Using intelligent agents to repurpose administrative data in fostering disease prevention in an outpatient context: the case of pneumococcal vaccination

Abstract

The use of intelligent agents is proposed as an economical way to repurpose administrative data in order to foster a program of disease prevention in an outpatient context. A retrospective computerized search was conducted using administrative hospital discharge data to identify patients admitted to a medical teaching unit who met the Canadian Immunization criteria for pneumococcal vaccination over a one-year period. For identification of persons eligible for pneumococcal vaccination, administrative discharge data was shown to have a sensitivity of 83%, (confidence interval [CI] 0.73-0.92) and a specificity of 78% CI (0.64-0.91), with a positive predictive value [PPV] of 87%, CI (0.83- 0.90) and a negative predictive value [NPV] of 72%, CI (0.58-0.86). This study demonstrates that administrative data appear promising as the basis for certain clinical applications. Specifically, the reasonably high specificity and sensitivity of diagnostic codes in administrative data could be utilized to trigger appropriate pneumococcal vaccination after hospital discharge among eligible patients who might otherwise never receive this efficacious intervention. Reminder systems in a hospital setting have received mixed results although positive results have been shown in several outpatient settings but using clinical data. Therefore, before a reminder system using administrative data in an outpatient context is implemented it seemed prudent to investigate this issue further.