Slides - AcademyHealth
Transcription
Slides - AcademyHealth
Data Linkage Training Project for Vital Statistics and Medicaid Claims Data: Getting to Know Medicaid and CHIP Data March 17, 2015 2:00-3:00 PM ET Teleconference Instructions Please make sure your computer is linked to your phone: • A box should appear on your screen when you log-in with dial-in instructions. OR • Click on the phone symbol in the toolbar at the top of your screen. Enter your phone number and click “join,” the system will call you directly. OR • Call in directly: – Dial 1-866-244-8528 – Enter the access code 602144 and press pound (#) – Once you have dialed in, click on the information symbol in top right corner of the meeting room and dial the “Telephone Token” number into your phone. – The system will link your phone with your computer. 2 •2 View the Slides in Full Screen • If you would like to view the slides in full screen, click the four way arrow button on the top right corner of the slides. 3 •3 Technical Assistance • If you have technical questions during the event, please type them into the chat box in the lower left-hand corner of the screen. • Live technical assistance is also available: – Please call Adobe Connect at (800) 422-3623 4 •4 Discussion To raise your hand: 1. Participants can use the hand raise button at the top of the screen to signal to the presenter that they would to speak To submit a question: 1. Click in the chat box on the left side of your screen 2. Type your question into the dialog box and click the Send button. Please mute your phones if you are not speaking to reduce background noise. 5 •5 Agenda • Welcome and Introductions • Medicaid Data 101 • Considerations for Linking Medicaid Data and Vital Records • Available Resources • State Example: Iowa • Closing Comments 6 Welcome and Introductions • Ellen O’Brien and Stephanie Kennedy, AcademyHealth • Danielle T. Barradas, Ph.D., Division of Reproductive Health, NCCDPHP, CDC • Lekisha Daniel-Robinson, M.S.P.H., Division of Quality, Evaluation and Health Outcomes, CMCS, CMS • Russell S. Kirby, Ph.D., University of South Florida • Craig A. Mason, Ph.D., University of Maine • Vivian Byrd, Keith Kranker, Michaela Vine, Researchers, Mathematica Policy Research • Debbie Kane, MCH Epidemiologist-CDC Assignee, Bureau of Family Health, HPCDP, Iowa Department of Public Health 7 Welcome and Introductions Phase II States • Connecticut • Delaware • Nebraska • New Jersey • Virginia • Washington, DC Phase I States • Georgia • Indiana • Kentucky • Maine • Massachusetts • Michigan • Mississippi • New Mexico • Nevada • West Virginia • Wyoming 8 Medicaid Data 101 Vivian Byrd, MPP Researcher Mathematica Policy Research 9 Three Medicaid/CHIP Core Set Measures Can Be Calculated Using Linked Data Core Set Measure Numerator Data Source Denominator Data Source Low Birth Weight Vital records Vital records alone or linked with Medicaid data to identify payer source Cesarean Delivery Vital records alone or linked with Medicaid data to identify Cesarean delivery and medical record review Vital records or linked with Medicaid data to identify payer source and medical record review PC-01: Elective Delivery Medicaid data and medical record review Medicaid data and medical record review; vital records may be used to measure gestational age PC-03: Antenatal Steroids Medicaid data and medical record review Medicaid data and medical record review; vital records may be used to measure gestational age Note: Measure-eligible population includes Medicaid and CHIP enrollees. 10 10 Medicaid Data 101: Obtaining Data for Use Medicaid Management Information System (MMIS) • • • • Eligibility Providers Claims adjudication Administrative costs for Medicaid • Data may be transferred to an easily accessible data warehouse 11 Medicaid Statistical Information System (MSIS) Medicaid Analytic eXtract (MAX) • Federal reporting required of all • Uses MSIS data states to create • Many identifiers are removed research-ready (names, addresses, phone data file numbers, etc.) • Person summary • Every eligible individual for every file for each state month in fiscal year by calendar year • Every service rendered, reported (based on date of by adjudicated quarter in fiscal service) year • 5 files: EL, IP, LT, OT, RX • MSIS data dictionary specifies formats • “Validated” on a quarterly basis Medicaid Data 101: Structure Medicaid data systems typically contain two types of files • Eligibility/Enrollment • Eligibility determinations use different system than claims adjudication • Point-in-time information • Temporary ID numbers • Claims/Encounters • • • • 12 Information required for a provider to receive payment Fee for service (FFS) vs. Managed Care data Data received and stored by states Timeliness and quality vary Medicaid Data Fields • Eligibility – Enrollment • Patient/beneficiary information • • • • Medicaid ID Date of birth Address Pathway to Medicaid Eligibility • Claims - Payments • • • • • • Diagnoses Procedures Provider IDs Facility Dates of Service Billing • Patient information • Payment information 13 Fields for Medicaid-Vital Records Data Linkage • • • • • • • • • • • Mother’s Eligibility Record(s) Social Security Number (SSN) Medicaid/CHIP ID Date of birth Name Address* Phone Race/Ethnicity Mother’s “Delivery Claim” Medicaid payment for the claim (verify yes/no) Date of delivery** Hospital/facility of delivery Other • • • • • • • • Infant’s Eligibility Record(s) Date of birth Name Address* Phone Gender Infant’s “Birth Claim” Medicaid payment for the claim (verify yes/no) Hospital/facility of birth Other * Street Address, City, State, ZIP code, and County ** Linked to infant’s date of birth in vital records 14 Medicaid Data 101: Common Issues • • • • • 15 Global billing for maternity care Mother and infant IDs Managed care encounter data Timing/rollout and completeness/quality Personnel needed to support extraction, manipulation, and analysis Major differences between Medicaid and Vital Records data Medicaid Vital Records Births/deliveries included in file Medicaid-covered births (in-state or out-of-state) All in-state births Identifying Medicaidcovered births/deliveries Yes: accurate history of Medicaid enrollment Underreported (especially in states with managed care) Variables for linkage availability Yes: when Medicaid data is collected from state Medicaid agencies (but not from CMS) Yes: when Medicaid data is collected from state Vital Records agencies (but not from CDC) Demographic/ characteristics availablea • • • • • • • • • • Data release a Both 16 Family income Aged, blind, and/or disabled eligibility Fee-for-service or managed care SNAP and TANF participation Rolling basis (claims may take up to a year to be submitted) Medicaid and Vital Records age, gender, race, ethnicity, and location Pregnancy history Pre-pregnancy weight Education Marital status Foreign-born WIC participation Annually: usually the summer or fall following each calendar year Major differences in outcome measures between Medicaid and Vital Records data Medicaid Vital Records Prenatal care Medicaid-covered prenatal care only; issues with bundled payments Dates of first/last visit and number of visits; may be inaccurate Gestation, birthweight, smoking during pregnancy, maternal weight gain, and breastfeeding at discharge Noa Yes (some variables more accurate than others) Method of delivery, maternal health complications, and NICU admissions Yes: diagnoses, procedures, and other fields in claims Yes: checkboxes on birth certificate Mortality Could be underreported Yes, by linking birth certificates to death certificates Medicaid-paid costsb and utilization (office visits, days in hospital, etc.) Yes No Outcomes after the birth /delivery Yes No a Diagnosis codes in claims only for gestation less than 36 weeks (765.2x) and birthweight less than 2,500 grams (765.0x, 765.1x). b Costs are typically available for fee-for-service beneficiaries only 17 Questions? 18 Considerations in Linking Medicaid Data and Vital Records 19 Key Decisions • Are you … 1. Linking mothers’ Medicaid data to Vital Records, 2. Linking infants’ Medicaid data to Vital Records, or 3. Both? • Should you link mothers’ and infants’ Medicaid data before linking to Vital Records, or link the mothers and infants separately? • Which data fields should be used? • How should the data files be prepared and structured? • How should the data be linked? 20 Fields for Medicaid-Vital Records Data Linkage • • • • • • • • • • • Mother’s Eligibility Record(s) Social Security Number (SSN) Medicaid/CHIP ID Date of birth Name Address* Phone Race/Ethnicity Mother’s “Delivery Claim” Medicaid payment for the claim (verify yes/no) Date of delivery** Hospital/facility of delivery Other • • • • • • • • Infant’s Eligibility Record(s) Date of birth Name Address* Phone Gender Infant’s “Birth Claim” Medicaid payment for the claim (verify yes/no) Hospital/facility of birth Other * Street Address, City, State, ZIP code, and County ** Linked to infant’s date of birth in vital records 21 Models for Medicaid-Vital Records Data Linkage: (1) Using the Vital Records “as a Hub” Mothers’ Medicaid/CHIP Data Vital Records Infants’ Medicaid/CHIP Data • • 22 Linkage to Vital Records twice (not once) Each linkage uses the mothers’ or infants’ information, but not both Models for Medicaid-Vital Records Data Linkage: (2) Linking Mothers’ to Infants’ Medicaid Data Mothers’ Medicaid/CHIP Data Mothers Linked to Infants Infants’ Medicaid/CHIP Data 23 • • Vital Records Relies on household IDs Linkage to Vital Records with mothers’ and infants’ information ID Numbers for Medicaid-Vital Records Data Linkage • Social Security Numbers • When available, SSN is a great, although imperfect, linking variable • Mothers’ SSNs routinely collected in Vital Records • Most Medicaid agencies collect most mothers’ SSNs • Infants’ SSNs not collected in Vital Records, and not always collected by Medicaid • Some agencies cannot release SSNs • Medicaid/CHIP ID Numbers • Collected in Vital Records in some states, but not all • Some beneficiaries may have more than one ID • Best to de-duplicate Medicaid IDs before matching with other datasets • Medicaid Household ID Number (or Case ID Number) • When available, can be used to link mothers’ and infants’ Medicaid data (for some mother-infant dyads, but not all) • Some Medicaid agencies do not routinely link mothers to their infants 24 Other Fields from the Medicaid Enrollment Records • Dates of birth • Usually complete in administrative data • Not a unique identifier—multiple births occur each day—but important for ruling out bad matches (e.g., mothers with multiple children) • Names • Usually available in administrative data • Not a unique identifier; can be entered differently in separate data systems • Phonetic algorithms and other techniques can address some challenges • NYSIIS and SOUNDEX algorithms for last names and first names, respectively • Address and phone number • When these fields match, good indication of a match • Fields often change: two records could match even if they do not have the same address or phone number • Other fields discussed in Technical Assistance Brief #4 • See list of Available Resources for link to brief 25 Fields from Medicaid Claims/Encounter Data • Available fields for linking • Type of provider (e.g., hospital or birth center) • Hospital/provider names and ID numbers • Date of delivery • Estimate delivery date from mothers’ claims to within a few days • Rare outcomes (e.g., multiple births or NICU admissions) • Plan ahead to include additional fields for quality measures or QI efforts • Identifying the birth/delivery claim • Identifying Cesarean deliveries • Other fields related to diagnoses or procedures • Summarize the claims data and merge onto the enrollment files • One row per delivery (for mothers) • One row per infant 26 Preparing and Cleaning the Data Before Linkage • Assign unique numbers to the rows in each data set (before doing anything) • Restrict data to the time period and population of interest • Use care to avoid inadvertently dropping records that should be included • Consider dropping multiple births • Drop variables that will not be used for matching • Some beneficiaries will have more than one enrollment record • Remove completely redundant (identical) rows • Consider keeping multiple rows per person • Requires more care than data with one row per person • Increases the likelihood of a match 27 Preparing and Cleaning the Data Before Linkage (cont’d) • Re-code variables to have the same structure in all files • For example, dates stored in one field or three • Use common coding schemes in all files • For example, gender coded as 0/1 or M/F • Give the same field the same name in all files • Assess data quality, in particular the extent of missing values or coding issues on variables used for matching 28 Additional Resources • Kranker, Keith, So O’Neil, Vanessa Oddo, Miriam Drapkin, and Margo Rosenbach. “Strategies for Using Vital Records to Measure Quality of Care in Medicaid and CHIP Programs.” Medicaid/CHIP Health Care Quality Measures, Technical Assistance Brief no. 4. Cambridge, MA: Mathematica Policy Research, January 2014. Available at http://www.medicaid.gov/Medicaid-CHIPProgram-Information/By-Topics/Quality-ofCare/Downloads/Using-Vital-Records.pdf 29 Additional Resources (2) • CDC/NCHS publications on LBW and C-section with technical appendices and benchmark data http://www.cdc.gov/nchs/data/nvsr/nvsr62/nvsr62_09.pdf http://www.cdc.gov/nchs/data/nvsr/nvsr63/nvsr63_06.pdf http://www.cdc.gov/nchs/data/nvsr/nvsr63/nvsr63_01.pdf 30 Additional Resources • Mathematica prepared a TA Brief on strategies for accessing vital records for quality measurement and improvement efforts (http://www.medicaid.gov/Medicaid-CHIP-Program-Information/ByTopics/Quality-of-Care/Downloads/Using-Vital-Records.pdf) • Medicaid partners can access a TA Mailbox with questions about Core Set measures ([email protected]) • Information on Medicaid Analytic eXtract (MAX) available at (http://www.medicaid.gov/Medicaid-CHIP-Program-Information/By-Topics/Dataand-Systems/MAX/MAX-General-Information.html) • Information on Medicaid Statistical Information System (MSIS) available at (http://www.medicaid.gov/Medicaid-CHIP-Program-Information/ByTopics/Data-and-Systems/MSIS/Medicaid-Statistical-InformationSystem.html) 31 Questions? 32 State Example: Iowa Debbie Kane, PhD MCH Epidemiologist-CDC Assignee, Bureau of Family Health, HPCDP Iowa Department of Public Health 33 Iowa Project Overview • Iowa has conducted linkage since 1989 legislative mandate – Senate File 538, 1989 General Assembly • Collaboration between the Department of Health Services and the Department of Public Health • Linkage process transitioned from mainframe to personal computer in 2010 – – – – Changes in claims data processing and where stored More timely data Upgrades in birth certificate (BC) data collection Began using LinkPlus 34 Iowa’s Medicaid Data Request – Maternal • Time frame – 01/01/2014 through 12/31/2014 • Include the following fields: – Date of service – Maternal DRG codes – All UB and HCFA 1500 with the above diagnoses (V22, V23, V24, V27, V27.2, V27.3, V27.5.) – The file layout (see excel spread sheet) • Include institutional claims files – Filter through DRG codes described above 35 Iowa’s Linkage Process • Why LinkPlus? – Can use CSV files/easy to move between SAS – Ability to create output files and re-link files – Ability to examine “uncertain” matches • Data cleaning and import – Import text files and clean data in SAS • Deterministic match – Requires common identifying fields • BC – Mother’s last name, first name, date of birth (DOB), county of residence • Medicaid claims – Mother’s last name, first name, DOB, county of residence – Multiple passes • BC – Mother’s maiden name, first name, DOB, county of residence • Medicaid claims – Mother’s last name, first name, DOB, county of residence 36 Iowa’s Lessons Learned and Next Steps Lessons • Be careful what you ask for – Be aware of global billing – Consistent use of procedure codes • Carve out time to prepare the data and do the actual linkage – At least a week for the linkage (after data cleaning) – Play with the linkage – try different blocking variables Next Steps • Document and develop protocol for how to handle “uncertain” matches • Need to examine how to handle multiple births • Need to develop validation process 37 Questions? 38 Save the Date • Tuesday, April 21 from 2-3 PM (ET) – Web Training: “Preparing for the In-Person Training” • Tuesday, April 28 from 2-3 PM (ET) – Advanced Web Training: “Obtaining Data from MCOs and Submitting Medicaid and CHIP Data to CARTS” • In-Person Training – May 4-5, 2015 in Washington, DC Thank You! • Please fill out the evaluation questions on screen for today’s web training • Contact Information: – Ellen O’Brien, AcademyHealth • [email protected] – Stephanie Kennedy, AcademyHealth • [email protected]