Purpose Many epidemiological studies in pediatric heart failure (HF) use administrative databases, defining patient cohorts by the presence of a single HF ICD code. However, the precision of ICD codes to identify true HF patients is unknown in pediatrics. The study aim is to describe the accuracy of HF ICD-10 codes alone and in combination with medication billing codes, in identifying pediatric patients with HF through electronic health records. Methods Clinical notes were reviewed for 378 patients 18 years old and younger in a pediatric HF clinic at a tertiary care medical center within 01/2016-12/2020. The initial outpatient note for each patient was reviewed. Patients were adjudicated as having HF if the provider note: 1) stated the patient had current/prior HF or was receiving HF-specific treatment or 2) described the patient as NYHA/Ross Class II-IV HF or ACC/AHA HF Stage C or D. ICD-10 codes for HF and medication billing codes during the month before and 6 months after the clinic encounter were obtained. Algorithms were developed to identify HF patients based on the presence of 1 or more inpatient or outpatient HF ICD codes, in combination with HF medications. Using the clinical note review as the adjudicated positive HF diagnosis, algorithms were assessed by sensitivity, specificity, positive and negative predictive values. For inpatient algorithms that could be applied to administrative databases, receiver operating characteristics curves were generated and accuracy reported. Results Median age was 10.1 years (IQR 4.5,14.8) with 39/378 adjudicated HF patients. Algorithm performance is shown in Table 1. Conclusion This is the first study to describe characteristics of ICD code algorithms for case ascertainment in pediatric HF. Results suggest that a combination of an ICD code and carvedilol can identify a cohort of HF patients however many patients will be missed. Natural language processing approaches may improve the accuracy of patient identification for pediatric HF outcomes research, and should be tested. Many epidemiological studies in pediatric heart failure (HF) use administrative databases, defining patient cohorts by the presence of a single HF ICD code. However, the precision of ICD codes to identify true HF patients is unknown in pediatrics. The study aim is to describe the accuracy of HF ICD-10 codes alone and in combination with medication billing codes, in identifying pediatric patients with HF through electronic health records. Clinical notes were reviewed for 378 patients 18 years old and younger in a pediatric HF clinic at a tertiary care medical center within 01/2016-12/2020. The initial outpatient note for each patient was reviewed. Patients were adjudicated as having HF if the provider note: 1) stated the patient had current/prior HF or was receiving HF-specific treatment or 2) described the patient as NYHA/Ross Class II-IV HF or ACC/AHA HF Stage C or D. ICD-10 codes for HF and medication billing codes during the month before and 6 months after the clinic encounter were obtained. Algorithms were developed to identify HF patients based on the presence of 1 or more inpatient or outpatient HF ICD codes, in combination with HF medications. Using the clinical note review as the adjudicated positive HF diagnosis, algorithms were assessed by sensitivity, specificity, positive and negative predictive values. For inpatient algorithms that could be applied to administrative databases, receiver operating characteristics curves were generated and accuracy reported. Median age was 10.1 years (IQR 4.5,14.8) with 39/378 adjudicated HF patients. Algorithm performance is shown in Table 1. This is the first study to describe characteristics of ICD code algorithms for case ascertainment in pediatric HF. Results suggest that a combination of an ICD code and carvedilol can identify a cohort of HF patients however many patients will be missed. Natural language processing approaches may improve the accuracy of patient identification for pediatric HF outcomes research, and should be tested.