Methodologies for CRN Caryn Etkin, PhD, MPH American Joint Replacement Registry Jialin Mao, MD, MSc Weill Cornell Medical College
Specific Aim Implement an approach to link registries and claims data; first using AJRR then applying to all ICOR-USA registries to allow distributed analyses of outcomes Using AJRR, we will link with Medicare, commercial and state (SPARCS) claims data, apply methodology to other five registries and demonstrate the value of the enhanced (linked) data for conducting post-market research
American Joint Replacement Registry Founded in 2009 Originally affiliated with the American Academy of Orthopaedic Surgeons Stand-alone non-profit since January 2015 Located in Orthopaedic Headquarters in Rosemont, IL Third attempt by profession at hip and knee NATIONAL Registry Multi-stakeholder governance model Profession (Surgeons) Medical Device Industry Health Plans Hospitals The Public
AJRR Data Elements LEVEL I Patient Name (Last, First) Date of Birth Social Security Number Diagnosis (ICD-9/10) Gender Race/Ethnicity Hospital Name and Address Surgeon Name (National Provider Identifier) Procedure Type (ICD-9/10) Date of Surgery Laterality Implants LEVEL II Comorbidities (ICD-9/10) CJR Risk Variables Height + Weight/Body Mass Index Length of Stay American Society of Anesthesiologists Score Operative and Post-operative Complications LEVEL III (Patient-reported Outcomes) Recommended: PROMIS-10 Global VR-12 HOOS/KOOS, JR. Also Available: SF-36 v1 HOOS/KOOS Oxford Hip and Knee Scores Knee Society Knee Scoring System Harris Hip Score WOMAC (Modified via HOOS and KOOS) SF-12, EQ-5D, WOMAC (only accepting final scores)
Statewide Planning and Research Cooperative System (SPARCS) Patient information Race/ethnicity, insurance, etc. Historical comorbidities Long term follow-up Complications, revision/ reoperation Main Presentation Title Edit In Slide Master
Linkage Methodology AJRR Layout: Procedure record Elements: year and month of birth, gender, procedure site (hip/knee), procedure date, hospital name, patient zip code Linkage (sequential) Step 1: Link with all variables Patients year and month of birth, gender, zip code, Exact procedure date, Hospital id Administrative Data Layout: Discharge per record Elements: year and month of birth, gender, procedure site (hip/knee) (by ICD-9/10), procedure date, admission/ discharge date, hospital ID, patient zip code Step 2: Link with Patients year and month of birth, gender, not including zip code, Exact procedure date, Hospital ID Step 3-4: Link with Patients year and month of birth, gender, Flexible procedure date (+/- 3 days), Hospital id Remove zip code Flexibility for procedure date Leave out sex or month of birth Step 5-6: Link with Hospital id, Patients year and month of birth + procedure date, or Patients year of birth + sex +procedure date Duplicate matches will be excluded Analysis Main Presentation Title Edit In Slide Master
Sequential matching algorithm for anonymous matching and cases matched between AJRR and SPARCS Matching algorithm Matching (AJRR case=4063) Facility ID year of birth month of birth sex procedure date zip Matched cases Cumulative cases matched % AJRR cases matched % correctly matched Step 1 x x x x x x 2666 2666 65.62% >99.9% Step 2 x x x x x 1033 3699 91.04% >99.9% Step 3 x x x x (+/-1) 24 3723 91.63% >99.9% Step 4 x x x x (+/-3) Step 5 x x x x 16 3739 92.03% >99.9% Step 6 x x x x To protect patient privacy, numbers of <11 cannot be reported. Number of cases matched during steps 4-6 were combined for this reason. And exact proportion of correctly matched cases cannot be shown. Main Presentation Title Edit In Slide Master