American Journal of Medical Quality · 2004 · 166 citations · 17 references
MedicineCardiologyInternal MedicineClinical CardiologyClinical Information SystemsDiagnostic ReasoningBlood Pressure ManagementCardiovascular DiseaseComorbidityComorbidity AssessmentComorbidity CaptureHypertension ManagementAdministrative Data SourcesComorbidity AnalysisComorbidity Coding AccuracyComorbidity Measurement MethodologyHypertension DiagnosisHypertension ScreeningTobacco Use ComorbidityTobacco Use Patterns
The study aimed to identify the optimal approach for detecting hypertension‑related comorbidities among VA outpatients using administrative data. Researchers reviewed 1,176 outpatient charts from 10 VA sites in 1999, treating 11 hypertension‑related diagnoses as the gold standard, and evaluated agreement, sensitivity, and specificity of multiple administrative algorithms. The most effective case‑finding strategy required at least two diagnoses within a two‑year window, yielding high agreement (0.85–0.98), sensitivity ranging 38–97%, and specificity above 91% for most conditions.
The objective was to determine the best strategy for identifying outpatients with hypertension-related diagnoses using Veterans Affairs (VA) administrative databases. We reviewed 1176 outpatient charts from 10 VA sites in 1999, taking the presence of 11 diagnoses relevant to hypertension management as the "gold standard" for identifying the comorbidity. We calculated agreement, sensitivity, and specificity for the chart versus several administrative data-based algorithms. Using 1999 data and requiring 1 administrative diagnosis, observed agreement ranged from 0.98 (atrial fibrillation) to 0.85 (hyperlipidemia), and kappas were generally high. Sensitivity varied from 38% (tobacco use) to 97% (diabetes); specificity exceeded 91% for 10 of 11 diagnoses. Requiring 2 years of data and 2 diagnoses improved most measures, with minimal sensitivity decrease. Agreement between the database and charts was good. Administrative data varied in its ability to identify all patients with a given diagnosis but identified accurately those without. The best strategy for case-finding required 2 diagnoses in a 2-year period.
17
The Measurement of Observer Agreement for Categorical Data
J. Richard Landis, Gary G. Koch · Biometrics · 1977 · 78.9K citations · Full text