Co-Payment Policies and Breast and Cervical Cancer Screening in Medicaid

Abstract

Objectives:

This study investigates the relationship between state Medicaid co-payment policies and cancer screening for Medicaid-enrolled women.

Study Design:

Cross-sectional analysis of administrative claims and enrollment data.

Methods:

Our data included Medicaid Analytic eXtract outpatient claims files across 43 states in 2003, 2008 and 2010, the years for which both MAX data and state cost-sharing data were available. Data on enrollee demographics and screening services from enrollment and claims files are merged with state-year data on co-payment policies and county-level controls from the Area Health Resources File. Participants were non-elderly, non-disabled, non-pregnant women in the recommended age range for each screening service (50–64 years for mammograms; 21–64 years for Pap tests) enrolled in fee-for-service Medicaid. The main independent variable is whether an enrollee faces cost-sharing for preventive services. We examine three categories of cost-sharing: co-payments for all visits, including for preventive services; co-payments for outpatient visits, but waived for preventive services; and no co-payments. The main outcome measure was receipt of mammogram or Pap test within a 12-month period.

Results:

Medicaid enrollees with copayments for preventive services were less likely to receive both screening mammograms and Pap tests than enrollees in states not requiring cost-sharing for preventive services.

Conclusions:

Co-payments for preventive services discourage breast and cervical cancer screening among Medicaid enrollees. The effect is larger for breast cancer screening, which is costlier and requires an additional visit. Considering this evidence, cost sharing for preventive services may lead to adverse health consequences and greater long-term costs.

Précis:

Co-payments for preventive services can discourage breast and cervical cancer screening among Medicaid enrollees, particularly breast cancer screening, which is more costly and time consuming.

Introduction

Morbidity and mortality from breast and cervical cancer can be reduced through screening.^1,2 Breast cancer is the second leading cause of cancer death among US women.³ Although the prevalence of cervical cancer is low in the US, it could be practically eliminated through screening and treatment for pre-cancerous lesions.⁴ However, cancer-related disparities persist among women of low socioeconomic status, for whom out-of-pocket costs are often a barrier to seeking screening and treatment for cancer.^5–7

While cost-sharing in Medicaid is typically low, most states require co-payments from adult enrollees^8,9 and in some states, cost-sharing is on the rise.¹⁰ Beginning in 2013, the Affordable Care Act (ACA) incentivized states to eliminate cost-sharing for preventive care, offering a 1% increase in the federal match rate for states that cover recommended preventive services without cost-sharing.^11,12 In contrast, to curb health care utilization, the Centers for Medicare and Medicaid Services encouraged states to increase cost-sharing for certain services for non-disabled adults in Medicaid.¹³ While the policy debate over the role of cost-sharing in the Medicaid program continues, evidence regarding the impact of cost-sharing on utilization of different types of services is needed to inform policy and programmatic decisions.¹⁴

Preventive care such as cancer screening is potentially high-value and has benefits for population health.¹⁵ For example, since its implementation, routine screening has led to a substantial decrease in deaths from cervical cancer.¹⁶ However, cost-sharing may discourage the use of these services, particularly among low-income women who have few resources and who do not habitually seek cancer screening, often leading to late-stage diagnosis and poor prognosis.^17–19

Several studies have examined the role of cost-sharing on healthcare utilization. The seminal RAND Health Insurance Experiment, which randomized families to insurance plans with various levels of cost-sharing, found that those with greater out-of-pocket costs had significantly lower healthcare utilization.²⁰ Recent studies suggest negative effects of co-payments on health care utilization for Medicaid and other low-income populations across multiple types of care but few have focused on preventive services.^21–23 Within the Medicare population, even co-payments as low as $10 are associated with significantly lower rates of mammography in managed-care plans.²⁴ While these studies suggest co-payments reduce utilization, evidence is needed on the effect of co-payments on preventive care for Medicaid enrollees, who are vulnerable due to their low-income, risk factors (e.g., smoking, obesity, comorbid conditions), and lack of access to care. One recent study examined the association of multiple Medicaid policies (not focusing primarily on cost-sharing) with receipt of cancer screenings using a single year of data and found that enrollees in states requiring a co-payment for physician services were less likely to receive certain screening services, including mammograms or Pap tests.¹⁸ The current study uses multiple years of data and detailed information on cost-sharing policies to evaluate how state-level Medicaid cost-sharing policies influence breast and cervical cancer screening among non-elderly, non-disabled adult enrollees.

Methods

Data and Study Sample

We used three datasets to examine the association between state Medicaid cost-sharing policies and breast and cervical cancer screening. Information on state Medicaid cost-sharing policies came from Kaiser Family Foundation (KFF) surveys of states on Medicaid policy benefits.^8,9,25–28 Data on Medicaid enrollees and their use of screening services came from the Medicaid Analytic eXtract (MAX) files. The MAX personal summary (PS) file provides enrollee’s eligibility pathway, demographics and type of Medicaid enrollment (managed care or fee-for-service (FFS)). Utilization data for breast and cervical cancer screening came from the MAX other therapy (OT) claims file. We use data for years 2003, 2008 and 2010, years for which both MAX data and state cost-sharing data were available. We supplement these datasets with county-level variables from the Area Health Resources File (AHRF), linked to the MAX data using county-level FIPS codes to control for area-level sociodemographic characteristics (as a proxy for individual-level socioeconomic variables not in the claims files) and for local health care provider supply.

The study sample comprises non-elderly, non-disabled, non-pregnant women in the recommended age range for each screening service (50–64 years for mammograms and 21–64 years for Pap tests) enrolled in Medicaid. We excluded males; women younger than 21 years and older than 64 years; women who had claims for pregnancy or labor and delivery within the calendar year; and individuals with dual eligibility or who were enrolled under the disabled eligibility category, had missing enrollment length, or appeared in the MAX data in more than one state within a calendar year. We restricted the sample to enrollees in FFS Medicaid for two reasons: first, managed care data may not be reliable for all states and years;^29,30 and second, cost-sharing requirements may differ for managed care plans compared to FFS plans. We excluded Maine and Kansas because MAX data were unavailable for the study years. Given that MAX data are based on standardization of state reports that vary in completeness, particularly for states with many enrollees in managed care who may, for example, only be in FFS Medicaid upon initial enrollment in the program, we focus on states with robust FFS claims during our study period. To gauge data completeness by state, we examined the count of FFS claims available in the MAX data per state and per year. We excluded states (N=4) with a FFS claim count for each study year (i.e., 2003, 2008 and 2010) 1.5 standard deviations below the average number of FFS claims that year. We chose the 1.5 standard deviation threshold to exclude the extreme outliers based on examination of the distribution of claims by state. We also excluded states in which the number of eligible women was less than 100 in each year for both samples. Consequently, we excluded 6 states (HI, MD, AZ, NM, TN, and RI). Thus, the final sample comprised data from 43 states.

Study Measures

Claims were used to measure the primary outcomes: receipt of mammography or Pap test within the given calendar year, based on procedure and diagnosis codes (Appendix Table 1). Explanatory variables of interest indicate whether the state required co-payments for different types of outpatient visits. Specifically, states fall into one of three categories in any year: those requiring co-payments for all visits, including preventive services; those requiring co-payments for outpatient visits, but with co-payments waived for preventive services; or those without co-payments, regardless of visit type. We construct copayment policy variables that compare states with co-payments for preventive services to states without co-payments for preventive services and then compare enrollees’ utilization of breast and cervical cancer screening in each of the three co-payment policy groups.

We control for potential confounders that may differ across states or years and may be associated with receipt of screening, including individual and area-level variables identified in models of health care access and shown in previous literature to be associated with preventive services and health care access.^18,31–34 Individual-level controls include age, race/ethnicity, basis of Medicaid eligibility (i.e., 1115 waiver adult, medically needy, parent, poverty, or other eligibility pathway), number of months in Medicaid that year, and whether the woman was enrolled in a primary care case management (PCCM) program, since those in PCCM may experience more coordinated primary care, which could increase use of preventive services.³⁵ County-level sociodemographic variables include percent of population age 25 and older with less than a high school diploma, percent of population that is white non-Hispanic, percent of population living in an urban area, percent unemployed and median household income. We also controlled for the availability of health care providers in the county of residence, measured per 1000 population in the county: primary care physicians (including OB/GYNs), specialists, hospital beds, Federally Qualified Health Centers (FQHCs) and Rural Health Centers (RHCs).

Analysis

We estimate two sets of multivariable logistic regression models adjusting for all covariates as well as state fixed effects, to control for time-invariant state characteristics, and year fixed effects to control for trends in screening over time. In the first model, we compare enrollees for whom co-payments apply to preventive services to those without co-payments for preventive services (regardless of co-payments for other types of visits). In the second model, we compare enrollees across the three co-payment groups described above. We estimate all models for a sample of women enrolled in Medicaid for the entire calendar year as well as the overall sample with at least one month of Medicaid enrollment. Using the estimated coefficients, we derive predicted probabilities of receiving screening. Because most states increased enrollment in managed care over time, we also estimate models including an indicator for the percentage of enrollees in managed care. Analyses were conducted using SAS software version 9.4.

Results

Table 1 describes co-payment policies for our sample states. Twelve states had no co-payments. Of the 26 states that required co-payments, two states (MN, NE) waived preventive co-payments over the entire study period. Cost-sharing policies changed over the study period for five states. DE, KY, MI, SC had no co-payments required for any services in 2003 but in 2008 and 2010 required co-payments for all visits, including those for preventive services. Additionally, co-payments for preventive visits were waived in MO in 2003, but these waivers were discontinued in 2008 and 2010.

Table 1.

Co-payment policies for general and preventive visits by state in 2003, 2008, 2010

Co-payment Policy	Number of States	List of States
No co-payments	12	AR, ID, IN, LA, MA, NH, NV, OH, TX, VT, WA, WV
Co-payments apply to preventive services	24	AK, AL, CA, CO, CT, DC, FL, GA, IA, IL, MS, MT, NC, ND, NJ, NY, OK, OR, PA, SD, UT, VA, WI, WY
Co-payments waived for preventive services	2	MN, NE
Co-payment policies changed over the study period	5	DE, KY, MI, MO, SC

Notes: DE, KY, MI, SC had no co-payments in 2003 but changed requirements in 2008 and 2010, implementing co-payments for all services (including preventive care). In 2003, MO required co-payments for most services but waived them for preventive visits, though MO policy changed to require co-payments for all services (including preventive care) in 2008 and 2010.

Table 2 reports results for mammogram receipt; 19% of women enrolled for 12 months and 11% enrolled for any number of months during the calendar year had a Medicaid claim for screening mammography. The low observed screening rates are similar to those observed in other research using MAX data to examine cancer screening.¹⁸ In Panel A, we report results from models of mammography receipt among women ages 50–64 years comparing screening rates among women with co-payments for preventive services visits to those without a co-payment for preventive services (either because the co-payment is waived for preventive care or because the state does not require a co-payment for any visit). Women with co-payments for preventive services are less likely to receive a screening mammogram than those without a co-payment for preventive care (aOR 0.81, 95% CI 0.71, 0.94 for 12-month enrollment sample). Results are similar when the sample includes women with any number of months of enrollment during the year.

Table 2.

Odds Ratios from Logistic Regressions of Receipt of Mammogram on Cost-sharing Policies (ages 50–64)

	12-Month Enrollment		Any Enrollment
N =	190,155		549,541
Received screening mammogram (%)	18.5		10.6
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Model A
Co-payments apply to preventive services vs. no co-payment for preventive visits (waived or no co-payments)	1.01 (0.98,1.03)	0.81 (0.71,0.94)	0.83 (0.82,0.85)	0.83 (0.74,0.92)
Model B
Co-payments apply to preventive services vs. no co-payments	1.01 (0.99,1.04)	0.84 (0.72,0.97)	0.83 (0.81,0.84)	0.85 (0.76,0.95)
Co-payments waived for preventive services vs. no co-payments	1.26 (1.10,1.45)	1.20 (0.80,1.79)	0.77 (0.70,0.85)	1.19 (0.89,1.61)
Co-payments apply to preventive services vs. co-payments waived for preventive services	0.80 (0.70,0.92)	0.70 (0.48,1.01)	1.08 (0.98,1.18)	0.71 (0.54,0.94)

Abbreviations: OR, odds ratio; CI, confidence interval. Models based on MAX data from 43 states. The 12-month enrollment sample includes women enrolled in Medicaid for the entire calendar year; the any enrollment sample includes women with at least one month of Medicaid enrollment in a given calendar year. Control variables included in adjusted models are age, race/ethnicity, basis of Medicaid eligibility (1115 waiver adult, medically needy, parent, poverty, or other eligibility pathway), number of months enrolled in Medicaid during the year, enrollment in a primary care case management (PCCM) program, percent of the county population 25 and older with less than a high school diploma, percent of the county population that is white non-Hispanic, percent of the county population living in an urban area, county unemployment rate, county median household income, primary care physicians (including OB/GYNs) per 1000 population in the county, specialists per 1000 population in the county, hospital beds per 1000 population in the county, Federally Qualified Health Centers per 1000 population in the county, Rural Health Centers per 1000 population in the county, year indicators, and state indicators.

Panel B reports results from models comparing all three co-payment policies. Women for whom co-payments applied to preventive visits had a lower likelihood of receiving a mammogram than women without co-payments for any visits (aOR 0.84, 95% CI 0.72, 0.97). In the 12-month enrollment sample, receipt of a mammogram among women with co-payments for preventive services did not differ significantly from women with co-payments for other outpatient visits but for whom co-payments were waived for preventive services (aOR 0.70, 95% CI 0.48, 1.01); the corresponding estimate for the sample with any length enrollment suggests women with co-payments for preventive services were less likely to receive a mammogram than women without co-payments for most visits but no co-payment for preventive services (aOR 0.71, 95% CI 0.54, 0.94). Among the sample with 12 months continuous Medicaid enrollment, the group with co-payments for most visits but not preventive services had the highest predicted probability (24%) of receiving a screening mammogram, whereas those who had co-payments for all visits had the lowest predicted probability (18%) of receiving a mammogram (see Appendix Figure 1).

Table 3 presents model estimates for cervical cancer screening; 30% of women enrolled for a full 12 months and 22% enrolled for any number of months during the calendar year had a Medicaid claim for a Pap test. Women with co-payments for preventive services are slightly less likely to receive a Pap test during the year (aOR 0.96; 95% CI 0.94, 0.98 for 12-month enrollment sample) (Panel A). Panel B compares women under the three co-payment policies. Women with co-payments for preventive services were slightly more likely to receive cervical cancer screening compared to women who had no co-payments for any type of visit (aOR 1.05; 95% CI 1.03, 1.07). In contrast, women with co-payments for preventive services were less likely to receive a Pap test than women with co-payments for most visits but not preventive visits (aOR 0.70; 95% CI 0.68, 0.73). Based on these results, we estimate that among the sample with 12 months’ continuous Medicaid enrollment, those with co-payments waived for preventive visits had the highest predicted probability (29%) of receiving a Pap test, whereas those without co-payments for any visits and with co-payments for preventive services had similar rates of Pap tests (Appendix Figure 1).

Table 3.

Odds ratios from regressions of receipt of Pap Test on cost-sharing policies (ages 21–64)

	12-Month Enrollment		Any Enrollment
N =	4,647,977		14,259,137
Received Pap test (%)	30.4		21.9
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Model A
Co-payments apply to preventive services vs. no co-payment for preventive visits (waived or no co-payments)	1.15 (1.14,1.16)	0.96 (0.94,0.98)	1.30 (1.29,1.31)	0.99 (0.97,1.00)
Model B
Co-payments apply to preventive services vs. no co-payments	1.18 (1.17,1.19)	1.05 (1.03,1.07)	1.31 (1.30,1.32)	1.05 (1.04,1.07)
Co-payments waived for preventive services vs. no co-payments	1.36 (1.33,1.38)	1.49 (1.43,1.55)	1.21 (1.20,1.23)	1.38 (1.34,1.42)
Co-payments apply to preventive services vs. co-payments waived for preventive services	0.87 (0.85,0.88)	0.70 (0.68,0.73)	1.09 (1.07,1.10)	0.76 (0.74,0.79)

Abbreviations: OR, odds ratio; CI, confidence interval. Models based on MAX data from 43 states. The 12-month enrollment sample includes women enrolled in Medicaid for the entire calendar year; the any enrollment sample includes women with at least one month of Medicaid enrollment in a given calendar year. Control variables included in adjusted models are age, race/ethnicity, basis of Medicaid eligibility (1115 waiver adult, medically needy, parent, poverty, or other eligibility pathway), number of months enrolled in Medicaid during the year, enrollment in a primary care case management (PCCM) program, percent of the county population 25 and older with less than a high school diploma, percent of the county population that is white non-Hispanic, percent of the county population living in an urban area, county unemployment rate, county median household income, primary care physicians (including OB/GYNs) per 1000 population in the county, specialists per 1000 population in the county, hospital beds per 1000 population in the county, Federally Qualified Health Centers per 1000 population in the county, Rural Health Centers per 1000 population in the county, year indicators, and state indicators.

Summary statistics, full regression results, and sensitivity analyses are presented in the Appendix. Appendix Table 2 reports descriptive statistics for the full sample and stratified by whether state Medicaid policy requires co-payments for any visits. Characteristics of the mammogram and Pap test eligible samples were similar, except that most Pap test eligible cases were younger than age 40 (82%). Full regression results for the mammography sample are presented in Appendix Tables 3 and 4. Conditional on all other covariates, women enrolled in earlier years were more likely to receive screening than those enrolled in 2010. Across all models, receipt of mammography is positively associated with FQHC, RHC, and specialist density in the enrollee’s county, but negatively associated with PCP density in the county. Receipt of Pap tests is also positively associated with FQHC and RHC density in the enrollee’s county, and negatively associated with PCP density in the county conditional on other controls. In contrast with mammography, specialist density is not consistently associated with cervical cancer screening across models.

We also tested whether our results changed when including an indicator for the percentage of enrollees in comprehensive managed care (CMC) in each state year using a range of thresholds to define this variable. Results remained virtually the same (Appendix Tables 5 and 6).

Discussion

This paper considers how state co-payment policies impact breast and cervical cancer screening among low-income non-disabled women enrolled in FFS Medicaid. For both mammograms and Pap tests, women were less likely to receive screening when their state required co-payments for these services. We also compared women across states and years with different co-payment structures to understand whether there was a difference in screening among those required to pay co-payments that apply to all visits; those required to pay co-payments for most outpatient visits, but for whom co-payments are waived for preventive services; and enrollees without co-payments. For both types of screening, the group that had co-payments for some visits but had them waived for preventive care had the highest rates of screening. Counterintuitively, women with co-payments for preventive services are somewhat more likely to receive Pap tests compared with those who face no co-payments for any services. This difference in findings between Pap tests and mammograms may arise from differences in these screening procedures’ complexity and time costs; this warrants further investigation in future research.

Our results are strongest for mammography, with a bigger difference in screening rates observed for mammograms compared to Pap tests between those with co-payments for preventive services versus with those without co-payments for preventive care. Mammograms are likely to require a separate visit that may entail an additional co-payment, while a Pap test could be completed during a visit to an OB/GYN or other primary care physician. Therefore, co-payment policies appear to restrict services that require separate visits and additional costs for the patient.

Our findings suggest that Medicaid policies should consider the impact co-payments have on utilization of potentially high-value services such as screening. While concerns have been raised regarding over-screening in older and higher-income populations,^36,37 among the non-elderly Medicaid population screening rates are far below population-wide target levels.³⁸ Thus, the influence of Medicaid policies, including cost-sharing requirements, on utilization of high-value services needs careful consideration before implementation. These policies could result in future higher utilization and costs.

Limitations

Notably, annual screening rates in Medicaid claims data are lower than survey estimates for the general population, low-income population or Medicaid enrolled women. This might reflect a combination of low screening rates among Medicaid enrollees, inability to measure screenings not paid for by Medicaid, and overestimates of screening in survey data. Our study does not capture screenings completed outside the Medicaid program. We estimate all models for two samples: those continuously enrolled for all 12 months of a calendar year and those with at least one month of enrollment within the year. The first is a consistently enrolled Medicaid sample whereas the second includes those who may have dropped out of Medicaid for a period of time and thus may have had a screening test outside of Medicaid reimbursement, but may also be more representative of a population known to churn in and out of Medicaid enrollment. Further, this study examines changes in annual screening for breast and cervical cancer over time in the context of changes in cost-sharing requirements; we do not capture whether the women in the sample receive guideline recommended screening and although our study uses multiple years of data, we only capture 3 non-adjacent calendar years, limiting the observation period. Therefore, we cannot assess longitudinal screening behavior outside of these timeframes. Other study limitations include the fact that we cannot control for all state-year differences in policy and health systems factors that may influence whether women receive recommended screenings. Further, the number of states that require co-payments for most services but waive them for preventive care is small. Finally, the data used are from the pre-ACA period, and the characteristics of non-elderly adult enrollees may have changed under Medicaid expansions.

Conclusions

Our results offer insight for policy, practice and future research. First, state Medicaid programs that do not require co-payments for preventive services have higher rates of screening among non-elderly, non-disabled women. Second, rates of annual screening among women enrolled in Medicaid are low. Third, we find suggestive evidence that the full range of cost-sharing policies (including for other types of services) may impact utilization of preventive care in the Medicaid population. The use of longitudinal data could improve our understanding of cost-sharing policies and allow us to assess how these policies facilitate or create barriers to preventive services in low-income insured populations over time.

Take-Away Points:

Co-payments for preventive services can discourage breast and cervical cancer screening among Medicaid enrollees. Such cost sharing may lead to adverse health consequences and greater costs.

• Analysis of outpatient claims files across 43 states from 2003, 2008 and 2010 indicated that Medicaid enrollees facing copayments for preventive services were less likely to receive both screening mammograms and Pap tests than those without co-payments.

• Ongoing changes to Medicaid policy across multiple states that emphasize cost sharing and consumer-directed health care principles should carefully consider the impact of out of pocket cost on preventive care and other high-value services.