Congenital heart defects (CHD) are the most common congenital anomaly with a prevalence of 8.6/1000 births in the US. 1 More than 2 million U.S. children and adults are living with CHDs, 2,3 with an estimated cost of care between 5 and 7 billion dollars.4 Despite knowledge that adult CHD prevalence is increasing, 4 and specialty care improves adult survival, 5 there is little reliable data on age-specific mortality, healthcare utilization, comorbidities, or late outcomes among individuals with CHD. Health claims data may not accurately identify the population of interest.6,8,9 We described an age-variable positive predictive value (PPV) for CHD for ICD-9-CM code 745.5 (secundum atrial septal defect),7 based on 2008-2010 data, with implications for the accuracy of claims based surveillance of the CHD population. We propose building on our existing infrastructure of a well-functioning population-based CHD surveillance database, developed under NOFO# DD12-1207 8 and NOFO# DD15-1506, to improve understanding of validity and utility of healthcare claims data for surveillance of individuals of all ages with CHDs, overall and by individual or healthcare characteristics; and to improve accuracy of surveillance of CHDs using healthcare claims data. Our source population is residents of all ages of the metro-Atlanta 5-county area. Data for 5-county metro Atlanta area will be used for estimating prevalence by capture recapture methodology. 9 We plan to validate a minimum of 1500 cases (at least 300 per severity group) identified in healthcare claims data who meet the NOFO# DD19-1902 case definition, through medical record review. Our data sources are healthcare systems that serve the majority of children and adults with CHDs in the 5-county metro-Atlanta area. These data sources have contributed cardiac and non-cardiac encounter level claims data from 2008-2013 to a linked, de-duplicated database of > 13,000 cases with claims data linked to medical rec
ords including cardiac and non-cardiac encounters, of all ages in 5-county metro Atlanta, allowing immediate access. For NOFO# DD19-1902, we propose adding encounter level data for 2014-2017 to the existing dataset. We propose developing algorithms to maximize PPV and sensitivity, by employing machine learning techniques in collaboration with Georgia Institute of Technology to develop, test, and improve algorithms for accurate detection of CHDs in healthcare claims data. Algorithms will be tested in additional data sources such as Georgia Medicaid, and Kaiser Permanente. We will also work in collaboration with the CDC in using Truven Health MarketScan© and Medicaid datasets.
Through validation of healthcare claims data in an integrated surveillance system for CHDs, we aim to: i) improve understanding of the validity and utility of healthcare claims data for surveillance of CHDs; ii) improve the accuracy of identification and surveillance of CHDs; and iii) increase awareness among the public and stakeholders. Achieving these aims should improve decision-making among stakeholders, and ultimately lead to more effective secondary prevention strategies to reduce the public health impact of CHDs.