Abstracts

Rationale: Infantile spasms are seizures that frequently occur in the setting of West syndrome – a severe infantile epileptic encephalopathy typically affecting infants three to 12 months of age. Currently, patient identification using claims information has not been performed. Development and validation of a case finding algorithm for infantile spasms in the claims data can be helpful in order to carry out research and quality improvement initiatives on this important patient population. Therefore, we tested different algorithms in a claims database as a pilot study in hopes of identifying patients with infantile spasms with validation performed via chart review. Methods: This study was conducted using Medicaid claims data incurred between July 1st, 2013 and June 30th, 2016 from a pediatric accountable care organization (ACO) affiliated with Nationwide Children’s Hospital (NCH) to identify patients with Infantile Spasms (IS) under 2 years old. We proposed five possible methodologies to identify Medicaid IS patients who received medical care at NCH during the study period. Each method used a different combination of four clinical concepts. The four clinical concepts were (a) “IS Claim” if a patient had an inpatient claim with an IS specific diagnosis code (ICD-9 codes: 345.60 and 345.61 and ICD-10 codes: G40.821, G40.822, G40.823 and G40.824); (b) “EEG” if a patient had an EEG billing code (95816, 95817, 95818, 95819, 95820, 95821, 95822, 95823, 95824, 95825, 95826, 95827, 95828, 95829, 95830, 95831, 95832, 95833, 95834, 95835, 95836, 95837, 95838, 95839, and 95840); (c) “Rx” if a patient had filled a prescription for a medicine typically used for IS (“H.P. Acthar”, “Sabril”, and “Pednisolone”); and (d) “Epilepsy” if a patient had an inpatient claim with any epilepsy diagnosis code (ICD-9 codes 345.xx and ICD-10 codes G40.xxx). The five algorithms tested were (1) IS Claim + EEG + Rx, (2) IS Claim + EEG, (3) IS Claim + Rx, (4) Epilepsy + EEG + Rx, and (5) Epilepsy + Rx.Patients met the following criteria were studied: 1) enrolled in the ACO for at least 6 months during the study period, 2) 0-2 years old at the time of receiving IS-related medical care, 3) patients seen at NCH during the study period. A standard IS patient list was generated using data from the NCH IS patient registry and the hospital electronic medical records. This patient list was used as the gold standard to test the reliability and accuracy of the five proposed methods. Three independent researchers reviewed the standard IS patient list to ensure the accuracy for data validation. We analyzed the sensitivity, specificity, positive predictive, and negative predictive values for the five methods specified above to test each method’s reliability and accuracy. All analyses were conducted in SAS Enterprise Guide 4.3. Institutional Review Board approval at Nationwide Children's Hospital was granted for the purposes of this project. Results: The database searched had 426,041 unique pediatric patients during the study period, and 59,139 of these patients enrolled in the ACO for at least 6 month and received medical care at NCH while they were under 2 years old were included in this study. A total of 32 IS patients confirmed by using the NCH IS registry data and confirmation using the electronic medical records. The number of IS patients identified by the five proposed methods are shown in Table 1. The second method has high sensitivity (91%) and specificity (100 %) than other methods. The other four methods had low sensitivity values (56-59%) although the specificity values were high (100%). The low sensitivity suggests that these methods were unable to identify true positives accurately. For example, the first method only identified 18 IS patients out of 32, the rest of the 14 true IS patients were missed by this method. Therefore, we conclude that the second method, which identifies IS patients through IS specific claim and EEG claim, is the reliable and validate approach to identify IS patients. Conclusions: Using a claims database can accurately identify patients with infantile spasms. The algorithm that had the highest sensitivity was one using the IS specific ICD diagnosis code plus EEG. We chose the codes for ICD and the CPT codes for EEG to allow ease for replication in any claims database. Limitations include the use of a small cohort of patients from only one institution. Further work is needed in a larger set of data from multiple geographic locations to confirm these findings. Using electronic health record to validate findings using claims data is imperative ensure success in similar work. Funding: No funding was received to support this abstract.