Professor and Director Center for Pharmacoepidemiology and Treatment Science, Rutgers University New Brunswick, United States
Background: The validity of claims-based algorithms for ischemic and hemorrhagic stroke in US Medicare is not well established for the International Classification of Diseases, 10th Revision (ICD-10) coding system, particularly for hemorrhagic stroke.
Objectives: To determine the positive predictive values (PPV) of ICD-10 algorithms for ischemic stroke and hemorrhagic stroke in US Medicare older adults.
Methods: ICD-10 algorithms were constructed by cross-walking established ICD-9 definitions with clinical input (JBS, SS, BLS). Ischemic stroke: Primary discharge code of I63.x (cerebral infarction). Hemorrhagic stroke: Primary discharge diagnosis of I60.x (nontraumatic subarachnoid hemorrhage) or I61.x (nontraumatic intracerebral hemorrhage). Using these claims-based definitions, new stroke events were identified in a 50% national sample of fee-for-service US Medicare older adults (≥age 66) with Part D prescription drug benefits during 2019. Medical charts were retrieved and redacted by a contract research organization and then adjudicated against established clinical diagnostic guidelines (CoreValve and SURTAVI trials definitions) by 3 trained physicians (MK, BL, AT) overseen by the faculty expert (JBS). PPVs and 95% confidence intervals (CIs) were calculated.
Results: Among 38,580 events meeting the definition for ischemic stroke, 103 charts were returned from a random sample of 428 charts sent for retrieval (24% retrieval rate). To date, 44 charts have been adjudicated with a PPV of 81% (70%-93%). We further identified 7,591 patients with primary hemorrhagic stroke codes I60.x or I61.x. 113 charts were returned from a random sample of 478 codes requested (24% retrieval rate). To date, 24 charts have been adjudicated with a PPV of 83% (68%-98%). Finally, we identified an additional 1,539 charts with primary diagnosis code I62.9 (nontraumatic intracranial hemorrhage, unspecified). 105 charts were returned from a random sample of 521 codes requested (20% retrieval rate). To date, 19 charts have been adjudicated with a PPV of 89% (76%-100%). A combined hemorrhagic stroke algorithm (I60.x, I61.x, or I62.9) weighted to account for the population distribution of the component events had a PPV of 84% (70%-99%). Adjudication is ongoing for the remaining charts retrieved.
Conclusions: ICD-10 algorithms for ischemic and hemorrhagic stroke perform well for US Medicare older adults and are adequate for pharmacoepidemiologic studies. We recommended including code I62.9 (nontraumatic intracranial hemorrhage, unspecified) in the algorithm for hemorrhagic stroke.
