Informatics in Primary Care 2004;12:139–45 © 2004 PHCSG, British Computer Society Introduction A major focus of practice-based research networks (PBRN) is the translation of existing evidence into clinical practice. Through the rigorous evaluation of methods of translating research into practice in community settings, PBRNs have the potential to identify and assess implementation strategies that are most likely to be effective and sustainable. 1 Of Refereed papers A first step towards translating evidence into practice: heart failure in a community practice-based research network Mihai Onofrei PharmD Clinical Pharmacy Specialist Jacquelyn Hunt PharmD Director of Primary Care Research Joseph Siemienczuk MD Chief Medical Officer Providence Health System, Portland, Oregon, USA Daniel R Touchette PharmD MA Assistant Professor, Oregon State University College of Pharmacy, Portland, Oregon, USA Blackford Middleton MD MPH MSc Director of Clinical Informatics Research and Development, Partners HealthCare System and Assistant Professor of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, USA ABSTRACT Objective To determine the validity of an electronic health record (EHR) in the identification of patients with left ventricular dysfunction in a primary care setting. Design A cross-sectional study. Setting Nine clinics participating from the Providence Research Network (PRN) comprising 75 physicians serving approximately 200 000 patients. All clinics utilise the Logician™ EHR for all patient care activities. Patients The study included all PRN patients with an active chart. Interventions All patients with a heart failure diagnosis in the problem list were identified by database query. Left ventricular ejection fraction (LVEF) data were identified through query of local cardiology and hospital echocardiography databases. Additional LVEF data were sought in a manual search of paper charts. Measurements and main results To determine the problem list coding accuracy for a heart failure (HF) diagnosis we evaluated sensitivity, positive predictive value and related derived statistical measures using documented LVEF as the ‘gold standard’. Of 205 755 active PRN patients, 1731 were identified with a problem list entry of HF. Based on comparison with documented LVEF, the sensitivity for problem list entry was 43.9% and 54.4% when HF was defined as an LVEF 55% and 40%, respectively. Conclusion The validity of an EHR problem list entry of HF was poor. The problem list validity could be enhanced through reconciliation with other data sources. Inaccurate EHR problem lists may have clinical consequences, including under- prescribing of beneficial therapies. Keywords: data quality, heart failure, medical record system, quality improvement