Missing Data Matters in Participatory Syndromic Surveillance Systems: Comparative Evaluation of Missing Data Methods When Estimating Incidence Rate

• Activity Number: 442 - Disease Prediction, Statistical Methods for Genetic Epidemiology and Mis
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Kristin Baltrusaitis* and Craig Dalton and Sandra Carlson and Laura White
• Companies: Harvard T.H. Chan School of Public Health and Hunter New England Population Health and Hunter New England Population Health and Boston University School of Public Health
• Keywords: Missing Data; Missing Not at Random; Multiple Imputation; Digital Epidemiology; Participatory Syndromic Surveillance; Influenza Surveillance

Abstract

Traditional surveillance methods have been enhanced by the emergence of online participatory systems that collect health-related digital data. However, not every volunteer consistently completes surveys. We assess how five missing data methods: available case, complete case, assume missing is non-case, multiple imputation (MI), and delta (?) MI, which is a flexible and transparent MI method under Missing Not at Random (MNAR) assumptions, affect Influenza-Like Illness (ILI) incidence rate (IR) estimates. We evaluate these methods using simulated and Flutracking data. In simulations, the ?-MI method has the smallest normalized root mean square error under MNAR models (NRMSE range: 0.8-12.5), and in sensitivity analyses, the ?-MI method outperforms other methods, under modest changes in the degree of MNAR. For Flutracking, 2018 IR estimates range from 30 to 35 ILI reports per 1000 person-weeks. Missing data is an important problem in participatory systems, and we show that accounting for missingness using statistical approaches leads to different inferences from the data.