University of North Carolina Chapel Hill, United States
Background: Since the transition from the 9th to 10th revision of the International Classification of Diseases coding system in the United States on October 1, 2015, many previously validated claims-based definitions of diseases have been adapted to ICD-10 using General Equivalence Mappings (GEM). However, there is still a lack of data on the performance of adapted definitions in databases used for pharmacoepidemiologic research.
Objectives: To estimate the prevalence of ICD-based definitions for common conditions examined in claims database studies and to compare the consistency of these definitions between the ICD-9 and -10 coding eras.
Methods: Using a 20% national random sample of Medicare claims (2013-2017), we identified monthly cohorts of patients ≥ 65 years old with Part A and B coverage with ≥1 diagnosis code recorded. Forward-backward mapping methods were used to map previously published ICD-9-based definitions to equivalent ICD-10 codes for 16 conditions: Alzheimer’s Disease, Non-Alzheimer’s dementia, chronic kidney disease, acute renal failure, hypertension, anxiety, depression, psychotic disorders, chronic obstructive pulmonary disorder, heart failure, myocardial infarction, dyslipidemia, type 2 diabetes mellitus, diabetic nephropathy, neuropathy, and retinopathy. The monthly prevalence was calculated as the number of individuals with at least one claim for each condition divided by the number of eligible enrollees for that month. Estimates were standardized to the age and sex distribution of December 2017 using direct standardization.
Results: Shifts in the trends of monthly period prevalence between the ICD-9 and -10 coding eras were observed for 4 of the conditions assessed. The monthly prevalence of chronic kidney disease increased from 5.43% in September 2015 to 7.32% in October 2015. The monthly prevalence for CKD ranged from 4.61% to 5.79% in the ICD-9 era but shifted to 6.90% to 8.59% in the ICD-10 era. Diabetic retinopathy also increased from 0.70% in September 2015 to 2.44% in October 2015, with ranges of 0.55% to 0.81% in the ICD-9 era and 2.24% to 3.10% in the ICD-10 era. Psychotic disorders decreased from 0.70% to 0.59% between September and October 2015, with ICD-9 and -10 era ranges of 0.62-0.79% and 0.49-0.63%, respectively. Myocardial infarction increased from 0.14% to 0.24%, with the range shifting from 0.10-0.21% to 0.24-0.41%.
Conclusions: GEM produced unstable prevalence estimates across the ICD-9 to ICD-10 transition for a quarter of algorithms we examined. Changes in prevalence across eras that were observed indicate that refinement of these mappings is a critical step to ensure that findings from claims-based studies are robust across calendar time, as more years of data continue to be released for large administrative databases.
