Accurate Identification of Sickle Cell Disease Cases in a Large Genotyped Cohort Using Electronic Health Record Data

The Food and Drug Administration recently approved sickle cell disease (SCD) gene therapies that require a contemporaneous cohort to compare their outcomes to those of individuals who did not receive gene therapy. Previously, we developed an automated algorithm to create a contemporaneous cohort of children and adults with SCD using electronic health record (EHR) data. We tested the hypothesis that the updated VUMC algorithm can identify individuals with SCD with sensitivity and specificity greater than 95%. In the Vanderbilt University Medical Center (VUMC) healthcare system, we identified a cohort of 33,141 adults and children of primarily African ancestry who underwent HBB sequencing, with a 1.4% prevalence of SCD (473/33,141). For SCD, the performance of the VUMC algorithm using ICD and laboratory data showed a sensitivity of 97.6% (462/473), specificity of 99.9% (32,663/32,668), positive predictive value (PPV) of 98.9% (462/467), and negative predictive value (NPV) of 99.9% (32,663/32,674). Three SCD phenotypes account for 98.5% of the individuals in the cohort. The cohort had sensitivities of 96.8% (272/281), 93.7% (136/145), and 85% (34/40) for HbSS/HbSß0, HbSC, and HbSβ+, respectively. The cohort had a specificity of 99.9% across all three phenotypes: HbSS/HbSß0, HbSC, and HbSβ+, respectively. We have developed an updated algorithm using ICD and laboratory data that accurately identifies individuals with SCD within a large healthcare system, and distinguishes HbSS/HbSß0, HbSC, and HbSβ+.