INTRODUCTION

Controlling growth in Medicaid spending is a persistent challenge for states, and lowering provider fees has been a common approach for reducing Medicaid spending.1 Many states set Medicaid provider fees at levels that are lower than Medicare and private insurance payments, and low fees have been commonly cited by providers as a reason for not accepting Medicaid patients.2, 3 In 2016, Medicaid paid primary care providers (PCPs) an average of 72% of Medicare fees and there is wide variation across states ranging from 38 to 126%.2 The Affordable Care Act (ACA) included a temporary Medicaid fee bump for primary care providers (PCPs) to Medicare levels in 2013–2014 to address potential shortages of PCPs for Medicaid enrollees, especially as many states adopted Medicaid expansion for low-income adults (37 states to date). As of 2018, only 17 states have continued this fee increase fully or partially, with most other states returning to lower pre-ACA fee levels.

A number of cross-sectional studies have found that in states with higher Medicaid payment rates, provider participation levels are higher, for example, as measured by their willingness to accept new Medicaid patients in their practice.3,6,7,8,9,8 Cross-sectional studies also suggest that access to care could be greater for enrollees living in higher- vs. lower-fee states, such as enrollees being more likely to report having a usual source of care9 and having higher rates of outpatient visits and receipt of certain preventive services.6, 9, 10 To the extent that higher fees increase the likelihood that enrollees receive appropriate medical care in the outpatient setting, raising fees could also improve health outcomes and reduce Medicaid spending, such as by decreasing inappropriate use of the emergency department or preventable hospitalizations. Studies that rely on cross-sectional comparisons across states with higher vs. lower fee levels, however, could be confounded by other differences, such as differences in state Medicaid coverage policies, benefit generosity, and program eligibility criteria.

We sought to systematically review the literature that examines the longitudinal impact of increasing or decreasing fees for outpatient providers, which provides stronger evidence on the potential causal effects of fee changes on providers and enrollees. We focused on three major categories of outcomes: provider participation, enrollee access to care, and service utilization.

METHODS

Search

We iteratively developed a search query that broadly identified studies through 2018 that examined the role of Medicaid provider fees. Our final search term was “Medicaid AND (provider* OR physician* OR doctor* OR (primary and care) OR specialist) AND (reimburs* OR pay OR payment OR fee OR charge OR fees OR charges OR payments),” which we conducted in two primary databases: PubMed/Medline and JSTOR. Our initial search of these databases identified 7081 unique titles (Fig. 1). We examined bibliographies of the studies selected for full article review, which yielded no additional articles for review through January 2019.

Figure 1
figure 1

PRISMA flow diagram. Figure 1 contains poor quality of text inside the artwork. Please do not re-use the file that we have rejected or attempt to increase its resolution and re-save. It is originally poor, therefore, increasing the resolution will not solve the quality problem. We suggest that you provide us the original format. We prefer replacement figures containing vector/editable objects rather than embedded images. Preferred file formats are eps, ai, tiff and pdf.A new file with better quality has been attached here for Figure 1.

Full size image

Title and Abstract Review

Two study team members independently screened the titles identified from the search for relevance to Medicaid fees and the following outcomes: provider participation, access to care, utilization or quality, health outcomes, and Medicaid spending. Because of the small number of articles that examined the latter two outcomes, we subsequently excluded these studies from our review after the full article review stage. At the conclusion of title review, there was a high level of agreement, with any discrepancies (60 of the 7081 titles) resolved by consensus. A total of 652 articles were selected for abstract review (Fig. 1); during this stage, two study team members reviewed the abstracts to identify empirical studies that assessed the impact of Medicaid fees on the outcomes of interest and identified 161 studies for full article review.

Inclusion Criteria

All articles selected for full article review were abstracted to determine whether they examined one or more of our outcomes of interest and matched other inclusion criteria. We excluded studies that used cross-sectional vs. longitudinal designs and those that focused on the impact of changing the payment mechanisms (e.g., fee-for-service vs. capitation or pay-for-performance) vs. changes in provider fee levels. We also excluded non-empirical studies (e.g., perspectives, commentaries) and those not published in peer-reviewed journals, including reports and working papers. We further limited our sample to studies focused on Medicaid and changes in provider payment in the outpatient setting and excluded those examining dental care, as there is greater variability across states in coverage of dental services.

We identified a total of 22 studies that examined the impact of changes in fees over time. Among these studies, we further restricted our review to studies that met criteria set by the Cochrane Collaborative for observational studies, which we slightly modified.11 The Cochrane Collaborative recommends requiring pre-post studies to have at least two intervention and two control groups; however, we relaxed this criterion slightly to include studies that had just a single intervention and control state. For interrupted time series studies, we excluded four studies that did not have at least three pre- and three post-intervention time periods but retained studies with only a single pre- or post-fee change year if their primary unit of analysis was a quarter or month and thus included three or more pre- and post-intervention time periods.4, 12,15,14 Because the studies included in the review varied in their measurement of fee changes, outcome definitions, and study designs, we did not attempt to combine and summarize the results using a meta-analysis. Our final sample included 18 studies (Fig. 1).

Content Abstraction and Synthesis

We abstracted the following information from the studies in the review using a structured form: research question and hypotheses, study design, primary data sources, predictor variable measurement (i.e., operational definition of fees and fee changes), outcome measures, findings, implications, and limitations. We classified studies as having one of four basic study designs, which vary in their methodological approaches and ability to address potential unobserved confounding. From most to least susceptible to potential confounding, the designs included: (1) interrupted time series designs with no concurrent control group (ITS); (2) dose-response (DR) model designs, which leveraged differential changes in provider fees over time across states to identify the effects of fee changes; (3) difference-in-difference (DD) designs, which compared enrollees in states with fee changes vs. a control group without fee changes (e.g., Medicaid enrollees in other states without fee changes, or individuals in the same state with different insurance types); or (4) triple difference or difference-in-difference-in-difference (DDD) designs, which leveraged both comparisons of Medicaid enrollees in states with different levels of fee changes over time and within state comparisons of Medicaid enrollees to those with different insurance types that did not face fee changes (e.g., privately insured or Medicare).

In cases where studies reported both univariate and multivariate findings, we report the multivariate findings. Where studies perform cross-sectional and longitudinal analyses, we report only the longitudinal results.

RESULTS

Study Characteristics

Table 1 describes the 18 studies included in the review. Six were published between 1980 and 1999,15,18,19,20,21,20 five between 2000 and 2009,9, 21,24,25,24 and seven between 2010 and 2018.25,28,29,30,31,32,31 In total, seven studies examined the effects of payment changes on provider participation in Medicaid, five on enrollee access, and fourteen on service use. All studies used quasi-experimental methods, of which five used a triple difference (DDD) design. Two studies used a difference-in-difference (DD) design, six studies used a dose-response (DR) model that examined changes in outcomes for states with larger vs. smaller fee changes over time, and five studies used an interrupted time series (ITS) design without a concurrent control group.

Table 1 Summary of Studies Examining the Effects of Changing Medicaid Provider Fees
Full size table

Provider Participation

Of the studies examining the effects of Medicaid fees on provider participation,17, 18, 20,23,22, 30, 31 there were two primary measures (Table 2): changes in the likelihood of accepting any or new Medicaid enrollees (“any participation”) and changes in the number of Medicaid enrollees or appointments per provider (“caseload”). Four studies examined changes among PCPs specifically.17, 20, 30, 31 Among the five studies examining the probability of any Medicaid participation by providers, only one study found positive effects of fee increases on the probability of participation17 and one found a slightly negative effect.18 Three studies found no significant association between fee changes and participation, including two studies that examined the effects of the ACA PCP fee bump specifically.30, 31

Table 2 Studies Examining Medicaid Fee Changes and Provider Participation
Full size table

Of the six studies that examined changes in Medicaid caseload, two found positive associations. Using a national survey of young physicians, Baker and Royalty (2000) estimated that a 10% increase in Medicaid-to-private insurance fee ratio for obstetricians was associated with a 2.5 percentage point (pp) increase in the percent of poor patients per physician panel.21 Adams (2001) estimated that a 10pp increase in the Medicaid-to-private fee ratio was associated with an increase of eight Medicaid children per physician per year using Medicaid claims data on children in four states.19 The remaining studies found no significant association.

Enrollee Access to Care

We classified measures of access to care into three primary categories (Table 3): having a usual source of care, having unmet need (e.g., reports of going without or delaying care), and appointment availability (probability of getting a new appointment and wait times until the next available appointment).

Table 3 Studies Examining Medicaid Fee Changes and Enrollee Access to Care
Full size table

Two studies by the same research team used an audit (“secret shopper”) approach to assess changes in appointment availability after the ACA PCP fee bump. Polsky et al. (2015) found a significant 8.3pp increase in the probability of getting an appointment for Medicaid enrollees vs. privately insured enrollees across 10 states following the implementation of the ACA fee bump (2014 vs. 2012–2013), but no changes in wait times.27 In a follow-up study, Candon et al. (2018) used a DR approach (with no comparison group of privately insured enrollees) and estimated that $10 increases in PCP fees were associated with a significant 1.7pp increase in the probability for getting a new appointment for Medicaid enrollees.29

The three remaining studies used earlier data (pre-ACA fee bump) and enrollee reports from surveys; these studies also found positive effects of fee increases on at least one measure of access.9, 25, 28 For example, Shen and Zuckerman (2005) used a DDD design and found an increase of 1.5pp in the proportion of Medicaid vs. privately insured enrollees reporting a usual source of care with a 1 unit change in the county-level Medicaid-to-national capitation rate; however, this study did not find significant changes in the proportion of Medicaid enrollees reporting unmet need.9 Callison and Nguyen (2018) used a similar DDD design and found that a 10pp increase in the Medicaid-to-Medicare fee ratio was associated with a 1.9pp increase in the proportion of Medicaid vs. low-income privately insured enrollees having a usual source of care, but not with other measures of unmet need, including delaying use of medical care or prescription drugs.28

Enrollee Service Use

We classified service use outcomes into four main subcategories (Table 4): (1) any use (e.g., at least one office visit per year); (2) service volume (e.g., total number of visits or procedures or number of visits or procedures per enrollee or provider); (3) use by site of care (e.g., office vs. emergency department); and (4) use of preventive care as a quality process measure (e.g., receipt of mammogram).

Table 4 Studies Examining Medicaid Fee Changes and Service Use
Full size table

Of the seven studies that examined whether fee changes affect the probability of having any service use (e.g., outpatient visits), four studies found a positive association,9, 16, 23, 24 and three studies using weaker study designs (DR and ITS) did not find significant associations.17, 20, 25 Two of eight studies that examined the impact of fee changes on service volume found positive associations,24, 28 and six found no significant association.15, 18, 20, 23, 25, 31

Three of seven studies examining the association between fee changes and shifts in site of care for Medicaid enrollees found increases in visit rates in outpatient office settings and decreases in visits rates in more expensive sites of care, such as emergency or hospital outpatient departments.18, 19, 24 In contrast, Callison and Nguyen (2018) found significant increases in utilization across all sites of care (ED, hospital outpatient departments, and office-based visits) associated with increases outpatient Medicaid-to-Medicare fee ratios between 2008 and 2012.28 The three remaining papers found no significant associations across any site of care.9, 20, 25

Three of the four studies that examined changes in receipt of preventive care used survey data and a DDD study design; these studies did not find a significant effect of changes in Medicaid fees on the likelihood of receiving preventive services. Both Atherly and Mortensen (2014) and Callison and Nguyen (2018) used the Medical Expenditure Panel Survey to examine changes in receipt of blood pressure and cholesterol testing, cancer screenings, and flu vaccinations.26, 28 Shen and Zuckerman (2005) used three waves of the National Survey of America’s Families to examine the probability of receiving a clinical breast examination, pap smear, or making at least one well-child visit in the last 12 months.9 In contrast, Adams (2001) used Medicaid claims data for children in four states in 1989 and 1992 with a DR study design, and found a positive association between Medicaid fees and the likelihood that physicians provided any preventive care services to children during the year; the magnitude was not stated.22

DISCUSSION

Our systematic review of the literature on the impact of Medicaid fees changes yielded a number of findings. First, our review suggests that the evidence that increasing Medicaid fees leads to increases in provider participation in the program is weak. Nevertheless, there was consistent evidence that fee increases were associated with improvements in certain measures of enrollee access to care, such as having a usual source of care or appointment availability. Although many studies have investigated changes in outpatient visits associated with fee changes, the evidence is largely mixed, and it is difficult to make generalizable conclusions with respect to the effect of fee changes on utilization. There was also little evidence that increasing primary care fees was associated with increases in receipt of preventive care.

The lack of strong evidence that increasing provider fees influences provider participation in the program is concerning given that this is the underlying mechanism through which potential improvements in care access and outcomes in Medicaid are posited to be achieved. It is notable that the only study to find an impact of fee increases on extensive participation, or the probability that more providers accept Medicaid enrollees, was in the context of a single state (Maryland) that raised their fees for obstetrical care to the level of private insurers.18 This is consistent with the economic model developed by Sloan et al. (1978) that suggests that when providers have a choice of accepting patients from multiple markets (e.g., private insurance and Medicaid), they prefer patients from market from which expected revenues are greater.32 Thus, the magnitude of Medicaid fee changes relative to other local payers including Medicare and private insurers is likely to moderate the impact of Medicaid fee changes. A number of studies measured changes in provider fees by indexing Medicaid fees relative to private insurance or Medicare fees, but most did not examine variation in the effects of fee changes that occurred at the lower vs. higher end of the index.

In contrast, two studies that investigated the impact of the ACA PCP fee bump, which raised Medicaid fees to Medicare levels, found no effect on participation. The temporary nature of the fee increase, which was federally funded for only two years (2013–2014), could have reduced providers’ willingness to accept new Medicaid patients, especially if there were administrative hurdles or burdens associated with entering the program. Reports also note operational issues in many states with implementing the payment bump, including delays in notifying or reimbursing providers, and lack of provider awareness of the program, which could have further discouraged participation.33, 34

A research team that used a secret shopper approach to assess the effect of the ACA PCP fee bump on appointment availability for Medicaid enrollees found increases in the probability of getting a new appointment, but not appointment wait times, for Medicaid enrollees. Polsky et al. focused on physicians in 10 states that were already accepting Medicaid enrollees, so increases in appointment availability reflect increases in physician participation on the intensive margin. This contrasts with findings from Decker (2018), who used a nationally representative survey to examine changes in both the extensive and intensive participation of providers before vs. after the ACA bump and found no changes. One possible explanation for these seemingly divergent findings is that Medicaid physicians increased their time to see patients (and therefore could offer more appointments) with the fee bump but did not change the overall composition of their patient panels. Growth in the workforce of nurse practitioners and physician’s assistants could also help increase primary care appointment availability.35 Importantly, most studies examining provider participation focused on physician participation and not other provider types.

Similarly, despite the weak evidence linking Medicaid fees with provider participation, two studies found positive effects of Medicaid fee increases on having a usual source of care. Having a usual source of care has been associated with increased likelihood of receiving recommended care, such as preventive care or guideline-consistent chronic care.36, 37 However, in both of these studies, they did not find significant changes in unmet need as measured by reports of avoiding or delaying needed care. This could suggest that there are other important barriers that mediate Medicaid enrollees’ care access outside of provider availability, such as lack of transportation, difficulties getting time off from work, or poor health literacy.38

There was not strong evidence among the studies in this review that overall utilization increased with fee increases. The mixed findings could be related to heterogeneity in the study time periods, settings, study designs, and underlying data sources of the studies included in the review, although we did not find obvious patterns in study findings related to these factors. There is also substantial heterogeneity across state Medicaid programs in their eligibility criteria, benefits (e.g., coverage for or limits on certain services), and local area characteristics including local demand for services relative to the supply of providers, which could moderate the effects of fee changes on outcomes for Medicaid enrollees.

There was a small body of evidence on the impact of fee changes on use of preventive services. The findings from these studies suggest that fee increases, on their own, could be insufficient to increase use of high-value or cost-effective preventive care, especially in the context of traditional fee-for-service (FFS) reimbursement. A number of states are currently implementing Accountable Care Organizations as a means of reforming payment and care delivery within their Medicaid programs, which could provide greater incentives for quality improvement.

Importantly, the majority of Medicaid enrollees nationally are currently enrolled in comprehensive managed care plans, which could also have greater incentives for increasing use of high-value preventive services (65% in 2015).39 The fee changes that are the focus of this review primarily applied to Medicaid FFS payments, which now represent a minority of enrollees but the majority of Medicaid spending because FFS enrollees are more likely to be elderly and disabled.40 Moreover, many states require that rates within Medicaid managed care match FFS rate changes, at least in part.41 We did not find evidence, however, on whether FFS fee changes had spill-over effects on Medicaid managed care enrollees. This could be, in part, due to the difficulty obtaining reliable and comprehensive encounter data for Medicaid managed care enrollees.42, 43

This review highlights the broader challenge of studying the Medicaid program and the critical need for timely and consistent data. There is about a 4-year lag to availability of Medicaid Analytic Extract (MAX) data from the Centers for Medicare and Medicaid Services (CMS) for many states, in part due to the effort needed to standardize and validate data across states. CMS efforts to improve the quality, comparability, and timeliness of Medicaid data through the Transformed Medicaid Statistical Infrastructure System (T-MSIS) could serve as an important building block for improving the Medicaid research infrastructure. Improving the ability to link these data with additional information, such as vital statistics, local and national disease registries, and electronic health records would improve our ability to study the impact of Medicaid policy on health outcomes, for which there is currently limited evidence.

Limitations

Our review is subject to limitations. We did not include a start date for our review, which spans nearly four decades. During this time the Medicaid program has undergone numerous changes, including large increases in the number of beneficiaries enrolled in managed care plans and expansions in the populations eligible for Medicaid; these changes could impact the effects of fee changes on provider participation and beneficiary access to care.

Conclusion

In summary, this review did not find strong evidence that increasing Medicaid provider fees positively affects provider participation in the program, although some measures of enrollee access improved with higher fees. Fee changes should be accompanied with careful monitoring of changes in access to care and downstream outcomes. Attention is also needed to identify and implement other strategies that could have more consistent effects on care utilization and quality.