Validation studies of ICD-10 diagnostic codes for identifying common causes of hospitalization: a scoping review

International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes are widely used to identify clinical conditions in administrative health data for epidemiologic research, surveillance, and health services evaluation. However, the validity of ICD-10-coded diagnoses varies across conditions and healthcare settings, and validation studies differ substantially in design, case definitions, and reporting practices.

To characterize validation studies of ICD-10-based case definitions for eight common causes of hospitalization and identify methodological factors affecting the interpretation and comparability of reported performance estimates.

PubMed and Embase (2005-2025) were searched for systematic reviews and primary studies validating ICD-10-coded diagnoses for the following conditions: heart failure, osteoarthritis, pneumonia, diabetes mellitus, acute myocardial infarction, chronic obstructive pulmonary disease, acute kidney injury, and stroke. Data on study characteristics, case definitions, reference standards, and performance metrics were extracted and descriptively summarized.

We identified 116 unique primary publications through 11 systematic reviews and our database search. Stroke and heart failure had the largest evidence base, whereas few validation studies were identified for several conditions. More than one-third of studies reported only positive predictive value, typically reflecting a study design that verified algorithm-identified cases and therefore precluded estimation of sensitivity and specificity. When reported, specificity was generally high, whereas sensitivity varied considerably; however, differences in study design limited comparability of estimates. Incorporating laboratory results, procedures, medications, or other electronic health record data often improved the performance of ICD-10-based case definitions.

Validation studies of ICD-10-coded diagnoses are highly heterogeneous in design and reporting, limiting direct comparison across studies. Many rely on designs that only estimate positive predictive value and therefore provide incomplete assessments of validity. Standardized, population-based validation studies measuring both sensitivity and specificity are needed to improve the reliability and reproducibility of administrative health data for epidemiologic research, surveillance, and quality improvement.