Preventive care and insurance coverage.
The choice to engage preventive tests and treatments depends on the many factors that influence the perceived marginal benefits and marginal costs of these services. This article examines several of the factors that affect the perceived costs and benefits of these treatments. In particular, the influence of having a choice of insurance plans, the type of insurance, and the individual's health status on the decision to access preventive services is examined.
One of the arguments for managed care as a cost saving insurance model is that it uses more preventive care than indemnity insurance (Rolnick et al. 1996). Health maintenance organizations (HMOs) are likely to pursue preventive care measures in an effort to reduce future health care costs. However, this strategy may be offset by the high turnover rates in managed care. For example, estimates of the turnover rates for managed care enrollees range from 16% to 19% (Association of American Physicians and Surgeons 1995). Thus over a five-year period, a managed care company may expect to lose up to 58% of their original enrollees and all of its investments in these enrollees' health. This may reduce the HMO's incentive to provide preventive tests because these future cost savings may accrue to some other insurer.
If HMOs are more likely to provide preventive care, then individuals who value this care should be more likely to enroll in an HMO, other things equal. Most studies that examine the effect of HMO enrollment in preventive care usage equations include as an independent variable a measure for enrollment or an indicator of whether tests are paid for (for a few examples see, Zapka et al. 1999; Eisen et al. 1999; Weisman and Henderson 2001).
This methodology is fine if the worker does not have a choice of plans at his or her place of employment. In this case, the benefits manager chooses the type of insurance plan, and the worker either chooses to be covered by this insurance or forgoes insurance cover-age. The effect of being in an HMO on the use of preventive care, in this case, is exogenous. However, when the worker has a choice of insurance plans, the effect of HMO enrollment on the use of preventive care becomes endogenous. This study uses the Medical Expenditure Panel Survey (MEPS) to examine how the endogeneity of the choice of an HMO over a fee-for-service plan affects the use of preventive treatments when a worker has a choice of insurance plans. The results are compared to those of workers who do not have a choice of plans and to a model where HMO is assumed to be exogenous.
II. LITERATURE REVIEW
Research has examined the effect of insurance type on the use of preventive treatments. Phillips et al. (2000) reviewed the literature from 1990 to 1998 on whether managed care enrollees received more preventive treatment than non-managed care enrollees. About 60% of the studies found no difference, whereas 37% found that managed care enrollees were significantly more likely to receive the preventive care services. HMO enrollees are more likely to be diagnosed in the earlier stages of cancer (and Lee-Feldstein et al. 2000; Riley et al. 1994) and are more likely to receive Pap smears, mammograms, and fecal occult blood tests (Gordon et al. 1998; Weisman and Henderson 2001). Eisen et al. (1999) find a positive association with HMO enrollment and prostate cancer screening. Carrasquillo et al. (2001) find that elderly women in HMOs were more likely to have mammograms.
In a recent study, Haas et al. (2002) find that Hispanics who are enrolled in managed care plans are more likely than fee-for-service Hispanics to receive mammography screenings, Pap smears, and breast exams. Likewise, whites with managed care are more likely than their fee-for-service counterparts to receive mammography and cholesterol screenings. Wang and Pauly (2003) find that managed care coverage generally supports more preventive care, but in some cases the preventive care is not cost-effective. Other studies have found income, insurance coverage, job type, and race to be important determinants of the use preventive care treatments (for recent studies see Almeida et al. 2001; Bednarek and Schone 2003; DeVoe et al. 2003; Friedman et al. 2002). Having a choice of insurance plans gives a worker more options over the types of care that will be covered. Davis et al. (1995) have examined how a choice of plans affects an enrollee's perception of the quality of the plan. Plan choice is a strong indicator of plan satisfaction. Managed care plans received lower ratings for quality when enrollees did not have a choice of plans. Ullman et al. (1997) use a choice of plans as a variable to examine satisfaction with specific plans. They find that having a choice of plans results in more satisfaction with the plan that is chosen, but that having choices within a plan, as indicated by the difference between a point of service plan and network only plans, does not have a significant effect on satisfaction when enrollees do not have a choice of plans.
Enrollment in HMOs is examined by Kemper et al. (1999/2000). They find that HMO enrollees are "younger, more likely to be single, and more likely to be from a minority group." However, to some extent this result may reflect job attributes and the type of plans offered in those jobs. Kemper and colleagues also find that HMO enrollees have lower incomes. They argue that this reflects the trade-off enrollees are willing to make between reduced provider choice and lower costs.
Hsieh and Lin (1997) investigate how health information affects the demand for preventive care. They find that increased health information is positively related to the demand for preventive care. Individuals compare the perceived marginal benefit of an investment in preventive care to the marginal cost. As more information about the benefits of preventive care becomes known, individuals are more likely to invest in this care. To the extent that workers with a choice are more likely to examine their different insurance options, having a choice of plans should be positively related to the demand for preventive care.
Many workers do not have a choice of plans; generally by default they are insured by the plan that their employer offers. Kemper et al. (1999/2000) find that 51% of their respondents have a choice of plans and 48% are in HMOs. A consequence of having a choice is that empirical models examining the effect of HMO enrollment on preventive care demand may be biased by the active choice of plans.
III. EMPIRICAL STRATEGY
Preventive care recommendations vary among different medical groups and associations. Table 1 shows the preventive care recommendations of various medical groups and associations. The recommendations are general recommendations and exclude special cases. For example, the recommendation for blood pressure check by the American College of Physicians indicates that these checks should be conducted every 1 to 2 years with the provision "Normotensive patients should have blood pressure measurements at least yearly if any of the following pertains: (1) diastolic blood pressure between 85 and 89 mm Hg: (2) African American heritage" (National Library of Medicine 2001) and three other special cases.
The receipt of preventive care is measured in the MEPS as the time since last test, exam, or treatment. Possible responses are "within the past year," "within past 2 years," "within past 5 years," "more than 5 years," and "never." Being constrained by these responses and given the recommendations in Table 1, the following measures are used as acceptable levels of treatment for these preventive services: blood pressure check--every 2 years for all patients 20 years and older (90.5% of the sample used in this study responding to the blood pressure check question had it within the recommended guidelines); cholesterol check--every 5 years between ages 35 to 65 for males (75.2% of the sample had cholesterol check within guidelines) and between ages 45 to 65 for females (84.6%); mammography--every 2 years examined separately for females between ages 40 to 49 (68.6%) and from age 50 to 70 (80.0%); Pap smear--annual ages 18 to 65 (66.2%); breast exam--annually for women age 40 and over (83.6%) and every 2 years for women under age 40 (67.9%); prostate-specific antigen--annually for all men age 50 and over (51.9%).
The effect of HMO enrollment on the use of the preventive treatments depends on whether the worker has a choice of plans. To estimate this effect, the empirical model examines three issues: (1) the determinants of plan choice; (2) the determinants of HMO enrollment for workers who have a choice of plans; (3) the influence of HMO enrollment on the use of preventive services for those with and without a choice of plans.
A. Basic Model
Previous studies of the effect of HMO enrollment on preventive care usage have used a dummy variable indicating HMO enrollment in equations predicting the use of preventive care. However, for individuals who have a choice of plans, HMO enrollment is endogenously determined. This implies that the effect of HMO enrollment on preventive care usage should be estimated separately for individuals with and without a choice of plans. For workers with a choice of plans, an instrumental variables approach can be used. For workers without a choice of plans the HMO variable can be included as an exogenous variable. In both equations the potential of sample selection bias based on whether an individual has a choice of plans must be accounted for.
It is assumed that the worker makes his or her choice to use preventive treatment based on the expected utility of that preventive treatment, EUP. The expected utility of preventive treatment is specified as
EU[P.sub.i] = g([X.sub.pi], HM[O.sub.i])
where [X.sub.pi] is a vector of variables affecting the choice of using preventive treatment for individual i, and HM[O.sub.i] is a variable indicating HMO enrollment. The latent variable, EU[P.sub.i], is not directly observed. Instead, the actual decision to use a preventive treatment is observed. For each worker, define
[Y.sub.pi] = 0 if EU[P.sub.i] < 0
(does not use preventive treatment)
[Y.sub.pi] = 1 if EU[P.sub.i] [greater than or equal to] 0
(uses preventive treatment)
where [Y.sub.pi] is a dummy variable indicating that worker i uses preventive treatment, p.
For workers who have a choice of plans, the HMO enrollment decision is based on the expected utility from this enrollment, EUH. As in the preventive treatment model, the expected utility is not observed, but the choice of plans is. As such, define for those workers who have a choice of plans,
[Y.sub.hi] = 0 if EU[H.sub.i] < 0
(does not choose an HMO)
[Y.sub.hi] = 1 if EU[H.sub.i] [greater than or equal to] 0
(chooses and HMO)
where [Y.sub.hi] is a dummy variable indicating that worker i is enrolled in an HMO.
Finally, workers are observed choosing an HMO only if they have a choice of insurance plans. Let [Y.sub.ci] be a dummy variable indicating that worker i has a choice of plans. This choice variable will also indicate whether the HMO variable is endogenous or exogenous.
B. No Choice of Plans
The probability that an individual uses preventive treatments depends on several independent variables as well as whether the individual is enrolled in an HMO. If a worker does not have a choice of insurance plans, [Y.sub.ci] = 0, then enrollment in an HMO occurs when that is the only insurance plan the worker is offered. As such, estimating the probability of preventive care for this group of workers must account for the fact that this sample is selected for workers who do not have a choice of plans. This model applies a sample selection correction based on Heckman's (1976) procedure applied to a probit model using a method similar to Van De Ven and Van Praag (1981). Let the choice equation, [Y.sub.ci] be specified as
(1) [Y.sub.ci] = [X.sub.ci][[beta].sub.c] + [[epsilon].sub.ci]
and the use of preventive treatments as
(2) [Y.sub.pi] = [X.sub.pi][[beta].sub.p] + HM[O.sub.i][[beta].sub.h] + [[epsilon].sub.pi], if [Y.sub.ci] = 0
[[epsilon].sub.p] [approximately] N(0, 1); and corr([[epsilon].sub.c], [[epsilon].sub.p]) = [rho],
where [X.sub.ci] and [X.sub.pi] are vectors for individual characteristics affecting the choice of plans, c, and whether the worker uses preventive treatment, p, HM[O.sub.i] is an indicator of HMO enrollment. Lee (1979) has shown that equation (2) can be estimated by including an additional variable [[lambda].sub.ni] where the n indicates no choice of plans. Let f and F denote the normal density and distribution function. Then
[[lambda].sub.ni] = -f([X.sub.ci][[beta].sub.c])/F([X.sub.ci][[beta].sub.c]).
The values of [[beta].sub.c] are estimated using a probit model and these estimates are used to predict [[lambda].sub.ni]. This predicted value of [[lambda].sub.ni] is then used in equation (2) to estimate the probability of using preventive treatment as follows:
(3) [Y.sub.pi] = [X.sub.pi][[beta].sub.p] + HM[O.sub.i][[beta].sub.h] + [alpha][[lambda].sub.ni] + [[eta].sub.ni],
where E([[eta].sub.ni] | [Y.sub.ci] = 0) = 0.
C. Choice of Plans
The selection model also applies for workers who have a choice of plans, but now the choice of HMO enrollment is endogenous to the system. Because workers must choose the insurance plan prior to choosing to use preventive treatments, the model is a recursive model. The preventive care usage equation is a function of HMO choice, but HMO choice is not a function of the usage of preventive care. Because all of the dependent variables are dichotomous, a maximum likelihood estimator must be used to predict the probability of receiving preventive care as a function of an endogenously determined HMO choice for workers with a choice. The choice variable is the selection criterion. The empirical strategy therefore is to estimate the following set of equations for individuals with a choice of plans:
(4) [Y.sub.ci] = [X.sub.ci][[beta].sub.c] + [[epsilon].sub.ci]
(5) [Y.sub.hi] = [X.sub.hi][[beta].sub.h] + [alpha][[lambda].sub.ci] + [v.sub.ci], if [Y.sub.ci] = 1
(6) [Y.sub.pi] = [X.sub.pi][[beta].sub.p] + [Y.sub.hi][[beta].sub.h] + [alpha][[lambda].sub.ci] + [[eta].sub.ci], if [Y.sub.ci] = 1,
where [Y.sub.hi] is the HMO indicator variable, [[lambda].sub.ci] = f([X.sub.ci][beta]c)/[1 - F([X.sub.ci][[beta].sub.c])] is the sample selection correction variable, and the error terms [v.sub.ci] and [[eta].sub.ci] have an expected value of zero when [Y.sub.ci] = 1. The variable [Y.sub.ci] can be predicted using instrumental variables. Identification variables are discussed in the data section.
The MEPS was conducted to provide nationally representative estimates of health care use and expenditures among the U.S. population (Agency for Healthcare Research and Quality 2001). Participants came from a probability weighted sample of noninstitutionalized U.S. civilians. The data in the article are from the 1996 full year consolidated data file HC012, which contains 22,601 respondents of which 15,745 are age 18 or older. For this study, only individuals 18 or older who are working and insured were examined, giving a sample size of 8,279. The uninsured (3,021 or 19.7% of the 15,745 respondents over age 18) or nonworkers (4,881 or 31.9%) were excluded because they may receive their health care under other circumstances or from other sources that may influence the demand for preventive treatments. The data include information on insurance coverage, type of insurance, perceived health status, demographic characteristics, questions on access and quality of care, and time since last preventive treatment.
The survey design for the MEPS data uses a stratified and weighted sample. Variables are included in the MEPS to allow estimation of a Taylor series variance. To perform this estimation on this stratified data, Stata (Stata Corporation 1999) was used. This program estimates standard errors that are adjusted for the complex survey design (Cohen et al. 1996). Cohen (1996) finds that Stata and SUDAAN produced identical point estimates and standard errors.
An explanation of the variables used to estimate each equation (4)-(6) is given in the following. Briefly, the choice equation (4) includes for the independent variables [X.sub.ci]: the number of people employed at the respondent's firm, union status, 9 occupation dummy variables, 10 industry variables, usual weekly hours of work, hourly wage, and demographic variables age, age squared, and dummy variables indicating male, married, black, Hispanic, educational level of a bachelor's degree, and region of residence South, West, and Northeast.
The HMO equation includes for [X.sub.hi]: a selection bias adjustment for choice of plans, the same demographic variables in the choice equation, self-assessed health and mental health dummy variables, log of income, and instruments of family size and dummy variables indicating the respondent lives in a metropolital statistical area (MSA) and has sick pay available from work.
The nine different preventive treatment equations include the selection bias variable for plan choice, an indicator of HMO coverage, the demographic variables in the choice equation, self-assessed health, and dummy variables, indicating their usual physician is a family practitioner or an internist.
A. The Determinants of Choice
Employers often offer their workers a choice of health plans. In this case, the worker compares the costs and benefits of each plan to determine which plan to enroll in. Having a choice of plans is likely to depend on the type of job a worker has. In particular, workers at larger companies may be more likely to have a choice of plans because the HMO Act of 1973 requires large employers who offer insurance to offer an HMO plan. To capture the effect of firm size, the number of workers employed at the respondent's company is included. Whether a firm offers a choice of plans may differ by industry or occupation because competition for workers in different industries and occupations may affect the benefit packages offered. For example, white-collar workers may be more likely than blue-collar workers to be offered a choice of plans. To account for these differences, 10 industry and 8 occupation dummy variables are included. Unionized workers may be more likely to have a choice of plans because this may be negotiated in the contract. A dummy variable indicating union status is included. Workers who work more hours and are paid a higher wage should be more likely to have a choice of plans because higher paid workers generally receive more fringe benefits as a percentage of their earnings. To capture this overall effect, usual weekly hours of work and the hourly wage are included. Higher values of both hours of work and wages should lead to a greater probability of choice. Total income is not included because workers may have nonlabor sources of income, and these sources are unlikely to influence whether the worker's firm offers a choice of health plans.
Individual worker characteristics may influence whether a worker has a choice to the extent that these characteristics affect a worker's choice of jobs. Thus, the model includes age and age squared as well as dummy variables indicating that the worker has at least a bachelor's degree, is married, male, black, and Hispanic. Also, the area of the country may influence whether workers have a choice among plans. The included regional variables are Northeast, South, and West; the excluded region is the Midwest.
The dependent variable is equal to 1 if the worker has a choice of insurance plans. This variable is constructed from a combination of variables. If the employer offers more than one insurance plan, the worker will have a choice of plans at their place of employment. Also a worker may have a choice of plans if the worker is offered only one insurance plan at his place of employment but reports that his insurance coverage is from some other source. In this case, the worker obviously chose another insurance plan over the plan that was offered at his place of employment. One way this can occur is if the worker receives insurance through his spouse's employment even though he is offered insurance coverage at his own work. Thus, married individuals may have a choice of health plans when both spouses are offered health insurance at their place of employment. Around 42% of the workers in the sample are given a choice of plans at work, and an additional 6% of workers who are offered only one plan at their work are enrolled in another plan from some other source. This implies that 48% of the workers have a choice of plans, which is consistent with Kemper at al. (1999/2000). Having a choice of plans regardless of the source of the choice may make workers more likely to examine the benefits of their plans and take advantage of the preventive care provided by those plans.
B. The HMO Enrollment Choice
The HMO enrollment choice only occurs for workers who have a choice of insurance plans. For those workers who have a choice of insurance plans, 58% choose an HMO. For workers without a choice, 40% are enrolled in HMOs. Overall, 49% of the sample is enrolled in an HMO.
The determinants of a worker choosing to enroll in an HMO include age, health status, and other demographic variables. Younger, healthier individuals are generally pursued by HMOs because the cost of covering these individuals is lower. Likewise, because younger individuals are unlikely to consume many health care services, they are more likely to seek out the lower cost insurance plan, which is often an HMO (Altman et al., 2003).
Healthier individuals are less costly to cover. HMOs may use advertisements showing healthy individuals pursuing outdoor activities, instead of showing people in poor health being treated. From the consumer side, less healthy individuals may be reluctant to join an HMO because they may have a limited choice of providers and may not be able to keep their current provider. The physical health variables in the model are dummy variables from a perceived health status question indicating poor health, fair health, good health, very good health, and excellent health. The omitted category is excellent health. Poor health and fair health are combined into one variable because less than 1% of the respondents were in poor health and only 5% reported fair health.
Following Kemper et al. (1999/2000), the author includes a variable indicating mental health status. The variable is an indicator that the worker has poor to fair mental health. Individuals with poorer mental health may be more likely to see their physicians and therefore receive preventive treatments.
Other respondent characteristics that may influence the type of plan a worker is enrolled in include race, marital status, income, and education. Previous studies have found that blacks are more likely to be enrolled in HMOs as are lower income workers (e.g., see Kemper et al., 1999/2000). The income variable is the log of total income for the respondent. This includes all sources of income and will capture any income effects on the choice of type of health insurance plan. Education is measured by a dummy variable equal to 1 if the respondent has a bachelor's degree. This education measure may influence whether a worker chooses an HMO through his or her ability to better understand their insurance options.
It can be argued that most of these variables should also be included in the preventive care treatment equations. In this case, the HMO variable would not be identified. Several additional variables are included in the HMO equation to identify it. The first variable is an indicator variable that the worker lives in an MSA. As the size of the population increases, HMOs become more viable because it is easier for the HMO to organize and negotiate with physician groups and hospitals. Furthermore, these hospitals and physician practices are likely to be located closer to the enrollee, making enrollment in the HMO more convenient. A second identifying variable is a dummy variable indicating the availability of sick pay at work. The ability to receive sick pay may affect a worker's choice of health plans but is unlikely to affect their decision to use preventive care. A third variable is family size. Larger families are more likely to use more health care, so workers with larger families may be more likely to choose relative cheaper HMO plans that have lower costs for routine care.
To test the validity of the instruments a test similar to Bound et al. (1995) was performed. They suggest a joint test on the excluded variable to ensure that the endogenous variable is correlated with the instruments. A Wald test of the instruments having no explanatory power was rejected with an F-statistic of 10.72 (p-value < 0.0001).
C. Demand for Preventive Care
If the expected utility of obtaining a particular preventive health services is positive, then the worker will be observed using this service. Variables that affect the costs and benefits of these treatments will influence the decision to use the treatment.
The probability of receiving treatment is modeled as a function of sociodemographic variables, regional measures, HMO enrollment, individual health status indicators, and the choice of plans sample selection correction variable. The sociodemographic variables included are dummy variables for black and Hispanic, age, age squared, the log of income, a dummy variable for married, an indicator that the worker has at least a bachelor's degree, self-assessed health status, regional dummy variables, and for blood pressure tests a dummy variable for male. Similar sociodemographic variables have been included in studies by Sudano and Baker (2003), Greene et al. (2001), and Carrasquillo et al. (2001). Weisman and Henderson (2001) find that black women are more likely than white women to receive preventive services.
Health investments and educational investments are likely to be positively related. Individuals with higher levels of education generally have higher earnings so that the opportunity cost of missing work is higher. This gives any investment in health and preventive care a greater return. The sign on the bachelor's degree variable should be positive.
Self-assessed physical health is included to determine if health status affects the use of these preventive treatments. Individuals with poor health status may be more likely to receive preventive care because they will be more likely to seei a physician on a regular basis. In this case, the physician may encourage the patient to have the preventive test. On the other hand, those in good health may be more likely to take preventive measures to catch many illnesses before their health deteriorates. The same physical health status variables used in the HMO model are included here. Negative coefficients on the health variables support the hypothesis that individuals in good health are proactive in their care, whereas positive coefficients may indicate that those in poorer health receive the preventive treatment as part of their routine care for their health problems.
The type of physician an individual sees may influence the decision to use preventive treatments. Because practice patterns can differ by specialty and training, two dummy variables are included for family practitioner and internist. The variables are equal to 1 if the worker has indicated that he or she has a usual source of care and that source of care is a family practitioner or an internist, respectively. Given the general nature of these two types of practitioners, they may be more likely to recommend that their patients have the preventive test.
When workers have a choice of plans, they must determine which insurance plan to use. Workers who have no choice of plans but are offered only one plan must either accept or reject the health plan offered by their employer. In this case, the worker may have less knowledge about all of the benefits of the plan because, other things equal, the benefits of obtaining additional plan information are not as high as they are in the case when a comparison of plans can be made. As such, workers with a choice of plans may be more likely to choose a plan that offers preventive treatments, because they have more knowledge about what is covered in their plan. Furthermore, workers may purposefully be choosing the plan to obtain specific types of services. Because HMOs are structured with an emphasis on prevention, they should be more likely to offer preventive treatment (Rolnick et al. 1996). Then other things equal, one would expect that workers with a choice of plans who wish to obtain these treatments should more likely choose to enroll in an HMO and use these treatments. Similarly, workers without a choice of plans may also be more likely to obtain preventive treatments if they are enrolled in an HMO, as long as HMO plans are more likely to offer preventive treatments.
Table 2 shows the means for the dependent variables. The first column shows the means for the full sample, and the next two columns show the means for workers without and with a choice of plans. A t-test for a difference in means by choice shows that in all cases, except mammogram for older women, those workers with a choice of plans are more likely to receive the preventive treatment and also more likely to be enrolled in an HMO. The means and t-tests for the independent variables used in the preventive treatment equations are shown in Appendix Table A1. These t-tests show significant differences for many of the variables.
Table 3 shows the estimated coefficients for predicting the probability that a worker has a choice of plans using a probit function. Workers at larger companies are more likely to be given a choice of plans, as are unionized workers. Workers in the white-collar occupations of professional, technical, and kindred; and clerical and kindred workers are more likely to have a choice of plans. Relative to the omitted industries of professional services, public administration, and industry unknown, workers in transportation, communication and utilities and finance, insurance and real estate are more likely to have of choice of health care plans. Most of the other included industries are less likely to have a choice of plans.
The probability of having a choice increases with age at a decreasing rate, reaching a maximum at age 47. Workers with higher wages and more hours of work per week are also more likely to have a choice of plans. Hispanics are less likely than whites to have a choice of plans. Relative to the Midwest, workers in the West are more likely to have a choice of plans. Workers with at least a bachelor's degree are more likely to have a choice of plans.
Table 4 shows the probit results for HMO enrollment for workers who have a choice of plans. The Heckman selection bias correction coefficient is significant indicating that sample selection is present. The significance of the family size coefficient likely reflects the lower costs of office visits and prescription drugs that generally prevail in HMOs. Other things equal, larger families are more likely to have more office visits and prescriptions in a given year and may therefore be more likely to choose the HMO.
A worker's health status is only marginally significant for those in very good health. Living in an MSA increases the probability of enrolling in an HMO, as does receiving sick pay at ones job. Consistent with Kemper et al. (1999/2000), workers with higher incomes are less likely to choose HMOs and Hispanics and blacks are more likely to choose HMOs. Married workers and workers with a bachelor's degree are less likely to choose to enroll in an HMO. Relative to the Midwest, workers in other regions of the country are more likely to enroll in HMOs.
The effect of HMO enrollment on the probability that the respondent receives the particular preventive treatment in the recommended time interval is shown in Tables 5-7. Each table has a set of results for each preventive treatment. The first column shows the probability of receiving the preventive treatment when the worker does not have a choice of plans. In this case, the HMO variable is exogenous. The second column is for those workers who have a choice of plans. The HMO variable is endogenous in this case. The final column shows the full sample results that would occur if having a choice of plans is ignored and HMO is exogenous. The estimated coefficients for married, age, age squared, male (in the blood pressure check model), and the regional dummy variables are not presented although they were included in the model.
Important for this study is the effect of HMO enrollment on the use of preventive services. In six out of the nine preventive health services, the coefficient on the HMO variable for workers without a choice of plans is significant. For workers with a choice of plans the HMO coefficient is significant in only two of the equations, cholesterol check for males age 35-65 and prostate exams. In both of these equations a Hausman (1978) test indicated that the instrumental variable estimates were appropriate (p-value < 0.05). The Hausman test was not significant for any of the other equations. An informal exclusion restriction test similar to that performed by Winter-Ebmer (2003) was also performed. The instruments should only influence the use of preventive treatments through their impact on the HMO choice. Thus the sample for those individuals with a choice of plans was split into those who chose an HMO and those who did not. For each preventive treatment, the excluded instruments should have no explanatory power. In this informal model, prostate exam, mammography for older women, and cholesterol check for women had no significant coefficients on any of the instruments.
Note that the preventive tests where the HMO coefficient is significant for workers without a choice are similar to those in the full model in the last column where the choice of plans is not accounted for and HMO is assumed exogenous. This indicates that in general HMOs are more likely to offer preventive treatment. For workers with a choice of plans, the choice of an HMO has little influence on whether preventive care is used. Thus overall, these results indicate that HMOs are generally more likely to offer preventive care, but when a choice of plans is available, preventive care is no more likely for individuals who choose HMOs.
The choice sample selection variable is significant in the blood pressure and the female cholesterol check equations for workers without a choice of plans and in the breast exam equation for older women workers with a choice of plans. Although the selection variable is not significant in many equations, the few significant HMO coefficients in the choice equations and the preponderance of significant HMO coefficients in the no choice equations indicates that choice is important.
The worker's race has little effect on the use of preventive treatments. Workers with higher incomes and a choice of plans are generally more likely to use preventive care. Being married is positively related to breast exams for younger women and Pap smears. Although not shown, males have a significantly lower probability of having their blood pressure checked.
The coefficients on the health status variables are generally positive and significant for the blood pressure test and cholesterol check and negative and significant for the prostate exam for males with a choice of plans. The positive coefficient for the more routine blood pressure is not surprising because it is often administered at each office visit, and workers in poorer health are more likely to have more office visits. Likewise, individuals in poor health may also be more likely to receive cholesterol tests because high cholesterol is a fairly common problem. The negative coefficient for the health status variables in the prostate exam equation for workers with a choice of plans indicates that workers in excellent health are more likely to have the prostate exam than workers in good to very good health. Perhaps these healthier individuals are more likely to seek out this test as a preventive measure.
Having a usual source of care as a family practitioner or an intern increases the probability of receiving many of the preventive treatments relative to not having a usual source of care or having a usual source of care from some other specialty.
VI. CONCLUSION AND DISCUSSION
Having a choice of plans is strongly influenced by the type of job a worker has. Workers who have a choice of plans are more likely to be enrolled in an HMO. Correcting for choice in the preventive care usage equations shows that in most cases, workers who choose an HMO are no more likely to receive preventive treatments than workers who did not choose the HMO. The influence of HMO enrollment on the use of preventive treatments is more likely to be significant for individuals who do not have a choice of plans. Several other methods of estimation were tried to determine if this results was consistent. In all cases, the HMO coefficient was more likely to be significant for those not having a choice of insurance plans. Thus the influence of HMO enrollment on preventive tests does not come from workers choosing an HMO to receive the preventive treatments, but instead comes from those workers who have no choice of plans and are in an HMO because that is the type of plan their employer offers.
The fact that the HMO coefficients in the preventive care equations for workers without a choice are similar to those in the full model where the choice of plans is not accounted for indicates that models that ignore plan choice should be interpreted carefully. For example, a positive HMO coefficient in a preventive care equation might be used as an argument for requiring employers to offer workers a choice of health care plans. The argument would be that this requirement could lead to more use of preventive care and lower health care costs. However, given the results from this study, the use of preventive care might not increase, but instead the unintended effect would be that premium costs might increase as the workers' risk pool is spread over several plans. Further studies on this topic are warranted because information about the number and types of choices was unavailable in this data set.
Agency for Healthcare Research and Quality. Puf Main Data Results. Rockville, MD, March 2001. Available online at www.meps.ahrq.gov/puf/pufdetail.asp.
Almeida, R. A., L. C. Dubay, and G. Ko. "Access to Care and Use of Health Services by Low-Income Women." Health Care Financing Review, 22(4), 2001, 27-45.
Altman, D., D. Cutler, and R. Zeckhauser. "Enrollee Mix, Treatment Intensity, and Cost in Competing Indemnity and HMO Plans." Journal of Health Economics, 22, 2003, 23-45.
Association of American Physicians and Surgeons. AAPS News, 51(5), 1995.
Bednarek, H. L., and B. S. Schone. "Variation in Prevention Service Use among the Insured and Uninsured: Does Length of Time without Coverage Matter?" Journal of Health Care for the Poor and Underserved, 14(3), 2003, 403-19.
Bound, J., D. A. Jaeger, and R. M. Baker. "Problems with Instrumental Variables Estimates When the Correlation between the Instruments and the Endogenous Explanatory Variable Is Weak." Journal of the American Statistical Association, 90, 1995, 443-40.
Carrasquillo, O., R. A. Lantigua, and S. Shea. "Preventive Services among Medicare Beneficiaries with Supplemental Coverage versus HMO Enrollees, Medicaid Recipients, and Elders with No Additional Coverage." Medical Care, 39(6), 2001, 616-26.
Cohen, S. "An Evaluation of Alternative PC Based Software Packages Developed for Analysis of Complex Survey Data." Agency for Health Care Policy and Research, Rockville, MD. 1996.
Cohen, S. B., R. DiGaetano, and H. Goksel. "Estimation Procedures in the 1996 Medical Expenditure Panel Survey Household Component." MEPS Methodology Report no. 5. ACHPR Pub. No. 99-0027. Agency for Health Care Policy and Research, Rockville, MD. 1999.
Davis, K., K. S. Collins, and C. Schoen. "Choice Matters--Enrollees' Views of Their Health Plans." Health Affairs, 14(2), 1995, 99-112.
DeVoe, J. E., G. E. Fryer, R. Phillips, and L. Green. "Receipt of Preventive Care among Adults: Insurance Status and Usual Source of Care." American Journal of Public Health, 93(5), 2003, 786-91.
Eisen, S. A., B. Waterman, C. S. Skinner, J. F. Scherer, J. C. Romeis, K. Bucholz, A. Heath, J. Goldberg, M. J. Lyons, M. T. Tsuang, and W. R. True. "Sociodemographic and Health Status Characteristics Associated with Prostate Cancer Screening in a National Cohort of Middle-Aged Male Veterans." Urology, 53(30), 1999, 516-22.
Friedman C., F. Ahmed, A. Franks, T. Weatherup, M. Manning, A. Vance, and B. L. Thompson. "Association between Health Insurance Coverage of Office Visit and Cancer Screening Among Women." Medical Care, 40(11), 2002, 1060-67.
Gordon, N. P., T. G. Rundall, and L. Parker. "Type of Health Coverage and the Likelihood of Being Screened for Cancer." Medical Care, 35(5), 1998, 636-45.
Greene, J., J. Blustein, and K. A. Laflamme. "Use of Preventive Care Services, Beneficiary Characteristics, and Medicare HMO Performance." Health Care Financing Review, 22(4), 2001, 141-53.
Haas, J. S., K. A. Phillips, D. Sonneborn, C. E. McCulloch, and S. Y. Liang. "Effect of Managed Care Insurance on the Use of Preventive Care for Specific Ethnic Groups in the United States." Medical Care, 40(9), 2002, 729-31.
Hausman, J. A. "Specification Tests in Econometrics." Econometrica, 46, 1978, 1251-71.
Heckman, J. J. "The Common Structure of Statistical Models of Truncation, Sample Selection, and Limited Dependent Variables and a Simple Estimator for Such Models." Annals of Economic and Social Measurement, 5, 1976, 475-92.
Hsieh, C., and S. Lin. "Health Information and the Demand for Preventive Care among the Elderly in Taiwan." Journal of Human Resources, 32(2), 1997, 308-33.
Kemper, P., J. D. Reschovsky, and H. T. Tu. "Do HMOs Make a Difference? Summary and Implications." Inquiry--Blue Cross and Blue Shield Association, 36(4), 1999/2000, 419-25.
Lee, L. F. "Identification and Estimation of Binary Choice Models with Limited (Censored) Dependent Variables." Econometrica, 47(4), 1979, 977-96.
Lee-Feldstein, A., P. J. Feldstein, T. Buchmueller, and G. Katterhagen. "The Relationship of HMOs, Health Insurance, and Delivery Systems to Breast Cancer Outcomes." Medical Care, 38(7), 2000, 705-18.
National Library of Medicine. Health Services/Technology Assessment Text. Online document available at http://hstat.nlm.nih.gov/hq/hquest/screen/textbrowse/t/1014219517607/s/64790, 2001.
Phillips, K. A., S. Fernyak, A. L. Potosky, H. H. Schauffer, and M. Egorin. "Use of Preventive Services by Managed Care Enrollees: An Updated Perspective." Health Affairs, 19(1), 2000, 102-16.
Riley, G. F., A. L. Potosky, J. D. Lubitz, and M. L. Brown. "Stage of Cancer at Diagnosis for Medicare HMO and Fee-for-Service Enrollees." American Journal of Public Health, 84(10), 1994, 1598-604.
Rolnick, S., J. L. LaFerla, D. Wehrle, E. Trygstad, and T. Okagaki. "Pap Smear Screening in a Health Maintenance Organization: 1986-1990." Preventive Medicine, 25, 1996, 156-61.
Stata Corporation. Stata Statistical Software: Release 6.0. College Station, Texas, 1999.
Sudano, J. Jr., and D. Baker. "Intermittent Lack of Health Insurance Coverage and Use of Preventive Services." American Journal of Public Health, 93(1), 2003, 130-37.
Ullman R., J. W. Hill, E. C. Scheye, and R. K. Spoeri. "Satisfaction and Choice: A View from the Plans." Health Affairs, 16(3), 1997, 209-17.
Van De Ven, W., and B. Van Praag. "The Demand for Deductibles in Private Health Insurance." Journal of Econometrics, 17, 1981, 229-52.
Wang, Y. R., and M. V. Pauly. "Preventive Care in Managed Care and Fee-for-Service Plans: Is It Cost Effective?" Managed Care Interface, 16(2), 2003, 47-50.
Weisman, C. S., and J. T. Henderson. "Managed Care and Women's Health: Access, Preventive Services, and Satisfaction." Managed Care, 11(3), 2001, 201-15.
Winter-Ebmer, R. "Benefit Duration and Unemployment Entry: A Quasi-Experiment in Austria" European Economic Review, 47(2), 2003, 259-73.
Zapka, J. G., E. Puleo, M. Vickers-Lahti, and R. Luckman. "Healthcare System Factors and Colorectal Cancer Screening." American Journal of Preventive Medicine, 23(1), 2002, 28-35.
STEPHAN F. GOHMANN*
*The author thanks Robert Ohsfeldt and the participants at the WEAI conference in Seattle, 2002, for their useful comments.
Gohmann: Professor of Economics, Department of Economics, College of Business and Public Administration, University of Louisville, Louisville, KY 40292. Phone 1-502-852-4844, Fax 1-502-852-7672 E-mail firstname.lastname@example.org
HMO: Health Maintenance Organization
MEPS: Medical Expenditure Panel Survey
MSA: Metropolitan Statistical Area
TABLE 1 Recommendations for Preventive Services Blood Cholesterol Cholesterol Pressure Male Female AAFP Age:>20 35-65 periodically 45-65 periodically periodically ACOG Yearly or 5 years 19-64 and 5 years 19-64 and appropriate 3-5 years >64 3-5 years >64 ACP Adults every 35-65 45-65 1-2 years CTFPHE Age 21-84 30-59 any visit certain cases NHLBI 2 years 5 years >20 5 years >20 USPSTF Periodically 35-65 periodically 45-65 periodically ACS ACPM NCI AMA ACR AUS Used in Every 5 years 35-65 5 years 45-65 this study 2 years >20 % of sample 0.905 0.752 0.864 receiving Mammography Mammography Pap over 50 under 50 Smear AAFP 1-2 years to 40-49 3 years 18+ age 70 counseling ACOG Yearly 1-2 yrs 40-49 Annual* 18+ ACP 1-2 years No screening 3 years 20-65 until 75 under age 50 CTFPHE Yearly until 70 No screening Annual* 18-69 under age 50 NHLBI USPSTF 1-2 years Not enough 3 years 18+ until 70 evidence ACS Yearly Annual 40+ Annual* 18+ ACPM 1-2 years Not enough Annual* 18-65 evidence NCI 1-2 years 1-2 years 40-49 Annual* 18+ AMA 1-2 years 40-49 ACR AUS Used in 2 years until 70 2 years 40-49 Annual 18-65 this study % of sample 0.800 0.686 0.662 receiving Breast Exam Breast Exam Age <40 Age [greater than or equal to]40 AAFP ACOG Yearly over 18 Annual ACP Annual CTFPHE None Annual 50-69 NHLBI USPSTF ACS 20-39 every Annual 3 years ACPM NCI AMA ACR AUS Used in 2 years Annual this study % of sample 0.679 0.836 receiving PSA AAFP Counsel age 50-65 ACOG ACP Counsel CTFPHE NHLBI USPSTF ACS Annual age [greater than or equal to]50 ACPM NCI AMA ACR Annual age [greater than or equal to]50 AUS Annual age [greater than or equal to]50 Used in Annual age [greater than or equal to]50 this study % of sample 0.519 receiving Notes: AAFP: American Academy of Family Physicians; ACOG: American College of Obstetrics and Gynecology; ACP: American College of Physicians; CTFPHE: Canadian Task Force on Periodic Health Examination; NHLBI: National Heart, Lung and Blood Institute; USPSTF: U.S. Preventive Services Task Force; ACS: American Cancer Society; ACPM: American College of Preventive Medicine; NCI: National Cancer Institute; AMA: American Medical Association; ACR: American College of Radiation; AUS: American Urological Association. *After three normal tests, they suggest less frequent testing, ACPM and CTFPHE suggest two annual tests and then every three years after that TABLE 2 Means of the Dependent Variables Full Sample No Choice Choice 0.48 (0.50) 6,622 Enrolled in HMO 0.49 (0.50) 6,622 0.40 (a) (0.49) 3,469 Blood pressure check w/in 2 0.91 (0.29) 6,311 0.89 (a) (0.32) 3,232 years > 20 Cholesterol check w/in 5 years 0.75 (0.44) 1,916 0.69 (a) (0.46) 914 male 35-65 Cholesterol check w/in 5 years 0.87 (0.34) 1,068 0.83 (a) (0.37) 521 female 45-65 Mammography w/in 2 years 50-70 0.80 (0.40) 687 0.78 (0.41) 355 Mammography w/in 2 years 40-49 0.68 (0.47) 903 0.62 (a) (0.49) 415 Pap smear w/in 1 year 18-65 0.65 (0.48) 3,183 0.61 (a) (0.49) 1,623 Breast exam w/in 2 years < 40 0.82 (0.39) 1,635 0.77 (a) (0.42) 881 Breast exam w/in 1 year 0.68 (0.47) 1,640 0.64 (a) (0.48) 806 [greater than or equal to] 40 Prostate exam w/in 1 year 0.50 (0.50) 688 0.46 (b) (0.50) 324 [greater than or equal to] 50 Choice Choice Enrolled in HMO 0.58 (0.49) 3,153 Blood pressure check w/in 2 0.93 (0.26) 3,079 years > 20 Cholesterol check w/in 5 years 0.79 (0.41) 1,002 male 35-65 Cholesterol check w/in 5 years 0.89 (0.31) 547 female 45-65 Mammography w/in 2 years 50-70 0.83 (0.38) 332 Mammography w/in 2 years 40-49 0.73 (0.45) 488 Pap smear w/in 1 year 18-65 0.69 (0.46) 1,560 Breast exam w/in 2 years < 40 0.87 (0.33) 754 Breast exam w/in 1 year 0.71 (0.45) 834 [greater than or equal to] 40 Prostate exam w/in 1 year 0.53 (0.50) 364 [greater than or equal to] 50 Notes: SDs are in parentheses. Sample size is in italics. (a) p-value < 0.01 for t-test of difference of means by choice. (b) p-value < 0.10 for t-test of difference of means by choice. TABLE 3 Choice of Plans Probit Results Variable Coefficient Number employed 0.002 (a) (0.000) Union 0.254 (a) (0.057) Professional, technical, and kindred 0.342 (c) (0.188) Managerial and administrative 0.317 (0.199) Sales workers 0.218 (0.198) Clerical and kindred workers 0.382 (b) (0.184) Craftsmen and foremen 0.181 (0.192) Operatives 0.116 (0.201) Transport operatives -0.009 (0.203) Service workers 0.255 (0.187) Laborers, not farming 0.283 (0.223) Agriculture, forestry, fisheries -0.436 (b) (0.196) Mining -0.567 (b) (0.235) Construction -0.409 (a) (0.105) Manufacturing -0.070 (0.068) Transportation, communications, utilities 0.163 (b) (0.075) Sales -0.109 (0.072) Sample size 6,625 Variable Coefficient Finance, insurance, real estate 0.295 (a) (0.081) Repair services -0.077 (0.084) Personal services -0.244 (c) (0.135) Entertainment and recreation -0.435 (a) (0.146) Age 0.054 (a) (0.010) Age squared/1,000 -0.576 (a) (0.116) Male -0.103 (b) (0.040) Married 0.029 (0.043) Hours 0.015 (a) (0.002) Wage 0.013 (a) (0.005) Black 0.056 (0.064) Hispanic -0.110 (c) (0.057) South 0.068 (0.056) West 0.247 (a) (0.059) Northeast 0.088 (0.062) Bachelor's degree 0.202 (a) (0.049) Constant -2.610 (a) (0.259) Notes: SEs are in parentheses. (a) p-value [less than or equal to] 0.01. (b) p-value [less than or equal to] 0.05. (c) p-value [less than or equal to] 0.10. TABLE 4 HMO Probit Equation for Workers with a Choice of Plans Variable Coefficient Plan choice correction -0.512 (a) (0.101) Family size 0.038 (c) (0.023) Age -0.005 (0.018) Age squared/1,000 -0.003 (0.217) Male -0.043 (0.046) Fair to poor health 0.074 (0.118) Good health 0.101 (0.077) Very good health 0.095 (c) (0.058) Fair to poor mental health 0.045 (0.185) Lives in MSA 0.352 (a) (0.116) Sample size 3,153 Variable Coefficient Receives sick pay 0.336 (a) (0.075) Log income -0.056 (c) (0.030) Hispanic 0.282 (a) (0.097) Black 0.197 (b) (0.085) Married -0.121 (c) (0.073) Bachelor's degree -0.156 (a) (0.057) South 0.195 (b) (0.082) West 0.499 (a) (0.104) Northeast 0.331 (a) (0.100) Constant 0.502 (0.492) Notes: SEs are in parentheses. (a) p-value [less than or equal to] 0.01. (b) p-value [less than or equal to] 0.05. (c) p-value [less than or equal to] 0.10. TABLE 5 Use of Preventive Treatment Probit Equations Blood Pressure Check w/in 2 Variable No Choice Choice HMO* 0.145 (c) (0.081) 0.758 (0.500) Choice selection** 0.393 (a) (0.138) -0.256 (0.192) Black 0.062 (0.144) 0.022 (0.135) Hispanic -0.107 (0.088) 0.010 (0.143) Log income 0.057 (c) (0.030) 0.046 (0.045) College degree 0.139 (b) (0.084) 0.219 (a) (0.084) Fair-poor health 0.612 (a) (0.143) 0.447 (b) (0.198) Good health 0.176 (c) (0.092) 0.186 (c) (0.095) Very good health 0.245 (a) (0.083) 0.195 (b) (0.099) Usual source of care: 0.160 (c) (0.083) 0.195 (b) (0.077) family practitioner Usual source of 0.486 (a) (0.186) 0.585 (a) (0.179) care: intern Sample size 3,232 3,079 Blood Pressure Check w/in 2 Cholesterol Check w/in 5 years Male 35-65 Variable Full No Choice HMO* 0.183 (a) (0.054) 0.058 (0.103) Choice selection** 0.462 (c) (0.266) Black 0.100 (0.109) 0.312 (0.191) Hispanic -0.069 (0.071) 0.254 (0.172) Log income 0.069 (a) (0.022) 0.060 (0.042) College degree 0.255 (a) (0.058) 0.343 (a) (0.128) Fair-poor health 0.533 (a) (0.113) 0.509 (a) (0.181) Good health 0.182 (a) (0.063) 0.256 (c) (0.142) Very good health 0.235 (a) (0.065) 0.200 (0.123) Usual source of care: 0.184 (a) (0.057) 0.105 (0.108) family practitioner Usual source of 0.529 (a) (0.130) 0.506 (b) (0.215) care: intern Sample size 6,311 913 Cholesterol Check w/in 5 years Male 35-65 Variable Choice Full HMO* 2.147 (a) (0.727) 0.105 (0.078) Choice selection** 0.078 (0.264) Black 0.187 (0.212) 0.345 (b) (0.143) Hispanic -0.442 (b) (0.199) -0.005 (0.110) Log income 0.079 (c) (0.038) 0.081 (a) (0.027) College degree 0.208 (c) (0.110) 0.381 (a) (0.078) Fair-poor health 0.254 (0.230) 0.377 (a) (0.139) Good health 0.025 (0.141) 0.149 (0.096) Very good health -0.082 (0.121) 0.103 (0.079) Usual source of care: 0.245 (b) (0.107) 0.160 (b) (0.073) family practitioner Usual source of 0.798 (a) (0.205) 0.533 (a) (0.140) care: intern Sample size 1,002 1,915 Cholesterol Check w/in 5 years Female 45-65 Variable No Choice Choice HMO* 0.159 (0.155) 0.988 (1.120) Choice selection** 0.712 (b) (0.310) -0.593 (0.438) Black 0.386 (0.388) -0.427 (0.277) Hispanic 0.116 (0.245) 0.075 (0.403) Log income 0.118 (0.074) 0.005 (0.080) College degree 0.268 (0.182) 0.009 (0.217) Fair-poor health 0.721 (b) (0.356) 0.220 (0.369) Good health 0.578 (a) (0.194) 0.134 (0.204) Very good health 0.079 (0.193) 0.154 (0.178) Usual source of care: 0.393 (b) (0.177) -0.273 (0.171) family practitioner Usual source of 0.919 (a) (0.238) 0.257 (0.258) care: intern Sample size 521 547 Cholesterol Check w/in 5 years Female 45-65 Variable Full HMO* 0.125 (0.112) Choice selection** Black 0.022 (0.202) Hispanic 0.079 (0.217) Log income 0.125 (b) (0.056) College degree 0.261 (b) (0.130) Fair-poor health 0.459 (c) (0.273) Good health 0.380 (a) (0.139) Very good health 0.157 (0.123) Usual source of care: 0.032 (0.121) family practitioner Usual source of 0.514 (a) (0.172) care: intern Sample size 1,068 Notes: SEs are in parentheses. Age, age squared, married, and regional dummy variables were also included in the model. (a) p-value [less than or equal to] 0.01. (b) p-value [less than or equal to] 0.05. (c) p-value [less than or equal to] 0.10. *When workers have a choice of plans, the HMO variable is endogenous and the predicted value of HMO is used in place of the actual value. **The selection variable is the inverse of the Mills ratio predicted from the no choice (choice) equation. TABLE 6 Use of Preventive Treatment Probit Equations Mammography w/in 2 years 40-49 Variable No Choice of Choice of Plans Plans HMO* 0.271 (c) (0.154) -0.472 (0.893) Choice selection** -0.053 (0.257) -0.537 (0.366) Black -0.171 (0.205) 0.046 (0.209) Hispanic -0.308 (0.244) 0.495 (c) (0.280) Log income 0.070 (0.072) 0.182 (b) (0.072) College degree 0.302 (c) (0.161) 0.055 (0.152) Fair-poor health -0.073 (0.310) 0.064 (0.296) Good health -0.099 (0.207) -0.040 (0.175) Very good health 0.036 (0.210) -0.128 (0.165) Usual source of care: -0.084 (0.156) 0.340 (b) (0.160) family practitioner Usual source of care: 0.156 (0.240) 0.292 (0.246) intern Sample size 415 488 Mammography w/in 2 years Mammography w/in 2 40-49 years 50-70 Variable Choice No Choice of Excluded Plans HMO* 0.169 (c) (0.096) -0.180 (0.183) Choice selection** -0.324 (0.391) Black -0.018 (0.139) 0.043 (0.288) Hispanic 0.016 (0.165) 0.505 (c) (0.284) Log income 0.129 (b) (0.052) 0.136 (0.084) College degree 0.211 (b) (0.104) 0.552 (b) (0.226) Fair-poor health 0.003 (0.207) 0.047 (0.352) Good health -0.110 (0.139) -0.178 (0.242) Very good health -0.108 (0.125) -0.156 (0.211) Usual source of care: 0.157 (0.110) 0.216 (0.179) family practitioner Usual source of care: 0.279 (0.182) 0.774 (a) (0.274) intern Sample size 903 355 Mammography w/in 2 years 50-70 Variable Choice of Choice Plans Excluded HMO* 1.490 (1.261) 0.015 (0.140) Choice selection** -0.461 (0.460) Black -0.291 (0.276) -0.045 (0.167) Hispanic -0.035 (0.413) 0.297 (0.248) Log income 0.387 (a) (0.130) 0.245 (a) (0.071) College degree -0.058 (0.224) 0.321 (b) (0.148) Fair-poor health -0.137 (0.349) 0.086 (0.232) Good health -0.398 (0.252) -0.221 (0.172) Very good health -0.293 (0.197) -0.212 (0.154) Usual source of care: -0.297 (c) (0.179) -0.045 (0.124) family practitioner Usual source of care: 0.033 (0.263) 0.355 (c) (0.184) intern Sample size 332 687 Pap Smear w/in 1 year 18-65 Variable No Choice of Choice of Plans Plans HMO* 0.153 (b) (0.068) -0.283 (0.521) Choice selection** 0.194 (0.151) -0.176 (0.201) Black 0.047 (0.100) 0.134 (0.126) Hispanic -0.110 (0.110) 0.111 (0.152) Log income 0.000 (0.038) 0.126 (a) (0.039) College degree 0.335 (a) (0.080) 0.176 (b) (0.085) Fair-poor health -0.030 (0.173) -0.242 (0.155) Good health 0.031 (0.100) -0.191 (c) (0.104) Very good health 0.076 (0.099) -0.133 (0.104) Usual source of care: -0.042 (0.087) 0.108 (0.078) family practitioner Usual source of care: 0.054 (0.121) -0.014 (0.139) intern Sample size 1,623 1,560 Pap Smear w/in 1 year 18-65 Variable Choice Excluded HMO* 0.145 (a) (0.051) Choice selection** Black 0.099 (0.080) Hispanic -0.044 (0.087) Log income 0.071 (a) (0.026) College degree 0.303 (a) (0.055) Fair-poor health -0.137 (0.104) Good health -0.084 (0.075) Very good health -0.045 (0.071) Usual source of care: 0.029 (0.056) family practitioner Usual source of care: 0.047 (0.091) intern Sample size 3,183 Notes: SEs are in parentheses. Age, age squared, married, and regional dummy variables were also included in the model. (a) p-value [less than or equal to] 0.01. (b) p-value [less than or equal to] 0.05. (c) p-value [less than or equal to] 0.10. *When workers have a choice of plans, the HMO variable is endogenous and the predicted value of HMO is used in place of the actual value. **The selection variable is the inverse of the Mills ratio predicted from the no choice (choice) equation. TABLE 7 Use of Preventive Treatment Probit Equations Breast Exam w/in 2 years < 40 Variable No Choice of Plans Choice of Plans HMO* 0.279 (a) (0.108) 0.013 (0.781) Choice Selection** 0.257 (0.293) -0.425 (0.371) Black 0.191 (0.218) 0.620 (a) (0.222) Hispanic -0.171 (0.156) -0.172 (0.192) Log income 0.065 (0.053) 0.153 (a) (0.049) College degree 0.276 (b) (0.125) 0.298 (0.191) Fair-poor health -0.104 (0.234) 0.025 (0.302) Good health 0.152 (0.153) 0.150 (0.169) Very good health -0.076 (0.132) 0.091 (0.190) Usual source of care: -0.139 (0.132) 0.256 (c) (0.155) family practitioner Usual source of care: -0.270 (0.222) 0.190 (0.291) intern Sample size 881 754 Breast Exam w/in Breast Exam w/in 1 year 2 years < 40 [greater than or equal to] 40 Variable Choice Excluded No Choice of Plans HMO* 0.243 (a) (0.091) 0.194 (c) (0.100) Choice Selection** 0.043 (0.194) Black 0.383 (b) (0.152) 0.288 (c) (0.153) Hispanic -0.218 (c) (0.115) -0.015 (0.174) Log income 0.133 (a) (0.039) 0.007 (0.063) College degree 0.368 (a) (0.100) 0.516 (a) (0.135) Fair-poor health -0.025 (0.177) 0.053 (0.199) Good health 0.151 (0.121) -0.099 (0.152) Very good health -0.052 (0.115) 0.092 (0.134) Usual source of care: -0.015 (0.104) 0.175 (0.116) family practitioner Usual source of care: -0.058 (0.164) 0.412 (b) (0.174) intern Sample size 1,635 806 Breast Exam w/in 1 year [greater than or equal to] 40 Variable Choice of Plans Choice Excluded HMO* 0.578 (0.694) 0.182 (b) (0.072) Choice Selection** -0.693 (a) (0.257) Black -0.162 (0.171) 0.134 (0.114) Hispanic -0.127 (0.223) -0.103 (0.132) Log income 0.213 (a) (0.070) 0.108 (b) (0.046) College degree 0.003 (0.108) 0.302 (a) (0.081) Fair-poor health -0.195 (0.207) -0.033 (0.144) Good health -0.098 (0.135) -0.113 (0.099) Very good health -0.053 (0.119) -0.018 (0.083) Usual source of care: -0.002 (0.110) 0.085 (0.077) family practitioner Usual source of care: 0.109 (0.162) 0.298 (a) (0.113) intern Sample size 834 1,640 Prostate exam w/in 1 year [greater than or equal to] 50 Variable No Choice of Plans Choice of Plans HMO* 0.315 (c) (0.161) 3.516 (a) (0.959) Choice Selection** 0.050 (0.269) 0.220 (0.390) Black 0.085 (0.258) -0.489 (0.307) Hispanic -0.199 (0.290) -0.729 (b) (0.309) Log income 0.002 (0.056) 0.019 (0.052) College degree 0.250 (0.184) -0.307 (c) (0.186) Fair-poor health 0.428 (0.301) -0.097 (0.305) Good health 0.278 (0.220) -0.423 (b) (0.192) Very good health -0.170 (0.233) -0.358 (b) (0.166) Usual source of care: 0.283 (c) (0.168) 0.135 (0.173) family practitioner Usual source of care: 0.833 (a) (0.280) 0.661 (a) (0.207) intern Sample size 324 364 Prostate exam w/in 1 year [greater than or equal to] 50 Variable Choice Excluded HMO* 0.237 (b) (0.103) Choice Selection** Black -0.011 (0.177) Hispanic -0.302 (0.192) Log income -0.009 (0.037) College degree -0.062 (0.129) Fair-poor health 0.157 (0.209) Good health -0.093 (0.141) Very good health -0.197 (0.134) Usual source of care: 0.202 (c) (0.110) family practitioner Usual source of care: 0.582 (a) (0.140) intern Sample size 688 Notes: SEs are in parentheses. Age, age squared, married, and regional dummy variables were also included in the model. (a) p-value [less than or equal to] 0.01. (b) p-value [less than or equal to] 0.05. (c) p-value [less than or equal to] 0.10. *When workers have a choice of plans, the HMO variable is endogenous and the predicted value of HMO is used in place of the actual value. **The selection variable is the inverse of the Mills ratio predicted from the no choice (choice) equation. APPENDIX TABLE A1 Means of the Independent Variables Full Sample No Choice Continuous variables Age 40.13 (12.19) 39.13 (a) (13.15) Log income 9.99 (1.14) 9.77 (a) (1.18) Dummy variables Black = 1 0.12 (0.33) 0.11 (a) (0.31) Hispanic = 1 0.13 (0.34) 0.15 (a) (0.35) Married = 1 0.64 (0.48) 0.61 (a) (0.49) Bachelor's degree = 1 0.35 (0.48) 0.27 (a) (0.44) South = 1 0.35 (0.48) 0.36 (0.48) West = 1 0.21 (0.41) 0.20 (c) (0.40) Northeast = 1 0.20 (0.40) 0.19 (b) (0.39) Fair to poor health = 1 0.07 (0.26) 0.08 (a) (0.27) Good health = 1 0.24 (0.42) 0.24 (0.43) Very good health = 1 0.35 (0.48) 0.34 (c) (0.47) Usual source of care: family 0.32 (0.47) 0.31 (0.46) practitioner = 1 Usual source of care: 0.09 (0.29) 0.08 (a) (0.28) intern = 1 N 6,622 3,469 Choice Continuous variables Age 41.23 (10.94) Log income 10.23 (1.03) Dummy variables Black = 1 0.14 (0.34) Hispanic = 1 0.11 (0.31) Married = 1 0.68 (0.47) Bachelor's degree = 1 0.44 (0.50) South = 1 0.34 (0.47) West = 1 0.22 (0.41) Northeast = 1 0.21 (0.41) Fair to poor health = 1 0.06 (0.24) Good health = 1 0.23 (0.42) Very good health = 1 0.36 (0.48) Usual source of care: family 0.33 (0.47) practitioner = 1 Usual source of care: 0.11 (0.31) intern = 1 N 3,153 Notes: Omitted categories are white for race, Midwest for region, excellent health for the health variables, and all other source of care or no usual source of care for the type of physician. (a) p-value < 0.01 for t-test of difference of means by choice. (b) p-value < 0.05 for t-test of difference of means by choice. (c) p-value < 0.10 for t-test of difference of means by choice.
|Printer friendly Cite/link Email Feedback|
|Author:||Gohmann, Stephan F.|
|Publication:||Contemporary Economic Policy|
|Date:||Oct 1, 2005|
|Previous Article:||Welfare reform, earnings, and incomes: new evidence from the Survey of Program Dynamics.|
|Next Article:||Health insurance coverage and reemployment outcomes among older displaced workers.|