Evaluating Part D Sponsors' Pharmacy Networks: Cost Access for Urban, Suburban, & Rural Be, Slides of Pharmacy

Download Evaluating Part D Sponsors' Pharmacy Networks: Cost Access for Urban, Suburban, & Rural Be and more Slides Pharmacy in PDF only on Docsity! Page 1 of 20 Analysis of Part D Beneficiary Access to Preferred Cost Sharing Pharmacies (PCSPs) April 28, 2015 Page 2 of 20 This page is intentionally left blank. Page 5 of 20 analysis. We used approved plan benefit and service area data from the Health Plan Management System (HPMS), pharmacy network data submitted by plans for posting in the Medicare Plan Finder, and the Medicare Part D eligible beneficiaries per ZIP Code file, to calculate access levels to PCSPs overall and by geographic area (urban, suburban, and rural) at the plan benefit package (PBP) level for both PDPs and MA-PDs. As a point of reference, we used the convenient access standard for entire networks to better understand beneficiary access to PCSP in their plans. Key findings on PCSP networks and the levels of access to preferred cost sharing provided by Part D sponsors are summarized below. Most standalone Prescription Drug Plans have PCSP networks, while only about one out of six Medicare Advantage Prescription Drug Plans have PCSP networks. In 2014, there were 1,186 plans under 62 PDP contracts. Of these, 849 plans (72%) under 37 PDP contracts (60%) offered PCSP networks. There were also 2,244 plans under 592 MA-PD contracts. PCSP networks were offered by 354 plans (16%) under 110 MA-PD contracts (19%). The average PCSP network (including both MA-PDs and PDPs) includes only 24 percent of the pharmacies that are included in the entire retail network, but PCSP networks range from only 1 network pharmacy to 99 percent of the entire network. Among plans with PCSP networks, the average size of the entire network (preferred and standard cost sharing pharmacies) is 65,395 pharmacies, with an average of 15,480 pharmacies in the PCSP network. Hence, the average PCSP network includes 24 percent of the pharmacies available in the entire network. Compared to PDP PCSP networks, the MA-PD PCSP networks are slightly smaller in both number and proportion of entire network pharmacies. As expected, PCSP networks offer less access than entire pharmacy networks, with the biggest differences in urban areas. On average, entire pharmacy networks meet or exceed the TRICARE access standard. That is, they provide 99 percent of urban beneficiaries access to a pharmacy within 2 miles, 99 percent of suburban beneficiaries access within 5 miles, and 97 percent of rural beneficiaries access within 15 miles (Table I).3 In contrast, the average PCSP network meets the convenient access benchmarks for suburban and rural beneficiaries but not for urban beneficiaries. On average, PCSP networks offer 79 percent of urban beneficiaries access to a PCSP within 2 miles, which is below the 90 percent standard set for entire networks. One in ten PCSP networks offers 2-mile access to fewer than 40 percent of urban beneficiaries in their plans’ service areas. 3 Of these plans, 3 parent organizations have waivers, covering 4 contracts and 10 plans, allowing them to offer a network that does not meet the convenient access standard. Page 6 of 20 Table I: Population Proportions of Beneficiaries within the Convenient Access Benchmarks for Entire Pharmacy Networks and PCSP Networks PCSP Network Mean Proportion of Beneficiaries within Distance Benchmarks Urban - Within 2 Miles Suburban - Within 5 Miles Rural - Within 15 Miles Entire Plan Networks 99% 99% 97% PCSP Networks 79% 94% 88% Number of Plans* 1,186 1,192 1,199 Note: *The totals are less than 1,203 as 17 MA-PD plan areas have no urban ZIP Codes, 11 have no suburban ZIP Codes and 4 have no rural ZIP Codes. Almost half of all PCSP networks meet the entire network convenience access benchmark for their plans’ service areas. While PCSP networks are not required to meet the convenient access standard set for entire pharmacy networks, we find that these networks often do meet it. Of the 1,203 PCSP networks, 548 (46%) meet the convenient access benchmark for their plans’ service areas (Table II). Table II: PCSP Networks Meeting the Convenient Access Benchmark Plans Meeting Convenient Access Benchmark Across Whole Service Areas Plans Meeting Convenient Access Benchmark by Population Density Urban (within 2 miles) Suburban (within 5 miles) Rural (within 15 miles) No. of Plans offering Preferred Cost Sharing 1,203 1,186 1,192 1,199 No. of Plans Providing Convenient Access 548 545 1,040 1,134 % of Plans Providing Convenient Access 46% 46% 87% 95% PCSP networks largely meet suburban and rural convenient access benchmarks but are less likely to meet the urban convenient access benchmark. PCSP networks are most likely to meet the rural convenient access benchmark, with 95 percent providing 15-mile access to a PCSP for at least 70 percent of the rural beneficiaries in their plans’ service area. Eighty-seven percent of PCSP networks provide 5-mile access to a PCSP for at least 90 percent of suburban beneficiaries in their plans’ service areas. In contrast, less than half of PCSP networks meet the urban access benchmark. If the urban distance benchmark were changed from 2 miles to 4.5 miles, 90% of plans would meet the revised benchmark in 2014. Since a number of plans did not meet the convenient access distance benchmark, we calculated the distance required for the vast majority of plans to meet the 90% urban/90% suburban/70% rural beneficiary access benchmarks. Distance simulations are shown in Table III. In order for 90% of plans to provide their urban and suburban beneficiaries with convenient access to a PCSP, the distance benchmark would need to be changed to 4.5 miles (instead of 2 miles) and 10 miles (instead of 5 miles), respectively. Page 7 of 20 Travelling 4.5 miles in an urban area could take an exceedingly long time, or become inaccessible for persons relying on public transportation. We found six plans that require beneficiaries residing in urban areas to travel an average distance of 30 miles or more to reach the nearest PCSP. Table III: Simulation of Different Distance Benchmarks for PCSP Networks Additional findings include (not discussed in the body of the report; detailed tables not shown): • While PCSP networks provide similar levels of access in suburban and rural areas across states, there is substantial variation in urban access across states. For example, among PDPs, the percent of PCSP networks meeting the urban benchmark ranged from 12 percent in Virginia to 86 percent in Arkansas. • PCSP networks that are largely comprised of only major retail chain pharmacies are less likely to meet the convenient access benchmarks compared to those networks with a mix of chain or independent pharmacies. Additionally, PCSP networks with a single pharmacy chain rarely meet the convenient access benchmark and account for most of the PCSP networks with the lowest access for urban beneficiaries. For example, of 190 PCSP networks composed of a single pharmacy chain, only 7 meet the convenient access benchmark for their service areas. Of the 127 PCSP networks that provide 2-mile access to fewer than 40 percent urban beneficiaries, 106 use single-chain networks. • There do not appear to be any notable differences in access by beneficiary race/ethnicity, age, gender, or low income subsidy (LIS) status. We conclude that many beneficiaries live in areas where Part D plans provide reasonably robust PCSP networks. However, some beneficiaries in all areas, but particularly those in urban areas, face limited, or in some instances, no access to PCSPs. These beneficiaries may be unable to obtain the lower cost sharing afforded to other members of the same plan, based solely on where they reside within the service area. While even the TRICARE standards reflect the understanding that there may always be some beneficiaries without convenient access to a nearby pharmacy, these results indicate a larger concern. As CMS evaluates policy options for addressing access to PCSPs, we will carefully weigh numerous competing considerations, including the needs of beneficiaries, burden on plans, and effects on Part D costs. 90% of Urban Beneficiaries within: 2 miles 2.5 miles 3 miles 3.5 miles 4 miles 4.5 miles 90% of Suburban Beneficiaries within: 5 miles 6 miles 7 miles 8 miles 9 miles 10 miles 70% of Rural Beneficiaries within: 15 miles 17 miles 19 miles 21 miles 23 miles 25 miles % PCSP Networks Meeting Plans’ Service Areas 46% 68% 79% 83% 87% 90% % PCSP Networks Meeting Urban 46% 68% 80% 84% 88% 90% % PCSP Networks Meeting Suburban 87% 93% 95% 95% 96% 97% % PCSP Networks Meeting Rural 95% 95% 96% 97% 97% 98% Page 10 of 20 III. Data and Methods This analysis is based on 2014 Medicare Advantage Prescription Drug (MA-PD) and stand-alone Prescription Drug Plan (PDP) plans offering PCSP networks in their benefit packages. According to HPMS, the system of record for approved plan benefit packages, in 2014 there were 2,244 plans offered by 592 MA-PD contracts, and 1,186 plans offered by 62 PDP contracts. A contract may offer multiple plans within or across service areas. Plans differ under a contract by service area, formulary tiers, or other benefit design choices, including the inclusion of PCSP networks. Of the 592 MA-PD contracts, 110 offer 354 plans with PCSP networks; 37 PDP contracts offer 849 plans with PCSP networks. Plans with PCSP networks are found in each of the 50 states, Washington DC, and Puerto Rico. To understand access in PCSP networks, this analysis measures beneficiary access for each of the PCSP networks offered by plans under both MA-PD and PDP contracts using data from the spring of 2014. As a comparison to the PCSP beneficiary access results, we also measure access to entire pharmacy networks across plans’ entire service areas. This provides a necessary control for variations in the supply of providers across plan service areas, as well as the differences in the distributions of beneficiaries across urban, suburban, and rural areas. Below, we review the data sources and the methodology used in developing the PCSP access results. A. Data Sources Our access analysis draws on five data sources. We use these data to identify: the plans offering PCSPs; each plan’s CMS-approved service area; the network pharmacies at which beneficiaries can access preferred cost sharing; and geographic location information for pharmacies and beneficiaries. They are described below. Medicare Plan Finder (MPF) Output Data: We use MPF data submitted on March 17, 2014 and displayed on MPF from March 31, 2014 through April 13, 2014. This data, which assists beneficiaries in choosing plans, provides a list of all pharmacy National Provider Identifiers (NPIs) in each plan, and show each pharmacy’s preferred cost sharing status. Health Plan Management System (HPMS): HPMS is the system of record for the Medicare Part D program, including the approved plan benefit packages, and plan service areas. The MPF data is validated against approved benefit package features in HPMS to ensure that plans offering a PCSP benefit are submitting both preferred and standard cost share pharmacies in their MPF lists, and that plans without the PCSP structure submit only standard cost share pharmacies. HPMS also provides ZIP Code level service area data for each plan. Page 11 of 20 National Council for Prescription Drug Programs (NCPDP): The NCPDP provides the business address for each pharmacy, matched by NPI. This database is used by the Medicare Plan Finder website and is expected to have the most accurate and up-to-date address information for pharmacies. Prescription Drug Coverage Contracting Beneficiary Count Data: These files provide the count of Medicare Part D-eligible beneficiaries at the ZIP Code level and are used by Part D sponsors to calculate geographic access measures for evaluating compliance with the convenient access to network retail pharmacies standards. In addition to total counts of eligible Medicare Part D beneficiaries, CMS provided ZIP Code counts by different beneficiary characteristics. Specifically, we use counts by race, age, gender, and low income subsidy (LIS) status. These counts allow us to assess access for different groups of beneficiaries. National Plan and Provider Enumeration System (NPPES): We use the NPPES to augment the address information from the NCPDP data, as not all pharmacies listed in the MPF data are identified in the NCPDP database. For those pharmacies that are not in the NCPDP database, we use the business address from the NPPES. B. Access Measurement and Methods This analysis was conducted for each PDP and MA-PD plan offering a preferred cost sharing benefit, using CMS-approved service areas in HPMS. The results tell us the percent of beneficiaries residing within certain distances from a network pharmacy, defined by each plan’s service area. There are two key access measures: • The percent of plans’ beneficiaries residing within the convenient access distance benchmark to a PCSP, by geographic area (urban, suburban, rural), by beneficiary group (age, gender, race, LIS status), and by group and geographic area. • The plans with PCSP networks meeting the convenient access benchmark, by geographic area type (urban, suburban, rural), by beneficiary group (age, gender, race, LIS status), and by group and geographic area. A key distinction for this analysis is the population density of the geographic areas where beneficiaries reside, specifically urban, suburban, and rural population densities. These densities are measured at the ZIP Code level and are defined as: • Urban: more than 3,000 persons per square mile • Suburban: 1,000 to 3,000 persons per square mile • Rural: under 1,000 persons per square mile. In addition to population density definitions, there are several definitions used to categorize beneficiary characteristics. Gender is defined as male and female beneficiaries. Age in years is defined by three groups: below 65, 65 to 74, and 75 and above. Race/ethnicity groups are categorized by: African American, Hispanic, Asian, White, Other, and Unknown. Finally, LIS Page 12 of 20 status, used as a proxy for income level, is a dichotomy: LIS eligible or not LIS eligible. The definition of LIS eligible is any beneficiary that is classified as eligible for full Medicaid benefits, partial Medicaid benefits, or the low income subsidy based on income and resources. We use the geocoding software package Quest Analytics® to calculate geographic access measures for the groups identified above. Quest Analytics® geocodes pharmacy addresses and beneficiary residence at the ZIP Code level by creating longitude and latitude measures for pharmacies and beneficiaries. These geocoded locations enable us to determine distances from beneficiaries to network pharmacies, which in turn facilitate the establishment of geographic access measures for different beneficiary groups. For pharmacies, we geocode locations using the pharmacy addresses identified in the NCPDP. Invalid pharmacy addresses in the NCPDP are flagged by Quest Analytics, which then estimates the pharmacy location within the ZIP Code area using information from the US Postal Service and the US Census Bureau to mimic population patterns within the ZIP Code area. If the ZIP Code is invalid, we replace the NCPDP address with the business address provided in the NPPES. We also use the NPPES to identify retail pharmacy addresses for those pharmacies in the MPF data that are not found in the NCPDP database. If the NPPES address is also not valid, we drop the pharmacy location from the analysis. Using this approach, we are able to geocode all but 107 of the 69,487 pharmacies that were identified as offering preferred cost sharing in at least one plan. Of all the 1,203 PCSP networks, these non-geocoded pharmacies represent no more than 0.3 percent of any single network. Geocoding beneficiary locations differs from the above process, as beneficiary addresses are not used. Instead, Quest Analytics® applies a proprietary “distributive geocoding” method to map beneficiary counts by ZIP Code to locations within a ZIP Code area using population locations from US Postal and US Census Bureau information. Consequently, Quest Analytics® approximates beneficiary locations to generally match the distribution of the overall population patterns in an area (as opposed to approximating all beneficiaries at the population centroid in the ZIP Code, for example). The Quest Analytics® process of estimating beneficiary locations used in this analysis mirrors the approach taken by sponsors when assessing network access, as beneficiary addresses are not available for such analysis. However, a key difference in our method is the use of complete beneficiary counts; sponsors often use a representative sample (approximately 5%) of the beneficiary population. Using the representative 5 percent sample of the total Medicare Part D eligible population greatly reduces the processing time required to measure distances between beneficiary and pharmacy locations without great cost to the reliability of calculated access measures. However, because we split beneficiary populations by different characteristics for this analysis, we use the complete beneficiary population. With longitude and latitude identified for each pharmacy and beneficiary, Quest Analytics® calculates the estimated driving distance to the nearest network pharmacy for each beneficiary and produces a beneficiary-level file for each plan. This output file is used to construct a plan- Page 15 of 20 their suburban and rural beneficiaries. These plans provide an average of 94 percent of suburban beneficiaries access to a PCSP within 5 miles, which exceeds the 90 percent benchmark for suburban beneficiaries. The average proportion of beneficiaries with access to a PCSP within 15 miles in rural areas is 88 percent, well above the 70 percent benchmark. But whereas PCSP networks tend to meet the suburban and rural benchmarks, they fall short of meeting the urban benchmark. The PCSP networks offer, on average, 79 percent of urban beneficiaries access to a PCSP within 2 miles, well below the 90 percent standard set for entire networks. Table 2: Population Proportions of Beneficiaries within the Convenient Access Standard for Entire Pharmacy Networks and within the Convenient Access Benchmark for PCSP Networks Network Mean (Std. Deviation) Proportion of Beneficiaries within Distance Standard/Benchmark Urban - Within 2 Miles Suburban - Within 5 Miles Rural - Within 15 Miles Convenient Access Level 90% 90% 70% Entire Plan Networks 99% (4%) 99% (3%) 97% (4%) PCSP Networks 79% (25%) 94% (14%) 88% (13%) Number of Plans* 1,186 1,192 1,199 Note: The numbers in parentheses are standard deviations. * The totals are less than 1,203 as 17 MA-PD plan areas have no urban ZIP Codes, 11 have no suburban ZIP Codes, and 4 have no rural ZIP Codes. C. Percent of PCSP Networks Meeting Convenient Access Benchmarks In addition to assessing the average percent of beneficiaries with access to a PCSP within established distances, we also examine whether PCSP networks meet the convenient access benchmark. Although the convenient access standard applies to only a plan’s entire network, this standard can be used as a benchmark for understanding beneficiaries’ access to PCSP networks and for comparing access for different groups of beneficiaries. Table 3 shows the percent of plans that meet the benchmark across the three geographic areas, holding constant both the mileage and access percentages. At 46 percent, almost half of all PCSP networks meet the convenient access benchmark for their whole service area. That is, 46 percent of plans have PCSP networks offering 2 mile access to 90 percent or more of urban beneficiaries, 5 mile access to 90 percent or more of suburban beneficiaries, and 15 mile access to at least 70 percent of suburban beneficiaries. In examining results by population density, we see that the vast majority of PCSP networks meet the suburban or rural benchmarks for the entire network, with PCSP networks most likely to meet the rural benchmark. However, plans are more likely to fall below the urban Page 16 of 20 benchmark (i.e., providing 90 percent of urban beneficiaries access to a pharmacy within 2 miles), with only 46 percent of PCSP networks meeting this benchmark. In fact, our analysis shows that if a PCSP network does not meet the benchmark overall, it nearly always falls below the urban one. In comparing MA-PDs and PDPs, a larger proportion of MA-PD plans meet the convenient access benchmarks: 55 percent of MA-PD PCSP networks meet these benchmarks compared to 42 percent of stand-alone PDPs.11 Table 3: PCSP Networks Meeting the Convenient Access Benchmark, by Population Density12 Plans (number and percent) Meeting Convenient Access Benchmark Geographic Subsets Urban (within 2 miles) Suburban (within 5 miles) Rural (within 15 miles) No. of Total Plans with PCSP Networks 1,203 1,186 1,192 1,199 No. Meeting Convenient Access Benchmark 548 545 1,040 1,134 % Meeting Convenient Access Benchmark 46% 46% 87% 95% No. of PDP Plans with PCSP Networks 849 849 849 849 No. Meeting Convenient Access Benchmark 353 363 743 808 % Meeting Convenient Access Benchmark 42% 43% 88% 95% No. of MA-PD Plans with PCSP Networks 354 337 343 350 No. Meeting Convenient Access Benchmark 195 182 297 326 % Meeting Convenient Access Benchmark 55% 54% 87% 93% Figure 1 shows the number of plans providing convenient access to different proportions of beneficiaries by urban, suburban, and rural status. The figure demonstrates that plans provide better access to a PCSP for rural beneficiaries than their urban counterparts, even when considering the difference between the urban and rural benchmarks (90 percent versus 70 percent, respectively). A total of 624 plans offer convenient access to a pharmacy offering preferred cost sharing to 90 percent or more of rural beneficiaries, while 545 plans offer convenient access to 90 percent or more of urban beneficiaries. However, some PCSP networks 11 PDP service areas by statute cover at least an entire state and may cover several states. In contrast, MA-PD plans may cover a portion of a county, several counties, or larger areas. Thus it may be easier for an MA-PD plan to meet the benchmark. 12 Since MA-PD plan service areas consist of counties and partial-counties and there are counties with no urban, suburban, or rural ZIP Codes, MA-PD plan service areas can consist of combinations of: urban, suburban, and rural beneficiaries; urban and suburban beneficiaries; urban and rural beneficiaries; suburban and rural beneficiaries; urban beneficiaries only; suburban beneficiaries only; rural beneficiaries only. The maximum number of networks available by population density, therefore, is not necessarily the maximum number of networks overall. Page 17 of 20 are providing access to a very low proportion of resident beneficiaries in each population density area. For urban areas the outlier or the lowest 10% of PCSP networks provide access to less than 40% of urban beneficiaries. On the other hand, for urban and rural areas the outlier PCSP plans provide convenient access to 87% and 77% of their beneficiaries, respectively. Figure 1: Number of Plans Providing Convenient Access to Different Proportions of Urban, Suburban, and Rural Beneficiaries D. Simulation of Different Distance Benchmarks for PCSP Networks As shown in the previous section, most PCSP networks do not meet the convenient access standard established for entire pharmacy networks. This largely results from PCSP networks falling short of providing 2-mile access to 90 percent of urban beneficiaries. In contrast, an overwhelming majority of PCSP networks meet the convenient access benchmark for suburban and rural beneficiaries, where the distance benchmarks are more generous. To better understand this further, we examined distances to a PCSP, and conducted a simulation analysis that changes the mileage component of the convenient access benchmark.