Lead/Presenter: Jonathan Shaw,
COIN - Palo Alto
All Authors: Shaw JG (Center for Innovation to Implementation, VA Palo Alto; Stanford University, Division of Primary Care & Population Health), Veldanda S (Center for Innovation to Implementation, VA Palo Alto), Saechao F (Center for Innovation to Implementation, VA Palo Alto) Romodan Y (Center for Innovation to Implementation, VA Palo Alto) Berg E (Center for Innovation to Implementation, VA Palo Alto) Phibbs CS (Center for Innovation to Implementation and Health Economics Resource Center, VA Palo Alto; Stanford University, Department of Pediatrics) Frayne SM (Center for Innovation to Implementation, VA Palo Alto; Stanford University, Division of Primary Care & Population Health)
Objectives:
Women's Health Evaluation Initiative (WHEI) aggregates ICD-9-CMs into 202 "conditions," to characterize women Veterans' health profiles. We developed/tested an algorithm to address the 2016 transition to ICD-10, mapping all 70,000+ ICD-10s.
Methods:
"Backward" ICD-10 to ICD-9 mapping guided assignment of every ICD-10 to the original 202 conditions. CMS's General Equivalency Mappings (GEMs) was our definitive cross-walk. For one-to-many translations, if the translation alternatives (multiple ICD-9s) mapped to discrepant WHEI conditions, we algorithmically mapped the ICD-10 concordant with the ICD-9 with highest frequency among FY2015 women. Two physicians manually reviewed algorithm results for ICD-10s with frequency > = 100 among FY2016 women to resolve any inconsistencies with original WHEI mappings. To test the algorithm, we compared condition frequencies in adjacent years/coding systems (FY2015/FY2016). Although temporal change is expected, prior work showed most condition frequencies did not change by > 5% across a longer (5-year) period; thus we examined any > 5% absolute change to identify conditions substantially impacted by ICD-9/10 translation.
Results:
Of 72,157 ICD-10s backwards mapped to ICD-9s, 7,961 (11%) had one-to-many GEMs mapping; of those, 3,583 had ambiguous WHEI condition mapping and were re-mapped per the algorithm. Manual review of algorithmically-mapped ICD-10s with frequency > 100 (n = 74) resulted in re-mapping of 28. Another 735 without GEMs ICD-9 equivalents were manually mapped. Comparing across the ICD-9/10 (FY2015/FY2016) divide, only 5 of 202 WHEI conditions frequencies changed > 5%. Review of these 5 showed, for two related conditions, ICD-10's greater specificity led to fewer cases assigned to the non-specific condition (substitution effect). It also revealed limitations of GEMs backwards mapping, with the most notable example being ICD-10 Z72.0 (tobacco use) mapping one-to-one to ICD-9 V69.8 (lifestyle problems) resulting in initial assignment to "Residual Codes" rather than appropriately to "Tobacco use." Such review prompted reassignment of another 102 codes.
Implications:
In VA research and surveillance, existing general equivalency mapping resources can facilitate identifying ICD-10 groupings to replace prior ICD-9 categories, but such efforts must be complemented by careful consideration of the mapping effort's purpose and selective manual review.
Impacts:
Our successful adaptation of 202 WHEI conditions to ICD-10 will allow meaningful comparison of healthcare profiles and temporal changes across the ICD-9/10 eras.