Accuracy of ICD-9 Coding for SCD in Children and Adolescents: Results from the Georgia (GA) Rush Surveillance Project

Background: Health services and outcomes research in sickle cell disease (SCD) has been limited by inadequate population-based surveillance and reliance on ICD-9 codes to identify individuals with SCD. The accuracy of ICD-9 coding in administrative claims datasets and the extent of miscoding is unknown. GA was one of 7 states funded to participate in the Registry and Surveillance for Hemoglobinopathies (RuSH) Project. The RuSH case definition for confirmed cases required laboratory documentation of SCD by hemoglobin electrophoresis (HEP)/HPLC. The RuSH case definition for probable cases required identification from administrative datasets of ≥ 2 encounters with an SCD ICD-9 code plus ≥ 1 encounter with ICD-9 or CPT codes from a predetermined list of SCD-associated treatments, procedures, and complications (SCD-TPC): eg hydroxyurea, RBC transfusion, cholecystectomy, stroke. We report results of an ancillary study designed to evaluate the accuracy of using administrative claims data for SCD surveillance by reviewing the medical records (MR) of a large group of children and adolescents identified by SCD ICD-9 codes. Specific objectives were to determine the effect on accuracy of 1) the n of encounters with SCD ICD-9 code for each individual, 2) the addition of SCD-TPC to the case definition, and 3) the length of the surveillance period. We also sought characterize clinical circumstances in which SCD ICD-9 codes were used inaccurately for individuals without SCD.

Methods: GA Medicaid (MC) and State Health Benefits Plan (SHBP) databases were used to identify individuals with ≥1 encounter with SCD ICD-9 codes (282.60-282.69; 282.41-282.42) during 2004-2008. The total n of encounters with SCD ICD-9 codes and encounters with SCD-TPC were determined for both a 5-yr (2004-2008) and 1-yr (2008) period. The MR of the subset of children and adolescents seen at Children's Healthcare of Atlanta (CHOA) or Grady were reviewed. Diagnosis of SCD was confirmed by MR review of clinical and laboratory data, including results of newborn screening (NBS), CBC, reticulocytes (retic), and HEP/HPLC. Criteria used to exclude the diagnosis of SCD included 1) absence of any documentation of SCD in provider notes and 2) NBS, HEP/HPLC, and/or CBC/retic results inconsistent with SCD. Cases with MR inadequate to confirm or exclude SCD were categorized as indeterminate.

Results: During the 5-yr period 2004-2008, 1,998 children and adolescents with ≥ 1 encounter with SCD ICD-9 code in MC or SHBP datasets were seen at CHOA/Grady, 1,474 (72.8%) of whom had ≥ 1 SCD ICD-9 code during 2008. The Table shows the relationship between the length of the surveillance period and n of encounters with SCD ICD-9 codes in MC/SHBP data sets and SCD status determined by MR review.

For the 5-yr surveillance period, the accuracy of ≥ 2 SCD ICD-coded encounters (93.7%) increased to (97.0%) with addition ≥ 1 encounter with SCD-TPC (RuSH probable case definition), but the number of missed cases increased from 28 (1.6%) to 251 (14.2%). For the 1-yr surveillance period, the same comparison increased the accuracy from 97.1% to 98.4%, but the number of missed cases of SCD increased from 53 (3.8%) to 329 (23.7%). Of 196 patients inaccurately identified as SCD by ICD-9 coding, 65 (33.2%) were hemoglobinopathy carriers; 11 (5.6%) had non-sickle hemoglobinopathies (e.g. HbCC), 9 (4.6%) thalassemias, 18 (9.2%) other non-malignant hematologic disorders, and 19 (9.7%) malignant disorders.

Conclusions: Use of administrative claims data to identify children and adolescents with SCD based on ICD-9 coding has important limitations. Accuracy of identification correlated directly with the n of encounters with SCD ICD-9 codes and indirectly with the length of the surveillance period. The addition of ≥ 1 encounter with SCD-TPC to a case definition of ≥ 2 encounters with SCD ICD-9 codes minimally improved specificity but resulted in a large number of missed cases. False positive identification of SCD by ICD-9 coding occurred most commonly in hemoglobinopathy carriers and those with non-sickle hematologic and malignant disorders.