P-1982. Negative Predictive Value of Methods to Identify Underlying Medical Conditions with and without Use of a Lookback Period among Adults with a Healthcare Encounter for Acute Respiratory Illness, September 2023 – August 2024

Abstract

Background: Vaccine effectiveness (VE) studies are necessary to understand how well vaccines work in the real world. Many VE studies rely on health records to capture underlying medical conditions (UMCs) from a single acute respiratory illness- (ARI) associated encounter, which may bias VE if UMCs are not fully captured. We assessed capture of UMCs from a single acute encounter and a lookback period. Methods: Data were used from MarketScan® Treatment Pathways, a healthcare claims dataset, between September 1, 2023 – August 31, 2024. We included beneficiaries aged ≥18+ years with ≥1 inpatient or emergency department (ED) claim containing an ICD-10 code for ARI who had 3 years of continuous enrollment in a participating insurance plan prior to the date of their first ARI claim (i.e., index encounter). The prevalence of UMCs was calculated using ICD-10 codes from 1) the index encounter, and 2) the 1-year lookback period; and the difference was reported. Negative predictive value (NPV) with 95% exact binomial confidence intervals was calculated for identification of UMCs on the index encounter date, using the 1-year lookback period to define true negatives. Results were stratified by age group and encounter setting. Results: Among 65,056 beneficiaries with ≥1 inpatient ARI event and 162,943 with ≥1 ED ARI event, the most prevalent UMC categories were cardiovascular, endocrine/metabolic, and respiratory (Tables 1-4). Among beneficiaries aged 18–64 years, NPV was < 80% for cardiovascular, endocrine/metabolic, and obesity categories; median difference in prevalence was 9.5 percentage points (pp) (min=0, max=30) (Tables 1 and 3). Among beneficiaries aged ≥65+ years, NPV was ≤80% for respiratory, cardiovascular, neurological and musculoskeletal, endocrine/metabolic, renal, and obesity categories; median difference in prevalence was 13.5 pp (min=0, max=52) (Tables 2 and 4). NPV, regardless of age, was > 90% for UMC categories with the lowest overall prevalence (i.e., cerebrovascular, hematologic, and underweight categories) (Tables 1-4). Conclusion: NPV was < 80% for common UMCs when identified using a single ARI encounter compared to a 1-year lookback period. Misclassification may influence VE estimates if UMCs are confounders in VE studies.