Diagnostic validity of hospital discharge codes in the ascertainment of amyotrophic lateral sclerosis

Excerpt

In a recent publication, Pisa et al. [1] assessed the validity of hospital discharge data as a method of identifying cases of amyotrophic lateral sclerosis (ALS). After a review of hospital charts from 511 patients, they concluded that a discharge code for ALS had a sensitivity of 93% and a specificity of 99% for the diagnosis of ALS. A detailed exam of the study design, however, suggests that sensitivity and specificity could not be adequately estimated. Sensitivity of a diagnostic test can be defined as the probability that a diseased person in the population tested will be identified as diseased by the test [2]. In the study by Pisa et al., then, sensitivity should be calculated as the probability that a person with ALS in the population of study, probably the reference population for the hospital, had a hospital discharge code of ALS. Therefore, to estimate sensitivity, having an enumeration of all the true cases in the population is necessary. Unfortunately, Pisa et al. do not have this information, since no population-based registry of ALS cases was available. They could not calculate sensitivity (or specificity, for that matter). In fact, with their study design, the only measure of diagnostic accuracy properly calculated is the positive predictive value. To have an unbiased estimate of the negative predictive value, they should have selected a random sample of all hospital discharges, not only discharges with a selected group of diagnostic codes. …