MAX MEDICAID POLICY BRIEF • CENTERS FOR MEDICARE & ME D I C A I D S E RV I C E S Brief 7 July 2012 Assessing the Usability of MAX 2008 Encounter Data for Enrollees in Comprehensive Managed Care Vivian L. H. Byrd, Allison Hedley Dodd, Rosalie Malsberger, Ashley Zlatinov A s growing numbers of Medicaid enrollees receive health About This Series benefits through comprehensive managed care, researchers and The MAX Medicaid policy issue brief series highlights policymakers seeking to understand the service use of these the essential role MAX data can play in analyzing the enrollees must rely on encounter data that states receive from Medicaid program. MAX is a set of annual, person-level managed care plans. However, not all states report encounter data files on Medicaid eligibility, service utilization, and data submitted by their plans into the Medicaid Statisti- payments that are derived from state reporting of Medicaid cal Information System (MSIS) and, until recently, little was eligibility and claims data into the Medicaid Statistical Infor- known about the data’s usability for research. In this issue mation System (MSIS). MAX is an enhanced, research- brief, we report on the availability, completeness, and quality friendly version of MSIS that includes final adjudicated of encounter data for physician, clinic, and outpatient services claims based on the date of service, and data that have for the first time and update a recent assessment of the inpatient undergone additional quality checks and corrections. CMS and prescription drug files to judge the usability of the 2008 produces MAX specifically for research purposes. For Medicaid Analytical eXtract (MAX) encounter data, which are more information about MAX, please visit: http://www. derived from MSIS. cms.gov/Research-Statistics-Data-and-Systems/Com- puter-Data-and-Systems/MedicaidDataSourcesGenInfo/ Background MAXGeneralInformation.asp. As states expand their use of managed care arrangements to pro- The MAX encounter data briefs are meant to inform research- vide services to Medicaid enrollees, researchers and policymakers ers and policymakers as they decide whether and how to use will need to analyze additional types of data to assess their encounter data for their research. MAX was designed to enable service use. With 50 percent of all full-benefit Medicaid enrollees research on Medicaid enrollment, service utilization, and enrolled in comprehensive managed care in 2008, relying on expenditures by calendar year at the enrollee level. Analysis fee-for-service (FFS) data to determine the service use of the by calendar year is particularly important with encounter data Medicaid population is no longer sufficient (Borck et al. 2012). because some states that submit them do not do so in every To capture the service use of comprehensive managed care enroll- quarterly MSIS submission (Byrd et al. 2011). ees, encounter data—that is, claims records that contain informa- tion on utilization but none on Medicaid expenditures—must be In a previous issue brief, we examined MAX 2007 data to evaluated as well. To ensure that managed care enrollees receive assess the availability, completeness, and quality of encounter the same level and quality of services as FFS enrollees, several data for inpatient hospital (IP) and prescription drug (RX) data states perform comprehensive checks on the data that they receive from health maintenance organization (HMO)/health insur- from managed care plans; however, the quality of the encounter ing organization (HIO) plans (Dodd et al. 2012). This initial data submitted by the states to the Medicaid Statistical Informa- look at the usability of encounter data was focused on IP and tion System (MSIS) is not clear (Byrd et al. 2011). Encounter data RX claims because, according to actuaries and state Medicaid do not undergo the same validation and quality checks in either officials, encounter data for these services—which are provided MSIS or MAX processing that FFS data undergo. by a relatively small number of providers—are typically easier 1 to collect and may be more complete than “other services” Table 1. Overview of Encounter Data Available (OT) data. Most of the service use among Medicaid enrollees, in MAX 2007 and 2008 for HMO/HIO Enrollees, however, including physician, clinic, and outpatient services, by File Type is captured in the OT file. Therefore, while the results of the initial assessment of the quality of the IP and RX encounter Number of Number of Number of data were encouraging, the data analyzed would only cover a States with States with Encounter limited subset of enrollee service use. File Type Data, 2007a Data, 2008a Claims, 2008 IP 25 29 1,947,019 In this brief, we use MAX 2008 data to update the IP and RX LT 18 22 560,201 encounter data assessment, and we extend our analysis to OT 27 34 350,312,637 physician, outpatient, and clinic services in the OT file. To be RX 18 20 87,573,721 usable, the data needed to be of comparable completeness and Source: Mathematica’s analysis of MAX 2007 and 2008 data. quality to FFS data. The remainder of this issue brief explains a Includes all states that submitted encounter data regardless of the level of how we conducted the analysis and elaborates on the results. HMO/HIO participation in the state, the number of claims submitted, or whether prescriptions were covered as part of the comprehensive managed care program. We plan to assess the usability of 2009 MAX encounter data when they are available. Methods The OT file may contain up to 22 types of service, while IP We constructed our analysis using the information in enrollees’ may contain four, and RX two. For the OT analysis, we chose basis-of-eligibility (BOE) category—that is, each enrollee’s clas- physicians (type of service = 08), outpatient hospital (type of sification as adult, child, disabled, or aged—to facilitate more service = 11), and clinic (type of service = 12) because these accurate state-by-state comparisons. The average 2008 capita- are services routinely sought and covered under Medicaid in tion payment for enrollees in comprehensive managed care was all states, and managed care plans are accustomed to collecting much lower for adults and children than for aged and disabled and reporting these data for quality assurance, such as for the beneficiaries, an indication that the expected level of service use, Healthcare Effectiveness Data and Information Set (HEDIS). and therefore the expected volume of encounter claims, is lower We included “inpatient hospital” (type of service = 01) from among adults and children (Borck et al. 2011). As states vary the IP file because while the IP file may contain three other widely in terms of the mixture of Medicaid populations enrolled types of service, “inpatient hospital” represents the vast major- in capitated managed care programs, examining the volume of ity of claims and services in the inpatient setting. We included encounter data submissions for all groups within a state could be “prescribed drugs” (type of service = 16) from the RX file, but misleading. Many states rolled out comprehensive managed care did not include durable medical equipment. to children and adult enrollees first, and only some have enrolled Since analyzing them both individually and together did not the aged and disabled populations in it. yield substantial differences, we decided to present physician, The proportion of aged and disabled enrollees with dual eligi- outpatient, and clinic services as a whole in this brief. Other bility (that is, eligible for both Medicaid and Medicare) could types of services included in the OT file might not be as easily also affect the claims for service utilization within a BOE. comparable across states or as complete. For example, the We excluded dual-eligible enrollees from our 2008 analysis volume of rehabilitation or occupational therapy services relies because the volume of encounter data is lower than those for heavily on how a state counts units of service, which can range non-dual enrollees, since many services they receive are cov- on claims from 15-minute increments to hour-long visits. There ered by Medicare (Young et al. 2012). were too few LT encounter claims for a cross-state analysis.1 In MAX 2008, encounter data for comprehensive managed We considered a state to have managed care if at least one care enrollees were available for over half of the states in at percent of enrollees participated in comprehensive managed least one type of file: IP, long-term care (LT), OT, or RX; and care at some point during the year. For prescription drug the number of states with data had increased in each file type services, we excluded 12 states whose managed care arrange- from 2007 (Table 1). We limited our analysis to fully capitated ments did not include prescription drug benefits. Because (comprehensive) managed care arrangement HMO/HIO plans states with low managed care enrollment are less likely to because they cover the widest range of services and because we devote resources to producing high-quality encounter data, we anticipated they would have the highest quality encounter data. analyzed data for a particular BOE group only if 10 percent 2 or more of full-benefit2 Medicaid enrollees within that group Because managed care coverage varies by state and type of were enrolled in an HMO/HIO plan. We did not analyze data enrollee, we evaluated the completeness and quality measures for a particular BOE group in a state if it had fewer than 200 for OT, IP, and RX data separately for each BOE for each state. claims because measures based on a small number of records To create comparison metrics, we calculated the average 2008 could skew estimates. value and standard deviation for each completeness and quality metric for each BOE using the full-benefit, non-dual FFS popu- Metrics lation across all states with substantial FFS participation. For each comparison metric, we used the average FFS value as the To be usable, encounter data needed to be both complete and of midpoint of our reference range. We set the top of the reference comparable quality to FFS data for our analysis. We conducted range at two standard deviations above the FFS average, and our analysis in two phases to account for these two character- the bottom at two standard deviations below the FFS average. istics. To judge completeness, we looked at two measures that We considered the reference range to be the acceptable range of assessed the volume of encounter data—the average number of values for the 2008 encounter data for that metric. The state’s claims and the percentage of enrollees with claims. To evaluate encounter data value was considered “good” if it fell within the quality, we used metrics that assessed the amount or quality of reference range. For certain measures, state values were highly information on the encounter itself. For the analysis of the OT skewed but typically either close to 100 percent or 0 percent encounter claims, we chose to use two quality measures for both for both FFS and encounter data. Rather than use the reference the diagnosis code and procedure code fields—one indicating range based on the average value, we defined a “good” value as whether the field was filled and the second analyzing the format 90 percent or greater for these measures. of the data in the field. For diagnosis code, we expected the field to be filled at a high rate because few physician, outpatient, For each BOE that met the analysis criteria, we compared the and clinic services claims are paid without a diagnosis code. state’s value to the FFS reference metric to determine if it fell We wanted to determine, however, how the diagnosis codes on within the acceptable range; the ranges are presented in Table 2. encounter claims compare in the level of specificity to those The number of states that fell within the range is shown in reported on FFS claims. The more characters in the diagnosis parentheses for each measure. For example, 23 of the 24 states code (more than 3 characters), the more specific the diagnosis is that met the thresholds for our analysis of OT data for adults had on the claim or encounter. Similarly, we expected the procedure an average number of OT encounter claims per enrollee between codes to be filled at a high rate, but the heavy reliance of some 1.04 and 12.10, inclusively. For the OT, IP, and RX data, states on procedure codes specific to the state make a national “complete” was defined as having values within the acceptable analysis more complicated. We examined whether the proce- range for at least one of the two completeness metrics for that dure codes were filled and whether the reported data were in the data type. For the OT data, “comparable quality” was defined standard national format. For the IP file, we created one quality as satisfying at least four of the five quality measures. For the measure for each of four fields that undergo scrutiny during the IP data, “comparable quality” was defined as satisfying at least MSIS data quality and validation review process, and for the RX three of the four quality measures. For the RX data, “comparable file, we created one quality measure for each of two fields that quality” was defined as satisfying at least one of the two quality we expect to see routinely filled on FFS claims. The metrics used measures. A BOE within a state was considered to have “usable” for evaluation of completeness and quality are shown in Table 2. data if the encounter data for that BOE met both the “complete” and “comparable quality” criteria. 3 Table 2. Metrics Developed to Analyze Medicaid Encounter Data in MAX 2008 Reference Range (Number of States Meeting Metric) Data Element Adults Children Disabled Aged OT– Physician, Clinic, and Outpatient Visits Completeness Measures Average number of OT encounter claims per enrollee 1.04–12.10 1.23–9.46 8.35–27.96 0.91–19.54 (23 of 24) (22 of 25) (15 of 20) (13 of 16) Percentage of enrollees with OT encounter claims 34.33–92.45 36.15–93.40 66.35–92.39 19.57–92.26 (22 of 24) (23 of 25) (14 of 20) (15 of 16) Quality Measures Percentage of OT encounter claims with place of 83.87–100 76.16–100 81.89–100 84.22–100 service code (23 of 24) (25 of 25) (20 of 20) (16 of 16) Percentage of OT encounter claims with primary 98.17–100 86.09–100 94.84–100 97.02–100 diagnosis code (24 of 24) (25 of 25) (20 of 20) (16 of 16) Percentage of OT encounter claims with a primary 90.85–98.81 80.92–100 88.08–100 89.16–99.41 diagnosis code length greater than 3 characters (23 of 24) (25 of 25) (20 of 20) (16 of 16) Percentage of OT encounter claims with a 71.47–100 82.13–100 78.78–100 82.68–100 procedure (service) code (20 of 24) (21 of 25) (17 of 20) (13 of 16) Percentage of OT encounter claims with a 60.77–100 64.32–100 66.88–100 70.41–100 procedure code in CPT-4 or HCPCS format (21 of 24) (22 of 25) (18 of 20) (15 of 16) IP–Inpatient Hospital Completeness Measures Average number of IP encounter claims per enrollee 0.00–0.40 0.02–0.15 0.10–0.54 0.00–0.44 (22 of 24) (18 of 24) (16 of 20) (14 of 15) Percentage of enrollees with IP encounter claims 0.21–32.51 1.06–13.08 7.55–25.39 3.62–22.39 (23 of 24) (20 of 24) (15 of 20) (11 of 15) Quality Measures Average length of stay 2.01–3.90 2.04–6.48 5.35–8.61 3.32–10.49 (23 of 24) (22 of 24) (9 of 20) (14 of 15) Average number of diagnosis codes 2.42–6.43 1.89–4.38 3.09–9.76 3.19–10.72 (20 of 24) (20 of 24) (16 of 20) (12 of 15) Percentage of IP claims with procedure codes 48.17–100.00 18.72–76.39 30.70–71.13 25.05–73.55 (18 of 24) (23 of 24) (15 of 20) (13 of 15) Percentage of IP claims with UB accommodation Values of ≥ 90% Values of ≥ 90% Values of ≥ 90% Values of ≥ 90% codes (20 of 24) (20 of 24) (13 of 20) (11 of 15) RX–Prescription Drugs Completeness Measures Average number of RX encounter claims per enrollee 1.86–12.95 1.80–7.22 17.27–50.09 0–48.22 (13 of 14) (14 of 15) (8 of 10) (8 of 8) Percentage of enrollees with RX encounter claims 26.79–88.04 31.46–80.84 68.14–89.30 12.21–89.82 (13 of 14) (14 of 15) (9 of 10) (7 of 8) Quality Measures Percentage of RX claims with date prescribed Values of ≥ 90% Values of ≥ 90% Values of ≥ 90% Values of ≥ 90% (13 of 14) (14 of 15) (9 of 10) (7 of 8) Percentage of RX claims with quantity Values of ≥ 90% Values of ≥ 90% Values of ≥ 90% Values of ≥ 90% (8 of 14) (9 of 15) (6 of 10) (4 of 8) Source: Mathematica’s analysis of the MAX 2008 IP, RX, OT, and Person Summary (PS) files. Note: The parenthetical data show the number of states that had values within the acceptable range, out of the total number of states that had sufficient participation and encounter claims submitted for analysis. UB = uniform billing, CPT-4 = Current Procedural Terminology, 4th Edition, HCPCS = Healthcare Common Procedure Coding System. 4 Findings criteria for any BOE. The remaining 6 states met the criteria for some BOEs but not others. OT Encounter Data IP Encounter Data Table 3 summarizes the availability, completeness, and quality of the OT encounter data for each state by BOE. Figure 1 illus- Table 4 summarizes the availability, completeness, and qual- trates how the criteria applied at each step of the analysis elimi- ity of the IP encounter data for each state by BOE. Figure 2 nated states from meeting the usability criteria. For example, illustrates how the criteria applied at each step of the analysis 35 states had comprehensive managed care at some point dur- eliminated states from meeting the usability criteria. The com- ing 2008. At least 10 percent of adult enrollees participated in pleteness and the quality of the IP encounter data were high. comprehensive managed care in 32 of these 35 states. Of these They were considered usable for at least one BOE category for 32 states, 24 (75 percent) submitted OT encounter claims for 22 of the 25 states that submitted these data (88 percent). Thir- adults. The completeness of the adult OT encounter data was teen states (52 percent) provided usable data for all of the BOE high, with 23 of 24 states submitting complete data. The qual- categories for which they submitted data (Arizona, Hawaii, ity of the encounter data was high as well, with 21 of 24 states Indiana, Kansas, Kentucky, Missouri, Nebraska, New Jersey, submitting data of comparable quality to the FFS data. Because New Mexico, Oregon, Virginia, Washington, and Wisconsin). they met the criteria for both completeness and quality, the OT This is the same number as in 2007; however, Minnesota, encounter data for adult enrollees are considered usable for 21 which met this threshold in 2007, did not meet it in 2008, while states (88 percent) that submitted data. Virginia met the threshold in 2008, but not in 2007 (data not shown). Of the 25 states that submitted suitable IP encounter Data can also be considered usable for 22 of the 25 states (88 data for the analysis, three states did not meet the criteria for percent) submitting data for children. Fifteen of the 20 states usability for any BOE. The remaining 9 states met the criteria submitting data for disabled enrollees (75 percent) met both for some BOEs but not others. completeness and quality thresholds, and of the 16 states sub- mitting encounter claims for the aged, 14 submitted data that RX Encounter Data can be considered usable. Table 5 summarizes the availability, completeness, and quality of The OT encounter data were considered usable for at least the RX encounter data for each state by BOE. Figure 3 illustrates one BOE category for 24 (96 percent) of the 25 states that how the criteria applied at each step of the analysis eliminated submitted these data. Eighteen states (72 percent) provided states from meeting the usability criteria. Almost every state usable encounter data for all the BOE categories for which that submitted RX encounter data submitted data that were they submitted data (Arizona, Delaware, Georgia, Hawaii, complete and of comparable quality to FFS data for every BOE Illinois, Indiana, Kansas, Kentucky, Michigan, Missouri, group. Thirteen states provided usable data for every BOE group Nebraska, New Jersey, New Mexico, New York, Oregon, for which they submitted data (Arizona, California, Georgia, Tennessee, Texas, and Virginia). Only one state (Maryland) Indiana, Kansas, Kentucky, Maryland, Michigan, Missouri, New submitted OT encounter data that did not meet the usability Mexico, Rhode Island, Virginia, and Washington). 5 Table 3. Summary of the 2008 MAX Encounter OT Claims OT Encounter State Has Percentage of CMC OT Encounter Records Are of OT Encounter Data Comprehensive Enrollees Met State Submitted OT Records Are Comparable Quality Are Usable for Managed Care Thresholdb Encounter Claimsc Completed to FFS Datae Researchf (CMC)a A C D E A C D E A C D E A C D E A C D E Alabama Alaska Arizona X X X X X X X X X X X X X X X X X X X X X Arkansas California X X X X X X X X X X X X X X X X X Colorado X X X X X X X X X X X X X X Connecticut Delaware X X X X X X X X X X X X X X X X X X X X X District of Columbia X X X X Florida X X X X X X X X X X X X X X X X X Georgia X X X X X X X X X X X Hawaii X X X X X X X X X X X X X X X X Idaho Illinois X X X X X X Indiana X X X X X X X X X X X X X X X X Iowa X Kansas X X X X X X X X X X X Kentucky X X X X X X X X X X X X X X X X X X X X X Louisiana Maine Maryland X X X X X X X X X X Massachusetts X X X X Michigan X X X X X X X X X X X X X X X X X X X X X Minnesota X X X X X X X X X X X X X X X X X X X Mississippi Missouri X X X X X X X X X X X Montana Nebraska X X X X X X X X X X X X X X X X X X X X X Nevada X X X New Hampshire New Jersey X X X X X X X X X X X X X X X X X X X X X New Mexico X X X X X X X X X X X X X X X X X X X X X New York X X X X X X X X X X X X X X X X X X X X X North Carolina North Dakota Ohio X X X X X Oklahoma Oregon X X X X X X X X X X X X X X X X X X X X X Pennsylvania X X X X X Rhode Island X X X X X X X X X X X X X X South Carolina X X X X South Dakota Tennessee X X X X X X X X X X X X X X X X X X X X X Texas X X X X X X X X X X X X X X X X X X X X X Utah Vermont X Virginia X X X X X X X X X X X X X X X X X X X X X Washington X X X West Virginia X X X Wisconsin X X X X X X X X X Wyoming Total 35 32 33 25 18 24 25 20 16 23 23 16 15 21 24 19 15 21 22 15 14 (continued) 6 Table 3. (continued) Source: Mathematica’s analysis of the MAX 2008 PS and OT files. Note: A=Adults, C=Children, D=Disabled, E=Aged. a At least one percent of enrollees participated in HMO/HIO at some point during 2008. b At least 10 percent of enrollees in the BOE participated in HMO/HIO at some point during the year. c In addition to having at least 10 percent HMO/HIO participation, the state submitted at least 200 encounter claims for the BOE. d The BOE-specific metric was met for at least one of the two completeness measures: (1) percentage of enrollees with OT encounter claims (TOS = 08, 11, 12) and (2) average number of OT encounter claims per enrollee. e The BOE-specific metric was met for at least four of the five quality measures: (1) percentage of OT claims with place of service, (2) percentage of OT claims with a primary diagnosis code, (3) percentage of claims with a primary diagnosis code with a character length greater than 3, (4) percentage of claims with a procedure (service) code, and (5) percentage of claims with a procedure code in CPT-4 or HCPCS format. f Both the completeness and quality standards were met for the BOE. Figure 1. Summary of the MAX 2008 OT Encounter Claims by Basis of Eligibility Categorya 33 32 Number of States 25 25 24 24 23 23 22 21 21 20 19 18 16 16 15 15 15 14 Adult Child Disabled Aged Basis of Eligibility Category  CMC Enrollment Met Threshold  OT Encounters–Comparable Quality  Number of CMC Encounter Claims Met Threshold  OT Encounters–Usable  OT Encounters–Complete a See Table 3 footnotes for data category definitions. 7 Table 4. Summary of the 2008 MAX Encounter IP Claims State Has Percentage of CMC IP Encounter IP Encounter Records IP Encounter Data Comprehensive Enrollees Met State Submitted IP Records Are Are of Comparable Are Usable for Managed Care Thresholdb Encounter Claimsc Completed Quality to FFS Datae Researchf (CMC)a A C D E A C D E A C D E A C D E A C D E Alabama Alaska Arizona X X X X X X X X X X X X X X X X X X X X X Arkansas California X X X X X X X X X X X X X X X X X Colorado X X X X X X X X X Connecticut Delaware X X X X X X X X X X X X X X X X X District of Columbia X X X X Florida X X X X X X X X X X X X X X X X Georgia X X X Hawaii X X X X X X X X X X X X X X X X Idaho Illinois X X X X Indiana X X X X X X X X X X X X X X X X Iowa X Kansas X X X X X X X X X X X Kentucky X X X X X X X X X X X X X X X X X X X X X Louisiana Maine Maryland X X X X X X X X X X X X X X Massachusetts X X X X Michigan X X X X X X X X X X X X X X X Minnesota X X X X X X X X X X X X X X X X X X X Mississippi Missouri X X X X X X X X X X X Montana Nebraska X X X X X X X X X X X X X X X X X X X X X Nevada X X X New Hampshire New Jersey X X X X X X X X X X X X X X X X X X X X X New Mexico X X X X X X X X X X X X X X X X X X X X X New York X X X X X X X X X X X X X X X X X X X North Carolina North Dakota Ohio X X X X X Oklahoma Oregon X X X X X X X X X X X X X X X X X X X X X Pennsylvania X X X X X Rhode Island X X X X X X X X X X South Carolina X X X X South Dakota Tennessee X X X X X X X X X X X X X X X X X Texas X X X X X X X X X X X X X X X X Utah Vermont X Virginia X X X X X X X X X X X X X X X X X X X X X Washington X X X X X X X X X X X West Virginia X X X Wisconsin X X X X X X X X X X X Wyoming Total 35 32 33 25 18 24 24 20 15 23 20 16 15 21 23 12 12 20 19 10 12 (continued) 8 Table 4. (continued) Source: Mathematica’s analysis of the MAX 2008 PS and IP files. Note: A=Adults, C=Children, D=Disabled, E=Aged. a At least one percent of enrollees participated in HMO/HIO at some point during 2008. b At least 10 percent of enrollees in the BOE participated in HMO/HIO at some point during the year. c In addition to having at least 10 percent HMO/HIO participation, the state submitted at least 200 encounter claims for the BOE. d The BOE-specific metric was met for at least one of the two completeness measures: (1) percentage of enrollees with IP encounter claims and (2) average number of IP encounter claims per enrollee. e The BOE-specific metric was met for at least three of the four quality measures: (1) average length of stay, (2) average number of diagnosis codes, (3) percentage of claims with procedure code, and (4) percentage of claims with UB accommodation codes. f Both the completeness and quality standards were met for the BOE. Figure 2. Summary of the MAX 2008 IP Encounter Claims by Basis of Eligibility Categorya 33 32 Number of States 25 24 24 23 23 21 20 20 20 19 18 16 15 15 12 12 12 10 Adult Child Disabled Aged Basis of Eligibility Category  CMC Enrollment Met Threshold  IP Encounters–Comparable Quality  Number of CMC Encounter Claims Met Threshold  IP Encounters–Usable  IP Encounters–Complete a See Table 4 footnotes for data category definitions. 9 Table 5. Summary of the 2008 MAX Encounter RX Claims State Has Percentage of CMC RX Encounter RX Encounter Records RX Encounter Data Comprehensive Enrollees Met State Submitted RX Records Are Are of Comparable Are Usable for Managed Care Thresholdb Encounter Claimsc Completed Quality to FFS Datae Researchf (CMC)a A C D E A C D E A C D E A C D E A C D E Alabama Alaska Arizona X X X X X X X X X X X X X X X X X X X X X Arkansas California X X X X X X X X X X X X X X X X X X X X X Colorado X X X X X Connecticut Delaware District of Columbia X X X X Florida X X X X X X X X X X X X X Georgia X X X X X X X X X X X Hawaii X X X X Idaho Illinois Indiana X X X X X X X X X X X X X X X X Iowa Kansas X X X X X X X X X X X Kentucky X X X X X X X X X X X X X X X X X X X X X Louisiana Maine Maryland X X X X X X X X X X X X X X X X Massachusetts X X X X Michigan X X X X X X X X X X X X X X X X X X X X X Minnesota X X X X X X X X X X X X Mississippi Missouri X X X X X X X X X X X Montana Nebraska Nevada X X X New Hampshire New Jerseyg New Mexico X X X X X X X X X X X X X X X X X X X X X New Yorkg North Carolina North Dakota Ohio X X X X X Oklahoma Oregong Pennsylvania X X X X X Rhode Island X X X X X X X X X X X X X X X X South Carolina X X X X South Dakota Tennessee Texas Utah Vermont X Virginia X X X X X X X X X X X X X X X X X X X X X Washington X X X X X X X West Virginia Wisconsing Wyoming Total 24 23 23 18 11 14 15 10 8 13 14 9 8 13 14 9 7 12 13 9 7 (continued) 10 Table 5. (continued) Source: Mathematica’s analysis of the MAX 2008 PS and RX files. Note: A=Adults, C=Children, D=Disabled, E=Aged. a At least one percent of enrollees participated in HMO/HIO/PACE at some point during 2008. There were 12 states in which MC plans did not provide a pharmacy benefit: CT, DE, IA, IL, NE, NJ, NY, OR, TN, TX, WI, and WV (Bagchi et al. 2012). b At least 10 percent of enrollees in the BOE participated in HMO/HIO at some point during the year. c In addition to having at least 10 percent HMO/HIO participation, the state submitted at least 200 encounter claims for the BOE. d The BOE-specific metric was met for at least one of the two completeness measures: (1) percentage of enrollees with RX encounter claims and (2) average number of RX encounter claims per enrollee. e The BOE-specific metric was met for at least one of the two quality measures: (1) percentage of RX claims with date prescribed and (2) percentage of RX claims with quantity. f Both the completeness and quality standards were met for the BOE. g NJ, NY, OR, and WI submitted RX encounter data even though prescription drugs were not included in the HMO benefit package. Figure 3. Summary of the MAX 2008 RX Encounter Claims by Basis of Eligibility Categorya Number of States 23 23 18 15 14 14 14 13 13 13 12 11 10 9 9 9 8 8 7 7 Adult Child Disabled Aged Basis of Eligibility Category  CMC Enrollment Met Threshold  RX Encounters–Comparable Quality  Number of CMC Encounter Claims Met Threshold  RX Encounters–Usable  RX Encounters–Complete a See Table 5 footnotes for data category definitions. Changes in IP and RX Encounter Data from managed care, the increase in the volume of data and its quality 2007 to 2008 is expected and promising. From 2007 to 2008, the number of states with at least 10 per- As noted earlier, our analysis of MAX 2008 encounter data cent of enrollees participating in comprehensive managed care excluded dual-eligible Medicaid enrollees, while our previous at some point during the year and that submitted IP encounter analysis of MAX 2007 data did not. This change affected the claims increased for all four BOE categories (Table 6). The number of states enrolling the disabled and aged in compre- number of states with complete IP encounter data increased hensive managed care, and likewise the encounters included for all groups between 2007 and 2008. The number of states in our analysis. The number of states meeting our threshold of with complete RX encounter data increased for only the aged enrolling at least one percent of disabled and aged non-dual between 2007 and 2008. These increases were due both to an enrollees in 2008 was greater than the number enrolling at least increase in the number of states with at least 10 percent par- one percent of disabled and aged (dual- and non-dual) enrollees ticipation in comprehensive managed care and to more states in 2007, which was expected, given that dual-eligible enrollees submitting encounter data. Given states’ increasing reliance on are often excluded from managed care enrollment in most 11 Table 6. Summary of IP and RX Encounter Claims by BOE for MAX 2007 and 2008 2007 2008 A C D E A C D E IP Encounter Claims CMC Enrollment Met Thresholda 33 33 22 9 32 33 25 18 CMC Encounter Claims Met Thresholdb 22 22 15 6 24 24 20 15 IP Encounter Claims–Complete c 19 19 12 6 23 20 16 15 IP Encounter Claims–Qualityd 20 21 10 5 21 23 12 12 IP Encounter Claims–Usable e 17 18 8 5 20 19 10 12 RX Encounter Claims CMC Enrollment Met Thresholda 27 27 18 6 23 23 18 11 CMC Encounter Claims Met Thresholdb 17 17 11 3 14 15 10 8 RX Encounter Claims–Complete f 17 17 10 3 13 14 9 8 RX Encounter Claims–Quality g 16 16 11 2 13 14 9 7 RX Encounter Claims–Usablee 16 16 10 2 12 13 9 7 Source: Mathematica’s analysis of the MAX 2008 PS, IP, and RX files. Note: A=Adults, C=Children, D=Disabled, E=Aged. a At least 10 percent of enrollees in the BOE participated in HMO/HIO at some point during the year. b In addition to having at least 10 percent HMO/HIO participation, the state submitted at least 200 encounter claims for the BOE. c The BOE-specific metric was met for at least one of the two completeness measures: (1) percentage of enrollees with IP encounter claims and (2) average number of IP encounter claims per enrollee. d The BOE-specific metric was met for at least three of the four quality measures: (1) average length of stay, (2) average number of diagnosis codes, (3) percentage of claims with procedure code, and (4) percentage of claims with UB accommodation codes. e Both the completeness and quality standards were met for the BOE. f The BOE-specific metric was met for at least one of the two completeness measures: (1) percentage of enrollees with RX encounter claims and (2) average number of RX encounter claims per enrollee. g The BOE-specific metric was met for at least one of the two quality measures: (1) percentage of RX claims with date prescribed and (2) percentage of RX claims with quantity. states. For disabled enrollees, 22 states met the 2007 threshold or more for adults and that submitted encounter claims, the for IP encounter data, and 25 met the 2008 threshold for the number with usable data for research on OT encounters would same, a small difference. Eighteen states included IP services fall from 21 (88 percent) to 17 (71 percent) if all criteria for for non-dual aged enrollees in comprehensive managed care in FFS completeness and quality had to be met. Among the 24 MAX 2008, compared to 9 in 2007. Given that our 2008 analysis states that had comprehensive managed care participation of 10 includes enrollees receiving their medical services from Medicaid percent or more for adults and that submitted encounter claims, only, we expected to see more states meeting the thresholds for the number with usable data for research on IP encounters completeness, and this is particularly true for the aged population. would fall from 20 (83 percent) to 12 (50 percent) if all criteria for FFS completeness and quality had to be met. Among the 14 The volume of encounter data in the RX file meeting the thresh- states that had comprehensive managed care participation of 10 olds for analysis dropped because more states provided phar- percent or more for adults and that submitted encounter claims, macy benefits with FFS arrangements, increasing from 8 in 2007 the number with usable data for research on RX encounters to 12 in 2008 (See footnote a in Table 5). The completeness, would fall from 12 (86 percent) to 7 (50 percent) if all criteria quality, and usability of the data, however, remained very high. for FFS completeness and quality had to be met. In this brief, we used selected FFS-based metrics to make a Caveats preliminary judgment about the quality and completeness of the Because FFS data are not without completeness and quality data for inpatient hospitalization, physician services, outpatient issues, we did not require a state’s encounter data to meet all hospital services, clinic services, and prescription medica- completeness and quality measures to be considered usable. tion. This approach has been useful because it illustrates that a If we had, the number of states with usable data for at least reasonable quantity of encounter data is available in MAX and one BOE would have dropped. Among the 24 states that had that they appear to be of good quality on basic measures. We comprehensive managed care participation of 10 percent assume that, like the FFS data, the MAX data that fall within 12 acceptable ranges accurately depict what is happening in the same quality as FFS data for more states than the IP data. This state. This analysis is limited, however, by its assumption that analysis will aid researchers in determining which states with FFS data provide a reasonable benchmark for judging the notable comprehensive managed care enrollment may be rea- encounter data, which may not be the case, depending on the sonable to analyze. By knowing the usability of the encounter particular populations a state chooses to enroll in managed care. data for physician and clinic encounters, inpatient and outpa- While populations receiving services through comprehensive tient visits, and prescription drugs, researchers and policymak- managed care plans are likely to differ from FFS populations in ers can reasonably consider adding the analysis of encounter important ways, we used metrics within two standard deviations data when assessing Medicaid service utilization in states with to account for differences in utilization patterns that may reflect substantial enrollment in comprehensive managed care. differences in populations or inherent differences between the FFS and managed care delivery systems. Additionally, it may be References difficult to extend our analysis of selected OT measures to other Bagchi, A, J. Verdier, and D. Esposito. “Statistical Compendium: types of services in the OT file. It will depend on the type of Medicaid Pharmacy Benefit Use and Reimbursement in 2008.” service, whether the type of service is covered by managed care Table 1. Washington, DC: Centers for Medicare & Medicaid arrangements, and how consistently services are billed across Services, March 2012. Borck, R., A.H. Dodd, A. Zlatinov, S. Verghese, R. Malsberger, and states or within plan arrangements. C. Petroski. “The Medicaid Analytic eXtract 2008 Chartbook.” Washington, DC: Centers for Medicare & Medicaid Services, 2012. Conclusions Byrd, V., J. Verdier, J. Nysenbaum, and A. Schoettle. “Technical Assistance for Medicaid Managed Care Encounter Reporting to the This brief provides an assessment of selected OT, IP, and Medicaid Statistical Information System, 2010.” Washington, DC: Mathematica Policy Research. Final report submitted to Centers for RX encounter data included in the MAX 2008 data files. The Medicare & Medicaid Services, February 2011. results are encouraging. More states submitted encounter data Dodd, A.H., J. Nysenbaum, and A. Zlatinov. “Assessing the Usability in 2008 than in 2007, reflecting an increase in the provision of of the MAX 2007 Inpatient and Prescription Encounter Data for data for existing plans, as well as an increase in the percent- Enrollees in Comprehensive Managed Care.” Washington, DC: Centers for Medicare & Medicaid Services, April 2012. age of enrollees in comprehensive managed care, particularly Young, K, R. Garfield, M. Musumeci, L. Clemans-Cope, and among aged enrollees. Most states that have comprehensive E. Lawton. “Medicaid’s Role for Dual Eligible Beneficiaries.” managed care plans are reporting selected OT, IP, and RX Washington, DC: Kaiser Commission on Medicaid and the encounter data. Of those data, the majority are complete and Uninsured, April 2012. of comparable quality to the FFS data for adults, children, the Endnotes disabled, and aged populations. Although several actuaries and 1 Encounter claims in the LT file are clustered among very few states in state officials involved in Medicaid administration at the state MAX 2008 data. After imposing our analysis criteria, there were too level have hypothesized that IP and RX data might be more few encounters for a cross-state analysis of LT data. complete and of higher quality than OT data because they are 2 A full-benefit Medicaid enrollee is defined here as an enrollee with collected from fewer providers (Byrd et. al 2011), our analysis a restricted benefits flag equal to one for any month of enrollment in did not confirm that hypothesis. The data for physician, outpa- the calendar year, meaning the individual is eligible for Medicaid or CHIP and entitled to the full scope of Medicaid or CHIP benefits. tient, and clinic services appear to be complete and of about the For further information on this issue brief series, visit our website at www.mathematica-mpr.com Princeton, NJ • Ann Arbor, MI • Cambridge, MA • Chicago, IL • Oakland, CA • Washington, DC Mathematica is a registered trademark of Mathematica Policy Research, Inc. ® 13