Author(s)
Jeffrey S. Brown, PhD, Investigator, Director of Data Coordinating Center1, Kristen M. Moore, MPH, Research Associate1, Susan Andrade, ScD, Investigator2, Andrea Cassidy-Bushrow, PhD, Assistant Scientist3, Sascha Dublin, MD, PhD, Assistant Investigator4, Robert Greenlee, PhD, MPH, Epidemiologist / Associate Research Scientist5, Cynthia Nakasato, MD, Investigator6, Richard Platt, MD, MSc, Principal Investigator1, Marsha M Raebel, PharmD, Investigator7, Cheri Rolnick, PhD, MPH, MA, Associate Director of Research8, David H. Smith, PhD, Investigator91Harvard Medical School and Harvard Pilgrim Health Care, Boston, MA, USA; 2 Meyers Primary Care Institute and Fallon Community Health Plan, Worcester, MA, USA; 3 Henry Ford Hospital, Detroit, MI, USA; 4 Group Health Center for Health Studies, Seattle, WA, USA; 5 Marshfield Clinic Research Foundation, Marshfield, WI, USA; 6 Kaiser Permanente Hawaii, Honolulu, HI, USA; 7 Kaiser Permanente Colorado, Denver, CO, USA; 8 HealthPartners Research Foundation, Minneapolis, MN, USA; 9 Center for Health Research, Kaiser Permanente Northwest, Portland, OR, USA
OBJECTIVES Using routinely collected administrative health care claims and medical encounter data for observational studies for post marketing evidence generation holds considerable promise. These studies rely heavily on the validity and accuracy of diagnosis coding. The objective of this pilot study was to assess the extent of within and across health system variation in diagnosis coding. METHODS Two ICD-9 diagnosis categories were selected for review: 557 (vascular insufficiency of intestine) and 708 (urticaria). Each category contained specific codes that have been used in observational studies using administrative claims data. The annual frequency for all subcategories from 2000-2007 was collected from nine HMO Research Network health plans (sites). The frequency of each 4-digit code within each category was compared across years and sites. Specifically, we assessed within-site and cross-site coding variation of subcategories 557.0 (acute vascular insufficiency of intestine) and 708.0 (allergic urticaria). RESULTS From 2000-2007, 54,395 diagnoses within the 557 category and 419,428 diagnoses within the 708 category were identified. Each year, subcategory 557.0 accounted for 37-46% of all 557 coding and subcategory 708.0 accounted for 12-13% of all 708 coding; this varied widely within and across sites. In 2007 subcategory 557.0 accounted for a low of 24% of 557 diagnoses at one site and a high of 53% at a different site; similar variation was observed in all study years. In the same year, subcategory 708.0 accounted for a low of 5% of 708 diagnoses at one site and a high of 20% at a different site; similar variation was also observed in all study years. CONCLUSIONS Our results indicate substantial variation in ICD-9 coding across time and health systems. Studies using administrative encounter data should carefully consider the extent and effect of coding variation when using specific ICD-9 subcategories to define inclusion/exclusion criteria or study outcomes.
Conference/Value in Health Info
2009-05, ISPOR 2009, Orlando, FL, USA
Value in Health, Vol. 12, No. 3 (May 2009)
Code
PMC27
Topic
Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems
Disease
Multiple Diseases
