No Now wha hat? t? Presenter: Kate Mullins Co-Author: Hailey - PowerPoint PPT Presentation

We f e fou ound nd a d a dat ata a quali uality ty is issu sue. e. No Now wha hat? t? Presenter: Kate Mullins Co-Author: Hailey DuBreuil 1

MEE EET T OU OUR R DATA Q A QUAL ALIT ITY Y TEA EAM Yuan Zhang Ka Kate e Mullin lins Haile iley y DuBreuil reuil Ka Kati tie e Howar ard kmullins@hsri.org hdubreuil@hsri.org khoward@hsri.org yzhang@hsri.org Project Manager Project Coordinator Data Scientist Research Analyst 2

AGE GENDA 03 03 HSRI RI and nd Our ur 01 01 02 02 Presen esentat tation n Wh What are data a Approac oach Objec ectiv tives es quality ity issue ues? s? 06 06 04 04 Fut uture ure 05 05 Issue e Quest stions ions Resolutio tion n Directions ections Frame mework 3

HS HSRI RI AN AND OU D OUR R AP APPR PROACH CH 4

Human Services Research Institute We are a nonprofit, mission-driven organization. We use our data expertise — developed over 40+ years — and our understanding of the complete health and human services landscape to help agencies and communities improve the health, well-being, and economic stability of the populations they serve. Housing & Homelessness | Population Health | Aging & Disabilities Child, Youth & Family | Behavioral Health | Intellectual & Developmental Disabilities 5

Population Health Team: What We Do We develop and maintain nonproprietary data collection and reporting systems, custom analytics, state-level health data warehouses, data quality improvement procedures, and healthcare transparency websites. 6

Newly Payer Submissions Detected (Commercial/Medicaid) Internet SFTP Server 1. 1. Data File Unzip, Decrypt, Monitor Enclave Submi miss ssion ion Initial Storage Process Server Data Intake & Stagin ing Validation Medicare Ingestion Ingestion Batch Recommendations Release Ingestion Passed Files Decisions 2. Data ta Ware rehou ouse se Calculated Variables, Member ID, Process essing ng & Provider Processing Business Rule Enhancem ncemen ent Release Processing - EMPI, Staging Grouper, Provider Index Client 3. Extracts, tracts, Review Analysis ysis-Rea eady y Release Datase sets, ts, and Analytic Layer Post-Intake Quality Client (DED/Valid Views) Assessments & Data Mining Reporting ting Sign-off Reports 7

Data Quality Approach Continual Improvement and Flexibility • Feedback loop with internal and external stakeholders • Regular process improvement procedures with flexibility to address more immediate issues 8

HSRI Help Desk Provides Data Intake Support • Detailed technical support to resolve validation issues and ensure data are submitted in a timely manner • All support requests received during business hours (Mon. – Fri.) are responded to in 2 hours or less • Accessible via: • Toll-free phone number • Email • Web contact form in the HSRI Data Submission and Quality Portal 9

PR PRES ESENT ENTATIO TION N OB OBJEC JECTIVES TIVES

Today’s Objectives 1. Inform data users on the complexity and challenges of resolving APCD data quality issues 2. Provide a framework for states navigating the process: a) Where to focus limited resources b) How to approach decision making and resolve issues 3. Make recommendations for future directions 11

WH WHAT AR ARE D E DATA QU A QUAL ALIT ITY Y IS ISSUES? UES? 12

Newly Payer Submissions Detected (Commercial/Medicaid) Internet 1. Data SFTP Server File Unzip, Decrypt, Monitor Submission Enclave Initial Storage Process Server Data Intake & Staging Validation Medicare Ingestion Ingestion Batch Recommendations Release Ingestion Passed Files Decisions 2. Data Calculated Variables, Warehouse Member ID, Processing & Provider Processing Business Rule Enhancement Release Processing - EMPI, Staging Grouper, Provider Index Client 3. Extracts, Review Analysis-Ready Release Datasets, and Analytic Layer Post-Intake Quality Client (DED/Valid Views) Assessments & Data Mining Reporting Sign-off Reports 13

What are data quality issues? • Inconsistent claim and/or encounter volume over time • Inconsistent PMPM over time • Low match rates for Patient/Provider/Encounter identifiers • Inconsistent population of fields over time • Mismatch of results when compared with external sources 14

How do we find out about issues? • Data Submitter Self-Report • Post-Intake Quality Assessments & Reporting • Data Mining • Data Users (internal and external) • Implementation of Third-Party Tools 15

Why are the issues difficult to address? AP APCD AP APCD APCD Admini inist stra rator Data Vend ndor or Data Sub ubmi mitt tter er ● ● ● Limited Resources ● ● ● Competing Priorities ● ● ● Revolving Door of Data Quality Issues ● ● ● Staff Turnover and Training Varying Requirements Across States • Tolerance for Issues ● • Policies for Resubmission • Data Submission Platforms 16

IS ISSU SUE E RE RESOL OLUTION UTION FR FRAM AMEW EWORK ORK 17

Framework for Resolution Identify Prioritize Make Implement Resolution Issues Decisions Resolution Options Is this issue What questions How can we prevent What are options for a similar issue in the worth pursuing? should be asked to resolving the issue? future? How urgent is it? choose the best resolution option? How can we best communicate the resolution? 18

Considerations for Prioritizing Issues Priority Payers Key Fields Status of Data Time Periods Issue impacts a high Used for Member Data impacted are in use Issue impacts multiple number of covered lives Identification, Claim months or years or high percentage of Versioning, etc. APCD, payer types, etc. 19

Options for Resolving Issues PROS CONS NS Up Update e Document mentati tion on Future standardization across No immediate impact (e.g. .g.: : sub ubmi miss ssion ion gui uide e or rul ule) e) payers Lengthy approval process Resolution in future submissions Modi dify fy Data Qu Quality ty Future standardization across No resolution in data Iden entif tific icat ation ion Proces cesse ses payers Educ Ed ucate Sub ubmi mitt tter ers Resolution in future submissions Historical issues remain Request uest Resubm submiss ssion ion from om Potentially time- and Historical issues resolved Sub ubmi mitt tter er resource-intensive Remed mediat ate e Data by Historical issues resolved Patchwork code Admini inist stra rator/ / Vendor ndor Modi dify fy Data User ser Users can work around issue No resolution in data Docume menta ntati tion on based on use case 20

Decision-Making Questions • Is this a significant issue that makes the data unusable? • Can data users code around the issue easily? • Is there time sensitivity to resolving the issue? • How many payers does the issue affect? • Does the issue occur in recent data (last 5 years)? • Does the issue span more than a short amount of time (e.g.: 1 month)? 21

Decision-Making Questions • Can the submitter fix the issue and resubmit? • Will resubmission cause unintended consequences/other quality issues? • Is there other specialty code that needs to be considered when processing? • Will remediation cause unintended consequences/other quality issues? • Does the issue warrant the resources necessary for resolution? 22

Example 1: Insurance Product Type Al All claims ms are sub ubmit mitted ed with h the e same e Issue: ue: Insurance urance Produ duct ct Type e (IPT) ) code de whil ile e the e eligi gibility bility IPT has s variati tion on Identification Method Data analysis and reporting • Data released to users • Payer has a large number of covered lives and makes up large Decision-Making proportion of APCD Considerations • Issue occurs in recent years (past three years) • Time constraints to fix the issue for data users • Data resubmission may cause shifts/unintended consequences • Administrator/vendor data remediation in historical submissions • Submitter education & resolution in future submissions Resolution • Exploring modification of data quality identification processes • Timeline from issue identification to resolution implementation: 2 months 23

Example 2: Provider Network Unknown Many payers submitted a high volume of claims with “Unknown” for Issue: e: Provide ider r Netw etwor ork k Indi dicat cator or, , which ich indica icates s if the e servi vicing cing provi vide der r is partic icipat ipating ing in vs. s. out ut of net etwor ork Identification Method Data Mining • Data released to users Decision-Making • Targeted payers with the largest impact Considerations • Issue occurs in recent years of data • Relatively easy issue to fix if the payer has the data and time • Requested resubmission of historical data from high-impact payers • Payers without capacity to resubmit files corrected the issue in future submissions Resolution • Modified data intake validation processes for earlier issue detection in the future • Timeline from issue identification to resolution implementation: 8 months 24

FU FUTUR URE E DI DIRE RECTIONS CTIONS 25