Medicaid Primary Care Physician Fees and the Use of Preventive Services among Medicaid Enrollees

Abstract

Objective

The Patient Protection and Affordable Care Act (ACA) increases Medicaid physician fees for preventive care up to Medicare rates for 2013 and 2014. The purpose of this paper was to model the relationship between Medicaid preventive care payment rates and the use of U.S. Preventive Services Task Force (USPSTF)–recommended preventive care use among Medicaid enrollees.

Data Sources/Study Session

We used data from the 2003 and 2008 Medical Expenditure Panel Survey (MEPS), a national probability sample of the U.S. civilian, noninstitutionalized population, linked to Kaiser state Medicaid benefits data, including the state Medicaid-to-Medicare physician fee ratio in 2003 and 2008.

Study Design

Probit models were used to estimate the probability that eligible individuals received one of five USPSF-recommended preventive services. A difference-in-difference model was used to separate out the effect of changes in the Medicaid payment rate and other factors.

Data Collection/Extraction Methods

Data were linked using state identifiers.

Principal Findings

Although Medicaid enrollees had a lower rate of use of the five preventive services in univariate analysis, neither Medicaid enrollment nor changes in Medicaid payment rates had statistically significant effects on meeting screening recommendations for the five screenings. The results were robust to a number of different sensitivity tests. Individual and state characteristics were significant.

Conclusions

Our results suggest that although temporary changes in primary care provider payments for preventive services for Medicaid enrollees may have other desirable effects, they are unlikely to substantially increase the use of these selected USPSTF-recommended preventive care services among Medicaid enrollees.

One of the goals of the Patient Protection and Affordable Care Act (ACA) was to increase the provision of preventive care, particularly in low-income populations (Doherty 2010). This is based on the belief that American society is being “penny-wise and pound-foolish,” creating a situation where low-income individuals cannot afford to obtain relatively inexpensive preventive care, with the public then later paying for more expensive acute care. Preventive care can lead to earlier diagnoses of illnesses, particularly cancer, and can help prevent potentially fatal illnesses such as heart disease and stroke. Prevention is important for individuals with chronic illnesses, such as diabetes and hypertension.

Although previous research has found that insurance is a key factor in the provision of preventive care (Lurie et al. 1984; Manning et al. 1987; Beckles et al. 1998; Schiff et al. 1998; Ayanian et al. 2000, 2003; Broyles, Narine, and Brandt 2002; Adams et al. 2003; Clancy et al. 2003; Davidoff et al. 2003), state Medicaid programs generally are not required to cover preventive services, and federal Medicaid law does not define preventive care (United States Government Accountability Office 2009). Overall, the use of preventive care among Medicaid enrollees is lower than for commercially insured individuals, although higher than for the uninsured (Mortensen and Atherly 2012). The reason for the discrepancy between Medicaid enrollees and the privately insured is unclear. It could be that Medicaid enrollees place a lower value on preventive care than commercially insured individuals. Alternatively, it could be that Medicaid enrollees have poorer access to primary care providers due to program characteristics such as low provider payment rates.

Currently, there are substantial differences between Medicaid and Medicare physician fees in many states. On average, Medicaid fees are 72 percent of Medicare fees overall, but only 66 percent for primary care services. The Medicaid payment rate was at or above Medicare fees in 11 states and below in the other 39 (Zuckerman, Williams, and Stockley 2009). This ranges from a low of 37 percent in New Jersey to a high of 143 percent in Wyoming for overall fees and from a low of 36 percent in New York and Rhode Island for preventive care to a high of 140 percent in Alaska (with Alaska being an outlier). The ACA attempts to increase the use of preventive (and other) services by expanding eligibility for Medicaid, increasing payments to providers, encouraging public programs to cover preventive services, and limiting the use of cost sharing for preventive services. The ACA increases Medicaid payment rates for primary care providers for preventive services and immunizations from the current level to Medicare levels for 2013 and 2014.

A key factor in access to care is primary care physicians' (PCPs) willingness to see Medicaid patients. PCPs who serve Medicaid beneficiaries most actively are also the most willing to accept new Medicaid patients (Kaiser Family Foundation 2011). However, they also face capacity constraints to serve more of this population. To increase access, it is necessary to convince more doctors to accept Medicaid patients. Nationally, 42 percent of primary care physicians are willing to accept new Medicaid patients (in 2008), which is considerably lower than for Medicare (61 percent) or privately insured (84 percent) (Cunningham 2011).

Many factors influence physicians' decisions regarding whether or not to accept a Medicaid patient. For example, delays in reimbursement and high administrative burden can offset the effects of high Medicaid fees, thereby lowering participation in high payment states to levels that are closer to those in states with relatively low payment rates (Cunningham and O'Malley 2009). Indeed, high Medicaid acceptance rates by physicians in a community have been found to be more important than overall fee levels for Medicaid enrollees' access to medical care (Cunningham and Nichols 2005). Although high fee levels do increase the probability that individual physicians will accept Medicaid patients, they do not necessarily lead to high levels of physician Medicaid acceptance in an area. Other factors, such as physician practice structure, health system characteristics, and other community factors, are more important.

Yet there is support in the literature for the notion that higher payment rates will increase access; higher Medicaid-to-Medicare fee ratios increase both the fraction of Medicaid patients seen by physicians and the number of private physicians who see Medicaid patients (Shen and Zuckerman 2005). Physicians in states with the lowest Medicaid fees are less willing to accept new Medicaid patients (Long, Settle, and Stuart 1986). Higher payments have been found to increase the probability of having a usual source of care and the probability of having at least one visit to a doctor for Medicaid adults, and produce more positive assessments of the health care received by adults and children (Kenney et al. 2011).

Although the studies cited above do show a relationship between physician access and overall Medicaid payment rates, they do not look at service-specific payment rates. Indeed, although Kenney et al. (2011) found that higher payments had positive effects on access, the study also found no effect on the probability of receiving preventive care or having unmet health needs. It may be that lower physician reimbursement simply leads to a shift in location where care is delivered (Decker 2007; Cunningham 2011). There have also been several other papers finding no relationship between Medicaid payment levels and receipt of preventive care (Cunningham and Nichols 2005; Cunningham and O'Malley 2009). Decker (2009) found that payment level cuts shift care away from physician offices to emergency and outpatient departments. However, Decker (2009) also found that the baseline number of visits was far higher for Medicaid recipients than for the privately insured—potentially due to lower copayments—so that lower payment levels do not clearly reduce access to levels below the privately insured, depending on what type of services are foregone due to the payment cuts.

In this paper, we focus on the effect of state-level Medicaid policies in effect before the ACA payment increase on the use of U.S. Preventive Services Task Force (USPSTF) recommended A and B preventive services among Medicaid enrollees in the 10 largest states. USPSTF is an independent panel of non-federal experts in prevention and evidence-based medicine that conducts scientific evidence reviews of a broad range of clinical preventive health care services to develop recommendations for primary care clinicians and health systems. Services that are recommended grade A services have a high certainty that the net benefit is substantial, and B services have high certainty that the net benefit is moderate or moderate certainty that the net benefit is moderate to substantial (USPSTF, 2012). We empirically estimate the relationship between preventive care services payment changes on the use of USPSTF recommended A and B preventive services. This paper adds to the existing literature on the subject by looking at the effect of payment changes on the delivery of specific services. It is hoped that Medicaid primary care payment increases will lead to higher rates of delivery of preventive care; the indirect evidence in literature to date is mixed on whether this strategy is likely to succeed. By directly examining the link between payment rates and preventive care, we provide evidence on the viability of the strategy.

Methods

The underlying theory behind the Medicaid payment increases is that Medicaid recipients are not provided preventive services due to low payment rates. The precise mechanism is unclear; it could be that higher payment rates induce more physicians to accept Medicaid, thus increasing access and thereby leading to higher screening rate. Alternatively, it could be that physicians who could, but do not, perform preventive screenings among their Medicaid patients at lower payment rates would do so at higher payment rates. In either case, it is hoped that an increase in provider payments for primary care services will lead to an increase in the quantity of primary care services supplied. The approach is supported by both standard economic theory and by a substantial literature finding a relationship between Medicaid payment rates and outcome variables such as physician participation in Medicaid (Gruber, Adams, and Newhouse, 1997; Zuckerman et al. 2004; Decker 2009).

In our empirical model, the main explanatory variable is Medicaid payment rate (discussed below), with control variables included for individual attributes related to the individual's need and likelihood of seeking preventive care, plus year and state fixed effects to control for time-invariant state and temporal changes in the use of preventive care.

Data

Our primary data source is the Medical Expenditure Panel Survey (MEPS) Household Component, a nationally representative sample of families and individuals across the nation (Cohen et al. 1996). The survey is sponsored by the Agency for Health Care Research and Quality and the National Center for Health Statistics and contains data on a variety of health status measures, health insurance coverage, and health care usage, including the use of preventive services. The survey employs a panel design, with several rounds of interviews generating data for two full calendar years. MEPS provides variables relating to health status, chronic diseases, income, race, gender, and education primary plus our dependent variables—the use of different types of preventive services. The Medicaid physician fees and the Medicaid-to-Medicare fee index are published by the Kaiser Family Foundation, described in detail in Zuckerman, Williams, and Stockley (2009). Creation of the analytic variables is described below. We use data from two different years—2003 and 2008. Total sample size varies depending on the particular condition (described below).

Analytic Approach

The target population for our analysis is Medicaid enrollees who are eligible for the different preventive services. Because individuals were not assigned randomly to states (and thus to primary care payment rates—our variable of interest), we face the possibility of omitted variable bias—unobserved factors correlated with both the Medicaid primary care payment rate and the subsequent probability that a person in the target population received preventive care. Our main strategy to address this issue is to use the well-known difference-in-differences model.

The standard difference-in-differences model utilizes two dichotomous variables—typical, a pre-post time variable and a variable indicating the treatment and control groups. In our model, we follow previous literature and set the “treatment” group as enrollment in Medicaid, while the control group is individuals with private insurance (Decker 2009, 2011). The pre year is 2003 and the post is 2008. However, our intervention is a change in a continuous variable—the Medicaid primary care payment rate. Thus, our difference-in-differences model is somewhat akin to dose/response models in pharmaceutical studies. Our approach is similar to that used by Decker (2011).

In our results, the coefficient of the “Medicaid” variable represents differences in rates of use of preventive services associated with Medicaid enrollment. This coefficient thus captures differences between Medicaid enrollees and privately insured individuals in underlying propensity to use preventive services across the two time periods. The “Medicaid Payment Rate” variable controls for differences based on overall Medicaid payment levels (described below). The payment effect is represented by the coefficient on the interaction term (Medicaid Enrollment*Payment Rate). That coefficient is the “difference-in-differences” term.

We estimate the difference-in-differences coefficients with a probit model. We report the marginal effects of each of our explanatory variables on the dependent variable. Although the marginal effects represented by many types of interaction terms in nonlinear models (like our probit equation) vary with the values of the explanatory variables in the model (Ai and Norton 2003), Puhani (2008) has shown that the difference-in-difference model is a special case and the treatment effect is given by the coefficient on the interaction term.

The control variables represent factors that have been found to be associated with both the probability of receiving preventive care and the probability of being enrolled in Medicaid. We include variables in our model for the following:

• Health (self-rated and presence of chronic illness)

• Marital status

• Insurance (Medicaid, Medicare)

• Employment characteristics (employed full time, part time, self-employed)

• Individual sociodemographic characteristics (income, education, race, marital status and age)

• A fixed effect for our initial year (2003)

We also include a state fixed effect to capture time-invariant differences in state-level variables associated with the use of preventive services. Our model controls for the complex survey design of the MEPS by using longitudinal strata and primary sampling unit identifiers and survey weights. The estimated coefficients are representative of the U.S. civilian, noninstitutionalized, adult population.

The key independent variable—the Medicaid primary care payment rate—represents a ratio of the weighted average of state Medicaid payments to Medicare payments for the 12 most frequent office-based primary care services, such as office visit, new patient, 30 minutes; office visit, established patient, 15-minute office visit, established patient, 25 minutes (Decker 2009). Although this ratio does not include comprehensive inventory of all Medicaid payment rates, Medicaid payment rates for similar services within a state tend to be highly correlated (Decker 2009). The use of the Medicaid-to-Medicare payment ratio is relatively standard in the existing literature, even in examining differences in utilization between Medicaid and private insurance utilization (e.g., Decker 2009; Zuckerman et al. 2004; Cunningham and Nichols 2005; Cunningham and O'Malley 2009; Gray, 2001). There are a number of reasons for the use of the ratio, rather than a ratio of Medicaid to private insurance payments. Pragmatically, private insurance payment rates are often not available and are potentially endogenous due to reverse causation. One advantage of the Medicaid-to-Medicare ratio is that payment rate changes for Medicaid are often made in response to budget considerations, rendering them relatively exogenous to our model (e.g., Kaiser Health News 2012). However, Medicare payment rates are highly correlated with private insurance rates (Hogan, 2004).

Furthermore, because this mechanism—increasing Medicaid payments relative to Medicare—is the one envisioned and operationalized by the ACA to increase the use of preventive care in the Medicaid population, it is logical to examine the effect of increases in Medicaid payments relative to Medicare payments.

Defining Preventive Services

Although there are a wide variety of preventive services recommended for Americans by various health plans, government agencies, and physicians groups, we are focusing this analysis on grade “A” and “B” preventive services recommended by the USPSTF. These include tests for colorectal, cervical and breast cancer, hypertension, and high cholesterol.

For our analysis, we created five separate binary indicators to reflect whether the USPSTF recommendation was met. Our measures are all drawn from the MEPS interview. Because the sample varies depending on the particular preventive procedure, the sample size varies accordingly. It ranges from a low of 7,468 for breast cancer screening to a high of 24,446 for blood pressure screening

The cervical cancer screening measure (also known as a “Pap smear”) is recommended every 3 years for women aged 21– 65 with an intact uterus. Mammography screening (breast cancer) is recommended for women aged 40–74 every 2 years in the ACA. Colorectal cancer screening is recommended for men and women between ages 50 and 75. The guidelines for colorectal cancer screening include: fecal occult blood test every year, sigmoidoscopy every 5 years combined with a fecal occult blood test, or colonoscopy every 10 years.

USPSTF recommends checking blood pressure for adults age 18 and over every 1–2 years. Screening for lipid disorders in adults is strongly recommended every 5 years for men over age 35 regardless of health history. There is a B recommendation for men aged 20–35 if they are at increased risk for coronary heart disease (CHD, defined as the presence of diabetes, previous personal history of CHD or noncoronary atherosclerosis, a family history of cardiovascular disease before age 50 in male relatives or age 60 in female relatives, tobacco use, hypertension, and obesity [BMI ≥ 30]). For women, the screen is strongly recommended every 5 years for women over age 45 if they are at increased risk for coronary heart disease and for women aged 20–45 if they are at increased risk.

For lipid disorders, we are unable to determine accurately an individual's increased risk primarily due to a lack of family history in the data; thus, we include all men over age 35 and all women over age 45 as having met the guideline if they had a cholesterol test in the past 5 years. This is consistent with the recommendation during much of the study period which was to screen all women over 45 years of age.

Our measure of meeting the recommendation for colorectal cancer screening is an indicator for if the individual had a sigmoidoscopy or colonoscopy within the past 5 years, or if the individual had a fecal occult blood test in the last year. This is because the 2003–2008 MEPS data do not distinguish between sigmoidoscopy or colonoscopy (this is distinguished in the 2009 data), and the question is truncated at 5 or more years. Our blood pressure measure indicates whether these adults have had their blood pressure checked in the last 2 years, which is consistent with other guidelines.

Results

Overall, screening rates for four of the five conditions are quite high and were relatively unchanged between 2003 and 2008 (Table 1). Overall, 91 percent of those eligible received cervical cancer screening in both 2003 and 2008, as did 80 percent of those eligible for breast cancer screenings and 94 percent of those eligible for blood pressure screenings. The screening rates for cholesterol and colorectal cancer increased slightly—from 91 to 93 percent and 49 to 55 percent, respectively.

Table 1.

Screening Rates, 2003–2008 for Eligible Individuals

	Cervical Cancer Screening (Pap Smear)		Breast Cancer Screening		Cholesterol Screening		Blood Pressure Screening		Colorectal Cancer Screening
	2003 (%)	2008 (%)	2003 (%)	2008 (%)	2003 (%)	2008 (%)	2003 (%)	2008 (%)	2003 (%)	2008 (%)
Overall	91	91	80	80	91	93	94	94	49	55
Private insurance	92	92	81	81	89	91	93	93	46	52
Medicaid	87	86	69	60	85	91	93	96	27	30
Medicare			80	79	96	96	97	97	53	59
Employed full time	93	93	82	81	88	90	92	93	43	50
Employed part time	91	92	81	81	93	94	95	94	51	54
Not employed	88	87	78	77	95	96	96	96	52	58
Self-employed	92	87	76	80	90	91	93	92	45	54
Self-rated health: Excellent	91	92	81	79	87	89	91	91	47	51
Self-rated health: Very good	92	92	81	81	90	93	94	94	49	56
Self-rated health: Good	92	89	80	81	92	94	95	95	49	57
Self-rated health: Fair	90	90	76	75	95	96	97	97	47	55
Self-rated health: Poor	92	86	75	71	97	95	99	99	55	53
No chronic illness	91	91	76	76	83	85	90	90	38	46
Any chronic illness	93	92	83	82	96	96	98	98	53	57
Income: Less than 100% of FPL	90	86	71	73	89	93	94	95	41	49
Income: 100–200% of FPL	88	90	71	72	91	92	92	93	42	49
Income: 200–400% of FPL	91	89	78	74	89	90	93	93	49	51
Income: Over 400% of FPL	93	93	84	85	92	94	95	95	52	59
Race: Black non-Hispanic	92	94	82	83	92	94	96	95	44	56
Race: Other non-White	85	85	74	75	92	90	91	83	32	44
Race: Hispanic	90	90	73	78	87	91	89	90	34	42
Race White	92	91	80	80	91	93	94	94	51	56
Education: Less than high school	82	85	69	68	90	92	92	93	39	47
Education: High school grad	90	88	80	77	90	91	94	95	48	50
Education: Some college	90	89	81	79	90	93	93	95	51	58
Education: College graduate	94	94	84	85	93	95	95	94	54	60
Married	94	93%	81	82	92	93	94	94	50	58
Widowed/Divorced/Separated	92	87	77	76	91	94	95	96	45	50
Never married	85	87	74	74	86	89	90	90	48	44
Age: Under 21	67	65					89	88
Age: 21–34	92	92					90	90
Age: 35–39	93	93	70	70	73	79	92	93
Age: 40–44	93	94	81	77	79	86	93	94
Age: 45–49	93	89	84	84	88	91	95	94
Age: 50–64	91	90	79	80	94	93	96	95	45	51
Age: 65–74					95	95	97	96	55	62
Age: Over 74					96	97	98	98	50	55
Sample size	4,358	4,077	3,899	3,569	8,056	7,591	12,690	11,756	5,825	5,623

Medicaid recipients generally had lower screening rates than either privately insured individuals or Medicare beneficiaries. In 2008, Medicaid recipients had lower screening rates for cervical cancer (86 percent vs. 92 percent for private insurance), breast cancer (60 percent vs. 81 percent for private insurance and 79 percent for Medicare), cholesterol screening (91 percent [Medicaid and private] and 96 percent [Medicare]), and colorectal cancer screening (30 percent vs. 52 percent [private] vs. 59 percent [Medicare]). However, with the exception of colorectal cancer screening, none of the screening rates significantly changed 2003–2008 within insurance category. Colorectal cancer screening increased for all payers, including Medicaid (27–30 percent), private insurance (46–52 percent), and Medicare (53–59 percent).

In the multivariate analysis (Tables 2–6), the Medicaid payment rate is negative in all four models, with the effect reaching statistical significance in three of the model (breast cancer, cholesterol, and blood pressure screening). In contrast, Medicaid enrollment is insignificantly related to the use of screening in all five models. This suggests that, controlling for other factors, Medicaid enrollees are equally likely to receive preventive services as individuals in the reference group (privately insured) and that states with higher Medicaid primary care payment rates had lower use of preventive services. For the models including Medicare beneficiaries (all but cervical cancer), Medicare enrollment is also insignificantly related to the probability of receiving preventive services.

Table 2.

Marginal Effect of Medicaid Payment Rates on Cervical Cancer Screening (“Pap Smear”)

	Coefficient	Std Err	t Statistic	p-value
Medicaid payment rate	−0.048	0.0377	−1.28	.201
Medicaid enrollee	0.009	0.0238	0.39	.693
Payment × Enrollee	−0.012	0.0383	−0.31	.757
Year 2008	−0.005	0.0068	−0.69	.494
Age: 21–34	0.073	0.0334	2.19	.029
Age: 35–39	0.068	0.0324	2.08	.038
Age: 40–44	0.070	0.0332	2.10	.037
Age: 45–49	0.047	0.0288	1.63	.103
Age: 50–64	0.042	0.0275	1.53	.126
Race: Black	0.032	0.0145	2.17	.030
Race: Other non-White	−0.058	0.0241	−2.39	.017
Race: Hispanic	0.008	0.0088	0.86	.388
Income: 100–200% of FPL	−0.012	0.0110	−1.10	.273
Income: 200–400% of FPL	−0.020	0.0135	−1.49	.138
Income: Over 400% of FPL	−0.002	0.0107	−0.20	.840
Employed part time	−0.007	0.0074	−0.90	.369
Not employed	−0.030	0.0138	−2.16	.031
Self-employed	−0.027	0.0168	−1.61	.108
Widowed/Divorced/Separated	−0.026	0.0134	−1.95	.052
Never married	−0.064	0.0253	−2.51	.012
Self-rated health: Very good	0.005	0.0076	0.67	.506
Self-rated health: Good	−0.002	0.0072	−0.25	.805
Self-rated health: Fair	0.003	0.0111	0.31	.755
Self-rated health: Poor	0.004	0.0153	0.28	.779
Any chronic illness	0.021	0.0103	2.09	.037
Education: High school grad	0.035	0.0171	2.07	.039
Education: Some college	0.047	0.0208	2.25	.025
Education: College graduate	0.068	0.0277	2.46	.014
Alabama	−0.001	0.0367	−0.02	.987
Arizona	0.015	0.0337	0.45	.650
California	0.019	0.0136	1.43	.154
Colorado	0.034	0.0278	1.22	.223
Connecticut	0.059	0.0148	3.99	.000
Florida	0.009	0.0165	0.53	.598
Georgia	0.031	0.0221	1.43	.154
Illinois	0.005	0.0202	0.27	.789
Indiana	0.018	0.0211	0.85	.398
Kentucky	0.007	0.0224	0.31	.757
Louisiana	−0.006	0.0335	−0.18	.857
Maryland	0.040	0.0226	1.75	.080
Massachusetts	0.044	0.0197	2.25	.025
Michigan	0.039	0.0173	2.25	.025
Minnesota	0.031	0.0201	1.56	.119
Missouri	−0.001	0.0223	−0.02	.980
New Jersey	−0.004	0.0165	−0.22	.826
North Carolina	0.046	0.0211	2.16	.031
Ohio	0.035	0.0212	1.67	.096
Oklahoma	0.032	0.0255	1.25	.213
Oregon	0.049	0.0268	1.82	.069
Pennsylvania	0.002	0.0182	0.10	.918
South Carolina	0.054	0.0224	2.39	.017
Tennessee	−0.016	0.0222	−0.71	.480
Texas	−0.001	0.0209	−0.05	.960
Virginia	0.026	0.0218	1.21	.225
Washington	0.043	0.0240	1.81	.071
Wisconsin	0.033	0.0199	1.64	.102

Table 6.

Marginal Effect of Medicaid Payment Rates on Colorectal Cancer Screening

	Coefficient	Std Err	t Statistic	p-value
Medicaid payment rate	−0.215	0.1395	−1.54	.123
Medicaid enrollee	−0.081	0.1147	−0.71	.479
Payment × Enrollee	−0.152	0.1850	−0.82	.410
Medicare enrollee	−0.001	0.0428	−0.03	.975
Year 2008	0.063	0.0131	4.84	.000
Age: 65–74	0.074	0.0413	1.78	.075
Age: Over 74	0.017	0.0436	0.39	.698
Race: Black	0.009	0.0193	0.44	.658
Race: Other non-White	−0.147	0.0271	−5.41	.000
Race: Hispanic	−0.094	0.0209	−4.51	.000
Income: 100–200% of FPL	−0.020	0.0238	−0.83	.406
Income: 200–400% of FPL	0.042	0.0244	1.73	.084
Income: Over 400% of FPL	0.104	0.0243	4.29	.000
Employed part time	0.052	0.0235	2.21	.027
Not employed	0.101	0.0186	5.40	.000
Self-employed	0.008	0.0229	0.36	.718
Widowed/Divorced/Separated	−0.045	0.0133	−3.35	.001
Never married	−0.060	0.0263	−2.27	.023
Self-rated health: Very good	0.033	0.0177	1.89	.060
Self-rated health: Good	0.046	0.0180	2.54	.011
Self-rated health: Fair	0.033	0.0212	1.55	.122
Self-rated health: Poor	0.083	0.0268	3.11	.002
Any chronic illness	0.128	0.0142	9.04	.000
Education: High school grad	0.036	0.0189	1.93	.054
Education: Some college	0.099	0.0239	4.15	.000
Education: College graduate	0.127	0.0201	6.31	.000
Gender: Female	−0.022	0.0106	−2.08	.038
Alabama	−0.001	0.0788	−0.01	.989
Arizona	0.108	0.0993	1.09	.276
California	−0.004	0.0366	−0.11	.914
Colorado	0.153	0.0658	2.32	.021
Connecticut	0.107	0.0745	1.44	.150
Florida	0.081	0.0433	1.86	.064
Georgia	0.060	0.0712	0.84	.403
Illinois	−0.068	0.0509	−1.33	.185
Indiana	−0.079	0.0593	−1.32	.186
Kentucky	0.047	0.0759	0.62	.537
Louisiana	−0.024	0.0909	−0.27	.790
Maryland	0.133	0.0745	1.79	.075
Massachusetts	0.151	0.0729	2.07	.039
Michigan	0.065	0.0539	1.21	.229
Minnesota	0.033	0.0604	0.55	.584
Missouri	−0.053	0.0487	−1.09	.276
New Jersey	−0.006	0.0381	−0.17	.866
North Carolina	0.109	0.0895	1.22	.224
Ohio	−0.008	0.0616	−0.13	.897
Oklahoma	−0.085	0.0891	−0.96	.339
Oregon	0.012	0.0752	0.15	.877
Pennsylvania	−0.048	0.0536	−0.90	.367
South Carolina	0.111	0.0737	1.51	.131
Tennessee	−0.153	0.0809	−1.89	.060
Texas	−0.029	0.0527	−0.56	.576
Virginia	0.078	0.0719	1.09	.278
Washington	0.148	0.0778	1.90	.058
Wisconsin	0.043	0.0627	0.68	.494

Table 3.

Marginal Effect of Medicaid Payment Rates on Breast Cancer Screening (“Mammogram”)

	Coefficient	Std Err	t Statistic	p-value
Medicaid payment rate	−0.134	0.0542	−2.47	.014
Medicaid enrollee	−0.077	0.0580	−1.33	.185
Payment × Enrollee	0.064	0.0931	0.69	.494
Medicare enrollee	0.003	0.0377	0.07	.947
Year 2008	−0.002	0.0083	−0.27	.790
Age: 45–49	0.065	0.0183	3.55	.000
Age: 50–64	0.094	0.0224	4.22	.000
Age: 65–74	0.086	0.0407	2.12	.035
Race: Black	0.043	0.0138	3.09	.002
Race: Other non-White	−0.033	0.0195	−1.66	.097
Race: Hispanic	0.016	0.0135	1.15	.252
Income: 100–200% of FPL	−0.027	0.0183	−1.46	.145
Income: 200–400% of FPL	−0.014	0.0180	−0.80	.426
Income: Over 400% of FPL	0.034	0.0179	1.90	.058
Employed part time	−0.004	0.0134	−0.27	.786
Not employed	−0.019	0.0120	−1.60	.110
Self-employed	−0.031	0.0196	−1.58	.116
Widowed/Divorced/Separated	−0.033	0.0128	−2.58	.010
Never married	−0.054	0.0208	−2.61	.009
Self-rated health: Very good	0.002	0.0108	0.15	.878
Self-rated health: Good	0.009	0.0110	0.85	.394
Self-rated health: Fair	−0.021	0.0158	−1.33	.185
Self-rated health: Poor	−0.027	0.0283	−0.96	.340
Any chronic illness	0.060	0.0161	3.74	.000
Education: High school grad	0.061	0.0199	3.05	.002
Education: Some college	0.068	0.0240	2.83	.005
Education: College graduate	0.098	0.0265	3.68	.000
Alabama	−0.012	0.0485	−0.25	.803
Arizona	−0.021	0.0635	−0.33	.740
California	−0.016	0.0220	−0.72	.471
Colorado	−0.013	0.0478	−0.27	.789
Connecticut	0.095	0.0291	3.26	.001
Florida	0.001	0.0239	0.03	.977
Georgia	0.007	0.0411	0.17	.862
Illinois	−0.035	0.0290	−1.20	.229
Indiana	−0.035	0.0431	−0.82	.414
Kentucky	0.034	0.0352	0.97	.331
Louisiana	−0.008	0.0514	−0.15	.877
Maryland	0.012	0.0412	0.29	.771
Massachusetts	0.049	0.0351	1.40	.163
Michigan	0.029	0.0265	1.10	.273
Minnesota	0.028	0.0317	0.88	.378
Missouri	−0.089	0.0399	−2.23	.026
New Jersey	−0.025	0.0281	−0.90	.369
North Carolina	0.025	0.0484	0.52	.600
Ohio	0.000	0.0318	0.01	.995
Oklahoma	−0.006	0.0481	−0.13	.895
Oregon	−0.003	0.0418	−0.07	.944
Pennsylvania	−0.050	0.0345	−1.45	.149
South Carolina	0.038	0.0493	0.76	.445
Tennessee	−0.093	0.0300	−3.10	.002
Texas	−0.050	0.0371	−1.36	.175
Virginia	−0.050	0.0531	−0.94	.350
Washington	−0.011	0.0503	−0.23	.819
Wisconsin	−0.034	0.0412	−0.82	.414

Table 4.

Marginal Effect of Medicaid Payment Rates on Cholesterol Screening

	Coefficient	Std Err	t Statistic	p-value
Medicaid payment rate	−0.055	0.0091	−5.97	.000
Medicaid enrollee	−0.004	0.0171	−0.22	.822
Payment × Enrollee	0.001	0.0260	0.04	.969
Medicare enrollee	−0.015	0.0168	−0.91	.361
Year 2008	0.007	0.0032	2.24	.026
Age: 40–44	0.024	0.0131	1.83	.068
Age: 45–49	0.043	0.0174	2.49	.013
Age: 50–64	0.055	0.0212	2.60	.009
Age: 65–74	0.072	0.0313	2.30	.022
Age: Over 74	0.077	0.0326	2.38	.018
Race: Black	0.012	0.0054	2.22	.027
Race: Other non-White	−0.005	0.0077	−0.65	.513
Race: Hispanic	0.006	0.0046	1.35	.177
Income: 100–200% of FPL	0.009	0.0090	0.96	.336
Income: 200–400% of FPL	0.018	0.0102	1.74	.082
Income: Over 400% of FPL	0.035	0.0144	2.44	.015
Employed part time	0.005	0.0062	0.84	.402
Not employed	0.011	0.0060	1.80	.072
Self employed	0.002	0.0050	0.41	.679
Widowed/Divorced/Separated	−0.012	0.0055	−2.13	.034
Never married	−0.011	0.0065	−1.75	.082
Self-rated health: Very good	0.011	0.0054	1.97	.049
Self-rated health: Good	0.016	0.0068	2.40	.017
Self-rated health: Fair	0.023	0.0086	2.65	.008
Self-rated health: Poor	0.026	0.0115	2.29	.023
Any chronic illness	0.050	0.0169	2.94	.003
Education: High school grad	0.014	0.0078	1.74	.082
Education: Some college	0.020	0.0098	2.09	.037
Education: College graduate	0.035	0.0142	2.43	.015
Gender: Female	0.014	0.0060	2.31	.021
Alabama	−0.004	0.0171	−0.25	.800
Arizona	0.019	0.0153	1.25	.211
California	−0.004	0.0109	−0.33	.740
Colorado	−0.005	0.0170	−0.31	.759
Connecticut	0.025	0.0113	2.24	.025
Florida	0.003	0.0114	0.23	.820
Georgia	0.005	0.0145	0.32	.752
Illinois	−0.016	0.0140	−1.14	.255
Indiana	−0.033	0.0201	−1.62	.106
Kentucky	−0.005	0.0154	−0.30	.762
Louisiana	−0.010	0.0186	−0.55	.581
Maryland	0.036	0.0145	2.48	.014
Massachusetts	0.031	0.0188	1.62	.105
Michigan	−0.009	0.0135	−0.63	.529
Minnesota	−0.010	0.0134	−0.71	.477
Missouri	−0.023	0.0163	−1.41	.159
New Jersey	−0.007	0.0120	−0.58	.562
North Carolina	0.031	0.0125	2.44	.015
Ohio	−0.005	0.0133	−0.40	.689
Oklahoma	−0.001	0.0192	−0.05	.962
Oregon	0.001	0.0147	0.10	.924
Pennsylvania	−0.019	0.0138	−1.37	.171
South Carolina	0.021	0.0131	1.62	.105
Tennessee	−0.033	0.0150	−2.20	.029
Texas	−0.008	0.0133	−0.63	.531
Virginia	0.010	0.0152	0.64	.521
Washington	0.005	0.0159	0.34	.732
Wisconsin	−0.017	0.0169	−1.03	.305

Table 5.

Marginal Effect of Medicaid Payment Rates on Blood Pressure Screening

	Coefficient	Std Err	t Statistic	p-value
Medicaid payment rate	−0.040	0.0029	−13.83	.000
Medicaid enrollee	0.005	0.0084	0.64	.521
Payment × Enrollee	0.001	0.0127	0.05	.961
Medicare enrollee	0.009	0.0104	0.89	.376
Year 2008	−0.001	0.0017	−0.31	.759
Age: 21–34	−0.009	0.0050	−1.81	.071
Age: 35–39	−0.005	0.0044	−1.13	.261
Age: 40–44	−0.005	0.0042	−1.27	.204
Age: 45–49	−0.002	0.0036	−0.66	.510
Age: 50–64	−0.004	0.0037	−0.96	.335
Age: 65–74	−0.012	0.0130	−0.94	.346
Age: Over 74	0.001	0.0079	0.08	.933
Race: Black	0.006	0.0032	2.01	.045
Race: Other non-White	−0.009	0.0048	−1.96	.051
Race: Hispanic	−0.006	0.0035	−1.56	.120
Income: 100–200% of FPL	−0.005	0.0041	−1.26	.210
Income: 200–400% of FPL	0.001	0.0033	0.43	.668
Income: Over 400% of FPL	0.009	0.0050	1.90	.059
Employed part time	0.003	0.0028	1.16	.246
Not employed	0.000	0.0026	0.01	.992
Self-employed	−0.002	0.0032	−0.62	.535
Widowed/Divorced/Separated	−0.005	0.0030	−1.74	.082
Never married	−0.008	0.0038	−2.13	.033
Self-rated health: Very good	0.009	0.0041	2.28	.023
Self-rated health: Good	0.012	0.0050	2.32	.021
Self-rated health: Fair	0.017	0.0069	2.53	.012
Self-rated health: Poor	0.025	0.0099	2.48	.013
Any chronic illness	0.027	0.0101	2.65	.008
Education: High school grad	0.010	0.0048	2.08	.038
Education: Some college	0.017	0.0069	2.49	.013
Education: College graduate	0.019	0.0075	2.51	.012
Gender: Female	0.027	0.0098	2.74	.006
Alabama	0.010	0.0062	1.64	.101
Arizona	0.008	0.0090	0.88	.377
California	−0.004	0.0054	−0.72	.472
Colorado	0.005	0.0087	0.62	.533
Connecticut	0.018	0.0056	3.29	.001
Florida	−0.004	0.0071	−0.52	.605
Georgia	0.003	0.0085	0.36	.718
Illinois	−0.004	0.0067	−0.58	.564
Indiana	−0.001	0.0068	−0.14	.890
Kentucky	0.010	0.0053	1.84	.066
Louisiana	0.009	0.0085	1.07	.286
Maryland	0.011	0.0072	1.53	.127
Massachusetts	0.027	0.0086	3.20	.001
Michigan	0.006	0.0051	1.09	.277
Minnesota	0.003	0.0069	0.45	.654
Missouri	−0.003	0.0082	−0.31	.757
New Jersey	0.003	0.0052	0.63	.529
North Carolina	0.021	0.0055	3.76	.000
Ohio	0.004	0.0061	0.58	.561
Oklahoma	0.019	0.0070	2.66	.008
Oregon	−0.002	0.0089	−0.22	.824
Pennsylvania	0.000	0.0058	−0.08	.934
South Carolina	0.000	0.0103	−0.02	.981
Tennessee	−0.019	0.0057	−3.29	.001
Texas	0.002	0.0060	0.37	.709
Virginia	0.019	0.0045	4.15	.000
Washington	0.006	0.0074	0.86	.392
Wisconsin	0.002	0.0076	0.26	.797

The key variable in the analysis is the interaction term, which represents the differential effect of increased Medicaid payment rates on use of preventive services in the Medicaid population. In all five models, this coefficient is statistically insignificant; the coefficients are greater than zero in only three of the five models. This suggests that higher Medicaid payment rates had no association with the use of these five preventive services in the Medicaid population.

However, there are many significant factors that do predict having a screening test, even if Medicaid enrollment and Medicaid payment rates are not significant predictors of receiving a screening. For both cholesterol and blood pressure screening, self-reported health status was negatively related to screening probabilities. Relative to excellent health, the probability of receiving a cholesterol test increased from 1.1 percentage points more likely for very good health to 2.6 percentage points for poor health. For blood pressure, the increase in the probability of screening ranged from 0.9 to 2.5 percentage points. Similarly, any chronic illness increased the probability of all the screening tests, with the effect ranging from a low of 2.1 percentage points (cervical cancer) to a high of 12.8 percentage points (colorectal cancer). Finally, education was strongly and consistently positively associated with screening. For all five screening tests, the effect increased with greater education, with the lowest range (blood pressure) going from a low of 1.0 percentage points for high school graduate (relative to nonhigh school graduates) to 1.9 percentage points (college graduate) and the highest range (colorectal cancer) going from 3.6 percentage points for high school graduate to 12.7 percentage points (college graduate).

Other individual factors were not significant. Race was largely insignificant except for colorectal cancer, where minorities were less likely to receive screening. Marital status was largely insignificant, as was employment.

A second set of factors which was significant is the state fixed effects. Connecticut had significantly higher screening rates than the reference state (New York) for four of the five measures: cervical cancer (5.9 percentage points higher), breast cancer (9.5 percentage points higher), cholesterol screening (2.5 percentage points higher), and blood pressure (1.8 percentage points higher). Massachusetts had higher screening rates for three (cervical cancer, blood pressure, and colorectal cancer) as did North Carolina (cervical cancer, blood pressure, and cholesterol). In contrast, Tennessee was significantly worse than New York on three measures: cholesterol screening (3.3 percentage points lower), blood pressure (1.9 percentage points lower), and colorectal cancer (15.3 percentage points lower).

Discussion

This paper explored the relationship between Medicaid primary care payment rates and the use of five preventive care services (breast cancer, cervical cancer, cholesterol, blood pressure, and colorectal cancer) recommended by the USPSTF. The descriptive statistics suggest only a slight relationship between Medicaid enrollment and use of preventive services and no relationship between Medicaid payment rates and the same. The multivariate analysis reinforces this finding, with no statistically significant relationships between Medicaid enrollment or Medicaid primary care payment rates and the use of preventive services.

There are a number of potential threats to our causal inference. First, we use the ratio of Medicare to Medicaid payment rates. To test the sensitivity of our model to this specification, we estimated the models with the Medicaid sample only and estimated the effect of changes in payment rates on changes in the use of preventive services and found a similar result. We also limited our analysis to only the 10 largest states to ensure that the results were not associated with smaller sample sizes in smaller states. Finally, we limited the data to only individuals with physician visits to test whether the results were somehow driven by lack of use. Our findings were robust to all of those specifications.

One further complication is that many Medicaid programs use managed care, which is a potential confounder in our data. The MEPS data identified whether Medicaid enrollees are in a HMO/managed care plans. To address this concern, we controlled for managed care enrollment, both through indicator variables and also through stratification. Our results were robust to these specifications.

We also rely on self-report of the use of preventive care. Although MEPS self-report has been validated with claims data and is an accepted source for data on use of preventive services, our results could be biased if self-report of services is correlated with changes in Medicaid payment rates.

A final limitation of our analysis is that the difference-in-differences model assumes that there are not any other significant shocks that are correlated with the changes in payments. Such shocks seem unlikely, partly because payment rates increased, decreased, and remained unchanged and partly because of the lack of changes in overall prevention use (Table 1).

The results of this analysis appear, at first blush, to be counterintuitive. The conventional wisdom is that low payment rates are a key obstacle to Medicaid enrollees accessing care. What this paper suggests is that incremental changes in particular payment rates are insufficient to increase use of services. The idea behind the payment increases is straightforward: if you want more Medicaid enrollees to receive preventive care, pay doctors more to provide preventive screenings. However, this straightforward idea runs into two practical challenges. First, after controlling for demographic factors, there is not a statistically significant difference in the receipt of preventive services between Medicaid, Medicare, and privately insured individuals. Second, changes in payment rates for preventive services are unlikely to motivate providers who are not accepting Medicaid enrollees to suddenly change their practice and begin to accept Medicaid enrollees and provide preventive screenings. Indeed, it is difficult to convince providers who do not accept Medicaid patients to change their practice and accept them with broad-based Medicaid payment increases. Service-specific payment increases are unlikely to motivate these providers to accept patients and provide preventive services. The results of this paper suggest that piecewise increases in the Medicaid preventive care payment rate are not an effective strategy to increase the receipt of preventive services in the Medicaid population.

Supporting Information

Additional supporting information may be found in the online version of this article:

Appendix SA1: Author Matrix.