Objective: To compare the fidelity of administrative data with clinical data when researching Down syndrome (DS).
Methods: From outpatient, inpatient, and emergency department administrative claims within our institution, we identified 252 patients aged 18-45 years with encounters coded for DS by the ICD9=758.0 from 2000 to 2008. We evaluated these cases for false-positive errors-cases in which DS was not actually present in clinical descriptions. Subsequently, we identified false-negative errors (cases in which DS was present without encounters coded as such) by examination of the medical records for all patients within our study frame who had one of several common DS comorbidities, including congenital heart disease, hypothyroidism, and atlantoaxial instability.
Results: Among the 252 people with an administrative code for DS, 53 (21%) did not have DS documented in their medical record (false-positive error). While searching for false-negative errors, 29 additional patients were discovered with DS documented in the medical record who had not been previously identified. This led to a final cohort of 228 patients with DS. The presence of a billing code for DS had moderate sensitivity (87%) and positive predictive value (79%), but high specificity (99.9%).
Discussion: Administrative claims misclassify a sizeable proportion of patients with DS. Judgments about quality of care on the basis of samples identified using administrative claims may not accurately reflect the experience of patients with the conditions in question. When using administrative databases to study the quality of care for patients with DS, diagnostic verification within the clinical record is advisable whenever possible.