Validity study of International Classification of Diseases coding for viral hepatitis: a comparison of different algorithms

Abstract

Background: Many studies used different algorithms of the International Classification of Diseases (ICD) codes to identify patients with hepatitis B and C virus infection (HBV and HCV); yet relatively few validation studies on these algorithms and between Ninth Revision and Tenth Revision (ICD-9-CM and ICD-10-CM) Methods: We systematic randomly sampled 10 000 patients in a healthcare system in southern Taiwan in which 6000 people from the medical center 2500 from the regional hospital and 1500 from the district hospital We reviewed electronic medical record system for years from 2005 to 2019 to determine whether the patient had HBV or HCV based on the prescription records laboratory results and text diagnosis by physicians We then used the number of HBV or HCV ICD codes occurred in outpatient or inpatient claims data to define different algorithms Different indicators of validity such as sensitivity specificity positive predictive value (PPV) and negative predictive value were calculated Lastly we examined associated factors with these indicators and likelihood ratios Results: We identified 726 patients with HBV and 55 patients with HCV The PPV for algorithm HBV ICD codes occurred in outpatient claims data at least 1 2 and 3 times was 72% 77% and 80% respectively and the sensitivity was 72% 69% and 63% respectively The PPV for algorithm HBV ICD codes occurred in inpatient claims data once or outpatient claims data at least 2 3 and 4 times was 77% 80% and 81% respectively and the sensitivity increased to 76% 72% and 69% respectively The PPV for algorithm HCV ICD codes occurred in outpatient claims data at least 1 2 and 3 times was 88% 92% and 94% respectively and the sensitivity was 75% 72% and 68% respectively The PPV for algorithm HCV ICD codes occurred in inpatient claims data once or outpatient claims data at least 2 3 and 4 times was 91% 93% and 94% respectively and the sensitivity increased to 81% 77% and 73% respectively Using algorithm at least 3 times of ICD codes in outpatient claims data as example the PPV from ICD-9-CM to ICD-10-CM was 80% to 90% for HBV and 94% to 99% for HCV The higher the level of hospital and the older the patients and patients visited gastroenterologists had better performance of ICD codes Conclusion: The accuracy of HCV ICD coding was better than HBV coding For both HBV and HCV the accuracy was better in ICD-10-CM than in ICD-9-CM If researchers considered only PPV the algorithm using outpatient claims data with at least 3 times of HBV or HCV ICD codes could reach the relative high PPV level If researchers also consider the sensitivity we suggested of addition of at least 1 time of ICD code in inpatient claims in the algorithm

Date of Award	2021
Original language	English
Supervisor	Tsung-Hsueh Lu (Supervisor)