older subjects and females were

more likely to receive preventive care. However, race was

not a significant predictor for preventive care utilization

except for breast tests. Better-educated subjects generally

were more likely to receive preventive care, as indicated by

the negative signs of regression coefficients between the

education variable and preventive care variables, a result

consistent with findings reported by Kenkel (1994).

However, education served as the significant predictor only

for flu shots and blood pressure testing in our analysis; and

educated people tended to decrease the use of

mammography screenings. The latter may be due to the

controversy about the potential side-effect of radiation and

fairly high false-positive rates of the mammography test

(Rizzo, 2005). Subjects in worse health were more likely to

receive preventive care as indicated by the consistent

negative signs of regression coefficients between the

perceived health status variable and preventive care

variables; this is probably because they are perceived to be

JHHSA FALL 2008 277

at higher risk for medical complications. Finally, the

negative signs of regression coefficients between the

poverty status variable and preventive care variables

indicated that people with better financial status tended to

pay more attention to preventive care. However, poverty

served as a significant predictor only for cholesterol

checks, physical examination, and breast tests.

Our second question examines whether HMO

consumer-friendly characteristics significantly influence

preventive care utilization. In this section, we reported their

effects by the type of preventive care. Detailed

explanations of the results will be presented in the

discussion section by the type of consumer-friendly

characteristics.

As indicated in Table 4, cholesterol check was

significantly influenced by lacking information on how

plan works (LKINFOM) and the requirement of filling out

paperwork (PPRWRKM). Routine physical examinations

were significantly influenced by problems in getting a

personal doctor or nurse (GTDCPRBM) and the subjects’

evaluation of the experience with plan (RATPLANM).

Obtaining a flu shot was not significantly associated with

any characteristics variables. A possible reason may be that

people are already well informed about the benefits of flu

vaccinations since influenza is one of 10 leading causes of

death (“National vital statistics report”, 2005). Pap smear

test were significantly associated with the occurrence of

calling customer service to complain or report problems

(CUSTSVCM) and the requirement of filling out

paperwork (PPRWRKM). Breast tests were significantly

associated with concerns in getting a personal doctor or

nurse (GTDCPRBM) and the subjects’ evaluation of the

experience with plan (RATPLANM). Mammograms were

related to the delays and waiting for plan approval for care

(APRVDLYM) and the subjects’ evaluation of the

experience with plan (RATPLANM). Blood pressure tests

278 JHHSA FALL 2008

were significantly related to the concern about delays and

waiting for plan approval for care (APRVDLYM), lacking

information on how plan works (LKINFOM), and the

concern of getting help from customer service

(PRBSVCM).

DISCUSSION

Model Discussion

Two research questions motivated this study: first,

“Do HMOs differ in their consumer-friendly

characteristics?” and if so, “Are those consumer-friendly

characteristics significantly associated with different

outcomes of health care utilization?” The answer to the first

question is that HMOs can be distinguished by their

consumer-friendly characteristics, which is evidenced by

the ANOVA results (see Table 3).

In general, the relationships between control variables

and preventive care utilization were consistent with

existing literatures (e.g., Kenkel, 1994; Rizzo, 2005). Yet

we need to extend the following two points: First, as noted

earlier, the relationship between age and preventive care is

unclear a priori (Rizzo, 2005). On the one hand, older

persons are at greater risk of illness, increasing the returns

to directing preventive care at them; and at the same time,

physicians are well aware that the need for routine

preventive care increases for aging individuals, and are

more likely to request preventive health intervention for

their older patients. On the other hand, the shorter life

expectancy of older individuals limits the benefits of

preventive care. Likewise, there is no a priori reason why

preventive care should be higher among females. However,

this pattern reflects well known gender-specific differences

in preferences for such care. The literature on health care

seeking routinely shows that women are more likely than

men to seek care, especially preventive health care (e.g.,

JHHSA FALL 2008 279

Anderson, Gonen, & Irwin, 2001). Second, although the

higher known incidence of hypercholestemia among

African-Americans implies that cholesterol checkups

should be significantly higher among African-Americans,

and some published literature (see Rizzo, 2005) has

proposed that African-Americana are also more likely to

receive physical examinations to obtain information on

cholesterol counts and other cardiovascular risk factors in

this cohort, our findings demonstrated the opposite. This

finding, nonetheless, is similar to those reported by Hass

and colleagues (2002). Their study found that the benefits

of managed care associated with the greater use of some

preventive care were not apparent for black persons or

Asian/Pacific Islanders enrolled in HMOs (Haas, et al.,

2002).

As to the second question: “Are consumer-friendly

characteristics significantly associated with different

outcomes of health care utilization?” we concluded that

some consumer-friendly variables were significantly

associated with some preventive care variables. In general,

when the concerns represented by the eight consumer-

friendly variables were negative, subjects tended to

decrease the use of preventive care. Otherwise, these

consumer-friendly variables tended to facilitate the use of

preventive care.

To be specific, Table 4 shows that problems

concerning getting a personal doctor or nurse

(GTDCPRBM) had a significant negative impact on the

frequency of physical examinations (CHECK) and breast

tests (BRSTEX). Previous literature suggests that when

getting a personal doctor or nurse becomes a major

problem, people tend to decrease routine physical

examinations and breast tests. However, “need approval for

treatment” (APRVTRTM) displayed no significant effects

on preventive care utilization. Since all HMO members use

health care providers within their plan's network and need

280 JHHSA FALL 2008

approval for the outside-network providers, “need approval

for treatment” may not be a satisfactory discriminant

variable. Delays in waiting for plan approval for care

(APRVDLYM) had a significant negative impact on

mammograms (MAMOGR) and blood pressure tests

(BPCHEK). It is understandable that when care delays due

to waiting for plan approval turns out to be a serious

problem, the result is the decrease in the frequency of the

preventive care utilization such as mammogram and blood

pressure tests during a fixed period, since people have to

wait for longer time to get the care access. Similarly, lack

of information on how the health plan works (LKINFOM)

also had a significant negative influence on cholesterol

checks (CHOLCK) and blood pressure tests (BPCHEK). It

is understandable that information about the health plan is

positively associated with the preventive care utilization

since the plan process of getting care is facilitated by the

enrollees’ full knowledge; correspondingly, if an enrollee

does not realize a service is provided or how to obtain it, he

or she is less likely to request it. Calling customer services

to complain or report problems (CUSTSVCM) also exerted

significant negative effects on women obtaining pap smear

tests (PAPSMR). Likewise, problems getting help from

customer service (PRBSVCM) had significant negative

impacts on blood pressure tests (BPCHEK). Thus, when

occurrence of problem reporting is less than a problem; and

when people in trouble can get timely help from customer

service, people tend to increase their preventive care

utilization such as pap smear tests or blood pressure tests.

The “Need to fill out paperwork for plan” (PPRWRKM)

significantly impacted both cholesterol checks (CHOLCK)

and pap smear tests (PAPSMR), and when enrollees were

required to fill out paperwork for the plan, they tended to

increase such care utilization. Paperwork may have

involved a self-reported health history, increasing the

physician’s awareness of family-related illness and

JHHSA FALL 2008 281

prompting the use of preventive tests; in addition,

paperwork helps formalize and routinize the care access

processing, and thus facilitate the care utilization. Subjects’

evaluation of the experience with plan (RATPLANM) was

significantly associated with physical examination

(CHECK), breast test (BRSTEX), and mammogram

(MAMOGR). Higher ratings of experience with plan point

to the greater use of physical examinations, breast tests, and

mammograms.

Implications and Discussion

Outside of the two research questions, this study

suggests another question that needs to be addressed: Why

do consumer-friendly variables have a differential impact

on preventive care variables? This state of affairs can be

interpreted in at least three ways. First, specialist care

rather than preventive care may serve as the better

dependent variables that can be significantly explained by

the consumer-friendly variables, since the former demands

the plan approval, and more frequently relates to the cost,

quality, and access boundaries of HMOs. Second, the

measurement of HMO characteristics is flawed; that is, the

eight consumer-friendly variables as measured by

enrollees’ experience with the plans may not be good

proxies of characteristics that can discriminate among

HMOs. A third interpretation is that consumer-friendly

variables have the greatest impact on consumer behavior,

but the health care encounter is an interaction between

patient and providers. Therefore, HMO characteristics that

influence physicians and other health care providers should

also be considered when analyzing the impact of HMO

characteristics on preventive care utilization. This issue

encourages the direction for future research. On the one

hand, it is meaningful to extend the investigation to

specialist care variables; on the other hand, as we noted

before, characteristics of HMOs can be measured by either

282 JHHSA FALL 2008

‘provider-driven’ or ‘consumer-centered’ approach. We

adopted the ‘consumer-centered’ approach in our study, but

it would be interesting to contrast this study’s results with a

‘provider-driven’ approach that distinguishes HMOs based

on the relationship between HMOs and physicians. The

“provider-driven” approach results in four essential HMOs:

staff model HMO, group model HMO, IPA (Independent

Practice Association) HMO, and network HMO.

Our research tried to tap the idea that HMOs have

multidimensionality in order to resolve the issue of

incongruent conclusions about HMO performance. Earlier

published literatures (e.g., Pearson et al., 1994; Manning et

al., 1984) treated HMOs as a unitary health plan form,

evaluating HMOs effectiveness without regard for

variations in their forms. From that perspective, HMOs

appear to have, at best, a modest positive relationship with

preventive health care utilization. However, when HMOs

are viewed as multifaceted plan forms, it can be argued that

these overall modest results may be due to the aggregation

of some forms of HMOs that are very effective with other

forms that are relatively ineffective.

Our study is more exploratory than explanatory in its

nature. The results discussed above indicate that since

HMOs characteristics do pose direct effects on preventive

care utilization, different forms of HMOs are associated

with markedly different outcomes of health care utilization.

Thus, the practical questions “Do HMOs make

difference?”, or “Are HMOs good for health maintenance?”

have no simple answer. Rather, to assess HMOs’

performance accurately researchers should, at the very

least, control for the degree of HMOs’ consumer-

friendliness.

Future conceptual and empirical work on HMOs

should develop and extend this notion that HMOs’

structural characteristics, whether consumer- or provider-

oriented, are important drivers of desired health care

JHHSA FALL 2008 283

service utilization. For example, researchers should

consider what characteristics are associated with good

performance, and what pose the potential threats to

undermine the quality of care. Our study is a first step in

this direction, but the consumer experience with health

plans is but one way to measure HMOs’ structural

characteristics. We also believe that a ‘provider-driven’

approach is well worth the effort. Therefore, other

appropriate and precise measures of HMO characteristics

should be explored. In addition, the objective assessment of

HMO performance demands the consideration of possible

contextual or contingency factors. Our study took into

account both sociodemographic factors and health status.

However, Miller and Luft (1994) point out that

performance of managed care organizations differs

considerably depending on which local market areas are

used for analysis. That is, the characteristics of the markets

in which managed care organizations operate may

influence their performance significantly. In short, the

structural form of an HMO may account for only a portion

of the variance in outcomes of health care utilization; other

situational factors are involved as well. Therefore, a

promising research direction is to include market factors

into future analyses when evaluating the performance of

HMOs.

The differences in the effectiveness of various forms

of HMOs raise questions about the mechanisms through

which HMOs may operate. One possibility is that different

forms of HMOs operate through different mechanisms; yet

similar outcomes may arise from very different processes.

Another possibility is that different forms of HMOs are

associated with different outcome variables, and thus

separate models or theories for individual forms of HMOs

appear more feasible than the integrative framework.

However, to understand in depth even one form of HMOs

will require consideration of antecedents, consequences,

284 JHHSA FALL 2008

mediating processes, and contextual contingencies.

Therefore, we hesitate to call for multiple models of

HMOs. Instead, we only propose that researchers might

compare and contrast the separate emerging models of

HMOs, perhaps in terms of general dimensions of form, or

some common mechanisms or processes. In this way, some

convergence of theory and research on HMOs may be

achieved.

CONCLUSION

In a 2000 WHO global ranking of health care, the

United States was ranked 37th. The WHO report (2000)

considered two factors the U.S. government and health care

community ignore: Does everyone have access? Is the cost

distributed equitably across all of society? WHO reasoned

that a fairly financed health system ensures financial

protection for everyone. Health systems can be unfair by

either exposing people to large, unexpected costs they must

pay on their own or by requiring those least able to pay for

care to contribute more proportionately than wealthier

citizens. Since their creation in 1920s, HMOs have been

regarded as promising health benefit plans that are designed

to address both access and cost issues. In this sense, the

careful reexamination of HMO performance is of special

significance. This paper proposed that an in-depth study of

HMO characteristics would potentially explain the mixed

performance of HMOs documented in previous studies. As

an exploratory study, this research provides reasonable

support to advocate that research should move beyond the

simple dichotomy between HMO and non-HMO insurance

to differentiate the effects of specific types of insurance as

well as the effects of specific care management tools, given

the complexity of insurance products. To understand

differences in performance among HMOs, it will be

important for health services researchers to reach a

JHHSA FALL 2008 285

consensus about the HMO and benefit plan characteristics

that should be routinely collected, analyzed, and discussed

in reports. These are the logical next steps in understanding

more about the mechanisms through which HMOs have

their effects on population health.

REFERENCES

Anderson, L.A., Gonen, J.S., & Irwin, K. (2001). At work

with the Protecting women's health. Business and

Health. Montvale: March vol. 19(3): 47.

Barringer, B., & Bluedorn, A. (1999). The relationship

between corporate entrepreneurship and

strategic management. Strategic Management Journal,

20(5): 421-444.

Barlett, D.L., & Steel, J.B. (2004). Critical condition: How

health care in American became big business and bad medicine. New York: Doubleday.

Culter, D., McClellan, M., Newhouse, J. (2000). How does

managed care do it? RAND J Econ, 31(3): 526-548.

Deaths and percentages of total deaths for the 10 leading

causes of death, by race: United States, 2002. (2005).

National Vital Statistics Reports. 53(17), March 7.

Ellis, R. (1989). Employee choice of health insurance. Rev

Econ Stat, 71(2): 215-223.

Glied, S. (2000). Managed care. In A. Newhouse & J. Cuyler

(Eds.), Handbook of health economics (vol.1, pp.707-

753). North-Holland: Amsterdam.

Haas, J.S., Phillips, K.A., Sonneborn, D., McCulloch, C.E., &

Liang, S.Y. (2002). Effect of managed care insurance on

the use of preventive care for specific ethnic groups in

the United States. Medical Care, 40(9): 743-751.

Herzlinger, R.E. (2004). Consumer-driven health care:

Implications for providers, payers, and policymakers.

Cambridge: Harvard Business School.

286 JHHSA FALL 2008

Jackson-Beek, M., & Kleinman, J. (1983). Evidence for self-

selection among health maintenance organization

enrollees. J Am Med Assoc, 250(20): 2826-2829.

Kenkel, D. (1991). Health behavior, health knowledge, and

schooling. J Polit Econ, 99(2): 287-305.

Kenkel, D. (1994). The demand for preventive medical care.

Appl Econ, 26:313-325.

Langwell, K., & Hadley, J. (1989). Evaluation of the medicare

competition demonstrations. Health Care Finan Rev,

11(2):65-80.

MacCallum, R. C., & Browne, M. W. (1993). The use of

causal indicators in covariance structure models: Some

practical issues. Psychological Bulletin, 114, 533-541.

Manning, W.G., Leibowitz, A., Goldberg, G.A., Rogers,

W.H., & Newhouse, J.P. (1984). A controlled trail of

the effect of a prepaid group practice on use of services.

New England Journal of Medicine. 310(23): 1505-1510.

Merrill, R., Brown, M., Potosky, A. et al. (1999). Survival and

treatment for colorectal cancer Medicare patients in two

group/staff health maintenance organizations and the

fee-for-service setting. Med Care Res Rev, 56(2): 177-

196.

MEPS HC-070: 2002 full year consolidated data file (2004).

Center for financing, access and cost trends, agency for

healthcare research and quality. Rockville, MD.

Miller, R.H., & Luft, H.S. (1994). Managed care plan

performance since 1980: a literature analysis. Journal of

the American Medical Association. 271(19): 1512-1519.

Miller, R., & Luft, H. (1997). Does managed care lead to

better or worse quality of care? Health Affairs, 16: 7-25.

Miller, R., & Luft, H. (2002). HMO performance plan update:

an analysis of the literature 1997-2001. Health Affairs,

21(4): 63-86.

Pearson, S.D., Lee, T.H. et al. (1994). The impact of

membership in a health maintenance organization on

hospital admission rates for acute chest pain. Health

Services Research. 29(1): 59-74.

JHHSA FALL 2008 287

Pedhazur, E. (1982). Multiple regression in behavioral

research. Holt, Rinehart & Winston, Fort Worth, TX.

President’s Advisory Commission on Consumer Protection

and Quality in the Health Care Industry (1998). The

State of Health Care Quality: How Good is Care? Agency for Healthcare Research and Quality: Rockville,

MD.

Rizzo, J.A. (2005). Are HMOs bad for health maintenance?

Health Economics, 14(11):1117-31.

Robinson, R., & Steiner, A. (1998). Managed Health Care.

Open University Press: Buckingham.

Robinson, R. (2000). Managed care in the United States: a

dilemma for evidence-based policy? Health Economics,

9: 1-7.

Schauffler, H., & Rodriguez, T. (1993). Managed care for

preventive services: a review of policy options. Med

Care Rev, 50(2): 153-198.

citovsky, A., McCall, N., & Benham, L. (1978). Factors

affecting the choice between two prepaid plans. Med

Care, 16(8): 660-675.

Sultz, H.A., &Young, K.M. (2004). Health care USA:

Understanding its organization and delivery. Jones and

Barlett Publisher, Sudbury, MA.

The World Health Report. (2000). Retrieved December 23,

2005, from http://www.who.int/whr/en/

288 JHHSA FALL 2008

Appendix I

Characteristics Variables of Medicaid HMOs and Private HMOs

JHHSA FALL 2008 289

Preventive Care Variables

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.