This presentation is part of: Poster Session - Clinical Strategies; Judgment and Decison Making
ACCURACY OF DIAGNOSTIC CODES: ONLY AS GOOD AS THE DIAGNOSIS
Mendel E. Singer, PhD1, Kathryn Y. Barzilai, RN1, Robert A. Bonomo, MD2, and Marta N. Jakubowycz, MD2. (1) Case Western Reserve University School of Medicine, Epidemiology and Biostatistics, Cleveland, OH, (2) Louis Stokes Cleveland VA Medical Center, Medicine, Cleveland, OH
PURPOSE: Retrospective analyses rely on the accuracy of diagnostic codes, yet it is unclear in some cases, such as outpatient pneumonia, that the diagnosis itself is reliable. We examined the accuracy of outpatient diagnosis of pneumonia in a Veterans Affairs facility. METHODS: We used an encounter database to identify patients seen at a Cleveland area VA facility during 2001-2002 with a primary diagnosis of pneumonia, who were treated as outpatients. Patients were included if they filled a prescription for an oral antibiotic within 3 days of the index visit. Patients were excluded if they were hospitalized or filled an antibiotic prescription in the last 30 days, or had already been evaluated for the current episode prior to the index visit. Patient records were abstracted by a nurse, recording results of chest x-ray (CXR) and the following symptoms: cough, shortness of breath, pleuritic chest pain, temperature>38, crackles or rales on auscultation. Pneumonia was considered to be “probable” if CXR positive, or unclear with >=2 symptoms; “possible” if unclear CXR and <2 symptoms, or no CXR and >=2 symptoms; “unlikely” if negative CXR, or no CXR and <2 symptoms. Chi-square tests, t-tests were used to compare patients with negative CXR to others. RESULTS: 106 cases were eligible for the study. The mean age was 65 years. Interrater reliability showed agreement in at least 20/21 cases for result of CXR and each of the symptoms of pneumonia. 23/106 patients had a major comorbidity. 88/106 patients had a CXR, with 28 positive, 10 unclear and 50 negative. Likelihood of pneumonia: 31% probable, 17% possible, 52% unlikely. There were no statistically or clinically significant differences between patients with a negative CXR and everyone else. Mean number of pneumonia symptoms was 1.9 in the CXR negative group and 2.2 for everyone else (p=.08). Treatment for patients with negative CXR was 45% quinolones, 39% macrolides vs. 51% quinolones, 37% macrolides for others. A comorbidity was noted in 22% of CXR negatives and 21% of others. CONCLUSIONS: In this study of VA outpatients diagnosed and treated for pneumonia, approximately half were unlikely to have pneumonia. Retrospective analyses of pneumonia should be cautious in their conclusions as the diagnostic codes may accurately reflect the physician’s diagnosis, but may not reflect the true diagnosis.
