Back

Validation of International Classification of Diseases, 10th Edition, Clinical Modification (ICD-10-CM) Code I68.0 for Cerebral Amyloid Angiopathy

Bruce, S. S.; Murthy, S.; Kamel, H.

2024-01-17 neurology
10.1101/2024.01.16.24301394 medRxiv
Show abstract

IntroductionCerebral amyloid angiopathy (CAA) is a common cause of intracerebral hemorrhage and cognitive impairment in the elderly. Though definitive diagnosis requires post-mortem pathological analysis, clinical and radiographic criteria allow for noninvasive, in vivo diagnosis. We sought to validate the new ICD-10-CM diagnostic code for CAA with respect to the recently updated Boston criteria, version 2.0. MethodsWe conducted a retrospective study of inpatient and outpatient encounters at a single hospital center, Weill Cornell Medicine (WCM), from 10/1/2015 to 12/31/2018. We randomly selected 25 encounters with the ICD-10-CM code I68.0 and 25 encounters with a primary diagnosis of intracerebral hemorrhage or ischemic stroke, without code I68.0. A single board-certified neurologist, blinded to ICD codes, reviewed detailed medical records and images from brain MRI and CT scans from all 50 selected encounters and identified subjects with possible or probable CAA by Boston criteria 2.0. Sensitivity and specificity of ICD-10-CM code I68.0 was calculated. ResultsOf the 50 selected encounters, 21 (42%) met criteria for possible or probable CAA: 18 (36%) met criteria for probable CAA, 2 (4%) met criteria for probable CAA with supporting pathology, and 1 (2%) met criteria for possible CAA. The ICD-10-CM code I68.0 was found to have a sensitivity of 81% (95% CI, 58-95%) and specificity of 72% (95% CI, 53-87%) for identifying possible or probable CAA. ConclusionThe ICD-10-CM code I68.0 was found to have good sensitivity and moderate specificity for CAA as defined by current clinical and radiographic diagnostic criteria. Based on our results, this code may be useful for identifying patients with CAA in future research using administrative claims data.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.